Metallurgical Process Simulation and Optimization: Qing Liu and Jiangshan Zhang

Metallurgical
Process Simulation
and Optimization
Edited by
Qing Liu and Jiangshan Zhang
Printed Edition of the Special Issue Published in Materials
www.mdpi.com/journal/materials
Metallurgical Process Simulation and
Optimization
Optimization
Editors
Qing Liu
Jiangshan Zhang
MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tokyo • Cluj • Tianjin
Editors
Qing Liu Jiangshan Zhang
University of Science and University of Science and
Technology Beijing Technology Beijing
China China
Editorial Office
MDPI
St. Alban-Anlage 66
4052 Basel, Switzerland
This is a reprint of articles from the Special Issue published online in the open access journal
Materials (ISSN 1996-1944) (available at: https://www.mdpi.com/journal/materials/special issues/
metallurgical process).
For citation purposes, cite each article independently as indicated on the article page online and as
indicated below:
LastName, A.A.; LastName, B.B.; LastName, C.C. Article Title. Journal Name Year, Volume Number,
Page Range.
ISBN 978-3-0365-6430-2 (Hbk)

ISBN 978-3-0365-6431-9 (PDF)
Cover image courtesy of Jiangshan Zhang and Wuhu Xinxing Ductile Iron Pipes Co., Ltd.
© 2023 by the authors. Articles in this book are Open Access and distributed under the Creative
Commons Attribution (CC BY) license, which allows users to download, copy and build upon
published articles, as long as the author and publisher are properly credited, which ensures maximum
dissemination and a wider impact of our publications.
The book as a whole is distributed by MDPI under the terms and conditions of the Creative Commons
license CC BY-NC-ND.
Contents
About the Editors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
Preface to ”Metallurgical Process Simulation and Optimization” . . . . . . . . . . . . . . . . . ix
Jiangshan Zhang, Yuhong Liu and Qing Liu

Metallurgical Process Simulation and Optimization
Reprinted from: Materials 2022, 15, 8421, doi:10.3390/ma15238421 . . . . . . . . . . . . . . . . . . 1
Md Irfanul Haque Siddiqui, Ayidh Albaqami, Latif Arifudin, Khalid Alluhydan
andIbrahimAbdullahAlnaser
Simulation of Inclusion Particle Motion Behavior under Interfacial Tension in Continuous
CastingMold
ZhanpengTie, HaiyanTang, KaiminWang, HongshengMiao, SenCai, FenqiangXian

andJiaquanZhang
Effect of Flow Field Optimization of an Asymmetric Multi-Strand Tundish on the Quality
ConsistencyofCrackingCon-RodSteel
Peng Zhao, Rongxun Piao and Zongshu Zou

Mesoscopic Fluid-Particle Flow and Vortex Structural Transmission in a Submerged Entry
Nozzle of Continuous Caster
Guoliang Liu, Haibiao Lu, Bin Li, Chenxi Ji, Jiangshan Zhang, Qing Liu and Zuosheng Lei
Influence of M-EMS on Fluid Flow and Initial Solidification in Slab Continuous Casting
Michał Moskal, Piotr Migas and Mirosław Karbowniczek

Multi-Parameter Characteristics of Electric Arc Furnace Melting
Shaopei Duan, Baokuan Li and Wenjie Rong

Numerical Simulation Study of Gas-Solid Heat Transfer and Decomposition Processes of
Limestone Calcined with Blast Furnace Gas in a Parallel Flow Regenerative Lime Kiln
Xiaodong Deng, Jianli Li and Xiao Xie

Effect of Preheating Temperature on Thermal–Mechanical Properties of Dry Vibrating
MgO-Based Material Lining in the Tundish
Yutang Li, Linzhu Wang, Chaoyi Chen, Shufeng Yang and Xiang Li
New Insights into the Mechanism of Nucleation of ZrO2 Inclusions at High Temperature
Yizhe Du, Zhidan Huang, Mujun Long, Huamei Duan and Dengfu Chen
Unveiling the Effect of CaF2 on the Microstructure and Transport Properties of Phosphosilicate
Systems
v
Rongxun Piao, Wenjin Zhu, Lan Ma, Peng Zhao and Biao Hu
Characterization of Hot Deformation of near Alpha Titanium Alloy Prepared by TiH2 -Based
Powder Metallurgy
Tong Wang, David Wexler, Liangliang Guo, Yangfan Wang and Huijun Li
In Situ Observation and Phase-Field Simulation Framework of Duplex Stainless-Steel Slab
during Solidification
Lei Liu, Xiuli Han, Mingduo Li and Di Zhang

Influence of Mineralogical Structure of Mold Flux Film on Heat Transfer in Mold during
Continuous Casting of Peritectic Steel
Jun Fu, Qing Tao, Xiaoan Yang, Bogdan Nenchev, Ming Li, Biao Tao and Hongbiao Dong
The Effect of Heat Source Path on Thermal Evolution during Electro-Gas Welding of Thick Steel
Plates
Paweł Malinowski, Justyna Kasińska, Sławomir Rutkowski and Monika Madej

Exploratory Data Analysis for the Evaluation of Tribological Properties of Wear-Resistant
Surface Layers Modified with Rare-Earth Metals
Liping Wu, Jianguo Zhi, Jiangshan Zhang, Bo Zhao and Qing Liu
Effect of Cerium on the Microstructure and Inclusion Evolution of C-Mn Cryogenic Vessel Steels
Krzysztof Żaba, Tomasz Trzepieciński, Stanislav Rusz, Sandra Puchlerska and Maciej
Balcerzak
Full-Field Temperature Measurement of Stainless Steel Specimens Subjected to Uniaxial Tensile
Loading at Various Strain Rates
Xiaoshuang Liu, Peng Zhang, Baoyi Wang, Xingzhong Cao, Shuoxue Jin and Runsheng Yu
Study of Interaction Mechanism between Positrons and Ag Clusters in Dilute Al–Ag Alloys at
Low Temperature
Yichao Hu, Yinxuan Qiu, Jian Chen, Liangyuan Hao, Thomas Edward Rufford,
VictorRudolphandGeoffWang
Integrating a Top-Gas Recycling and CO2 Electrolysis Process for H2-Rich Gas Injection and
ReduceCO2 EmissionsfromanIronmakingBlastFurnace
vi
About the Editors
Qing Liu
Qing Liu is a Professor at the University of Science and Technology Beijing. His research
interests include the fine control of the solidification process in continuous casting, metallurgical
process engineering, simulation and optimization of metallurgical process, and highly efficient
utilization of metallic resources. He has published around 200 peer-reviewed papers.
Jiangshan Zhang
Jiangshan Zhang is an Associate Professor at the University of Science and Technology Beijing.
His research interests include advanced material manufacturing and process metallurgy simulation.
He has published around 50 peer-reviewed papers.
vii
Preface to ”Metallurgical Process Simulation and
Optimization”
Metallurgy involves the art and science of extracting metals from their ores and modifying the
metals for use. With thousands of years of development, many interdisciplinary technologies have
been introduced into this traditional and large-scale industry. In modern metallurgical practices,
modelling and simulation are widely used to provide solutions in the areas of design, control,
optimization, and visualization, and are becoming increasingly significant in the progress of digital
transformation and intelligent metallurgy.
This Special Issue (SI), entitled “Metallurgical Process Simulation and Optimization”, has been
organized as a platform to present the recent advances in the field of modelling and optimization
of metallurgical processes, which covers the processes of electric/oxygen steel-making, secondary
metallurgy, (continuous) casting, and processing. Both fundamental insights and practical foresights
are presented, including the topics of thermodynamics, kinetics, physical modelling, numerical
simulation, CFD, 3D visualization, microstructural evolution, quality control, artificial intelligence,
big data, and cloud computation. Eighteen articles have been included that cover various aspects of
the topic.
We would like to take this opportunity to express our most profound appreciation to the MDPI
Book staff; the editorial team of the Materials journal, especially Mr. Felix Guo; the Assistant Editor
of this Special Issue; the talented authors; and the hardworking and professional reviewers.
Qing Liu and Jiangshan Zhang

Editors
ix
materials
Editorial
Metallurgical Process Simulation and Optimization
Jiangshan Zhang *, Yuhong Liu and Qing Liu *
State Key Laboratory of Advanced Metallurgy, University of Science and Technology Beijing,
Beijing 100083, China
* Correspondence: [email protected] (J.Z.); [email protected] (Q.L.); Tel.: +86-10-8237-5255 (Q.L.)
Metallurgy involves the art and science of extracting metals from their ores and mod-
ifying them for use. Over thousands of years of development, many interdisciplinary
technologies have been introduced into this large-scale, traditional industry. In mod-
ern metallurgical practices, modeling and simulation have been widely used to provide
solutions for design, control, optimization, and visualization, and they are increasingly
significant in the progress of digital transformation and intelligent metallurgy. This Special
Issue (SI), “Metallurgical Process Simulation and Optimization”, presents recent advances
in the modeling and optimization of metallurgical process, and includes nearly twenty
articles covering various aspects of the topic. A second volume of this successful SI is
being organized in which researchers from both academia and industry are invited to
publish their new work. The purpose of the current editorial is to briefly summarize the
publications included in this SI.
Advanced modeling methods have been widely used to simulate the fluid flow, heat
transfer, and mass transfer behavior of metallurgical vessels, and are usually coupled
with species, multiphase, and electromagnetic fields to optimize their design/operation.
The effect of interfacial tension on alumina inclusion motion behavior was studied in
a continuous casting mold using a commercial CFD software package, as reported by
Siddiqui et al. [1]. It was found that inclusions were vulnerable to engulfment by the
solidification front under a high-surface-tension gradient. Tie et al. [2] shed light on the
Citation: Zhang, J.; Liu, Y.; Liu, Q.
fluid flow inside an asymmetric multi-strand tundish using both numerical and physical
simulation, and the flow field was optimized to improve the overall quality of bloom
Optimization. Materials 2022, 15, castings and con-rod products. In Zhao et al.’s [3] numerical study of mesoscopic fluid-
8421. https://doi.org/10.3390/ particle flow, a bifurcated pool-type SEN under steady operating conditions was employed
ma15238421 using the lattice Boltzmann method (LBM) coupled with the large eddy simulation (LES)
model with validation. Liu et al. [4] investigated the influence of M-EMS on fluid flow and
Received: 16 November 2022
initial solidification in slab continuous casting; a model coupled with an electromagnetic
Accepted: 22 November 2022
field was developed to simulate the transient turbulence flow and initial solidification, with
Published: 26 November 2022
special consideration of the effects of different electromagnetic stirring (EMS) currents and
Publisher’s Note: MDPI stays neutral casting speeds. The function of an Electric Arc Furnace is essentially to melt metals, and
with regard to jurisdictional claims in Moskal et al. [5] optimized the melting process using statistical-thermodynamic modeling
published maps and institutional affil- based on, among other things, multiple linear regression (MLR). In the work of Duan
iations. et al. [6], the gas–solid heat transfer and decomposition processes of limestone calcined
with blast furnace gas were studied in a parallel-flow regenerative lime kiln; a Porous
Medium Model (PMM) and a Shrinking Core Model (SCM) were used to examine the
feasibility of calcining limestone with low-calorific fuel gas. The software ANSYS was used
Copyright: © 2022 by the authors.
in Deng et al.’s [7] study to simulate the temperature field and stress–strain distribution
Licensee MDPI, Basel, Switzerland.
This article is an open access article
on the working layer of a four-strand tundish under three preheating stages through the
distributed under the terms and
indirect coupling method.
conditions of the Creative Commons State-of-the-art characterization techniques and experimental methods enable more
Attribution (CC BY) license (https:// detailed and accurate insights into the metallurgical process, and provide enriched in-
creativecommons.org/licenses/by/ formation for modeling and process optimization. Li et al. [8] examined the nucleation
4.0/). process of zirconium oxide inclusions in steel using classical nucleation theory and first
Materials 2022, 15, 8421. https://doi.org/10.3390/ma15238421 1 https://www.mdpi.com/journal/materials

Materials 2022, 15, 8421
principles, accompanied by SEM, TEM, and XRD characterization methods. An optimized

nucleation pathway of ZrO2 was achieved at a high temperature. Molecular dynamics
simulations (MD) were combined with experiments in Du et al.’s work [9], which revealed
the existence of the form and functional mechanism of CaF2 in phosphosilicate systems.
The hot deformation characteristics of Ti alloys were analyzed using the strain hardening
exponent, strain rate sensitivity, a processing map, and microstructure observation in Piao
et al.’s study [10], and an optimum processing temperature and strain rate were proposed.
Wang et al. [11] developed an in situ observation method to characterize the solidification of
duplex steel during solidification, and phase-field simulation was also carried out to study
its solidification and heat transfer behavior. The influence of the mineralogical structure
of mold flux film on heat transfer was studied during continuous casting of peritectic
steel in the work of Liu et al. [12], which covers the layered structure, crystallization ratio,
mineralogical species, and morphology features of flux films. Optimized mineral phase
structures of flux film were put forward to eliminate longitudinal cracks. An electro-gas
welding experiment on thickness E36 steel plates was conducted in Fu et al.’s study [13]. A
semi-ellipsoid heat source model was developed using linear, sinusoidal, or oscillate-stop
paths. Different heating paths were recommended for corresponding thicknesses of steel
plate. Malinowski et al. [14] investigated the beneficial effect of rare-earth metal oxides on
the wear resistance of surface layers applied to castings, which are intended for structural
elements of machinery and equipment in mining and recycling. Exploratory data analysis
was carried out for the evaluation of tribological properties, and improvements were made
to the modified surface wear resistance. The effects of another rare-earth metal, Cerium
(Ce), were studied on the casting slab quality, microstructure and inclusions of cryogenic
vessel steel in Wu et al.’s work [15]. Moreover, Zaba et al. [16] used full-field image corre-
lation and infrared thermography techniques to study the effects of strain rate, specimen
orientation, and plastic strain on the value and distribution of temperature in dog-bone
stainless-steel specimen deformation in uniaxial tensile tests. In addition, Liu et al. [17] used
positron annihilation lifetime spectroscopy (PALS), high-resolution transmission electron
microscopy (HRTEM), and positron annihilation Coincidence Doppler broadening (CDB)
techniques synergistically to study the microstructural evolution of alloys in their early
aging stages and at low temperatures. Moreover, an optimized measurement temperature
was proposed.
In addition, thermodynamic analysis of H2 behavior was conducted in an iron-making
blast furnace at different stages of the process of integrating top-gas recycling and CO2
electrolysis for H2 -rich gas injection, as discussed in Hu et al.’s study [18]; this is of
fundamental importance for better performance in H2 metallurgy.
Funding: This research was funded by the National Natural Science Foundation of China (No.
U21A20112, 52004024).
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Siddiqui, M.I.H.; Albaqami, A.; Arifudin, L.; Alluhydan, K.; Alnaser, I.A. Simulation of Inclusion Particle Motion Behavior under
Interfacial Tension in Continuous Casting Mold. Materials 2022, 15, 7458. [CrossRef] [PubMed]
2. Tie, Z.; Tang, H.; Wang, K.; Miao, H.; Cai, S.; Xian, F.; Zhang, J. Effect of Flow Field Optimization of an Asymmetric Multi-Strand
Tundish on the Quality Consistency of Cracking Con-Rod Steel. Materials 2022, 15, 3698. [CrossRef] [PubMed]
3. Zhao, P.; Piao, R.; Zou, Z. Mesoscopic Fluid-Particle Flow and Vortex Structural Transmission in a Submerged Entry Nozzle of
Continuous Caster. Materials 2022, 15, 2510. [CrossRef] [PubMed]
4. Liu, G.; Lu, H.; Li, B.; Ji, C.; Zhang, J.; Liu, Q.; Lei, Z. Influence of M-EMS on Fluid Flow and Initial Solidification in Slab
Continuous Casting. Materials 2021, 14, 3681. [CrossRef] [PubMed]
5. Moskal, M.; Migas, P.; Karbowniczek, M. Multi-Parameter Characteristics of Electric Arc Furnace Melting. Materials 2022, 15, 1601.
[CrossRef] [PubMed]
6. Duan, S.; Li, B.; Rong, W. Numerical Simulation Study of Gas-Solid Heat Transfer and Decomposition Processes of Limestone
Calcined with Blast Furnace Gas in a Parallel Flow Regenerative Lime Kiln. Materials 2022, 15, 4024. [CrossRef] [PubMed]
2
Materials 2022, 15, 8421
7. Deng, X.; Li, J.; Xie, X. Effect of Preheating Temperature on Thermal-Mechanical Properties of Dry Vibrating MgO-Based Material
Lining in the Tundish. Materials 2022, 15, 7699. [CrossRef] [PubMed]
8. Li, Y.; Wang, L.; Chen, C.; Yang, S.; Li, X. New Insights into the Mechanism of Nucleation of ZrO2 Inclusions at High Temperature.
Materials 2022, 15, 7960. [CrossRef]
9. Du, Y.; Huang, Z.; Long, M.; Duan, H.; Chen, D. Unveiling the Effect of CaF2 on the Microstructure and Transport Properties of
Phosphosilicate Systems. Materials 2022, 15, 7916. [CrossRef]
10. Piao, R.; Zhu, W.; Ma, L.; Zhao, P.; Hu, B. Characterization of Hot Deformation of near Alpha Titanium Alloy Prepared by
TiH2 -Based Powder Metallurgy. Materials 2022, 15, 5932. [CrossRef] [PubMed]
11. Wang, T.; Wexler, D.; Guo, L.; Wang, Y.; Li, H. In Situ Observation and Phase-Field Simulation Framework of Duplex Stainless-Steel
Slab during Solidification. Materials 2022, 15, 5517. [CrossRef] [PubMed]
12. Liu, L.; Han, X.; Li, M.; Zhang, D. Influence of Mineralogical Structure of Mold Flux Film on Heat Transfer in Mold during
Continuous Casting of Peritectic Steel. Materials 2022, 15, 2980. [CrossRef] [PubMed]
13. Fu, J.; Tao, Q.; Yang, X.; Nenchev, B.; Li, M.; Tao, B.; Dong, H. The Effect of Heat Source Path on Thermal Evolution during
Electro-Gas Welding of Thick Steel Plates. Materials 2022, 15, 2215. [CrossRef] [PubMed]
14. Malinowski, P.; Kasińska, J.; Rutkowski, S.; Madej, M. Exploratory Data Analysis for the Evaluation of Tribological Properties of
Wear-Resistant Surface Layers Modified with Rare-Earth Metals. Materials 2022, 15, 2032. [CrossRef] [PubMed]
15. Wu, L.; Zhi, J.; Zhang, J.; Zhao, B.; Liu, Q. Effect of Cerium on the Microstructure and Inclusion Evolution of C-Mn Cryogenic
Vessel Steels. Materials 2021, 14, 5262. [CrossRef] [PubMed]
16. Żaba, K.; Trzepieciński, T.; Rusz, S.; Puchlerska, S.; Balcerzak, M. Full-Field Temperature Measurement of Stainless Steel Specimens
Subjected to Uniaxial Tensile Loading at Various Strain Rates. Materials 2021, 14, 5259. [CrossRef] [PubMed]
17. Liu, X.; Zhang, P.; Wang, B.; Cao, X.; Jin, S.; Yu, R. Study of Interaction Mechanism between Positrons and Ag Clusters in Dilute
Al–Ag Alloys at Low Temperature. Materials 2021, 14, 1451. [CrossRef] [PubMed]
18. Hu, Y.; Qiu, Y.; Chen, J.; Hao, L.; Rufford, T.E.; Rudolph, V.; Wang, G. Integrating a Top-Gas Recycling and CO2 Electrolysis
Process for H2-Rich Gas Injection and Reduce CO2 Emissions from an Ironmaking Blast Furnace. Materials 2022, 15, 2008.
[CrossRef] [PubMed]
3
materials
Article
Simulation of Inclusion Particle Motion Behavior under
Interfacial Tension in Continuous Casting Mold
Md Irfanul Haque Siddiqui 1 , Ayidh Albaqami 1, *, Latif Arifudin 1 , Khalid Alluhydan 1
and Ibrahim Abdullah Alnaser 1,2
1 Department of Mechanical Engineering, King Saud University, Riyadh 11451, Saudi Arabia
2 Center of Excellence for Research in Engineering Material (CEREM), King Saud University,
Riyadh 11451, Saudi Arabia
* Correspondence: [email protected]
Abstract: Inclusions entrapped by the solidifying front during continuous casting adversely affect
the properties of the final steel products. In this study, we investigated the effect of the interfacial
tension due to surfactant concentration, particularly sulfur, on alumina inclusion motion behavior
during molten steel solidification in a continuous casting mold. A two-dimensional numerical
model was developed in Ansys Fluent software to simulate the inclusion motion in a continuous
casting mold. Further, the impacts of different values of the alumina inclusion diameter, sulfur
concentration, and melt temperature were studied to understand the inclusion motion behavior.
The inclusion diameter affected the inclusion distribution throughout the domain. The alumina
inclusion entrapment percentage varied in the case of sulfur mixing (using an empirical relationship
for modeling). It was found that the removal percentage varied according to the sulfur concentration.
The addition of sulfur at concentrations from 10 ppm to 70 ppm resulted in a 4% increase in the
removal of alumina inclusions (trapped in the solidifying shell), except for the 100-ppm case. Smaller-
Citation: Siddiqui, M.I.H.; Albaqami, sized inclusion particles had a 25% higher chance of entrapment at the top level of the mold. Under
A.; Arifudin, L.; Alluhydan, K.; the effect of a higher surface tension gradient between inclusions and the melt, the predicted findings
Alnaser, I.A. Simulation of Inclusion show that inclusions were vulnerable to engulfment by the solidification front.
Particle Motion Behavior under
Interfacial Tension in Continuous Keywords: alumina inclusion; interfacial tension; steelmaking; continuous casting mold; surfactant;
Casting Mold. Materials 2022, 15, sulfur; simulation
7458. https://doi.org/10.3390/
ma15217458
Academic Editors: Qing Liu and

Jiangshan Zhang
1. Introduction
Steel’s mechanical properties and chemical composition are continually changing, and
Received: 25 July 2022
worries about cost, energy, and the environment in the manufacturing process are becoming
Accepted: 30 August 2022
increasingly relevant. To fulfill demand, steel’s strength, ductility, durability, and corrosion
Published: 24 October 2022
resistance have all increased over time. A variety of processes, strongly influenced by
Publisher’s Note: MDPI stays neutral the equipment, modify the final product’s shape, appearance, and properties [1]. Due
with regard to jurisdictional claims in to its superior features, steel has been widely used for various applications since it was
published maps and institutional affil- commercialized at the beginning of the 1900s. The need for clean steel is an important
iations. aspect for steel producers in the context of greenhouse emissions, competition, customer
requirements, and sustainability. It has been reported widely that nonmetallic inclusions
significantly influence the properties of steel products. The mold remains the last stage in
the continuous casting process, where the proper implementation of processes can lead
to significant inclusion removal. To properly control nonmetallic inclusions during the
continuous casting process, it is of the utmost importance to understand the inclusion
motion behavior in the continuous casting mold [2–4].
conditions of the Creative Commons
1.1. Inclusions in the Continuous Casting Mold
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/ There are two kinds of nonmetallic inclusions in steel, each with its unique mode of
4.0/). formation. Indigenous oxide inclusions, which form when the steel melt is deoxidized, are

Materials 2022, 15, 7458
one kind. The bulk of these oxides is removed from the melt during refining and degassing
in the ladle, but some nonmetallic oxide inclusions remain. When the deoxidized steel
melt is reoxidized by air or entrained slag is introduced into the melt during the melt
transfer from the ladle to the mold, exogenous inclusions occur. Exogenous inclusions are
generally much larger than native inclusions and hence more harmful. Inclusions cause
problems during the casting, rolling, and heat treatment processes, and they may cause the
steel to break during use. The size and nature of nonmetallic inclusions that decrease steel
properties are not constant but rather fluctuate according to the application [5–9].
Typically, in the secondary steelmaking process, oxygen impurities are removed by
aluminum deoxidation. However, alumina inclusions are formed during the deoxidation
process. The steel components and structure require control of alumina inclusions to
ensure their design performance. Inadequate removal of alumina inclusions during the
steelmaking process can induce micro-surface cracks during continuous casting and hot
rolling [10,11]. Generally, solid nonmetallic inclusions are not well-wetted with liquid
steel. Inclusions distributed in liquid steel are sometimes engulfed in the solidification
interface in the mold and become defects in steel products. Inclusions entrapped by the
solidifying front during continuous casting will adversely affect the properties of the final
steel products [1,12–14]. Therefore, the behavior of inclusions at the interface should be
investigated in detail.
1.2. Inclusion Motion Characteristics

The interfacial tension, which is also referred to as the pushing and engulfment phe-
nomenon, has an impact on the distribution of inclusion particles trapped in the solidifying
slab of the continuous casting mold. The interfacial characteristics are connected to the
pushing and engulfment of inclusion particles [15–17]. Particles move from one position
with higher interfacial tension to another position with lower interfacial tension when there
is a gradient of interfacial tension around a gas bubble or a small liquid inclusion. The
Marangoni force’s impact on inclusions at the solid–liquid border is seen in Figure 1. As is
the case for pure metals, surface tension typically decreases as temperature rises. The pres-
ence of a surface-active element in a metal is not necessarily a given, though. To regulate
Marangoni convection in the molten metal and the subsequent solidification in the mold,
the concentration of minor elements in steels is important [18]. Some researchers have used
bubbles to conduct extensive research on this phenomenon in water model experiments.
Mukai et al. [15–17] reported that the pushing and engulfment of small bubbles occur in
the solid–liquid interface due to the presence of a gradient of interfacial tension in the
boundary layer. Further, Shibata et al. [19] studied the inclusion/bubble motion behavior
at the solid–liquid interface of the solidifying metal surface. Some researchers empha-
size that the Marangoni effect will affect the entire process of microgravity experiments
due to temperature gradients or concentration gradients of surfactant elements [18,20].
Yin and Emi [20] reported that Marangoni flow was vigorously strong in molten steel, even
with low concentrations of oxygen and sulfur. J. Jeong et al. (2020) [11] carried out compre-
hensive experimental work to find the correlation of interfacial tension between alumina
inclusions and molten automotive steel. They studied the effect of surfactant concentration
(sulfur) and temperature on interfacial tension.
1.3. Background of the Research

As a structural material, steel has benefits such as strength, ductility, and durability.
Durability relates to resistance to wear, fatigue, hydrogen-induced cracking, and stress
corrosion cracking, whereas ductility is exemplified by properties such as deep drawa-
bility, cold formability, and low-temperature toughness. Large nonmetallic impurities
significantly reduce steel’s ductility and durability. These inclusions are known as “dirty”
inclusions, and they are seen in steel. The increasing need for enhanced ductility and
durability in demanding applications can only be met by “clean” steels, which have a
smaller number of evenly distributed, small-size impurities.
6
Materials 2022, 15, 7458
Figure 1. The effect of Marangoni force on an inclusion near the solidifying front (d = distance
between solidifying boundary and saturation-level interfacial tension; Fi = interfacial tension).
Oxides, sulfides, nitrides, carbides, and their compounds or composites are examples
of nonmetallic inclusions. During typical cooling of steel below the solidus temperature,
sulfides, carbides, and nitrides precipitate. To improve steel qualities, small particles of
certain oxide inclusions, sulfides, carbides, and nitrides have been used to manipulate
the microstructure. Most big oxide inclusions and certain sulfide inclusions, on the other
hand, occur when the steel is still liquid. They may create errors in casting goods, cause
processing problems and failures, lower productivity, impair product quality, and diminish
premium yield if they are not eliminated from the steel melt before solidification.
Improvement of the performance of continuous casting processes has been a subject
of extensive research in the metallurgical industry. High-speed casting in the continuous
casting operation is an essential objective of many of these studies, where the goal is to
improve productivity without sacrificing the quality of the casts. In the continuous cast-
ing process, the fluid flow phenomenon greatly affects the quality of steel and defects
in the final steel product [21]. It is part of a long-term effort to develop and apply com-
prehensive models of continuous casting and is supported by computational models and
experimental measurements.
The particles in the liquid melt can move toward or away from the freezing front
during the solidification of a liquid containing scattered second-phase particles; particles
close to the freezing front will either be consumed or rejected. During solidification, these
two events cause the particle distribution to change. The molten steel flows in the continu-
ous casting process include several interacting phenomena that interact with each other in a
complex form, such as heat transfer, solidification, multiphase turbulent flow, clogging, elec-
tromagnetic effects, complex interfacial behavior, particle entrapment, thermal-mechanical
distortion, stress, cracks, segregation, and microstructure formation [22–24]. In the contin-
uous casting mold, studies on fundamental phenomena, such as temperature, solidified
shell growth, turbulent fluid flow along with multiphase physical processes, electromag-
netic effects and particle transport, microstructure and grain structure, thermal properties,
distortion, and stress, have all made significant strides toward accurately predicting these
factors by computational models. However, the accurate prediction of the inclusion motion
7
Materials 2022, 15, 7458
behavior near the solidifying boundary layer in response to the surfactant concentration
needs much further study [25]. Mukai and Masafumi (2003) [15] studied the motion of
inclusion particles at the solidifying boundary layer. They reported that the movement of
inclusion particle motion due to the surfactant gradient is inextricably linked to some of
the phenomena in the casting process, such as bubble- or inclusion-related defects in steel.
This effect may be noticed due to the presence of a surface tension gradient around the
solidifying boundary induced by the surfactant concentration. Yin et al. (2018) [26] studied
inclusion motion and entrapment during full solidification in a curved billet caster. They
created a numerical model in three dimensions linking flow, solidification, and inclusion
motion. It was noted that the distribution of inclusions inside the mold’s liquid pool is not
exactly symmetrical. Additionally, the flow pattern of the molten steel and the inclusions’
buoyancy force have a significant impact on the mobility and trapping of tiny inclusions in
the mold.
Z. Liu and B. Li (2018) [27] developed a model to simulate the transient fluid flow,
heat transfer, and solidification processes in various vertical-bending continuous casting
casters. The enthalpy-porosity approach was used to simulate the heat transfer and so-
lidification of steel in the caster. The vertical length has less effect on the removal ratio
of inclusions from the top surface but determines the entrapment positions of inclusions.
More inclusions can be carried and captured deep in the mold cavity with a larger ver-
tical length. J. Jeong et al. (2020) [11] predicted the behavior of an alumina inclusion in
front of the solid–liquid interface during solidification, and the interfacial tension between
SPFH590 microalloyed steel and alumina was experimentally determined. According to
the research results, the surface stresses of steel samples decreased as the temperature rose.
Additionally, the contact angles of samples containing 11 to 72 ppm sulfur decreased with
rising temperature, but they rose for the sample containing 94 ppm sulfur. The estimated
interfacial tension values were used to anticipate the behavior of an alumina inclusion
in front of the solid–liquid interface in SPFH590 steel. According to estimates, inclusions
transitioned from being trapped to being driven away from the solid–liquid interface as
the sulfur concentration rose from 5 to 10 ppm.
Using the empirical results established by Ref. [11], Siddiqui et al. (2020) [28] used
a novel correlation of the interfacial tension of alumina inclusions in molten steel under
different conditions of sulfur concentration to study inclusion motion at the solid–liquid
boundary. It was reported that inclusions were prone to engulfment by the solidification
front under the influence of higher interfacial tension between the inclusions and the melt.
H. Yin and T. Emi (2003) [20] investigated the direction and velocity of the surface
flow of the steel melt in the vicinity of the solid–melt interface. It was found that the
solutal Marangoni flow was very strong during the solidification of the steel melt with
very low O and S contents. Further, many fine inclusion particles were brought to the free
surface near the interface by this flow. Fei et al. (2019) [29] investigated the surface flaws
in interstitial-free (IF) steel and carried out quantitative metallographic assessments of
near-surface inclusions as well as surface liquid flow detection using the nail-board tipping
method. The findings demonstrate that at casting rates of 0.8 and 1.0 m/min, a thin liquid
mold flux layer develops, non-uniform argon bubble floating causes the entrainment and
subsequent entrapment of the liquid flux, and small inclusion particles of Al2 O3 may also
collect at the solidification front.
Q. Wang and Zhang (2016) [30] studied the entrapment and final distribution of inclu-
sions in the solidified shell. Their results indicated the distribution of oxide inclusions along
the radial direction. Further, with a larger diameter, inclusions tended to be entrapped
toward the center area of the billet. Chen, Ren, and Zhang (2018) [31] investigated the fluid
flow, solidification, motion, and entrapment of inclusions in an actual vertical-bending con-
tinuous casting (CC) strand. It was reported that the inclusion removal fraction increased
significantly with the increase in inclusion diameter, while the percentage of large-size
inclusions in the entrapped inclusions was decreased. Kölbl and Harmuth (2019) [32]
reported that the quality of steels is mainly influenced by the infiltration of mold slag into
8
Materials 2022, 15, 7458
the gap between the strand and the mold and the temperature-dependent solidification
behavior of the slag. You et al. (2017) [2] reported that the formation of nonmetallic inclu-
sions in the solidification process can essentially influence the properties of steels. They
stated that microsegregation and inclusion formation are fundamental to simulating the
phenomenon in the solidification process. Wein Chen et al. (2019) [33] carried out work in
which a coupled three-dimensional model, the Lagrangian Discrete Phase Model, and a
VOF multiphase model were developed to investigate the transient two-phase flow and
bubble distribution in continuous casting strands. The influence of the lift force, bubble
diameter, and argon flow rate on the transient two-phase flow and bubble distribution were
investigated. Soumava Chakraborty et al. (2020) [34] studied the efficiency of the removal
of solid alumina inclusions by filtration and the distribution of inclusions in a stainless
steel casting. The study documented that inclusion floatation inside the mold cavity plays
a role in reducing the inclusion concentration in the casting.
It is noted from the literature review that several techniques have been implemented
to prevent inclusions in steel, particularly in a continuous casting mold. It is also evident
that the inclusion motion behavior and entrapment in the mold are dependent upon
several factors, such as thermo-physical properties and molten steel flow characteristics.
Amongst these variables, the surfactant concentration plays an important role in inclusion
motion near the solid–liquid boundary layer. The concentration level affects the interfacial
tension, and thus, Marangoni forces play an important role in the pushing and engulfment
phenomenon. The interfacial tension gradient along the interface between liquid and
inclusion particles will propel the particle in the direction of decreasing interfacial tension. It
is established in studies in the literature that inclusion particles are driven in the mold by the
interfacial tension gradient, especially in front of a solidifying interface and in the vicinity of
the interface between liquid steel [35]. Further, it is also evident that minor concentrations
of certain surfactants, such as oxygen, nitrogen, and sulfur, play an important role in
inducing interfacial tension at the solid–liquid boundary.
In most of the previous studies, inclusion motion has been characterized in terms
of fluid flow movement in a continuous casting mold. Very few experimental studies
have reported on the inclusion motion behavior under Marangoni forces induced by
the surfactant concentration in the continuous casting mold. None of the experimental or
numerical research works has reported on inclusion motion using experimental correlations
between interfacial tension, surfactant concentration, and temperature. Amongst the few
available numerical works, no research has been reported to identify the effect of sulfur
concentration on the pushing and engulfment of alumina inclusions near the solidifying
front of automotive steel in a continuous casting mold.
In this work, we investigated the inclusion motion behavior in a continuous casting
mold using two-dimensional turbulent modeling. The inclusion motion characteristic has
been studied with different parameters, such as the input temperature of molten steel in
the mold and the size of the inclusion particles. Further, we focused on inclusion motion in
the solidifying zone. This research study focused on alumina inclusion motion behavior
in molten automotive steel (SPFH590) in a continuous casting mold. Firstly, a full-scale
model of the mold was utilized to simulate the melt flow and inclusion flow characteristics
depending on some fundamental parameters. Secondly, interfacial correlation and thermo-
physical property data from the literature were used for the simulation. In this work, the
numerical simulation was simplified by utilizing only important phenomena, such as heat
transfer and the interfacial tension distribution in a two-dimensional domain, assuming
low to negligible velocity at the solid–liquid interface.
2. Numerical Modeling
A computational fluid dynamic-based numerical model was developed by using Ansys
Fluent to simulate the molten steel flow and alumina inclusion motion in the continuous
casting mold. Table 1 shows the alloying elements of the steel modeled in this study. The
simulation model considers the molten steel flow, heat transfer, solidification of molten steel,
9
Materials 2022, 15, 7458
and inclusion motion. Therefore, a two-dimensional model was used for the simulation.
The study was carried out using different parameters, such as the surfactant concentration
and the temperature of molten metal. Tables 2 and 3 show the details of thermo-physical
properties for numerical modeling.
Table 1. Chemical composition of selected steel (SPFH590) (wt%).
Element C Mn Si P Al Nb
Concentration ≤0.1 ≤3.0 ≤0.5 ≤0.1 ≤0.1 ≤0.1
Table 2. Thermo-physical properties of SPFH590 steel.
Parameters Values
Density of molten steel [36] ρ (kg m−3 ) = 8621.17 − 0.88T
* Viscosity of molten steel [36] μ (mPa s) = 0.2388 ∗ exp(47.44/(RT))
Specific heat [37] 750 J kg−1 K−1
Thermal conductivity [37] 41 W m−1 K−1
Surface tension (σL) and interfacial tension (σPL) Equations (9) and (10) {Ref: [10,11]}
Solidus temperature 1781 K
Liquidus temperature 1798 K
Alumina inclusion size 5, 100, and 300 μm
* where μ is the viscosity, R is the molar gas constant, and T is the absolute temperature (K).
Table 3. Process Parameters.
Process Parameters Values

Mold width 1500 mm
Mold length 3600 mm
SEN submergence depth 160 mm
Nozzle port downward angle 15 degree
Inlet velocity 2 m/s
Outlet Pressure outlet condition
Alumina inclusion density 2500 kg/m3
Shell surface temperature 1273 K
Mold conductivity 315 W/mk
Latent heat 272,000 (J/kg)
The experimental data for the interfacial tension of alumina inclusions and other
thermo-physical properties were from the research works of J. Jeong et al. (2020) [10] and
Siddiqui et al. [28].
The Governing Equations

The numerical model was developed to simulate the melt flow and predict the inclu-
sion motion behavior in a continuous casting mold. In the first stage of the research work,
we used a full-scale model of the continuous casting mold for the simulation. A transient
two-dimensional numerical model was developed using Ansys Fluent (Canonsburg, PA,
USA). The inclusion particles were incorporated using the discrete phase modeling method.
In addition, the computational fluid dynamics (CFD) model was based on a realizable
k-epsilon model and was initially treated as a single-phase model. A major advantage of
the realizable k-epsilon model is its ability to accurately predict the spreading rate of both
planar and round jets, which is highly related to this work.
In the second stage of the work, we used small geometry to understand the alumina
inclusion motion near the solid boundary. In order to investigate the impact of weld
parameters and other interfacial phenomena, a two-dimensional multiphase model was
built for this example. The governing equations for mass and momentum for the current
10
Materials 2022, 15, 7458
simulation model, which was created using a mathematical model based on CFD, are
as follows: →
∂ρ
+ ∇· ρ v = Sm (1)
∂t
∂ → →
=
→ →
ρ v + ∇· ρv v = −∇ p + ∇· τ + ρ g + F (2)
∂t
where p denotes the static pressure, g denotes the gravitational constant, Sm denotes the
mass source term, and F is the external gravitational force. The velocity field, pressure, and
temperature of the fluid in the specified domain are determined using the solutions of the
=
aforementioned Equation contains the stress tensor τ information (3).

= → →T 2 →
τ = μ ∇v + ∇v − ∇· v I (3)
3
where μ is the molecular viscosity, I is the unit tensor, and the second term on the right-hand
side is the effect of volume dilation. The energy equation is expressed in Equation (4):

∂ → → = →

(ρE) + ∇· v (ρE + p) = ∇· k e f f ∇ T − ∑ h j J j + τ e f f · v + Sh (4)
∂t j
where keff is the effective conductivity (k + kt , where kt is the turbulent thermal conductivity,
→
defined according to the turbulence model being used), and J j is the diffusion flux of
species j. Sh is a volumetric heat source term.
Melt pool (molten metal) solidification was quantified using the enthalpy-porosity
approach. The mushy zone, or the liquid portion of molten metal in the range of 0 to 1,
was regarded as a porous medium. The liquid fraction of each cell was determined after
taking the porosity into account. Additionally, the solidified cells were handled as if they
had a porosity percentage of one and were non-porous. As a result, it was assumed that the
velocities of fully solidified cells were zero. Additionally, a “pseudo”-porous medium was
developed to represent the mushy zone. This indicates that the porosity percentage in the
mushy zone ranges from 0 to 1. The sensible enthalpy, h, and the latent gas were combined
to compute the metal enthalpy in this model:
H = h + ΔH (5)
where T
h = hre f + C p dT
Tre f
Additionally, href is the reference enthalpy, Tref is the reference temperature, and Cp is
the specific heat at constant pressure.
Further, the liquid fraction, £, can be defined as:
£ = 0 i f Temperature ( To ) < Solidus Temperature( Ts)
£ = 1 i f Temperature ( To ) > Solidus Temperature( Ts)

To − Ts
£= i f Ts < T < Tl
Tl − Ts
The molten metal’s latent heat content is given as L, ΔH = £. For both solids and
liquids, the latent heat content ranges from 0 to 1. The value of latent heat can be obtained
from. Furthermore, the energy equation for solidification/melting problems is written as:
∂ →
(ρH ) + ∇· ρ v H = ∇·(k∇ T ) + S (6)
∂t
11
Materials 2022, 15, 7458
where S is the source term, and H is the enthalpy. The fundamental strategy used to
handle melting and solidification is to modify the thermal energy equation by including a
phenomenological heat source factor. The domain started with a certain temperature of
molten steel at the start of the calculation. The initial circumstances were used to determine
the heat source term.
The local mass fraction of molten steel, sulfur material, Yi , is predicted by the solution
of a convection–diffusion equation for the ith species. The conservation formula for all
liquid phases is as follows:
∂ → →
(ρYi ) + ∇· ρ v Yi = −∇· J j (7)
∂t
Therefore, one of the components of microalloyed steel is sulfur, and one goal of this
work is to comprehend how alumina inclusions move when there is interfacial tension. The
following equation can be used to determine the diffusion coefficient in this situation since
sulfur diffuses in steel:
1
kT m1 + m2 2
D= (8)
2πμd 2m2
where d is the diameter of atoms, and m1 and m2 are the atomic mass of the solute and
solvent, respectively. T is the temperature of the melt, μ is the viscosity of the molten metal,
and k is Boltzmann’s constant (1.38 × 10−23 J/K).
3. Numerical Details
The goal of the current effort was to create a numerical model that can simulate melt
flow and forecast inclusion motion behavior in a continuous casting mold. Using a full-scale
model of a continuous casting mold, CFD simulation was performed in the first part of the
research project. Utilizing Ansys Fluent, a transient two-dimensional numerical model was
developed (Academic version: 21.0, ANSYS, Inc., Canonsburg, PA, USA). With the use of
the discrete phase modeling technique, inclusion particles (1000 particles at t = 0 s) were
then injected. In this study, the aggregation phenomenon of inclusion particles was ignored
in favor of the spherical form assumption.
A schematic representation of the domain, information on the phases, and the meshed
zone are shown in Figure 2a, along with the dimensions of a full-scale mold. The mold’s
exterior sidewalls were regarded as stiff walls because of their heat transfer conductivity of
315 W/mk. The simulation’s thermo-mechanical characteristics were derived from earlier
work by Siddiqui et al. [28]. The domain’s exterior walls were regarded as convective
barriers with appropriate heat transfer rates. We determined the mean velocity along the
vertical length of the mold for the grid independence test.
The specifics of the grid independence test are shown in Figure 2b (the approximate
number of elements is illustrated). It was determined that, in comparison to a greater
number of elements, about 200,000 elements produced outcomes that were satisfactory (see
Figure 3). According to the grid independence test, the domain’s mid-vertical plane had
a mean velocity change of no more than 5%. However, adding more items resulted in a
noticeable increase in computing time. By taking an average temperature reading at the
mid-vertical plane of the domain, the convergence of the solution was investigated. From
Ref. [28], it was possible to determine the thermo-physical characteristics of molten steel
and alumina inclusions.
12
Materials 2022, 15, 7458
(a)
(b)
Figure 2. (a) Schematic of the domain and (b) domain meshing.
13
Materials 2022, 15, 7458
Figure 3. Grid independence test.
Equation (9) was employed in the simulation to account for the surface tension corre-
lation between molten steel and alumina. Equation (10) was also used in a computer model
to measure the interfacial tension between molten steel and alumina inclusions. We used
empirical relationships from Jeong et al. [11]. They conducted experiments to determine
the surface tension of SPFH590 steel and the interfacial tension of SPFH590 steel and an
alumina inclusion. The following formulas can be used to calculate the surface tension of
SPFH590 steel:
σL = (1511 + 0.08277 T )
−(1041 − 0.5156 T ) (9)
×{ln[1 + exp(−3.583 + 19846/T )(wt.% S)]}
where σL is the surface tension, T is the temperature in K, and S is the sulfur concentration
in ppm.
The following equation describes the interfacial stress between SPFH590 steel and an
alumina inclusion.

σPL = 3050.51 + 131437.97 × (wt.% S) − 1.544 × 107 (wt.% S)2

−3.378 × 109 (wt.% S)3
(10)
+ −0.8498 − 79.739 × (wt.% S) + 7655.06 × (wt.% S)2

+1.962 × 106 × (wt.% S)3 T
where σPL is the interfacial tension between SPFH590 steel and an alumina inclusion.
Furthermore, sulfur concentration was achieved across the domain by using the species
model. In this numerical model, the solidification model was also taken into account.
The species continuity equation was solved at each time step during the simulation’s
two transient periods. The numerical model for molten steel solidification was validated
with the numerical results of Cho et al. [38], as shown in Figure 4. The shell thickness
profile was compared with predicted numerical results. The predicted results seem to have
good agreement with the previous work of Cho et al. [38].
14
Materials 2022, 15, 7458
Figure 4. Comparison of solidifying steel shell thickness in the mold (Cho et al. [38]).
4. Results and Discussion

Mold and Inclusion Characteristics
In the first part of our work, we investigated the inclusion distribution in a continuous
casting mold. A two-dimensional discrete phase model (DPM) was used to analyze and
track the movement of inclusion particles. In addition to this solidification, a model was
employed to simulate the solidified zone in the mold. A suitable turbulence model, the
k- realizable model, was selected to properly configure the turbulence flow of molten
steel. Further, inclusions were assumed to be spherical in shape and non-conglomerating
to simplify the numerical modeling procedure.
Figure 5a,b shows the distribution of 5-micron-diameter inclusions at the mid-section
of the mold at different periods. It can be observed that inclusion concentrations (calculated
by the discrete phase model, DPM) increase over time at a given section. Secondly, it is
also observed that a major part of the inclusion distribution is available on the left and
right sides of the mold, which are solidifying boundary layers. Further, the midzone of the
mold has a lower concentration of inclusions. The inclusion motions are affected by several
parameters, such as the velocity of the local zone, buoyancy force, natural convection, and
interfacial tension. It is expected that inclusions move to the solidifying zone due to an
increase in the thermal gradient and a lower local velocity field.
Figure 5c,d show the distribution of 100-micron-diameter inclusions at the mid-section
of the mold at different periods. In this case, inclusion concentrations have different
distribution characteristics as compared with small (5-micron size) inclusions. It can be
noticed here that inclusions have a lump-sum uniform distribution over the x-axis length.
However, it cannot be denied that some part of the inclusion population is concentrated
towards the solidifying boundary. A significant portion of the inclusion distribution is
noticed on the left and right sides of the mold, which are hardening border layers. We
later study the larger-sized inclusion particles and their distribution in the mold zone. It
was found that inclusions are more evenly concentrated in the mold zone as compared
to the smaller sizes. This implies that larger inclusions can be more influenced by melt
flow velocity and buoyancy forces. Furthermore, the effect of temperature on the inclusion
distribution was investigated to analyze the impact of natural convection and interfacial
tension. Figure 6 shows the inclusion concentration on the same plane but with elevated
temperature conditions. It can be seen that the inclusion concentration is more evenly
distributed in the mold as compared to lower-temperature conditions.
15
Materials 2022, 15, 7458
Figure 5. Inclusion (DPM) concentration in the mold at 1853 K; sizes: (a,b) 5-micron diameter and
(c,d) 100-micron diameter.
Figure 6. Cont.
16
Materials 2022, 15, 7458
Figure 6. Inclusion (DPM) concentration with a 100-micron diameter in the mold at (a,b) 1853 K and
(c,d) 1893 K.
Figure 7 shows the velocity profiles of 5-, 100-, and 300-micron inclusion particles
at different sections of the mold. On the top side of the mold, we can see that inclusions
have a higher flow velocity especially at the left and right sides of the solidifying boundary.
Similarly, for the midzone, we can see that the velocity of inclusions has been reduced due
to a reduction in flow field velocity. The flow field velocity is reduced due to a reduction in
temperature and an increase in the viscosity of the melt. In the lower zone (Y = 0 m), we
can observe the minimal velocities of inclusion particles. The velocity profile of 100-micron
inclusion particles at different sections of the mold is shown in Figure 7. We can see that
inclusions have a larger flow velocity on the top side of the mold, especially on the left and
right sides of the solidifying border. Similarly, we can see that the velocity of inclusions
decreases at the midzone (Y = 1.95 m) due to a decrease in flow field velocity. Because of
the lower temperature and higher viscosity of the melt, the flow field velocity is lower. We
can see the lowest velocities of inclusion particles in the bottom zone (Y = 0 m).
Figure 7. Cont.
17
Materials 2022, 15, 7458
Figure 7. The inclusion velocity profile in the mold at 1853 K; size: (a) 5 microns, (b) 100 microns,
and (c) 300 microns.
Further, it depicts the velocity curve of 300-micron inclusion particles at various mold
sections. On the top side of the mold, especially on the left and right sides of the solidifying
border, inclusions have a higher flow velocity. Similarly, due to a drop in flow field velocity,
the velocity of inclusions is reduced at the midzone (Y = 1.95 m). The flow field velocity is
reduced because of the lower temperature and increased viscosity of the melt. The bottom
zone (Y = 0 m) has the slowest inclusion particle velocities.
The average inclusion concentration profile on the x-axis of the domain is shown
in Figure 8 with respect to inclusion size and time. Figure 8a represents the distribution
of small inclusions (5-micron diameter) over different timescales, namely, 200 s, 380 s,
and 560 s. It is observed that the average inclusion concentration remains unvaryingly
distributed throughout the domain. However, during a certain period of the simulation,
two peaks can be observed in the domain. This suggests that the inclusion distribution
increases near the middle zone of the domain. Further, when we changed the inclusion
size in the same conditions, it was observed that inclusions are concentrated in less time
(390 s), as shown in Figure 8b. A similar pattern is also observed in Figure 8c, where the
inclusion diameter is significantly increased to 300 microns. In another case, we tried to see
the effect of the inlet temperature on the average distribution of inclusions under the same
18
Materials 2022, 15, 7458
conditions. Figure 8d shows the effect of the inlet temperature change. It was observed that
the peak pattern of the average distribution has an impact on temperature, and inclusion
concentrations are formed on a larger timescale.
Figure 8. Inclusion concentration profile in the mold: (a) 5 microns, (b) 100 microns, (c) 300 microns,
and (d) 100 microns (at 1893 K).
In Figure 9, the inclusion concentration profile is shown in the whole domain of

the simulation by using different sizes of inclusions. It can be noted that the inclusion
distribution is mostly on the middle-left side of the mold. On the other hand, we can see
almost the same level of inclusion distribution. In addition to this, inclusions more or
less have the same concentration at six different time lengths. The inclusion distribution
density is depicted in Figure 9a–c throughout the domain for different cases. This density
concentration further confirms the inclusion aggregation towards the solidifying boundary
zone. Further, the inclusion motion with streamlines is shown in Figure 10. The position
and motions of inclusions are shown at a well-established flow time (1000 s) in the mold.
19
Materials 2022, 15, 7458
(a)
(b)
Figure 9. Cont.
20
Materials 2022, 15, 7458
(c)
Figure 9. Inclusion concentration profile in the mold at 1853 K; diameter size: (a) 5 microns,
(b) 100 microns, and (c) 300 microns.
Figure 10. Alumina inclusion motion in the mold.
21
Materials 2022, 15, 7458
Figure 11 shows the temperature contour of the domain at different time intervals.
The initial temperature of the melt inflow was 1853 K. The temperature of molten steel
decreases after some time, and the boundary layer of solidifying steel is visible in the domain.
Figure 12 shows the velocity contour in the domain at different time intervals. Figure 13
shows the total number of inclusion particles available in the domain as a function of
injection time using the discrete phase model. We studied the effect of sulfur concentration
on the entrapment of alumina inclusions near the solid–liquid boundary layer in the mold.
The sulfur concentration affects the interfacial tension at the solid–liquid boundary, and
thus, it impacts the motion of alumina inclusions due to interfacial tension. Figure 14
shows the trapped inclusions near the solid–liquid boundary in the mold at different sulfur
concentrations. It can be noted here that when there was no mixing of sulfur, the entrapment
percentage was low with respect to time. Because of that condition, we did not utilize the
empirical relationship in the simulation. However, the entrapment percentage varied in the
case of sulfur mixing (using the empirical relationship in modeling). It can be noticed that a
10 ppm sulfur concentration does not have much impact. However, as we increase the sulfur
concentration up to 70 ppm, we can see that there is a significant rise in the entrapment of
alumina inclusions in the solidifying boundary layer. Another important outcome that we
notice here is that the maximum entrapment of alumina inclusions occurs at 70 ppm, but
the entrapment percentage decreases at 100 ppm sulfur. The variation in entrapment can
be directly correlated with the experimental correlation between the interfacial tension and
sulfur concentration, as stated in Equations (9) and (10).
Figure 11. Temperature contour of the mold domain (initial melt temperature: 1853 K).
22
Materials 2022, 15, 7458
Figure 12. Velocity contour of the mold domain (initial melt temperature: 1853 K).
Figure 13. Injection rate and total particles in the mold using DPM modeling.
23
Materials 2022, 15, 7458
Figure 14. Trapped inclusions near the solid–liquid boundary in the mold at different sulfur concentrations.
Figure 15 shows the removal percentage of inclusions from the outlet (i.e., outlet
inclusion particles are assumed to have moved out or absorbed in the solidified shell). It
can be noted here that the removal percentage varies according to the sulfur concentration
(around 4%). We can observe that an increase in sulfur ppm up to 70 ppm leads to higher
removal percentages, but the results are the opposite at 100 ppm. Moreover, larger particles
are more prone to move out of solidified shells as compared to smaller particles.
Figure 15. Removal percentage of alumina inclusions in different conditions of sulfur concentration
and inclusion diameters.
We also studied the entrapment of inclusion particles at the top surface of the mold. In
a typical continuous casting mold, the slag layer is artificially created on the top surface for
many reasons. One of the advantages is that inclusion particles become stuck in the slag
layer. Thus, in this simulation model, we considered the entrapment of particles when they
24
Materials 2022, 15, 7458
reach the top surface. Hence, a particular inclusion particle’s velocity is frozen at zero, and
the discrete phase model stops tracking it. Figure 16 shows the entrapment of inclusion
particles on the top surface of the mold. We observed that the smaller-sized inclusion
particles have higher chances of entrapment (more than 25%) and are thus removed from
the molten metal. However, the larger-sized inclusion particles sustain their motion in the
melt for a longer period, and their percentage entrapment is lower. The other conclusion
that we made is that interfacial tension plays an important role in entrapment, as can
be seen in Figure 16. An increase in sulfur concentration up to 70 ppm leads to more
entrapment; however, an increase to 100 ppm does not make a more significant difference.
Figure 16. Entrapment (at the top surface) of alumina inclusions under different conditions of sulfur
concentration and inclusion diameters.
Further, we also studied the effect of melt inlet temperature and its effect on inclusion
entrapment and removal (from the outlet/solidified shell). Figure 17 shows the entrapment
and removal (from the outlet) percentage of alumina inclusions. It can be noted that
the inclusion entrapment percentage is increased as the inlet temperature is increased.
The temperature at the inlet and top surface zone affects the surface tension and melt
flow characteristics. Secondly, we can also observe that the removal percentage does not
vary significantly.
Figure 17. Entrapment and removal (from outlet) percentages of alumina inclusion.
25
Materials 2022, 15, 7458
5. Conclusions
A numerical model was developed to study alumina inclusion motion behavior in
a continuous casting mold. Initially, we investigated the inclusion motion behavior in
a continuous casting mold using a two-dimensional CFD model. The inclusion motion
characteristic was studied with different parameters, such as the input temperature of
molten steel in the mold and the size of the inclusion particles. Secondly, we focused
on inclusion motion under the influence of interfacial tension. The effect of surfactant
concentration (sulfur in this work) on the motion of alumina inclusions was studied. The
following are the main conclusions obtained from this research work:
• A transient, two-dimensional model of a continuous casting mold was developed
using Ansys Fluent software. Alumina inclusions in microalloyed steel were tracked
under the condition of solidification at the boundary side. Further, the empirical
relationship of interfacial tension was used to study the effects of temperature and
surfactant concentration (sulfur) on inclusion motion near the solidifying boundary
layer in the mold.
• The simulation findings indicate that the size of the inclusions and the inlet temper-
ature have significant effects on the distribution of inclusions. It was also observed
that a major part of the inclusion distribution was on the left and right sides of the
mold, which are solidifying boundary layers. Further, the midzone of the mold had
a lower concentration of inclusions. The inclusion motions were affected by several
parameters, such as the velocity of the local zone, buoyancy force, natural convection,
and interfacial tension.
• The inclusion diameter affected the inclusion distribution throughout the domain.
Larger-sized inclusions had different distribution characteristics as compared to small
(5-micron size) inclusions. It was found that larger inclusions were more evenly
concentrated in the mold zone as compared to the smaller sizes. This implies that larger
inclusions can be more influenced by the melt flow velocity and buoyancy forces.
• The effect of temperature on the inclusion distribution was investigated to analyze
the impact of natural convection and interfacial tension. It was noticed that the
inclusion concentration was more evenly distributed in the mold as compared to
lower-temperature conditions.
• The entrapment percentage of alumina inclusions was low over time when the interfa-
cial tension equation was not used. However, the entrapment percentage varied in the
case of sulfur mixing (using the empirical relationship in modeling). It was found that
a 10 ppm sulfur concentration did not have much impact. However, we observed that
there was a significant rise in the entrapment of alumina inclusions in the solidifying
boundary layer at sulfur concentrations up to 70 ppm. Another important outcome
was that the maximum entrapment of alumina inclusions occurred at 70 ppm, but the
entrapment percentage was lower at 100 ppm sulfur.
• It was found that the removal percentage varied according to the sulfur concentration.
The addition of sulfur at concentrations from 10 ppm to 70 ppm resulted in around
a 4% increase in the removal of alumina inclusions (trapped in the solidifying shell),
except for the 100-ppm case.
• The smaller-sized inclusion particles had a 25% higher chance of entrapment at the
top level of the mold (slag layer). However, the larger-sized inclusion particles sus-
tained their motion in the melt for a longer period, and their percentage entrapment
was lower.
• It can be noted that the inclusion entrapment percentage increased as the inlet tem-
perature was increased. The temperature at the inlet and top surface zone affected
the surface tension and melt flow characteristics. Secondly, we also observed that the
removal percentage did not vary significantly.
26
Materials 2022, 15, 7458
Author Contributions: Conceptualization, M.I.H.S. and A.A.; methodology, A.A.; software, L.A. and
A.A.; formal analysis, K.A., I.A.A. and L.A.; writing—original draft preparation, M.I.H.S. and A.A.;
writing—review and editing, K.A. and I.A.A. All authors have read and agreed to the published
version of the manuscript.
Funding: This research received no external funding.
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Available on request.
Acknowledgments: The authors would like to thank Deanship of scientific research in King Saud
University (Riyadh, Saudi Arabia) for funding and supporting this research through the initiative of
DSR Graduate Students Research Support (GSR).
References
1. Kovačič, M.; Župerl, U. Genetic programming in the steelmaking industry. Genet. Program. Evolvable Mach. 2020, 21, 99–128. [CrossRef]
2. You, D.; Michelic, S.K.; Presoly, P.; Liu, J.; Bernhard, C. Modeling Inclusion Formation during Solidification of Steel: A Review.
Metals 2017, 7, 460. [CrossRef]
3. Qin, X.; Cheng, C.; Li, Y.; Zhang, C.; Zhang, J.; Jin, Y. A simulation study on the flow behavior of liquid steel in tundish with
annular argon blowing in the upper nozzle. Metals 2019, 9, 225. [CrossRef]
4. Wang, Y.; Yang, S.; Wang, F.; Li, J. Optimization on reducing slag entrapment in 150 × 1270 mm slab continuous casting mold.
Materials 2019, 12, 1774. [CrossRef] [PubMed]
5. Merder, T.; Pieprzyca, J.; Warzecha, M.; Warzecha, P.; Hutny, A. Evolution of the numerical model describing the distribution of
non-metallic inclusions in the tundish. Materials 2021, 14, 2229. [CrossRef] [PubMed]
6. Yu, Q.; Yang, X.; Lai, C.; Tong, Z. Study on mns inclusion aggregation along continuous casting slab thickness of medium carbon
structural steel. Metals 2022, 12, 56. [CrossRef]
7. Lu, X.; Zhang, Z.; Lv, M.; Li, X.; Song, B.; Fang, M. Evolution of Non-Metallic Inclusions in 27SiMn Steel. Metals 2022, 12, 718. [CrossRef]
8. Tie, Z.; Tang, H.; Wang, K.; Miao, H.; Cai, S.; Xian, F.; Zhang, J. Effect of Flow Field Optimization of an Asymmetric Multi-Strand
Tundish on the Quality Consistency of Cracking Con-Rod Steel. Materials 2022, 15, 3698. [CrossRef]
9. Migas, P.; Śl˛ezak, M.; Karbowniczek, M.; Szcz˛ech, S.; Hornik, A. Numerical Investigation of Outflow of Non-Metallic Inclusions
during Steel Refining in the Ladle. Materials 2022, 15, 3039. [CrossRef]
10. Jeong, J.; Park, D.; Shim, S.; Na, H.; Bae, G.; Seo, S.-J.; Lee, J. Interfacial tension between SPFH590 microalloyed steel and alumina.
Metall. Mater. Trans. B 2019, 51, 690–696. [CrossRef]
11. Jeong, J.; Park, D.; Shim, S.; Na, H.; Bae, G.; Seo, S.J.; Lee, J. Prediction of Behavior of Alumina Inclusion in Front of Solid–Liquid
Interface in SPFH590 Steel. Metall. Mater. Trans. B Process Metall. Mater. Process. Sci. 2020, 51, 690–696. [CrossRef]
12. Yin, Y.; Zhang, J.; Dong, Q.; Zhou, Q.H. Fluid flow and heat transfer with nail dipping method in mould during continuous
casting process. Ironmak. Steelmak. 2019, 46, 855–864. [CrossRef]
13. Yin, Y.; Zhang, J.; Dong, Q.; Zhou, Q.H. Challenges in Special Steel Making. Ironmak. Steelmak. 2019, 46, 855–864. [CrossRef]
14. Siddiqui, M.I.H.; Maurya, A.; Ashraf, M.; Asiri, F. Modeling of Inclusion Motion Under Interfacial Tension in a Flash Welding
Process. In Advanced Computational Methods in Mechanical and Materials Engineering; CRC Press: Boca Raton, FL, USA, 2021;
pp. 91–110.
15. Mukai, K.; Zeze, M. Motion of fine particles under interfacial tension gradient in relation to continuous casting process. Steel Res.
2003, 74, 131–138. [CrossRef]
16. Wang, Z.; Mukai, K.; Lee, J. Behavior of fine bubbles in front of the solidifying interface. ISIJ Int. 1999, 39, 553–562. [CrossRef]
17. Matsushita, T.; Mukai, K.; Zeze, M. Correspondence between surface tension estimated by a surface thermodynamic model and
number of bubbles in the vicinity of the surface of steel products in continuous casting process. ISIJ Int. 2013, 53, 18–26. [CrossRef]
18. Scheller, P.R.; Lee, J.; Yoshikwa, T.; Tanaka, T. Treatise on Process Metallurgy Volume 2: Process Phenomena. In Treatise on Process
Metallurgy; Seetharaman, S., Ed.; Elsevier: Oxford, UK, 2013; Volume 2, pp. 119–139.
19. Shibata, H.; Yin, H.; Yoshinaga, S.; Emi, T.; Suzuki, M. In Situ Observation of Engulfment and Pushing of Nonmetallic Inclusions
in Steel Melt by Advancing Melt/Solid Interface. ISIJ Int. 1998, 38, 49–56. [CrossRef]
20. Yin, H.; Emi, T. Marangoni flow at the gas/melt interface of steel. Metall. Mater. Trans. B 2003, 34, 483–493. [CrossRef]
21. Thomas, B.G. Modeling of the continuous casting of steel—Past, present, and future. Metall. Mater. Trans. B Process Metall. Mater.
Process. Sci. 2002, 33, 795–812. [CrossRef]
22. Siddiqui, M.I.H.; Jha, P.K. Assessment of turbulence models for prediction of intermixed amount with free surface variation using
CLSVOF method. ISIJ Int. 2014, 54, 2578–2587. [CrossRef]
23. Mazumdar, D.; Evans, J.W. Modeling of Steelmaking Processes; CRC Press: Boca Raton, FL, USA, 2009.
24. Sahai, Y.; Emi, T. Tundish Technology for Clean Steel Production; World Scientific: Singapore, 2008.
27
Materials 2022, 15, 7458
25. Thomas, B.G. Review on Modeling and Simulation of Continuous Casting. Steel Res. Int. 2018, 89, 1700312. [CrossRef]
26. Yin, Y.; Zhang, J.; Dong, Q.; Zhou, Q.H. Modelling on inclusion motion and entrapment during the full solidification in curved
billet caster. Metals 2018, 5, 320. [CrossRef]
27. Liu, Z.; Li, B. Effect of vertical length on asymmetric flow and inclusion transport in vertical-bending continuous caster. Powder
Technol. 2018, 323, 403–415. [CrossRef]
28. Siddiqui, M.I.H.; Geleta, D.D.; Bae, G.; Lee, J. Numerical Modeling of the Inclusion Behavior during AC Flash Butt Welding. ISIJ
Int. 2020, 60, 2503–2511. [CrossRef]
29. Fei, P.; Min, Y.; Liu, C.; Jiang, M. Effect of continuous casting speed on mold surface flow and the related near-surface distribution
of non-metallic inclusions. Int. J. Miner. Metall. Mater. 2019, 46, 855–864. [CrossRef]
30. Wang, Q.; Zhang, L. Determination for the Entrapment Criterion of Non-metallic Inclusions by the Solidification Front during
Steel Centrifugal Continuous Casting. Ironmak. Steelmak. 2016, 46, 855–864. [CrossRef]
31. Chen, W.; Ren, Y.; Zhang, L. Large Eddy Simulation on the Fluid Flow, Solidification and Entrapment of Inclusions in the Steel
Along the Full Continuous Casting Slab Strand. JOM 2018, 46, 855–864. [CrossRef]
32. Kölbl, N.; Harmuth, H. Automated break temperature determination of mould slags for the continuous casting of steel based on
temperature-dependent viscosity data. Ironmak. Steelmak. 2020, 47, 899–902. [CrossRef]
33. Chen, W.; Ren, Y.; Zhang, L.; Scheller, P.R. Numerical Simulation of Steel and Argon Gas Two-Phase Flow in Continuous Casting
Using LES + VOF + DPM Model. Ironmak. Steelmak. 2019, 71, 1158–1168. [CrossRef]
34. Chakraborty, S.; O’Malley, R.J.; Bartlett, L. Removal of Alumina Inclusions from Molten Steel by Ceramic Foam Filtration. Int. J.
Met. 2021, 15, 1006–1020. [CrossRef]
35. Mukai, K.; Zeze, M.; Morohoshi, T. Motion of fine particles under interfacial tension gradient in relation to solidification of steel.
Mater. Sci. Forum 2006, 508, 211–220. [CrossRef]
36. Iida, T.G. RIL: The Physical Properties of Liquid Metals; Clarendon Press: Oxford, UK, 1993.
37. Maurya, A.; Jha, P.K. Influence of electromagnetic stirrer position on fluid flow and solidification in continuous casting mold.
Appl. Math. Model. 2017, 48, 736–748. [CrossRef]
38. Cho, S.M.; Thomas, B.G.; Hwang, J.Y.; Bang, J.G.; Bae, I.S. Modeling of inclusion capture in a steel slab caster with vertical section
and bending. Metals 2021, 11, 654. [CrossRef]
28
materials
Article
Effect of Flow Field Optimization of an Asymmetric
Multi-Strand Tundish on the Quality Consistency of Cracking
Con-Rod Steel
Zhanpeng Tie 1,2 , Haiyan Tang 1, *, Kaimin Wang 1 , Hongsheng Miao 2 , Sen Cai 1 , Fenqiang Xian 2
and Jiaquan Zhang 1, *
1 School of Metallurgical and Ecological Engineering, University of Science and Technology Beijing,
No. 30 Xueyuan Road, Haidian District, Beijing 100083, China; [email protected] (Z.T.);
[email protected] (K.W.); [email protected] (S.C.)
2 Technology & Quality Center, Xining Special Steel Co., Ltd., 52 Chaidamu West Road, Xining 810005, China;
[email protected] (H.M.); [email protected] (F.X.)
* Correspondence: [email protected] (H.T.); [email protected] (J.Z.)
Abstract: Cracking con-rod is an advanced high-precision connecting structure based on brittle

expansion, breaking and reconnection of steel, to solve the problem of assembly circle missing. High
carbon micro-alloyed steel C70S6, as a dominant material for the production of cracking con-rod,
has extremely strict requirements on non-metallic inclusions in steel and microstructure stability.
Continuous casting tundish plays an important role in removing large-sized inclusions and stabilizing
casting quality. Aiming at the inconsistent casting quality of C70S6 steel produced by a three-strand
asymmetric tundish and the frequent occurrence of slag entrapment problems in Xining Special
Steel, the tundish structure was optimized by means of physical modelling combined with numerical
simulation, and the quality of the bloom castings and subsequent hot-rolled products before and
after optimization were compared based on volume production. The results show that a new flow
Citation: Tie, Z.; Tang, H.; Wang, K.; control design to the tundish can effectively improve the consistency of its metallurgical effect for
Miao, H.; Cai, S.; Xian, F.; Zhang, J. each of the three strands and the following overall product quality, in which the flow field and
Effect of Flow Field Optimization of temperature field in the tundish are more uniform. This is due to the adoption of a vortex inhibitor
an Asymmetric Multi-Strand Tundish and an optimized wall structure according to the measured RTD curve, ink trajectory and numerical
on the Quality Consistency of simulation on the 3-D streamline contours and temperature distribution in the tundish. The peak
Cracking Con-Rod Steel. Materials concentration of outlet 1 is decreased from 6.5 before optimization to less than 2.0 after optimization,
2022, 15, 3698. https://doi.org/
which means the elimination or alleviation of the local short-circuit flow. The maximum temperature
10.3390/ma15103698
difference of C70S6 molten steel measured at the outlets of the tundish three strands is decreased
Academic Editor: Antonio Lamura from 2–5 ◦ C to 1–3 ◦ C, which is in good agreement with the numerical simulation results. The
difference in columnar crystal ratio of the corresponding bloom castings is decreased from 2.27–3.17%
Received: 11 April 2022
to 1.26–1.85%, and the consistency of central carbon segregation index is also significantly improved.
Accepted: 11 May 2022
Published: 21 May 2022
In addition, the difference in oxygen content among the three strand blooms is decreased from
1.7–3.5 ppm to 0.8–1.9 ppm. As a result, the overall mechanical properties and microstructure stability
Publisher’s Note: MDPI stays neutral
of the hot-rolled products are improved statistically, in which the hardness fluctuation is decreased
with regard to jurisdictional claims in
from 84 HBW to 60 HBW, the inclusion grade of types B + C + D + Ds is reduced to 1.105, and
published maps and institutional affil-
the occurrence rate of Ds dropped to 0.118%. Accordingly, the failure rate of the cracking con-rod
iations.
is controlled stably within 4‰, and the fracture is generally smoother than that before tundish
optimization. In summary, the flow field optimization to a multi-strand asymmetric tundish has a
clear effect on improving the overall quality of its bloom castings and rolled products, which should
Copyright: © 2022 by the authors. be paid more attention industrially. Meanwhile, the present study provides a reliable theoretical and
Licensee MDPI, Basel, Switzerland. experimental reference for the improvement of metallurgical effects of an asymmetric-typed tundish
This article is an open access article commonly used in special steel production.
conditions of the Creative Commons Keywords: high carbon steel; cracking con-rod; C70S6; tundish; flow field; water modelling;
Attribution (CC BY) license (https:// mathematical simulation; temperature field; bloom quality consistency
creativecommons.org/licenses/by/
4.0/).

Materials 2022, 15, 3698
1. Introduction
Cracking con-rods are important transmission components connecting crankshafts
and their pistons, which bear high requirements on load capacity, installation and service
precision. As an advanced connecting structure, a cracking con-rod is made based on
brittle expansion, breaking and reconnection, which is expected to solve the frequent issue
of assembly circle missing. To produce a precise cleavage fracture for tight joints and
avoid big head distortion of the reconnecting rod due to plastic deformation, very strict
demands for a uniform microstructure of steels needs to be met [1]. The non-quenched
and tempered free cutting steel C70S6 is a near eutectoid micro-alloyed steel invented in
Germany for cracking con-rod manufacture [2]. It is also a new type of energy-saving
and efficient steel, with features of high-carbon, micro-alloyed, high-sulfur content and
narrow chemical composition range, thus it is relatively difficult to produce by continuous
casting. In addition, non-metallic inclusions in steel destroy the continuity of the steel
matrix, causing stress concentration and inducing crack formation, thus improving the
cleanliness of molten steel is one of the key demands to solve this problem.
From the perspective of structure and performance requirements on finished products,
on the one hand, the high-carbon and high-sulfur steel has a large tendency to shrink and
crack in the process of continuous casting, which will easily result in the failure of ultra-
sonic detection. Additionally, its large segregation tendency due to its inherent chemistry
can easily cause poor stability of the microstructures and mechanical properties of the
subsequent rolling products [3–5]. Therefore, an in-depth understanding of the as-cast steel
microstructure and its defect characteristics are of great significance on developing a con-
tinuous casting practice of this type of high-class steel. For this reason, we systematically
characterized its solidification microstructures and composition segregation characteristics
in the present continuous casting condition based on industrial development tests [6]. In
this study, the as-cast genetic defects which may affect the requirements of the expanding
cleavage fracture during the processing of steel products were revealed, and the technolog-
ical approaches of optimizing continuous casting process were put forward. As a result,
the quality of the bloom castings has been improved to an applicable degree of the users.
However, due to the asymmetric three-strand tundish used in this steel plant, the tempera-
ture difference is large among different strands in tundish caused by an uneven flow field,
in which the maximum temperature difference of molten steel measured at the outlets near
the stopper reached 5 ◦ C. Moreover, the bloom qualities of C70S6 steel and subsequent
rolled products corresponding to each strand are not at a consistent level. Therefore, it is of
great necessity to control the consistency and stability of casting conditions from tundish,
at the very beginning of continuous casting.
Tundish, a transition container connecting ladle and mold in the casting process of
molten steel, not only plays the role of receiving liquid steel and distributing it to the molds,
but also has the metallurgical functions of homogenizing the composition and temperature
of liquid steel and removal of non-metallic inclusions [7–9]. For the continuous casting
of high-quality special steel, the consistency of the metallurgical effect of each strand in
the tundish is key to improving and stabilizing the overall quality of subsequent products,
representing a popular issue in recent years [10,11]. A reasonable tundish structure can
ensure a good flow-state of liquid steel in the tundish, which is beneficial to improving the
consistency of each strand, reducing the temperature difference between tundish outlets,
and prolonging the residence time of liquid steel in the tundish, thus promoting the removal
of inclusions.
In view of the high temperature characteristics of continuous casting tundish, the
optimization of its structure is mainly carried out by water modelling [12–14] and numerical
simulation [15–17]. For example, Tang et al. [18] optimized a 4-strand tundish in a steel
mill through the above method. After application in the steel production, it was found that
macroinclusions and microinclusions in the billet castings were reduced by 44.9% and 2.7%,
respectively, compared with the use of the prototype tundish, and the total oxygen and
inclusion contents in the four-strand billet castings were very consistent. Chen et al. [19]
30
Materials 2022, 15, 3698
used the same method to optimize the flow field of a two-strand symmetrical tundish
for slab production. By increasing the distance between dam and retaining wall, the
average residence time of water in the simulated tundish and volume fraction of the
piston zone were increased, and the fraction of dead zone was reduced. The total oxygen
content of billet was reduced by 30.32% after applying the optimized tundish in production.
Merder et al. [20] studied the motion behavior of non-metallic inclusions in a two-strand
tundish by using different simulation models and showed that the process of microparticle
distribution can be well described by the Euler–Lagrange method according to the DPM
model. In addition, their distribution characteristics and local concentrations can be
reflected with a high accuracy. Sheng et al. [21] applied a CFD model to study the effect of
thermal buoyancy on fluid flow and residence time distribution in a single-strand tundish.
The results showed that thermal buoyancy has a significant impact on the flow pattern
and temperature distributions of molten steel in the tundish. The increase of heat loss
through the top surface shortens the mean residence time of molten steel in the tundish,
and thus leads to an increase in dead volume fraction and a decrease in plug-flow volume
fraction. Sheng [22] also investigated certain key aspects regarding the CFD simulation of
a tundish, such as the choice of turbulence model, meshing, boundary conditions and a
discretization scheme. In his study, a mesh size of 0.006 m, realizable k-ε turbulence model
and second-order upwind discretization scheme are the recommended modeling settings
for tundish flow simulation.
In the present paper, a three-strand continuous casting tundish was studied for con-
tinuous casting of cracking con-rod steel in Xining Special Steel. The tundish has an
asymmetric geometric structure with a casting position of the ladle long nozzle close to its
No.1 and 2 strand, but far from the No. 3 strand. Usually, an asymmetric tundish is more
difficult to design than a symmetric tundish, and the uniformity of the flow field is also
poorer. The industrial big data found that under the prototype flow control device of the
tundish, the molten steel temperature at the first strand outlet and the occurrence rate of
inclusions in the final cast products were higher, while its equiaxed crystal ratio measured
lower than the other strands. Although there are many reports on the optimization of
tundishes, most studies focus on symmetric multi-strand tundishes, which are evidently
different from their asymmetric counterparts.
To meet the strict requirements of the cracking con-rod regarding the consistencies of
the microstructure and properties of the rolled products, the tundish structure is optimized
by physical and numerical simulation methods in the present study, and an overall quality
comparison of the bloom castings and the following hot-rolled products before and after
optimization is given based on a large number of industrial tests. These on-site volume
data will be of great value in evaluating the casting practice for the quality production of
special steel. Meanwhile, this study intends to provide a roadmap or benchmark for the
optimization of other asymmetric multi-strand tundishes.
2. Production Process and Basic Parameters of C70S6 Steel

The manufacturing process of C70S6 steel is as follows: 110 t Consteel EAF (elec-
tric arc furnace) → 75 t LF (ladle furnace) + 75 t VD (Vacuum degas) → three-strand
250 mm × 280 mm bloom continuous caster → hot rolling → magnetic plus ultrasonic flaw
detection. The chemical composition is shown in Table 1, indicating that this steel has the
characteristics of high C, S and N content, and low Si and Al content.
Table 1. Chemical composition of steel C70S6, wt%.
C Si Mn P S Cr Ni Mo Al V N
0.67–0.73 0.15–0.25 0.53–0.60 ≤0.045 0.060–0.070 0.10–0.15 0.04–0.08 ≤0.03 ≤0.010 0.030–0.040 0.012–0.016
Steel is cast by a three-strand tundish whose structure and original flow control device
is shown in Figure 1. The flow of molten steel is mainly controlled through a retaining wall
31
Materials 2022, 15, 3698
with a diversion hole measuring 140 mm in diameter and angled 8◦ upwards on every
side wall. After the molten steel is poured into the impact zone from a long nozzle, it is
transferred through the diversion hole in the side wall to three submerged entry nozzles
(SEN). The process parameters of the continuous casting of C70S6 steel are shown in Table 2.
Figure 1. Schematic diagram of the prototype tundish (a) and flow control device (b), unit: mm.
Table 2. Process parameters for the continuous casting of C70S6 steel.
Parameter Value
Tundish capacity 21 t
SEN inner diameter 40 mm
Inner diameter of ladle nozzle 85 mm
Depth of the molten bath in tundish 950 mm
Section size of the bloom caster 250 mm × 280 mm
Casting speed 0.75 m/min
3. Research Methods
3.1. Physical Modelling
Taking the industrial tundish shown in Figure 1 as the prototype, a model tundish
with a geometric similarity scale of 1:2 was designed based on the similarity principle,
and water was used to simulate the flow of molten steel in the tundish. According to
the similarity principle, the Froude (Fr) numbers of the prototype and model should be
equal [13], which can be expressed as:
Frm = Frp (1)
32
Materials 2022, 15, 3698
where m denotes the tundish model, and p is the prototype.

The Fr is defined as:
u2
Fr = (2)
gL
where u represents flow velocity, L is characteristic length of tundish and g is gravity acceleration.
Thus, the main geometrical and operation parameters between the model and proto-
type can be expressed with similarity ratio λ as follows:
Lm /Lp = λ, Vm /Vp = λ3 , um /up = λ0.5 , Qm /Qp = λ2.5 (3)
where V is tundish volume and Q is the volumetric flowrate of molten steel.

Substituting in the data from Table 2, Qm and V m can be obtained, respectively.
The stimulation–response method [13,23] was adopted in this experiment. When the
liquid level of tundish reached a given height, 100 mL saturated KCl solution was injected
as a pulse tracer from a side port mounted at the upper position of the ladle shroud. The
conductivity of KCl at the three outlets of tundish was monitored by three conductivity
probes. The conductivity curve with time (RTD curve) was obtained by DJ800 (IWHR,
Beijing, China) system.
The stagnation time tmin and peak time tpeak of the fluid in the tundish was directly
read from the RTD curve. Other flow characteristic parameters were obtained according to
the modified mixing model proposed by Sahai Y [24], as shown in Table 3, where C(t) is the
measured tracer concentration at the time t; j is a specific strand, j = 1, 2, 3; n is total strand
number, n = 3. The optimization effect of flow control device was evaluated by analyzing
the flow characteristic parameters of molten steel in tundish and the RTD curve.
Table 3. Calculation formula of flow characteristic parameters of molten steel in tundish.
Parameters Symbol and Unit Formula

Theoretical mean residence time tf , s ∞ Vm /Q m∞
Actual average residence time tav , s 0 tC ( t )dt/ 0 C ( t )dt
Volume fraction of plug flow θp , % 2 ( tmin + tpeak ) /tf
1
Volume fraction of dead zone θd , % 1 − tav /tf

Volume fraction of well-mixed zone θm , % 1 − θp − θd
n
Standard deviation between strands S, s [ ∑ ( x j − x )2 ] / ( n − 1)
j =1
To visually display the flowing trajectory of molten steel from ladle to mold, ink was
added to the water in the tundish from the same position where KCl was injected, and its
dispersion was recorded by a camera.
3.2. Mathematical Model

A mathematical model including four main governing equations was developed to
calculate the flow field and temperature profiles of molten steel in the prototype and
optimized tundishes, with the casting speed set as 0.75 m/min for a bloom section of
250 mm × 280 mm. These equations and their designations are given in Table 4, and details
of the model are described in refs. [10,13,25].
33
Materials 2022, 15, 3698
Table 4. Governing equations of the mathematical model.
Governing Equations Symbols
1. Continuity equation ρ is density of molten steel, ui

∂(ρui ) is fluid velocity, μe f f is effective viscosity,
∂xi =0 P is pressure.
2. Momentum conservation equation x is the spatial coordinates at i and
∂(ρui u j ) ∂p ∂u
∂x j = − ∂xi + ∂x∂ i [μe f f ( ∂u
∂x j + ∂xi )] + ρgi
i j
j directions.
3. Turbulent
equations
∂(ρkui ) μ ∂k
∂xi = ∂x∂ i (μe f f + σkt ) ∂x + G − ρε k is turbulence kinetic energy, ε is the
i
∂(ρεui ) μt ∂ε
= ∂xi (μe f f + σε ) ∂xi + C1 kε G − C2 ρ εk
∂ 2 turbulent kinetic energy dissipation rate, σk
∂xi and σε are the Schmidt numbers for k and ε,
∂u ∂u j
G = μt ∂x j ( ∂u
∂x +
i
∂xi ) G is generation rate of turbulence energy
i j
2 and μeff is effective viscosity.
μe f f = μ + μt = μ + ρCμ kε
C1 = 1.44, C2 = 1.92, Cμ = 0.09, σk = 1.0, σε = 1.3
4. Energy equation
T is temperature, Ke f f is effective
∂ ∂ ∂T
∂xi ( ρTui ) = Ke f f thermal conductivity.
∂xi ∂xi
3.3. Assumptions and Boundary Conditions

3.3.1. Main Assumptions
The following assumptions were made for simulation calculation: liquid steel in
tundish was incompressible single-phase turbulent flow; influence of the slag layer on flow
was ignored; the slag-steel surface was treated as a free surface.
3.3.2. Boundary Conditions and Solution Methods

The inlet of the molten steel in tundish adopted a velocity inlet. The inlet temperature
was set as 1773 K based on the casting temperature of steel C60S6. The direction of speed
was perpendicular to the section of the long nozzle inlet, and its value was obtained by the
size of the casting section and the casting speed according to the mass conservation, which
is 0.486 m/s in the present study. The tundish outlet adopted outflow boundary conditions,
the same casting speed for each strand; the melt surface was considered to be flat, and the
vertical gradient of all variables on the free surface was zero; the tundish wall surface was
regarded as a non-slipping condition, and the standard wall function was adopted to deal
with the near wall; the heat dissipation of each wall of the tundish adopted the second
type of boundary condition. The material parameters and boundary conditions of the
model are shown in Table 5, in which the turbulence kinetic energy and turbulence energy
dissipation rate were obtained by semi-empirical formulas k = 0.01v2 and ε = 2 × k1.5 /D,
respectively, (v represents the flow rate of molten steel and D represents the diameter of the
long nozzle), and the others originated from reference studies [10,16].
Table 5. Boundary conditions and material parameters.
Boundary Condition and

Value Boundary Condition and Material Parameter Value
Material Parameter
W/m2
Longitudinal wall heat flux, −4600 kg/m3
Density of liquid steel, 7014
Transversal wall heat flux, W/m2 −4000 Specific heat of liquid steel, J/(kg·K) 750
Free surface heat flux, W/m2 −15,000 Thermal conductivity of liquid steel, W/(m·K) 41
Bottom heat flux, W/m2 −1800 Viscosity of liquid steel, kg/(m·s) 0.0061
k, m2 /s2 0.00238 ε, m2 /s2 0.00274
The flow and temperature distribution of molten steel in the tundish were simulated
by commercial CFD software FLUENT 14.5 (ANSYS, Pittsburgh, PA, USA, 2012). Choosing
the pressure-based solver for calculation, and the pressure and velocity coupling were
34
Materials 2022, 15, 3698
based on the SIMPLE algorithm. The computational convergence residual was set to less
than 10−6 for the energy equation and 10−4 for the other equations.
4. Experimental Results and Discussion

4.1. Results of Water Modelling
In total, 30 groups of experimental schemes were designed and carried out, but only
the tundish prototype and the final optimization scheme are relevant to the comparative
quality analysis here, thus this paper only lists the two schemes for comparison. The
retaining wall structure of prototype case A was shown in Figure 1. The optimized case B
is shown in Figure 2, in which a diversion hole measuring 90 mm in diameter and upward
angle of 10◦ is displayed on the right side of the wall near outlet 1, and two diversion holes
are displayed on the left side near outlet 2, with each hole at the same diameter and upward
angle as on the right side but a different horizontal inclination of 3◦ and 5◦ , respectively.
Additionally, a vortex inhibitor is designed as shown in Figure 2b. The flow characteristic
parameters of the two schemes are shown in Table 6.
Figure 2. Retaining wall structure (a) and vortex inhibitor (b) in optimized case B and their geometry
sizes, unit: mm.
Table 6. Flow characteristic parameters of the two experimental schemes.
Case Outlet No. tmin /s tpeak /s tav /s θd /% θp /% θm /% Stmin /s Stav /s

1 17 20 267
A 2 16 19 369 53.08 2.38 44.53 19.20 25.56
3 50 64 434
1 30 41 651
B 2 31 71 637 21.10 5.59 73.31 11.12 15.22
3 50 60 591
As can be seen from Table 6, the proportion of dead zone (θ d ) in the tundish prototype
case A reaches as high as 53.08%, while the piston zone θ p is only 2.38%. A large proportion
of dead zone indicates that molten steel flows much slowly in some areas of tundish,
which renders present non-metallic inclusions difficult to collide and grow, and the local
molten steel temperature will also be relatively low due to weak heat exchange. The
stagnation times tmin of outlets 1 and 2 are 17 and 16 s, respectively, and the peak time tpeak
is 20 and 19 s, respectively. The difference between tmin and tpeak is only 3 s, suggesting that
the short-circuit flow is possibly formed at the two outlets. The stagnation time of outlet
3 is about 2.5 times that of outlets 1 and 2, and its average residence time is much longer
than that of the other two, indicating poor consistency among strands. This is because the
35
Materials 2022, 15, 3698
diversion holes are far from the No. 3 strand, while pointing to Nos. 1 and 2 in direction,
and the fluid flows out from outlets 1 and 2 first. Its kinetic energy is small when flowing
to outlet 3, resulting in larger standard deviations of the stagnation time and average
residence time among the three strands, Stmin and Stav , are 19.20 s and 25.56 s, respectively.
Compared with case A, the dead zone proportion of case B is only 21.1%, and the
stagnation times of outlets 1 and 2 are extended to 30 and 31 s, respectively. The standard
deviations of stagnation time and average residence time are reduced to 11.12 and 15.22 s,
respectively, indicating a clear improvement of the consistency among strands.
Figure 3 shows the RTD curves of the two schemes, where the ordinate is dimensionless
concentration and the abscissa is dimensionless time. As can be observed, after the tracer
KCl is added into the tundish in case A, the concentrations of outlets 1 and 2 rapidly reach
the peak value, and the peak concentration of outlet 1 is as high as 6.5, indicating that a
short-circuit flow has formed in the tundish. The short-circuit flow will make the molten
steel flow out of the tundish rapidly. In addition, its residence time in the tundish is short,
and there is inadequate time for inclusions to grow or float. This may be related to the
lower pass rate of ultrasonic detection of the rolled products corresponding to tundish
outlet 1 in industrial production.
Figure 3. Comparison of RTD curves between cases A and B.
Although the RTD curve of case B is not very smooth, the peak concentration is greatly
reduced to less than 2.0, and the consistency among different strands and the short-circuit
flow are also improved as compared with case A. In addition, the average residence time
of the three strands ranges from 591 to 651 s, much longer than the 267 to 434 s in the
original case. This will provide enough time for inclusions in the molten steel to float and
be removed, and thus lead to an improvement in the cleanliness of the steel.
Figure 4 compares the ink trajectory of the two cases. The ink tracer consists of 50 mL
black ink mixed into 150 mL water. From Figure 4a, it can be seen that the ink from
diversion holes in case A flows downwards and reaches outlets 1 and 2 at about 16 s,
then reaches outlet 3 at about 60 s. This is basically in agreement with the experimental
measurement. The ink moves to the side wall of the tundish and then flows back to the
middle area. The top region of outlet 2 mixes slowest, possibly representing the main dead
zone in the tundish. The local temperature of molten steel will possibly also be low. The
ink fills the whole tundish at about 180 s in this case. By comparison, the ink in case B
(Figure 4b) flows upwards due clearly to the increased upward inclination angle of the
diversion holes from 8◦ to 10◦ , with which the fluid will be delayed in arriving at outlets
1 and 2. However, because of the reduced hole size (from 140 mm to 90 mm) in the wall,
the fluid velocity will increase, which accordingly shortens the stagnation time of outlet
3 to 51 s. As a result, the flow of molten steel will become more uniform in case B than in
case A. The ink is completely mixed at 124 s, 56 s shorter than that in case A.
36
Materials 2022, 15, 3698
Figure 4. Ink trajectory comparison of cases A (a) and B (b).
4.2. Validation of CFD Modelling and Mesh Independence Test

The numerical simulation method has been widely used in the structural optimization
of continuous casting tundish, and the authors have also conducted an abundance of
research works on this topic [10,13,18] with its reliability confirmed by water simulation
experiments and industrial application results [10,18]. This paper adopts the same math-
ematical modeling and calculation method, and the reliability of the simulation will be
further verified through actual temperature measurements.
Meanwhile, a mesh independence test was carried out because of its importance to the
simulation results. As shown in Table 7, different numbers of grids were designed for case
A and case B, respectively. By applying the same boundary conditions and parameters,
the temperature of each tundish outlet was calculated and compared with industrial
measurement data. Data from the table indicate that the temperature of each outlet is
slightly changed and the calculation accuracy can be met when the number of grids is about
300,000~600,000. Therefore, 394,471 tetrahedral grids for case A and 559,800 tetrahedral
grids for case B are chosen for the following studies.
Table 7. Mesh independence study and model validation.
Maximum Temperature
Number Difference among Three
Case T outlet1 /K T outlet2 /K T outlet3 /K
of Grids Strands for Industrial
Measurement/◦ C
309,506 1768 1765 1767
394,471 1768 1765 1767
A 489,901 1768 1765 1767 2–5, average 3.2
740,025 1769 1765 1766
1,093,481 1768 1767 1765
321,107 1766 1766 1767
431,562 1766 1766 1767
B 559,800 1766 1766 1767 1–3, average 1.8
736,498 1766 1766 1767
1,031,898 1766 1767 1767
37
Materials 2022, 15, 3698
4.3. Results of Numerical Simulation

Figure 5 shows the 3D streamline contours of cases A and B. It is seen that in case A,
the velocity of molten steel in impact zone is great with most at 0.05 m/s. In addition, the
flow is active in the area near outlet 1 as the diversion hole on one side of retaining wall
points to it adjacently. The local velocity reaches over 0.02 m/s. However, the molten steel
at outlets 2 and 3 flows slowly at under 0.01 m/s. This is due to deviation of the diversion
hole on the right wall away from outlet 2 and further still from outlet 3. By comparison,
the flow field of case B is more uniform, and the flow velocity and streamline of molten
steel near outlets 2 and 3 are clearly intensified, due to the adoption of an optimized wall
structure. The area with flow velocity over 0.02 m/s increases, and the velocity of the
impact zone is greatly reduced with the help of a vortex inhibitor. The flow field in the
tundish is crucial to the temperature uniformity of the molten steel within.
Figure 5. Three-dimensional streamline contours of tundish in cases A (a) and B (b).
Figure 6 compares the temperature distribution of cases A and B in the vertical section
of the outlet center. Similarly, the temperature distribution in case B is more uniform than
in case A. The maximum temperature difference at three outlets of tundish is only 1 K, 2 K
lower than case A. The maximum temperature difference in the whole section is decreased
from 8 K (1769 minus 1761 K) to 6 K (1769 minus 1763 K), and the high (1768~1769 K) and
low temperature areas (1761~1763 K) are both also reduced in case B.
From the results given above, the consistency of flow and temperature distribution of
case B are both significantly improved compared with case A, which is expected to improve
the stability and consistency of steel product quality among different strands.
38
Materials 2022, 15, 3698
Figure 6. Comparison of temperature distribution of cases A (a) and B (b) in the vertical section of
outlet center.
5. Comparison of Industrial Application Effect before and after Tundish Optimization

Based on the results of the physical and numerical simulation, the optimized case
B was successfully adopted to industrial production of C70S6 steel. The temperature
difference of molten steel at the three outlets of the tundish, and the qualities of bloom
castings and subsequent rolled products before and after optimization were compared
under the same given refractory material and operating conditions.
5.1. Comparison of Temperature Difference of Molten Steel

The molten steel temperature in the tundish was measured comparatively before and
after tundish optimization, in which the 2nd, 3rd, 4th, 5th and 7th heats were selected
for each casting test, three times for each heat during the casting process. A temperature
detector (TC1200-S-AL, Heraeus, Hanau, Germany) was inserted into the molten steel from
the stopper holes between strands, and the maximum temperature difference among the
three strands was recorded. For better comparison, measurement of the three strands was
carried out at nearly the same casting time. There are 13 groups of effective data obtained
for the comparison, which are shown in the Figure 7.
As is seen, the maximum temperature difference between strands measures 2–5 ◦ C
before tundish optimization, and the average temperature difference is 3.2 ◦ C. After op-
timization, the maximum temperature difference decreases to 1–3 ◦ C, and the average
temperature difference drops to 1.8 ◦ C, indicating an improved temperature field unifor-
mity of the tundish. Additionally, the maximum temperature difference calculated by
numerical simulation was 3 ◦ C before optimization and 1 ◦ C after optimization, which also
verifies the reliability and accuracy of the numerical simulation from an industrial perspec-
tive. Meanwhile, the results also show that the flow field optimization has significantly
improved regarding temperature consistency in the tundish. As we know, uniformity of the
temperature field is helpful in stabilizing and unifying the superheat for the three-strand
castings, and thus stabilizing the as-cast solidification structure of the strands, along with
the identical quality of the as-rolled and subsequent products.
39
Materials 2022, 15, 3698
Figure 7. Maximum temperature difference among strands before and after tundish optimization.
5.2. Comparison of Erosion of Molten Steel on Refractory Material in Tundish

The change of flow field of the tundish may affect the erosion of its lining, and thus
affect the cleanliness of blooms and the following rolled products. For this reason, the
eroded appearance of the tundish working layer at three positions near the slag line was
compared at the end of the casting. The test positions (A, B, C) are illustrated in Figure 8.
The eroded appearances of the three tundishes were examined before optimization, and
five tundishes were examined at the same locations after optimization. To ensure the
comparability and effectiveness of erosion data before and after optimization, the material
of the working layer, amount and method of cover slag added, level height of molten steel
in tundish, fluctuation of casting speed and temperature measurement times were strictly
controlled identical under the two working conditions. The erosion width and depth are
shown in Table 8.
Figure 8. Schematics of erosion examination locations.
Table 8. Erosion state of the working layer in tundish.
Item Before Optimization After Optimization

Measured number of tundish, # 3 5
A 90–120 90–100
Erosion width, mm B 90–110 90–110
C 90–110 90–110
A 10–12 9–11
Erosion depth, mm B 10–12 9–11
C 10–12 9–11
40
Materials 2022, 15, 3698
As shown in the Table 8, the erosion width of the working layer at position A close to
strand No. 1 decreases from 90–120 mm before optimization to 90–100 mm after optimiza-
tion, and the depth decreases from 10–12 mm to 9–11 mm. This is because a diversion hole
in the retaining wall before optimization directly points to zone A, imparting significant
washing to the local lining material. After optimization, the diversion hole is provided with
an upward and horizontal deflection angle to avoid direct impingement on the tundish
wall. This comparison indicates that a good flow field can reduce erosion of the lining,
and thus hopefully reduce exogenous inclusions from erosion. The erosion widths of
positions B (near No. 2 strand) and C (No. 3 strand) are basically the same before and after
optimization, while the erosion depth is slightly reduced after optimization.
5.3. Comparison of Bloom Quality Consistency

The effect of tundish flow field optimization on bloom quality should be embodied
mainly by the improvement of bloom quality and that of its consistency among different
strands. For this reason, 10 batches of C70S6 steel blooms in total were compared, with
5 batches (1#–5#), respectively, before and after tundish optimization. For each batch, No.
1 to No. 3 strand blooms at the same casting time were taken and machined, then acid
etched with 1:1 HCl solution at 70 ◦ C for 20 min. The columnar, equiaxed crystal rates and
the center shrinkage were evaluated and their differences among the three strands were
revealed. Moreover, the rod samples were taken from the same position of each strand
bloom, and then the total oxygen content was measured with an oxygen–nitrogen analyzer
(TCH600, St Joseph, MI, USA). The results are shown in Table 9.
Table 9 shows that before flow field optimization, the columnar crystal percentage
of three-strand billets distributes from 43.86–58.19%, and the difference among different
strands ranges between 2.27–3.17%. After optimization, the difference of columnar crystal
rates among strands is reduced to 1.26–1.85%, and the difference of equiaxed crystal
rates decreases from 2.25–2.87% to 0.48–2.10%. There is no significant change in center
porosity before and after optimization. It is well known that equiaxed and columnar crystal
proportions are closely related to casting superheat. A high superheat means a greater
temperature gradient between the inner liquid core and the outer, along with a lower
solidification nucleation rate, and thus resulting in a more developed columnar crystal.
Meanwhile, the excessive growth of columnar crystals causes the enrichment of solute
elements in the center, generating center segregation. Before tundish optimization, the
temperature of molten steel near No.1 strand is higher than that of other two strands, thus
the superheat is relatively high, and accordingly the proportion of columnar crystals is
slightly higher than that of other strands from statistics. After optimization, the temperature
field is more homogeneous, thus the microstructure difference among strands is reduced.
Moreover, Table 9 shows that the oxygen content difference between three strands of
blooms is 0.8–1.9 ppm after flow field optimization, while it is 1.7–3.5 ppm before that, indi-
cating that the consistency of the bloom cleanliness has been effectively improved. Mean-
while, the oxygen content in the No. 1, No. 2 and No. 3 strands measures 11.2–14.3 ppm,
8.8–11.8 ppm and 8.5–13.1 ppm, respectively, before optimization, while it is 8.8–11.8 ppm,
9.2–10.8 ppm and 7.9–11.6 ppm, respectively, after optimization. The oxygen content after
optimization is lower than that before optimization, especially in the first and third strands.
This is because, as mentioned above, the stagnation time and average residence time of
molten steel in tundish after optimization are both improved, which provides sufficient
time for the inclusions in molten steel to float. In addition, the consistency among strands
significantly improves due to the more uniform flow field after optimization.
41
Table 9. Comparison of C70S6 bloom quality before and after tundish optimization.
Working Condition Before Optimization After Optimization

Item
Detected Sample No. 1# 2# 3# 4# 5# 1# 2# 3# 4# 5#
No. 1 49.25 46.27 56.1 48.75 53.81 47.80 52.56 51.00 49.96 53.88
Materials 2022, 15, 3698
Columnar No. 2 48.56 43.86 58.19 46.36 51.50 49.06 53.50 52.85 48.75 55.17
crystal rate, % No. 3 46.08 46.48 55.92 46.87 54.15 48.70 54.25 51.35 48.54 54.86
Δmax 3.17 2.62 2.27 2.39 2.65 1.26 1.69 1.85 1.42 1.29
No. 1 28.32 25.04 22.21 28.71 27.96 30.36 25.40 24.38 26.21 20.06
As-cast Equiaxed No. 2 28.41 26.84 24.55 27.91 28.13 30.17 26.36 23.28 26.43 21.72
macrostruc- crystal rate, % No. 3 30.87 27.74 23.66 30.78 30.21 29.88 27.24 24.49 27.81 19.62
ture
Δmax 2.55 2.70 2.34 2.87 2.25 0.48 1.84 1.21 1.60 2.10
No. 1 1.5 1.0 1.5 1.0 1.5 1.5 1.5 1.5 1.5 1.0
Central No. 2 1.5 1.0 1.5 1.0 1.5 1.5 1.5 2.0 1.0 1.0
shrinkage No. 3 1.5 1.0 1.5 1.5 2.0 1.5 1.5 1.5 1.5 1.5
Δmax 0.0 0.0 0.0 0.5 0.5 0.0 0.0 0.5 0.5 0.5
No. 1 0.97–1.03 0.98–1.03 0.98–1.08 - - 0.97–1.03 0.96–1.04 - - 0.98–1.03
Columnar
No. 2 0.96–1.04 0.95–1.01 0.95–1.03 - - 0.96–1.02 0.96–1.02 - - 0.96–1.03
crystal zone
No. 3 0.95–1.02 0.94–1.03 0.99–1.02 - - 0.99–1.02 0.98–1.01 - - 0.96–1.02
C segregation
42
index at No. 1 0.95–1.04 0.93–1.02 0.96–1.03 - - 0.95–1.03 0.96–1.04 - - 0.95–1.03
Equiaxed
different No. 2 0.97–1.05 0.96–1.03 0.95–1.01 - - 0.97–1.01 0.98–1.03 - - 0.97–1.02
crystal zone
regions of No. 3 0.95–1.03 0.98–1.03 0.97–1.03 - - 0.96–1.03 0.95–1.02 - - 0.96–1.3
bloom No. 1 1.11 1.05 1.07 - - 1.10 1.09 - - 1.08
Center point No. 2 1.03 1.12 1.12 - - 1.09 1.03 - - 1.09
No. 3 1.05 1.02 1.03 - - 1.10 1.08 - - 1.04
No. 1 12.3 14.3 13.2 11.8 11.2 10.8 8.8 11.2 9.9 11.8
No. 2 8.8 11.8 10.4 9.4 9.9 9.2 9.2 10.8 10.2 10.7
T. O. ppm
No. 3 10.2 10.9 13.1 8.5 9.5 8.9 7.9 11.6 10.8 10.5
Δmax 3.50 3.40 2.80 3.30 1.70 1.90 1.30 0.80 0.90 1.30
Materials 2022, 15, 3698
The as-cast pieces of blooms after acid etching were sampled by drilling with a ϕ5 mm
drill for chemical analysis as shown in Figure 9. A total of 25 points were drilled in each
bloom from No. 1 to No. 25, and their carbon contents were measured by an HCS-140
infrared carbon and sulfur meter. The local carbon segregation index was calculated based
on the ratio of its carbon content to the average carbon content at all points, namely:
Ci /(∑Ci /25), where i = 1, 2, . . . . . . 25. Three heats were taken respectively before and after
optimization, and each heat took three strands of blooms. Therefore, there are 18 pieces
(3 heats × 3 strands × 2 working conditions = 18) in total for the statistics. The C segregation
indexes of all points are shown in Table 9, and those along the cross-section center line
(from points 1 to 9) are shown in Figure 10.
Figure 9. Schematic diagram of sampling location for C segregation measurement.
Figure 10. Carbon segregation index before and after optimization.
43
Materials 2022, 15, 3698
As can be observed in Table 9, the carbon segregation indexes in equiaxed crys-

tal and columnar crystal zones for different batches of bloom pieces fluctuates between
0.95 and 1.05. There seems to be no clear change before and after optimization, but the
consistency among different strands is significantly improved as shown in Figure 10, and
the segregation index of the center point 5 in No. 2 strand is also reduced from 1.13 to 1.10.
This is because the element segregation is relevant to the as-cast structure, which is closely
related to the superheat of molten steel [26]. After optimization, the temperature distri-
bution of molten steel in tundish is more uniform as shown in Figure 6. The superheat
for each strand casting is at the same level, and thus their solidification and segregation
consistency are significantly improved accordingly, together with a greatly reduced center
carbon segregation index.
5.4. Comparison on the Quality of Rolled Products by Overall Stactistical Data

The mechanical properties of hot-rolled products before and after tundish optimization,
including tensile strength Rm, yield strength Rel, elongation A and reduction of area Z,
were measured using a universal material testing machine WAW-Y500C, and the hardness
measured with an HB-3000B Brinell hardness tester. Additionally, the ferritic phases and
band structures were observed using a Carl Zeiss optical microscope after samples were
etched with 2% nitric acid alcohol solution. In addition, the grain sizes were evaluated by
the standard method with water-quenched samples after insulating at 860 ◦ C for one hour.
Finally, the product quality indexes measured above are comparatively listed in Table 10.
According to the big data analysis, after the flow field of the tundish is optimized, the
pass rate of tensile strength of rolled products is 98.9%, which is higher than 98.3% before
optimization, while the yield strength, elongation and section shrinkage show no clear
change. The hardness fluctuation (maximum value minus minimum value) decreases from
84 HBW before optimization to 60 HBW after optimization. The ferrite content is stabilized
from 3–9% to 3–8%. The band structure decreases from 2.08 to 1.87 grade. Inclusions
B + C + D + Ds decrease from 1.143 to 1.105 grade, and the occurrence rate of Ds inclusions
decrease from 0.192% to 0.118%. Ds refers to single particle inclusion, which are large in
size and harmful to steel service property, thus requiring strict control in many special
steels. In this study, the occurrence rate of Ds-type inclusions in C60S6 steel is reduced by
the optimization of tundish, since the stagnation time and average residence time of strands
1 and 2 are prolonged, and the short-circuit flow is eliminated after optimization, which
gives the inclusions adequate time to float and be removed. In addition, as mentioned
above, the optimized tundish structure can alleviate the lining erosion in zone A and thus
reduce large-sized exogenous inclusions into the molten steel. The A-type inclusion was
not counted in this study because the steel contains S and no conspicuous related service
defects in its rod products from A-type inclusions. Moreover, no clear grain-size index
variation was observed in this study.
5.5. Products Customer Feedback

According to user feedback regarding the cracking con-rod products, the failure rate of
products was controlled within 4‰ during their forming processes, and the fracture of con-
necting rod was relatively smooth compared with that before casting practice optimization.
44
Table 10. Comparison of the quality of rolled products before and after tundish optimization.
Item Before Optimization After Optimization
Number of samples, # 1153 912

Measured item Rm, MPa Rel, MPa A, % Z, % Rm, MPa Rel, MPa A, % Z, %
Standard requirement 900–1050 >550 >10 20 900–1050 >550 >10 20
Materials 2022, 15, 3698
Mechanical
property Measured value 878–1050 485–599 8.5–16 13–29 849–1018 439–692 6.5–19.5 8–48
Average value 943 555 12.7 21.5 941 552 11.9 20.5
Pass rate, % 98.3 99.6 97.8 98.5 98.9 99.6 97.9 98.4
Measured value, HBW 229–313 225–285
Hardness
Average value, HBW 258 256
Hardness difference
(Maximum minus 84 60
minimum), HBW
Microstructure Microstructure F+P F+P
Ferritic content, % 3–9 3–8
Grain size
Measured value 7.0–8.0 7.0–8.0
45
Band structure
Measured value 1.0–2.5 1.0–2.5
index
Average value 2.08 1.87
Inclusions Grade of (B + C + D + Ds) 1.143 1.105
Ds occurrence rate, % 0.192 0.118
Materials 2022, 15, 3698
6. Conclusions
The flow field optimization to a multi-strand asymmetric tundish has been conducted
through water modelling combined with a numerical calculation method for improved
metallurgical behavior, and the beneficial results were revealed accordingly through volume
production. The conclusions are drawn as follows:
(1) The water modeling experiments show that dead zone proportion of the tundish
is reduced from 53.08% originally to 21.10% after optimization, and the stagnation
time of its outlets 1 and 2 are extended to about twice that of before, indicating clear
improvement in the consistency among strands. The average residence time of the
three strands measures 591 to 651 s, much longer than the 267 to 434 s in the original
case. This will provide sufficient time for inclusions in the molten steel to float and be
removed, and thus improve the cleanliness of steel.
(2) The numerical simulation results show that the flow field in the tundish is more
uniform due to the adoption of a vortex inhibitor and an optimized wall structure,
which is beneficial for temperature-field uniformity. The maximum temperature
difference at the three outlets of the tundish is only 1 K, 2 K lower than the original
case. The maximum temperature difference in the whole outlet vertical section is
decreased from 8 K to 6 K and both the high and low temperature areas are reduced
by the optimized case.
(3) The maximum temperature difference of molten steel measured among the outlets
around stoppers of the three strands is decreased from an initial 2–5 ◦ C to 1–3 ◦ C after
optimization; the lining erosion in region A near the No. 1 strand of the tundish is
clearly alleviated due to the disappearance of the local short-circuit flow.
(4) The differences in columnar crystal rate among different strands of the bloom castings
are decreased from 2.27–3.17% to 1.26–1.85% after tundish optimization; the consis-
tency of the strand central line carbon segregation index is significantly improved.
The total oxygen content difference among strands is reduced from 1.7–3.5 ppm to
0.8–1.9 ppm, with the No. 1 strand showing a more significant effect than the other
two strands.
(5) The main mechanical properties and microstructure stability of the rolled products
are statistically well improved owing to the generally identical as-cast primary quality,
in which the pass rate of tensile strength reaches 98.9%, the hardness dispersion is de-
creased from 84 HBW to 60 HBW, the non-metallic inclusion grade of B + C + D + Ds
is decreased from 1.143 to 1.105, and the occurrence rate of Ds type inclusion is de-
creased from 0.192% to 0.118%. Additionally, the band-structure level of the hot-rolled
products drops from 2.08 to 1.87.
In summary, flow-field optimization of the tundish reveals a clear beneficial effect
to its subsequent castings and hot-rolled products, especially for an improved identical
quality index while cast by different strands.
Author Contributions: Conceptualization, Z.T. and H.T.; methodology, Z.T and F.X.; software,
K.W. and S.C.; validation, Z.T., H.T. and K.W.; formal analysis, Z.T. and F.X.; investigation, Z.T.
and F.X.; resources, H.M.; data curation, H.M.; writing—original draft preparation, Z.T. and H.T.;
writing—review and editing, H.T. and J.Z.; visualization, K.W. and S.C.; supervision, H.T. and J.Z.;
project administration, H.T., J.Z. and H.M.; funding acquisition, H.T. All authors have read and
agreed to the published version of the manuscript.
Funding: This research was funded by National Natural Science Foundation of China (grant number
51874033) and Beijing Natural Science Foundation (grant number 2182038) to Haiyan Tang.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
46
Materials 2022, 15, 3698
References
1. Wang, Z.H.; Li, S.; Ge, Y.; Zhou, L.; Zhao, X.M.; Mao, X.Y.; Hu, Q.X. Research on Microstructure and Mechanical Properties and
Fracture Splitting Properties of Forged Connecting Rod of C70S6 Non-quenched and Tempered Steel. Hot Work. Technol. 2015, 44,
39–43.
2. Deng, X.Y.; Li, J.; Xie, J.B.; Hu, D.L.; Wang, Y.; Fu, J.X. Comparison of Quality between C70S6 Steels Produced by Chinese and
Overseas Steelmakers for Fracture Splitting Connection Rod. Iron Steel Vanadium Titan. 2018, 39, 148–154.
3. Zhang, Z.; Li, H.Y.; Zhou, L.; Liu, H.S.; Tang, H.Y.; Zhang, J.Q. As-Cast Spot Segregation of Gear Steel and Its Evolution in the
Rolled Products. Acta Metall. Sin. 2021, 57, 1281–1290.
4. Li, L.; Zhang, Z.H.; Luo, M.; Li, B.; Lan, P.; Zhang, J.Q. Control of Shrinkage Porosity and Spot Segregation in Ø195 mm
Continuously Cast Round Bloom of Oil Pipe Steel by Soft Reduction. Metals 2021, 11, 9. [CrossRef]
5. Yang, F.F.; Zhang, Z.H.; Liu, H.S.; Luo, M.; Li, G.; Zhang, J.Q. Banded segregation and its related band-typed mixed grain structure
in high strength oil well pipe steel. J. Iron Steel Res. 2021, 33, 979–986.
6. Xia, Y.; Li, L.; Wang, P.; Tie, Z.P.; Lan, P.; Tang, H.Y.; Zhang, J.Q. Characteristics of the as-cast high-carbon microalloyed continuous
casting bloom steel for expansion-break connecting rods. Chin. J. Eng. 2022, 44, 189–197.
7. Merder, T.; Warzecha, M. Optimization of a Six-Strand Continuous Casting Tundish: Industrial Measurements and Numerical
Investigation of the Tundish Modifications. Metall. Trans. B 2012, 43, 856–868. [CrossRef]
8. Li, S.X.; Tie, Z.P.; Zheng, B.A.; Kou, Y.S.; Yan, Q.Z.; Tang, H.Y.; Zhang, W.; Zhang, J.Q. Numerical simulation for an optimized
flow field in a three–strand tundish of bloom casting. Contin. Cast. 2018, 44, 47–53.
9. Ai, X.G.; Dong, H.; Li, S.L.; Zeng, H.B.; Li, H.Y. Optimization of flow uniformity control device for six–stream continuous casting
tundish. J. Iron Steel Res. Int. 2020, 27, 1035–1044. [CrossRef]
10. Tang, H.Y.; Li, X.S.; Zhang, S.; Zhang, J.Q. Fluid Flow and Heat Transfer in a Tundish with Channel Induction Heating for
Sequence Casting with a Constant Superheat Control. Acta Metall. Sin. 2020, 56, 1629–1642.
11. Deng, W.; Niu, S.; Yang, E.J.; Wang, D.J.; Wang, K.M.; Tang, H.Y. Study on an identical metallurgical effect for muti-strand tundish
during special steel bloom casting. J. Iron Steel Res. 2021, 33, 1144–1153.
12. Manish, M.; Tripathi, K.P.; Hosahali, T.; Kumar, D.S.; Konangi, R. Reduction in skull loss in billet caster tundish through water
modelling studies. Ironmak. Steelmak. 2021, 48, 803–810. [CrossRef]
13. Tang, H.Y.; Guo, L.Z.; Wu, G.H.; Xiao, H.; Yao, H.Y.; Zhang, J.Q. Hydrodynamic Modeling and Mathematical Simulation on Flow
Field and Inclusion Removal in a Seven–Strand Continuous Casting Tundish with Channel Type Induction Heating. Metals 2018,
8, 374. [CrossRef]
14. Ranjan, S.; Mukherjee, S.; Agnihotri, A.; Mazumder, D. Physical Modelling of Nitrogen Variation in Continuously Cast Blooms
Resulting from Atmospheric Exposure of Steel in Tundish during Initial Stages of Teeming. ISIJ Int. 2022, 62, 609–612. [CrossRef]
15. Sheng, D.Y.; Chen, D.F. Comparison of Fluid Flow and Temperature Distribution in a Single-Strand Tundish with Different Flow
Control Devices. Metals 2021, 11, 796. [CrossRef]
16. Yue, Q.; Zhang, C.B.; Pei, X.H. Magnetohydrodynamic flows and heat transfer in a twin-channel induction heating tundish.
Ironmak. Steelmak. 2017, 44, 227–236. [CrossRef]
17. Zhang, B.L.; Liu, F.H.; Zhu, R.; Zhu, J.F. Effects of Multiple-Hole Baffle Arrangements on Flow Fields in a Five-Strand Asymmetric
Tundish. Materials 2020, 13, 5129. [CrossRef]
18. Tang, H.Y.; Yu, M.; Li, J.S.; Bao, Y.P.; Jiang, J.; Ji, C.B.; Li, T. Numerical and physical simulation on inner structure optimization of a
continuous casting tundish and its metallurgical effect. J. Univ. Sci. Technol. Beijing 2009, 31, 38–42.
19. Chen, Y.; Ou, X.D.; Wei, H.R.; Xue, L.Q.; Yuan, P.F. Physical simulation and application of flow control devices optimization in
two-strand tundish. Contin. Cast. 2019, 44, 75–78.
20. Kumar, A.; Mazumdar, D.; Koria, S.C. Modeling of Fluid Flow and Residence Time Distribution in a Four-strand Tundish for
Enhancing Inclusion Removal. ISIJ Int. 2008, 48, 38–47. [CrossRef]
21. Merder, T.; Pieprzyca, J.; Warzecha, M.; Warzecha, P.; Hutny, A. Evolution of the Numerical Model Describing the Distributionof
Non-Metallic Inclusions in the Tundish. Materials 2021, 14, 2229. [CrossRef] [PubMed]
22. Sheng, D.Y.; Jönsson, P. Effect of Thermal Buoyancy on Fluid Flow and Residence-Time Distribution in a Single-Strand Tundish.
23. Sheng, D.Y. Synthesis of a CFD Benchmark Exercise: Examining Fluid Flow and Residence-Time Distribution in a Water Model of
Tundish. Materials 2021, 14, 5453. [CrossRef] [PubMed]
24. Sahai, Y.; Emi, T. Melt Flow Characterization in Continuous Casting Tundishes. ISIJ Int. 1996, 36, 667–672. [CrossRef]
25. Launder, B.E.; Spalding, D.B. The numerical computation of turbulent flows. Comput. Methods Appl. Mech. Eng. 1974, 3, 269–289.
[CrossRef]
26. Wang, Y.; Song, J.X.; Cheng, N.L.; Guo, Z.H.; Li, J.S.; Yang, S.F.; Zhao, M.J.; Wang, C. Application of Graphite Electrode Plasma
Heating Technology in Continuous Casting. Materials 2022, 15, 2590. [CrossRef]
47
materials
Article
Mesoscopic Fluid-Particle Flow and Vortex Structural
Transmission in a Submerged Entry Nozzle of
Continuous Caster
Peng Zhao 1 , Rongxun Piao 2, * and Zongshu Zou 3
1 School of Materials Science and Engineering, Shandong Jianzhu University, Jinan 250101, China;
[email protected]
2 Department of Mechanical Engineering, Anhui University of Science and Technology, Huainan 232001, China
3 School of Metallurgy, Northeastern University, Shenyang 110819, China; [email protected]
Abstract: Understanding the essence of the flow oscillations within a submerged-entry nozzle (SEN)
is essential to control flow patterns in the continuous casting mold and consequently increase the
superficial quality of steel products. A numerical study of the mesoscopic fluid-particle flow in a
bifurcated pool-type SEN under steady operating conditions is conducted using the lattice Boltzmann
method (LBM) coupled with the large eddy simulation (LES) model. The accuracy of the model has
been verified by comparing vortex structures and simulated velocities with published experimental
values. The LBM modeling is also verified by comparing the “stair-step” jet patterns observed in
the experiment. The geometrical parameters and operational conditions of physical experiments are
reproduced in the simulations. By comparing the time-averaged velocities of Reynolds-averaged
Navier–Stokes equations (RANS) with LBM models, transient mesoscopic fluid-particles and related
vortex structures can be better reproduced within the SEN. The visualization of internal flow within
the SEN is illustrated through the mass-less Discrete Phase Model (DPM) model. The trajectories show
Citation: Zhao, P.; Piao, R.; Zou, Z.
that the LBM–LES–DPM coupled model is good at predicting the transient vortical flow within the
Mesoscopic Fluid-Particle Flow and
SEN. A large vortex is found inside the exit port and continuously changes in shape and size therein.
Vortex Structural Transmission in a
The monitoring points and lines within the SEN are selected to illustrate the velocity variations
Submerged Entry Nozzle of
Continuous Caster. Materials 2022, 15,
and effective viscosity, which can reflect the oscillating characteristics even under stable operating
2510. https://doi.org/10.3390/ conditions without changes at the exit from the SEN. Furthermore, the formation, development,
ma15072510 diffusion, and dissipation of the vortex structures from the exit port of the SEN are also investigated
using the Q criteria. The comparison of the power spectrum with high-frequency components along
the exit port indicates that the flow oscillations must originate from within the SEN and are intensified
Jiangshan Zhang
in the exit port. The mesoscopic LBM model can replicate the fluid-particle flow and vortex structure
Received: 31 January 2022 transmission as well as their turbulence effects inside the SEN in detail.
Accepted: 23 March 2022
Published: 29 March 2022 Keywords: fluid-particle flow; submerged entry nozzle; continuous casting; lattice Boltzmann
Publisher’s Note: MDPI stays neutral method; large eddy simulation; vortex structures
iations.
1. Introduction
A common element in flow networks (industrial piping systems) is a T-junction that
splits the flow into two nearly symmetric streams. What these systems often have in
common is the presence of low-density particles or air bubbles, which get trapped at the
junction, accumulate, and ultimately change the flow distribution. The unanticipated
trapping mechanism can create malfunctions, and unexpected dangers in industrial envi-
ronments [1]. The flow behavior presented inside the SEN has similarities to the fluid flow
Attribution (CC BY) license (https:// behavior in other engineering fields. One of the most significant challenges in continuous
creativecommons.org/licenses/by/ casting is to obtain clean steel. The quality of steel produced by continuous casting depends
4.0/). mainly on the characteristics of the liquid steel flow pattern within the mold. This pattern

Materials 2022, 15, 2510
depends on the flow dynamics of the nozzle that is immersed in liquid steel. The behavior
of the transient flow within the SEN has a determinant influence on the flow pattern, which
was believed to be associated with the internal and superficial steel quality, such as its
effects on asymmetrical defects such as slip cracking, pinholes, and blisters [2–7]. Conse-
quently, these results show the importance of an accurate description of the behavior of the
flow. Therefore, a better understanding of the flow structures and development of vortices
within the SEN can be beneficial to quality control and improvements of the process such as
reducing energy consumption and the cost of high-quality slabs and foundry alloy (silicon
aluminum alloys) [8].
Since the effect of SEN flow on mold flow is difficult to detect under real operating
conditions, experimental models have been applied to investigate the flow of steel in
practice. As a supplement to experimental methods of investigation, the flow within the
SEN also has been subject to numerical simulations. To reproduce the flow pattern in an
SEN, the behavior of the fluid inside the SEN has been studied using the models based
on the Reynolds-averaged Navier–Stokes equations (RANS), such as the k-ε turbulence
model [9–12]: RANS models can reproduce time-averaged average flow patterns. The
symmetrical flow inside the SEN is obtained when the velocity inlet is aligned perfectly with
the SEN bore, however, they do not allow replication of transient dynamic flow behaviors.
The observations in steel-making and physical simulations show that the transient behavior
inside the SEN ports evolves with time. If gas bubbles or inclusions are introduced into
SEN flow, the simulations will become significantly more complicated. Common to these
works is the investigation of fundamental aspects that modify the flow pattern inside the
mold. For example, Calderon-Ramos et al. [13] studied the effect of both the shape of the
exit ports and the angle of inclination of the ports on symmetric jets. Similarly, Zhang
et al. [14] studied the influence of the port angle and immersion depth of the SEN exit ports
on the liquid steel flow pattern, which can reduce mold-level fluctuations.
Compared with the RANS model that provides time-averaged variables of turbulent
flows, the LES model can provide more detailed instantaneous turbulence to capture small
vortices better. In the LES model, large-scale eddies are solved directly with the filtered
Navier–Stokes equations, while small-scale eddies are modeled as sub-grid scale grids.
SEN flow and mold flow have been further investigated through the LES model on regular
grids. Similarly, the presence of unsteady flow inside the SEN has also been described
in previous work, allowing comparison of their results with physical experiments. For
example, Real et al. [15] studied the periodic flow behavior inside the SEN using the LES
model. Flow separation occurs in the upper region of SEN, where the area suddenly
increases in channels. There are differences in the influence that each of these geometric
properties has on the flow pattern of liquid steel inside the nozzle. Liu et al. [16] further
developed computational fluid dynamics (CFD) methods by coupling them with LES
models to simulate asymmetrical phenomena at both sides of the mold. Gonzalez-Trejo
et al. [17] characterized the fluid dynamics of two separate SENs without exit ports, one
with a well and the other without a bottom well, through LES with dynamic k-equation
filtering. The model reproduced the dynamic nature of the internal flow pattern seen in
physical experiments, however, only the bifurcation of the flow of steel in nozzle interior
was examined and the effects of the port shape were not considered in their research.
Recently, the mesoscopic LBM model has been used to demonstrate micro-fluidics
and complicated turbulent flows: this might be of interest to engineers running continuous
casting operations. Pirker et al. [18] predicted the horizontal secondary vortices and bubble
aggregation through the embedded lattice Boltzmann sub-region positioned in the bottom
SEN region. The author [19,20] predicted the complicated flow pattern and coherent
vortices inside a continuous casting mold. These studies have produced significant results,
reflecting the potential advantages that predict local complicated flow behaviors. Flow
structures play an essential role in complex flow regimes inside the SEN; however, previous
work has paid little attention to the exploration of interior asymmetrical flow structure and
their turbulent interaction mechanisms using the LBM–LES model inside the SEN.
50
Materials 2022, 15, 2510
The present work explores the fluid-particle flow and vortex structure transmission
in jets that emerge from the nozzle ports as well as their turbulent effects. The effects of
the tundish sliding nozzle and argon bubbles on liquid steel flow or any change in the
operational parameters are not included. The bifurcated pool-type SEN and rectangular
geometry of the SEN ports should be considered in this simulation. The fluid-particle flow
within the SEN can be reproduced using the LBM model. The accuracy of the model is
verified by comparing vortex structures and simulated velocities with results reported in
the literature. The transmission of the vortex structures along the exit port of the SEN is
also explored by the Q-criterion. The monitoring points and lines are finally selected to
study asymmetrical turbulent flow at different positions of the exit port inside the SEN.
2. Model Formulation
2.1. Lattice Boltzmann Method
Lattice Boltzmann Method is given by [21]:
1 →
f i ( x + ci , t + δt) − f i ( x, t) = ( f ( x, t) − f i eq ( x, t)) + F i (1)
τLB i
where ci represents the discrete lattice velocities; δ denotes the time step; f i ( x, t) is the
→
discretized distribution functions; τLB represents the dimensionless relaxation time; F is
the gravitational force adopted into the LBM model; f i eq ( x, t) is Maxwell distribution
equilibrium function given by:

c · u ( c · u )2 u2
f i eq ( x, t) = ρwi 1 + i 2 + i 4 − (2)
Cs 2Cs 2Cs2
√
where Cs is the sound speed, Cs = 1/ 3; wi is weight according to the lattice discretization,
and i range from 0 to 18, as follows:
⎧ ⎫
⎪
⎨
1
3 i=0 ⎪
⎬
wi = 1
i = 1–6 (3)
⎪
⎩
18 ⎪
⎭
1
36 i = 7–8
The model D3Q19 is applied for the LBM model, and the collision operator employs
the Bhatnagar-Gross-Krook (BGK) model with a single relaxation time [22].
The macroscopic quantities are calculated by taking the moments of distribution
functions:
ρ = ∑ fi (4)
i
ρu = ∑ ci f i (5)
i
where ρ is the macroscopic density; u is the macroscopic velocity.
2.2. Large Eddy Simulation

The subgrid-scale stresses τ ij is defined by Smagorinsky subgrid model [23]:
1
τij − δij τkk = −2υSij (6)
3
where δij is the Kronecker delta function, the isotropic part of the Reynolds stress term τkk
is included in the pressure term; the subscripts i and j represent three Cartesian directions.
51
Materials 2022, 15, 2510
Sij is expressed as the rate of large-scale strain tensor given by

1 ∂ui ∂u j
Sij = + (7)
2 ∂x j ∂xi
The viscosity υ in the Smagorinsky model is defined as:
υ = υ0 + υeddy (8)
where υ0 is the kinematic viscosity; υeddy is eddy viscosity given by:

υeddy = CΔ2 S (9)
where Δ is expressed as the minimum size; C is the Smagorinsky constant, which depends
on the size of the grid, and the value ranges between 0.1 and 0.2.
Relaxation time τLB can be expressed as:
1
τLB = 3 ν0 + CΔ2 S + (10)
2

where S is the intensity of the local filtered stress tensor given by:

υ02 + 18CΔ2 ∏i,j ∏i,j − υ0
S = (11)
6CΔ2

eq
where ∏i,j = ∑ cαi cαj f i − f i is the local non-equilibrium stress tensor.
α
3. Model Setup and Validation

3.1. Calculation Details
The computational domain includes a bifurcated SEN nozzle with the bottom pool.
The global computational domain comprises the incoming duct, the flow deflection region,
the two opposing port openings, and part of the emerging confined jets. Table 1 lists
the geometrical SEN, simulation conditions, and operating parameters considered in the
simulation.The lattice spacing is set to 2 mm. This computational domain is discretized by
approximately 500,000 elements in the LBM model. The time step for these simulations
is 0.02 s. The resolved scale is 0.002 m. The starting time step for these simulations is
0.001 s. The inlet and outlet boundary conditions in the model are based on the bounce-
back condition for the non-equilibrium portion of the distribution. The constant velocity
inlet boundary condition is based on the casting speed at the inlet of SEN. The pressure
outlet condition is applied in the exit outlet at both sides of the SEN. Figure 1 shows a
schematic diagram of the geometry and boundary conditions as well as monitoring points
and lines within the SEN considered in the study.
Table 1. Simulation conditions considered in the model.
Parameter Value
Mold width (m) 1.2
Mold thickness (m) 0.23
Casting speed (m·min−1 ) 1.3
SEN submergence depth (mm) 200
SEN port angle (deg) 15◦
SEN port shape Rectangle
SEN port height (mm) 45
SEN port width (mm) 35
Molten steel density (kg·m−3 ) 7000
Molten steel viscosity (Pa·s) 0.006
52
Materials 2022, 15, 2510
Figure 1. SEN geometry and two monitoring points.
3.2. Validation
The complex flow consists of multiple vortices in the mold. The behaviors of the jet
emerging from the exit port of the SEN play an essential role in the flow pattern in the
continuous casting mold. The jet vortices induced by the Q-criterion are investigated using
the LBM model. The Q-criterion is given by [24]:
1 ∂ui ∂u j 1
Q=− = Ωij Ωij − Ψij Ψij > 0 (12)
2 ∂x j ∂xi 2
where Ωij and Ψij are the symmetric and anti-symmetric components of ∇u, respectively.
Thus, the evolution of structures flow is deduced from the Q-criterion.
The models have been applied to a one-half scale water model for comparison, and
the details of geometry and operating conditions are treated as being under the same
conditions. Simulated velocity profiles are compared with previous experimental measure-
ments. Figure 2a–c illustrates a comparison of the “stair-step” jet outside the SEN through
(a) water experiment, (b) predicted velocity, and (c) vortex structures. The simulation
results of the patterns and vortex structures of the jets also agree with the observations
from water experiments, which reproduce the “stair-step” jets at two sides of the SEN
when black ink is injected into the upper SEN tube. By comparing the simulation and the
water experiment, the LBM model shows that it can be used to model the behavior of the
oscillating jets within the SEN.
53
Materials 2022, 15, 2510
Figure 2. Comparison of the “stair-step” jet inside the mold under the same conditions through
(a) the water experiment, (b) predicted velocity, and (c) vortex structures.
To verify the accuracy of the modeling the interior flow inside the SEN, the LBM
model also has been applied to the water experiments for a comparison of experimental
velocities and simulation results. Simulated velocity profiles are compared with those
measured results from the water experiment under the same conditions, and the details of
geometry and operating conditions are available in the literature [17]. Figure 3 illustrates
a comparison of the turbulent fluid flow (left) through the LBM model with the vortex
structures outlined by the fluid velocity vectors for Case A (right) inside the nozzle internal
prototype (NIP). The NIP is used to evaluate the vortex distribution. As can be seen from
the figures, there is a dominant vortex that occupies a large fraction of the volume of the
nozzle. Figure 3a shows the vortex structure (isosurface of v = 1.2 m/s) in this simulation
close to the exit port of the SEN. Vortex tube outlined by the velocity vectors is present in
the same projection in Figure 3b. This surface outlines the shape of the vortex that can be
produced inside the NIP. The comparison shows the vortex structures in this model are
consistent with the vortex surface along with the velocity vectors in the previous simulation
for Case A with the well pool. The LBM model can be validated to obtain internal flow and
structure inside the SEN.
54
Materials 2022, 15, 2510
Figure 3. Comparison of (a) turbulent fluid flow pattern through the LBM model (left) with (b) vortex
structure outlined by the fluid velocity vectors for Case A [17] (right) inside the NIP.
To compare the velocity of the flow inside the NIP, the velocity magnitudes along two
straight lines from the physical model are available in the literature [17]. Figure 4 illustrates
a comparison of experimental results along the path in Case A and simulated time-averaged
velocities that measured three times under the same conditions. The measurements are
undertaken along vertical lines close to the left side of the exit nozzle: these show the shape
of the velocity profiles changes at different heights of the SEN. The velocity fluctuates
significantly at the nozzle exit. The predicted velocity increases with the height of the exit
port of the nozzle until it reaches a maximum value and then decreases with the height
thereof. In the right-hand figure, the agreement between both methods is not as good as
the velocity profile across the height of the exit port; however, the result reproduces the
velocity profile as a function of SEN height, qualitative agreement is obtained.
Figure 4. Comparison of experimental results for Case A [17] and predicted velocities prediction at
the left-hand line of the SEN.

4.1. Fluid-Particle Flow
The LBM model for the fluid-particle flow is also conducted in this simulation. To show
the transient process of fluid-particle flow in the SEN under the steady-state conditions,
Figure 5a–d show the fluid-particle flow patterns at 0.1, 0.2, 0.25, 0.3, and 0.35 s, respectively.
The transient process of fluid-particles can be described as follows: transient fluid-particles
emerge from the inlet of the SEN bore. The upper SEN pipe is filled with the fluid-particles;
55
Materials 2022, 15, 2510
thereafter, the particle-based jet impinges onto the bottom of the SEN pool at a high velocity,
and flow towards two sides of the SEN port. Meanwhile, the fluid-particles at the SEN
pool interact with those from the upper tube; finally, the fluid-particles almost fill the entire
the SEN port, meanwhile, backflow is also found in the upper region of the SEN port,
as shown in Figure 5e. The fluid-particles at the SEN port interact with those within the
backflow, resulting in more complex transient flow within the SEN. The results show that
the mesoscopic LBM model can obtain the transient behavior of fluid-particle flow within
the SEN in detail.
Figure 5. Transient evolution of mesoscopic fluid-particle flow inside the SEN at different times of
(a) 0.1, (b) 0.2, (c) 0.25, (d) 0.3, and (e) 0.35 s.
Passive stream-tracers allow the simulation to track Lagrangian particles (mass-less)

along the fluid, which is thus useful to better observe the streamlines lines or flow trajec-
tories. The mass-less Discrete Phase Model (DPM) is used here to describe the vortical
flow within the SEN. The visualization of internal flow within the SEN is illustrated in the
following process (Figure 6a–d): Mass-less particles start to be separated at the corners
between the upper tube and the exit port of the SEN. Then, some particles flow directly
towards the exit port, while others move into the lower region until they impinge on the
bottom wall of the SEN. These mass-less particles finally flow along the bottom region
toward the exit port, accompanied by swirling particles-flowlines close to port of the SEN
(see enlarged area Figure 6d). The trajectories show that the LBM–LES–DPM coupled model
is good at predicting the transient vortical flow within the SEN because the inclusions are
found dispersed at a fairly low volume fraction (usually less than 20%) in the steel through
actual industrial experiment.
56
Materials 2022, 15, 2510
Figure 6. The evolution of passive particles flow and stream-tracers within the SEN at different stages:
(a) corner separation, (b) central diffusion, (c) bottom diffusion, and (d) vortical flow.
In order to better explain the mechanism of internal vortices structures inside SEN,
Figure 7a–c schematically shows the formation of internal flow corresponding coherent
vortices inside the SEN. The main processes are as follows: (a) Vortex structures are formed
at upper region, meanwhile, they interact with the backflow zone at the upper region of the
SEN; (b) vortex structures are concentrated near the concave shape, and they also interact
with the incoming flow from the upper tube; (c) vortices structures from the upper region
interact with those at the bottom concave, which can cause a more complex transient flow
inside the SEN. In short, fluid separation occurs at the upper separation and the lower
concave of the SEN. Vortex vortices split into small vortices and interact with each other,
leading to the accumulation in these regions and intensifying the oscillating jets from the
SEN port thereon.
To quantify the time-averaged velocity of this mesoscopic fluid, Figure 8a illustrates
the time-averaged velocity contour on the middle plane (XZ). The simulation results are
measured three times with an interval of 10 s. These maps indicate that the flow patterns
in both exit ports are identical. The symmetric flows on both sides of the SEN ports are
present. Additionally, the flow velocity contour and vectors at the exit port of the SEN
are also included in Figure 8b. The corresponding value of X for the plane is −0.03 m.
Note that the area has a very low velocity along the upper edges of the exit ports, while
fluid flow with high velocity is uniformly distributed at the bottom of the exit ports. A
close inspection of the low-velocity vectors directed towards the interior at the upper SEN
shows the existence of reverse flow (Figure 8c), which promotes the oscillation of the jet as
it emerges from the exit port of the SEN.
57
Materials 2022, 15, 2510
Figure 7. Schematic streamlines of the formation and development of internal flow corresponding
coherent vortices inside the SEN at different stages: (a) corner separation, (b) bottom diffusion, and
(c) vortical flow.
Figure 8. The time-averaged velocity contour and vectors at the exit port of the SEN from different
views: (a) the middle plane XZ, (b) the plane XY, and (c) the exit port.
The better to show the evolution of flow structure at the exit port, Figure 9a–c shows
transient vortex contour and trace streams perpendicular to the exit port of the SEN under
the same condition. The simulation takes, as an example, the situation where a large vortex
is generated inside the SEN with a well pool. The analysis of the simulation led us to the
following observations: the big vortex located close to the bottom of the SEN is always
present and is cylindrical, however, there is always a bias of the vortex axis with respect
to the central plane of the SEN. The rotation axis of the large vortex is not substantially
parallel to that of the exit ports. This large vortex oscillates continuously, changing in both
shape and size.
To quantify the turbulent flow close to the exit port, the effective viscosities (on the
monitoring lines) at both sides of the exit points are investigated. The locations of the two
reference lines are illustrated in Figure 1. Figures 10 and 11 show the effective viscosities
of the turbulent flow along the vertical monitoring lines close to exit port of the SEN at
different times (Figure 1). From the figures, the turbulent viscosity of flow near the wall of
the mold always exhibits fluctuation characteristics at both sides of the exit port. Figure 10
58
Materials 2022, 15, 2510
demonstrates the asymmetric viscosity at both sides, where the viscosity on the left upper
port is higher than that on the right. Figure 11 shows a similar mirror image of the viscosity
distribution, where the viscosity on the right upper port is higher than that on the left. It
can be inferred that the viscosity change is related to the wobbling jet behavior, reflecting
asymmetric changes in flow patterns from the exit port.
Figure 9. Transient vortex contours and trace streams on the center plane perpendicular to the exit
port of the SEN at different times: (a) the large vortex appears at the bottom, and (b) the vortex
gradually flows upward, (c) the vortex moves towards the middle output.
Figure 10. Asymmetric effective viscosities of the turbulent flow along the vertical monitoring lines
close to the exit port of the SEN.
Transient fluid-particle flow within the SEN is reproduced, the oscillation of fluid-
particle flow is quantitatively analysed in this section. The periodic behavior within the
SEN will be characterised mathematically here. The velocity magnitude fluctuations at
monitoring Points 1 and 2, located at the upper corner and the middle along the exit port
of the SEN (Figure 1), respectively. Figure 12 shows the comparison between the velocity
time series at Point 1 using different models. Numerical simulations using the RANS (k-ε
turbulent) model are conducted to compare the transient flow within the SEN through LBM
model. The solid bold lines correspond to simulations using the LBM model, and dotted
lines represent the RANS model. The RANS results illustrate that the velocity fluctuations
at the exit port are relatively stable, which are not as severe as those obtained from the
59
Materials 2022, 15, 2510
LBM results. The behavior of mesoscopic fluid-particles inside the SEN exhibits periodic
changes.
Figure 11. Mirror image of effective viscosities of the turbulent flow along the vertical monitoring
lines close to the exit port of the SEN.
Figure 12. Comparison between velocity time series at Point 1 using the LBM and RANS models.
All numerical simulations presented are performed under the same the conditions.
Figure 13 demonstrates the velocity change for Point 2 through the LBM and RANS
models. In the simulation of the RANS model, the dynamic change in the SEN is not
evident; however, the LBM simulation also demonstrates that the flow behavior in periodic
variation predicts a greater velocity oscillation close to the port (Point 2) than one at the
corner (Point 1) within the SEN, which promotes the oscillations of the jet thereon. From
the above comparison, the results show that the flow oscillations initially occur on the
interior bifurcation tip due to flow separation, where the area suddenly changes in the
shape of SEN geometry. The oscillations of the jet are intensified when the internal flow
develop along the exit port of the SEN.
60
Materials 2022, 15, 2510
Figure 13. Comparison between velocity time series at Point 2 using LBM and RANS models.
4.2. Flow Structures

The vortex structures play a significant role in the flow pattern on both sides of the
SEN. To clarify this, the process of vortices development within the SEN is schematically
divided into four stages, as shown in Figure 14a–c:
(I) Formation of the vortices: the vortices emerge at the upper corner of the SEN. The
area of a flow suddenly increases in channels, and separate at the boundary layer;
(II) Development of the vortices: the size of the vortices increases gradually. The vortices
develop along with the exit port and produce more vortices inside the SEN;
(III) Dissipation of vortices: the vortices interact with each other and disappear inside the
SEN. The vortices persist for a relatively short period.
Figure 14. Formation, development, and dissipation of the vortices at the channel of the SEN at
different stages: (a) formation of the vortices, (b) development of the vortices, and (c) dissipation
of vortices.
Differing from those studies, in this simulation, we propose that the evolution of the
vortex structures is induced in terms of satisfaction of the Q-criterion. Figures 15–18 show
the vortices induced by the Q-criterion and corresponding streamlines. To illustrate the
61
Materials 2022, 15, 2510
evolution of the vortices, Figure 15 illustrates that the vortices with the greatest strength
are mainly concentrated at the upper corner of the SEN. The intensities (driving force) of
the vortices are relatively large, which are of great importance to the evolution of vortex
structures. Next, the corresponding vortices propagate from the upper corner to the exit
port of the SEN, as shown in Figure 8. After that, the vortices then spread and deform
along with the exit port, as shown in Figures 16 and 17. The small vortices are finally
separated from their larger counterparts close to the exit port, increasing the instability of
the jet flow from the exit port. In addition, the separated vortices are also found (Figure 18),
accompanied by small swirling flows, which are consistent with the phenomenon of the
“stepped” jets that are observed in a water experiment.
Figure 15. Formation of vortex structures (a) at the interior corner of the SEN port and (b) corre-
sponding streamlines.
Figure 16. Development of vortex structures (a) at the upper edge of the SEN port (b) corresponding
streamlines.
62
Materials 2022, 15, 2510
Figure 17. Diffusion of vortices (a) along with the channel of SEN port and (b) corresponding
streamlines.
Figure 18. Separation of vortex structures (a) nearby the SEN port and (b) corresponding streamlines
as well as accompanying swirling flow.
To investigate the oscillating flow, the velocity time series obtained from the LBM
model is processed into a power spectrum to estimate the oscillation period of the circu-
lation. These samples are then processed into a power spectrum by using a fast Fourier
transform (FFT) routine as defined by [25]:
1
P( f k ) = | f |2 k = 1, 2, . . . , N/2 (13)
N2 k
N
Uk = ∑ ue2πijk/N k = 1, 2, . . . , N (14)
k
fk = k = 1, 2, . . . , N/2 (15)
NΔt N
63
Materials 2022, 15, 2510
where u is the velocity in the time series, tN denotes the times at the start and the end
of the sampling period; N represents the number of sampling points. Thus, a frequency
analysis of the velocity magnitude time series is conducted. In order to study the oscillating
flow along the exit port of the SEN, frequencies of velocity fluctuation are analyzed at
monitoring points P1, and P2 (Figure 1). Figures 19 and 20 show the power spectrum of
velocity magnitude with the SEN port at the corner (Point 1) and the exit port (Point 2),
respectively. The dominant peak represents the oscillation frequency. There are many
frequencies in the flow at the SEN exit port. The oscillations are not a simple combination
of modes inside the SEN. The maximum amplitude is 0.03 at frequency 1 Hz at Point 1,
however, this exceeds 0.07 at Point 2. It can be concluded that the flow oscillations have
been intensified close to the exit port of the SEN.
Figure 19. Power spectrum of velocity magnitude inside the exit point at Point 1.
Figure 20. Power spectrum of velocity magnitude inside the exit point at Point 2.
64
Materials 2022, 15, 2510
5. Conclusions
The fluid-particle flow inside a bifurcated SEN is studied to evaluate flow oscillations
through the LBM model. From the results, the following conclusions are drawn:
(1) The accuracy of the model has been verified by comparing vortex structures and
simulated velocities with published experimental values inside and outside the SEN.
Comparing simulations with RANS and LBM models, the results show that the
mesoscopic LBM can predict the transient fluid-particle flow along the exit port of the
SEN in detail, which can better reproduce the behavior of the oscillating jet within
the SEN.
(2) The transient flow behavior results from the dynamic interaction of fluid-particle flow
within the SEN. The velocities always exhibit strong high-frequency components. The
periodic changes of mesoscopic fluid-particles flow within the SEN can be reproduced
even under stable operating conditions. The visualization of internal flow within the
SEN show that the LBM–LES–DPM coupled model is good at predicting the transient
vortical flow within the SEN.
(3) Vortex structures are of great importance to the development of turbulent flow along
the exit port of the SEN. The turbulent viscosity of flow near exit output of the SEN
always fluctuates at both sides of the exit port. The asymmetric viscosities at both
sides become similarly mirror-like at different times. The formation, development,
and dissipation of detached vortices induced by the Q-criterion that identifies vortex
structures can cause oscillation of jet within the SEN.
(4) There are many frequencies (between 1 Hz and 3 Hz) seen in flow oscillations along
the channel of the SEN port. The oscillations are not a simple combination of modes
and flow oscillations have been intensified simultaneously. The jet oscillations must
originate from within the SEN interior and are intensified along the exit port.
This work represents our further efforts towards fluid-particles flow, vortices structures
and their effects on turbulence inside bifurcated SEN flow. Further studies are needed
to study the behaviour of introduced particles or bubbles under the framework of LBM
model, the interaction between vortex structure and bubbles as well as their interactions
(aggregation and breakdown).
Author Contributions: Software, P.Z.; validation, P.Z.; investigation, Z.Z.; writing—original draft
preparation, P.Z.; writing—review and editing, R.P.; funding acquisition, R.P. All authors have read
and agreed to the published version of the manuscript.
51174122), and Anhui University of science and technology introduction of talent research start-up
fund project (No. 13200456).
Data Availability Statement: Not applicable.
Acknowledgments: National Engineering Laboratory for Vacuum Metallurgy, Kunming University
of Science and Technology for the supports.
References
1. Vigolo, D.; Radl, S.; Stone, H.A. Unexpected trapping of particles at a T junction. Proc. Natl. Acad. Sci. USA 2014, 111, 4770–4775.
[CrossRef] [PubMed]
2. Yang, J.P.; Liu, Q.; Guo, W.D.; Zhang, J.G. Quantitative evaluation of multi-process collaborative operation in steelmaking-
continuous casting sections. Int. J. Miner. Metall. Mater. 2021, 28, 1353–1366. [CrossRef]
3. Wu, Y.; Liu, Z.; Wang, F.; Li, B.; Gan, Y. Experimental investigation of trajectories, velocities and size distributions of bubbles in a
continuous-casting mold. Powder Technol. 2021, 387, 325–335. [CrossRef]
65
Materials 2022, 15, 2510
4. Liu, G.; Lu, H.; Li, B.; Ji, C.; Zhang, J.; Liu, Q.; Lei, Z. Influence of M-EMS on Fluid Flow and Initial Solidification in Slab
Continuous Casting. Materials 2021, 14, 3681. [CrossRef]
5. Hibbeler, L.C.; Thomas, B.G. Mold slag entrainment mechanisms in continuous casting molds. Iron Steel Technol. 2013, 10, 121–136.
6. Li, X.; Li, B.; Liu, Z.; Niu, R.; Liu, Y.; Zhao, C.; Yuan, T. Large eddy simulation of multi-phase flow and slag entrapment in a
continuous casting mold. Metals 2019, 9, 7. [CrossRef]
7. Zhao, P.; Li, Q.; Kuang, S.B.; Zou, Z. Mathematical modeling of liquid slag layer fluctuation and slag droplets entrainment in a
continuous casting mold based on VOF-LES method. High Temp. Mater. Process. 2017, 36, 551–565. [CrossRef]
8. Kuz’min, M.P.; Chu, P.K.; Qasim, A.M.; Larionov, L.M.; Kuz’mina, M.Y.; Kuz’min, P.B. Obtaining of AleSi foundry alloys using
amorphous microsilica e crystalline silicon production waste. J. Aalloys Compd. 2019, 806, 806–813. [CrossRef]
9. Thomas, B.G.; Yuan, Q.; Mahmood, S.; Liu, R.; Chaudhary, R. Transport and entrapment of particles in steel continuous casting.
10. Dash, S.K.; Mondal, S.S.; Ajmani, S.K. Mathematical simulation of surface wave created in a mold due to submerged entry nozzle.
Int. J. Numer. Methods Heat Fluid Flow 2004, 14, 606–632. [CrossRef]
11. Wondrak, T.; Eckert, S.; Gerbeth, G.; Klotsche, K.; Stefani, F.; Timmel, K.; Yin, W. Combined electromagnetic tomography for
determining two-phase flow characteristics in the submerged entry nozzle and in the mold of a continuous casting model. Metall.
Mater. Trans. B 2011, 42, 1201–1210. [CrossRef]
12. Torres-Alonso, E.; Morales, R.D.; Garcia-Hernandez, S. Cyclic turbulent instabilities in a thin slab mold. Part II: Mathematical
model. Metall. Mater. Trans. B 2010, 41, 675–690. [CrossRef]
13. Calderón-Ramos, I.; Morales, R.D.; Servín-Castañeda, R.; Pérez-Alvarado, A.; García-Hernández, S.; de Jesús Barreto, J.; Arreola-
Villa, S.A. Modeling study of turbulent flow in a continuous casting slab mold comparing three ports SEN designs. ISIJ Int. 2019,
59, 76–85. [CrossRef]
14. Zhang, H.; Fang, Q.; Xiao, T.; Ni, H.; Liu, C. Optimization of the flow in a slab mold with argon blowing by divergent bifurcated
SEN. ISIJ Int. 2019, 59, 86–92. [CrossRef]
15. Real, C.; Miranda, R.; Vilchis, C.; Barron, M.; Hoyos, L.; Gonzalez, J. Transient internal flow characterization of a bifurcated
submerged entry nozzle. ISIJ Int. 2006, 46, 1183–1191. [CrossRef]
16. Liu, Z.Q.; Li, B.K.; Jiang, M.F. Asymmetric flow and bubble transport inside a slab continuous-casting mold. Metall. Mater. Trans.
B 2014, 45, 675–697. [CrossRef]
17. Gonzalez-Trejo, J.; Real-Ramirez, C.A.; Carvajal-Mariscal, I.; Sanchez-Silva, F.; Cervantes-De-La-Torre, F.; Miranda-Tello, R.;
Gabbasov, R. Hydrodynamic analysis of the flow inside the submerged entry nozzle. Math. Probl. Eng. 2020, 2020, 6267472.
[CrossRef]
18. Pirker, S.; Kahrimanovic, D.; Schneiderbauer, S. Secondary vortex formation in bifurcated submerged entry nozzles: Numerical
simulation of gas bubble entrapment. Metall. Mater. Trans. B 2015, 46, 953–960. [CrossRef]
19. Zhao, P.; Li, Q.; Kuang, S.B.; Zou, Z. LBM-LES simulation of the transient asymmetric flow and free surface fluctuations under
steady operating conditions of slab continuous casting process. Metall. Mater. Trans. B 2017, 48, 456–470. [CrossRef]
20. Zhao, P.; Yang, B.; Piao, R. Lattice Boltzmann Method Modeling of Flow Structures and Level Fluctuations in a Continuous
Casting Process. ACS Omega 2019, 4, 13131–13142. [CrossRef]
21. He, X.; Luo, L.S. Theory of the lattice Boltzmann method: From the Boltzmann equation to the lattice Boltzmann equation. Phys.
Rev. E 1997, 56, 6811–6817. [CrossRef]
22. Qian, Y.; D’Humières, D.; Lallemand, P. Lattice BGK models for Navier-Stokes equation. EPL Eur. Lett. 1992, 17, 479. [CrossRef]
23. Smagorinsky, J. General circulation experiments with the primitive equations: I. The basic experiment. Weather Rev. 1963, 91,
99–164. [CrossRef]
24. Hunt, J.C.; Wray, A.A.; Moin, P. Eddies, stream, and convergence zones in turbulent flows. In Studying Turbulence Using Numerical
Simulation Databases—II; Center for Turbulence Research: Stanford, CA, USA, 1988; pp. 193–208.
25. Lawson, N.J.; Davidson, M.R. Oscillatory flow in a physical model of a thin slab casting mold with a bifurcated submerged entry
nozzle. J. Fluids Eng. 2002, 124, 535–543. [CrossRef]
66
materials
Article
Influence of M-EMS on Fluid Flow and Initial Solidification in
Slab Continuous Casting
Guoliang Liu 1,2 , Haibiao Lu 3 , Bin Li 3 , Chenxi Ji 2 , Jiangshan Zhang 1 , Qing Liu 1, * and Zuosheng Lei 3, *
1 State Key Laboratory of Advanced Metallurgy, University of Science and Technology Beijing,
Beijing 100083, China; [email protected] (G.L.); [email protected] (J.Z.)
2 Shougang Group Co., Ltd., Research Institute of Technology, Beijing 100043, China;
[email protected]
3 State Key Laboratory of Advanced Special Steel & Shanghai Key Laboratory of Advanced Ferrometallurgy,
Shanghai University, Shanghai 200444, China; [email protected] (H.L.);
[email protected] (B.L.)
* Correspondence: [email protected] (Q.L.); [email protected] (Z.L.)
Abstract: A mathematical model coupled with electromagnetic field has been developed to simulate
the transient turbulence flow and initial solidification in a slab continuous casting mold under differ-
ent electromagnetic stirring (EMS) currents and casting speeds. Through comparing the magnetic
flux density, flow field with measured results, the reliability of the mathematical model is proved.
The uniform index of solidified shell thickness has been introduced to judge the uniformity of the
solidified shell. The results show that a horizonal recirculation flow has been generated when EMS
is applied, and either accelerated or decelerated regions of flow field are formed in the liquid pool.
Large EMS current and low casting speed may cause the plug flow near the mold narrow face and a
suitable EMS current can benefit to the uniform growth of solidified shell. Meanwhile, an industrial
test exhibits that EMS can weaken the level fluctuation and number density of inclusion. Overall, a
Citation: Liu, G.; Lu, H.; Li, B.; Ji, C.;
rational EMS current range is gained, when the casting speed is 1.2 m/min, the rational EMS current
Zhang, J.; Liu, Q.; Lei, Z. Influence of
is 500–600 A.
M-EMS on Fluid Flow and Initial
Solidification in Slab Continuous
Casting. Materials 2021, 14, 3681.
Keywords: continuous casting; electromagnetic stirring; flow behavior; uniform index of solidi-
https://doi.org/10.3390/ma14133681 fied shell
Academic Editor:
Francesco Iacoviello
1. Introduction
Received: 3 June 2021 Continuous casting has been widely developed as the most important production
Accepted: 26 June 2021 process in the steelmaking industry. During this process, molten steel flows into the mold
Published: 1 July 2021 through a submerged entry nozzle (SEN), solidifies against the water-cooled copper mold
walls in the presence of many complex metallurgical phenomena, including multiphase
Publisher’s Note: MDPI stays neutral flow, heat transfer, solidification and solute transport, and it finally forms a solidified
with regard to jurisdictional claims in shell [1]. Flow of molten steel in the mold and solidification in this process is of great
published maps and institutional affil- importance because it is responsible for many surface and internal defects [2], such as hooks
iations. and longitudinal cracks. Therefore, it is important to control the flow and solidification
within an acceptable process to avoid defects.
To enhance or even control the liquid steel flow, many techniques have been developed
and used, including argon blowing [3], electromagnetic brake [4,5] (EMBr), electromagnetic
Copyright: © 2021 by the authors. stirring [6–12] (EMS), etc. Together with these techniques, EMS has been shown to be
Licensee MDPI, Basel, Switzerland. one of the most effective countermeasures to improve the molten steel flow and initial
This article is an open access article solidification in the mold and most works focused mainly on the rotary stirring in billet or
distributed under the terms and round bloom.
conditions of the Creative Commons Li et al. [9] developed a mathematical model coupling the electromagnetic and flow
fields; it was developed to investigate the influence of the SEN clogging rate on the flow
field and the influence of electromagnetic stirring (EMS) on the asymmetric mold flow.
4.0/).

Materials 2021, 14, 3681
Fang et al. [13] and Ren et al. [14] simulated the flow, temperature and solidification field
in billet mold, and the results showed that EMS can reduce the impinging effect of jet flow
and eliminate molten steel superheat. Wang et al. [15] developed a 3-D coupled model
considering electromagnetic field, flow field, heat transfer and level fluctuation under
different EMS positions in continuous billet mold, and found that the velocity and wave
height at steel/slag interface decreases under lower stirrer position. Maurya et al. [16–18]
analyzed the influence of EMS on flow and solidified shell under different EMS currents
and frequencies. However, little research considered the influence of EMS on the flow and
initial solidification on slab.
In a series of works from Fujisaki [7,19,20] and his coworkers, a 3-D magnetohydro-
dynamic model was developed to evaluate the flow, heat transfer as well as free surface
in mold with EMS separately, and the results showed that EMS makes the solidified shell
uniform and the dynamic deviation of temperature stable. Li et al. [21] investigated the in-
fluence of EMS position on flow field, possibility of slag entrapment and inclusion removal
in slab mold, the flow pattern in the mold changes greatly under different EMS positions
and the low stirrer position is favorable to inclusion removal, but the solidified behavior
of shells are not considered in his study. In addition, most of them mainly considered the
influence of EMS current on the flow and heat transfer in the mold, while the influence of
EMS under other process parameters, such as casting speed, SEN depth, etc. has not yet
been investigated.
In this paper, the fluid flow and solidification in a continuous casting slab mold under
EMS were numerically simulated by a multi-physics model. By comparison with the
magnetic flux density data and flow field, the reliability of the mathematical model is
proved. The uniform index of solidified shell thickness has been introduced to judge the
uniformity or solidified shell. Afterwards, the metallurgical behaviors under different EMS
currents and casting speeds were investigated and compared to obtain optimal process
parameters with a relatively reasonable flow pattern and solidified shell. According to the
simulation results, the industry test was carried out to estimate the influence of EMS on
level fluctuation and number density of inclusion.
2. Mathematical Modeling
2.1. Assumption
In order to simplify the numerical simulation, the present work includes the following
assumptions and simplifications:
(1) The influence of flow field on the electromagnetic field is ignored due to the small
magnetic Reynolds number [22], and the electromagnetic field is assumed to be
quasi-static.
(2) The influence of Joule heat generated by currents is ignored in simulation of heat
transfer and solidification due to its low frequency.
(3) The liquid steel and the liquid slag behave as incompressible Newtonian fluids.
(4) The effects of mold oscillation and mold curvature are not taken into account [23].
2.2. Governing Equation

A transient three-dimensional mathematical model for slab continuous casting has
been developed which was coupled with electromagnetic field, flow and heat transfer. The
corresponding governing equations are written as follows:
2.2.1. Electromagnetic Model

Electromagnetic field was determined by solving Maxwell’s equations:
→
∇· B = 0 (1)
→ →
∇×H = J (2)
68
Materials 2021, 14, 3681
→
→ ∂B
∇× E = − (3)
∂t
! "
→ →
J =σ E (4)
→ →
where B is the magnetic flux density, σ is the electric conductivity, E is the electric field
→ →
strength, V/m; J is the induced current density and H is the magnetic field strength.
The time-averaged electromagnetic force can be calculated by:
→ 1 → →∗
F = Re( J × B ) (5)
2
→ →∗
where F is the time-average electromagnetic volume force, B is the complex conjugate of
→
B and Re denotes the real part of the complex quantity.
2.2.2. Fluid Flow and Solidification Model

Continuity equation:
→
∇·(ρ u ) = 0 (6)
Momentum equation (N-S):
→
∂(ρ u ) →→ → →
+ ∇·(ρ u u ) = −∇ p + ∇·(μe f f ∇ u ) + ρg + Sm + F (7)
∂t
→
where u is the fluid velocity, p is the pressure, ρ is the fluid density and μeff is the ef-
fective viscosity, the standard κ − ε turbulent model is applied to calculate the effective
viscosity, μeff .
Energy equation,

∂ → μt
(ρH ) + ∇·(ρ u H ) = ∇· (λ + C p )∇ T (8)
∂t Prt
where, T
H = hre f + c p dT + f l Ls (9)
Tre f
where λ is the thermal conductivity of the fluid, Cp is the specific heat capacity of the
fluid, Prt is the turbulence Prandtl number (0.85), H is enthalpy, href is the reference
enthalpy relative to temperature Tref , fl is liquid fraction at mushy zone and Ls is latent heat
of solidification.
The enthalpy-porous model is used to simulate the solidification of steel in a continu-
ous slab casting mold, the liquid-solid mushy zone is treated as a porous zone. The sink
(Sm ) is added to the momentum equations as a source term:
(1 − f l )2 →
Sm = A ( u − uc ) (10)
( f l 3 + x ) mush
where, Amush is a mushy zone constant, 108 [24] and uc is the casting speed.
2.3. Geometry Model and Boundary Conditions

2.3.1. Geometry Model
A three-dimensional mathematical model was created that consists of submerged
entry nozzle (SEN), Liu et al. [25] as the authors studied the chamfered slab mold in studied
metals, a chamfered slab mold with the section of 1600 mm × 230 mm, length of 800 mm;
considering the influence of lower recirculation flow on the flow field of molten steel in the
69
Materials 2021, 14, 3681
mold, we extended the mold model to 3000 mm. Figure 1a shows the geometry model of
an electromagnetic simulation, a pair of traveling-wave electromagnetic stirrers is designed
on both sides of the mold wide face. The schematic of fluid simulation is presented in
Figure 1b, and the other process parameters for numerical simulation are given in Table 1.
(a) (b)
Figure 1. Schematics of the calculation model: (a) model used for electromagnetic simulation (surrounding air cuboid is not
shown), (b) model used for fluid simulation.
Table 1. Process parameters used in mathematical simulation.
Parameters Value Parameters Value

Section size of mold (mm2 ) 1600 × 230 Casting speed (m/min) 0.8, 1.0, 1.2, 1.4
Size of SEN port (mm2 ) 80 × 60 Coil number of each stirrer 36
Outer diameter of SEN (mm) 120 Turn number of each coil 20
Stirrer center from
Inner diameter of SEN (mm) 80 75
meniscus (mm)
Inclination angle (◦ ) 15 EMS frequency (Hz) 4
Submergence entry depth (mm) 170 EMS current (A) 400, 500, 600, 700
2.3.2. Boundary Conditions for Electromagnetic Simulation

The whole geometry model of the electromagnetic field was taken to be surrounded by
an air cuboid (2.95 m × 1.2 m × 3.6 m) in which most of the magnetic flux lines are closed.
Boundary conditions are applied on the external surface of this cuboid with magnetically
flux parallel boundary [14].
2.3.3. Boundary Conditions for Fluid Field Simulation

(1) The inlet velocity of the SEN was calculated based on the mass conservation, and
turbulent kinetic energy and the energy dissipation rate are estimated by the semi-
empirical equations [26]. The casting temperature is set as 1827 K.
70
Materials 2021, 14, 3681
(2) The outlet boundary at the bottom of the calculation domain is a fully developed
outflow condition.
(3) The mold wall is treated with the no-slip boundary condition and the heat flux on
the wide and narrow faces is a function of distance toward the mold bottom, as
shown in Equation (11), which is similar to the form proposed by Savage [27]. The
convective heat transfer boundary condition is imposed on the extended region of the
continuous caster, and the average heat transfer coefficient for wide and narrow faces
is 320 W/(m2 ·K) and 360 W/(m2 ·K), respectively.

60L
q = 2.68 − ϕ (11)
uc
where q is the heat flux density, ϕ represents constants for wide and narrow faces
calculated according to the imported and exported temperature difference and the
cooling water flow rate in the mold, which is 0.288 and 0.292, respectively, and L is
the distance from the meniscus, m.
(4) The top surface is treated as a free-slipped boundary condition and considering the
heat insulation of mold flux, adiabatic condition is applied to it.
2.4. Numerical Solution Procedure

In this study, the numerical simulation is divided into two sections: first, the magnetic
field and Lorentz force are computed using the commercial software, ANSYS EMAG,
by solving the Maxwell equations, and then the time-averaged electromagnetic force
is interpolated into the momentum equation as a source term of ANSYS FLUENT. The
calculation domain of fluid field model is divided into 2 million finite volumes and grid-
independent verification was performed; the structured mesh is used to simulate the fluid
flow and heat transfer in the slab mold. The pressure-implicit with splitting of operators
(PISO) algorithm was used for the pressure-velocity coupling and first-order upwind is
used for the discretization of momentum and energy equations and the time step size is
0.004 s. The physical parameters for numerical simulation are given in Table 2.
Table 2. Physical parameters used in mathematical simulation [28,29].
Molten steel Parameters Value

Density (kg/m3 ) 7000
Specific heat (J/(kg·K)) 720
Thermal conductivity (W/(m·K)) 31
Viscosity (Pa·s) 0.0065
Latent heat (J/kg) 275,000
Solidus temperature (K) 1802
Liquidus temperature (K) 1812

3.1. Validation of Electromagnetic Field and Flow Field
Figure 2 displays the distribution of measured and numerical magnetic flux density
at 15 mm from the mold’s fixed side of the stirrer mid-plane with operating condition of
700 A/4 Hz, where magnetic flux density data was measured using a CT-3 Teslameter
within an empty mold. It can be observed that the numerical results matched well with the
measured data; both the predicted data and distribution tendency are similar, the biggest
numerical magnetic flux density is about 98 mT, which was located near the left side of
the mold.
71
Materials 2021, 14, 3681

)L[HGVLGH RSHUDWLQJFRQGLWLRQ
PP VWLUUHUPLGSODQH $+]
/RRVHVLGH 0HDVXUHGDWD

(OHFWURPDJQHWLFIOX[GHQVLW\P7
1XPHULFDOGDWD

'LVWDQFHIURPWKHFHQWHURIZLGHIDFHP
Figure 2. Distribution of measured and numerical magnetic flux density at stirrer mid-plane
(Z = −0.075 m).
Figure 3a depicts the distribution of time-averaged electromagnetic force with the

operating condition of 700 A/4 Hz, and it can be seen that the distribution of time-averaged
electromagnetic force is centrosymmetric, which resulted from the similar distribution
of magnetic flux density [9]. The tangential components of electromagnetic force are in
the same direction at each edge, but their directions are opposite to each other, and thus,
produce a horizontal recirculation. Four transverse swirls of electromagnetic force exist
in the interior of the cross-section, which correspond to the pole number acting on the
molten steel. Figure 3b illustrates the variation of electromagnetic force at 15 mm from the
mold’s fixed side of the stirrer mid-plane under different EMS currents. It can be seen that
electromagnetic force increases with the EMS current, as the EMS current increases from
400 to 800 A, the maximum electromagnetic force is 2133 N/m3 and 7950 N/m3 .
Figure 3. Distribution of (a) time-average electromagnetic force at the center of stirrer and (b) electro-
magnetic force under different EMS currents.
In order to validate the mathematical model, a 1/5th scaled physical model using
mercury when EMS is not applied was established; the fluid velocity in the mold was
measured by means of ultrasound Doppler velocimetry (UDV), and Li et al. [30] as the
authors published this apparatus’ description in ISIJ International. Meanwhile, a mathemat-
ical model was set corresponding to the physical model both on the geometry and material
properties, etc.
Figure 4a exhibits the flow pattern obtained from the physical model and Figure 4b
exhibits the flow pattern obtained from the mathematical model at a quarter longitudinal
plane near the fixed side of the mold. When EMS is not applied, a typical roll-flow pattern is
observed, both the measured flow pattern and numerical flow pattern are similar not only
in flow structure but also in characteristic points, for example, the impinging point at mold
narrow face. It can be seen that the numerical results matched well with the experimental
72
Materials 2021, 14, 3681
results. Overall, by comparison of the magnetic flux density and flow field between
simulation and experimental results, the mathematical model is proven to be reliable.
(a) (b)
Figure 4. Vector and contour of fluid flow obtained at quarter plane of mold fixed side under the condition of 1.2 m/min-0 A:
(a) physical model and (b) numerical model.
3.2. Effect of EMS Current on Fluid Flow in the Mold

In order to analyze the flow field inside the liquid pool, two planes and corresponding
characteristic lines were selected, as indicated in Figure 1b. Plane 1 is a transverse plane
which −5 mm distance from the meniscus, and line 1 is located at Y = 57.5 mm of plane 1,
while plane 2 is a longitudinal plane at Y = 57.5 mm iso-surface, and line 2 is located at
X = −700 mm of plane 2. The velocity distribution at plane 1 and plane 2 under different
EMS currents when the casting speed is 1.2 m/min are shown in Figure 5.
As shown in Figure 5a, when EMS is off, the flow pattern at plane 2 is a classical double-
roll structure as the jet travels to the narrow face and then splits into an upward (to the
free surface and back towards the SEN) and downward (to carry molten steel deep into the
mold) flow. For plane 1, molten steel impinges the free surface near the mold narrow face
and results in a large velocity at this region, whereas the flow pattern near the SEN is almost
stagnant. This nonuniform velocity distribution may affect the temperature distribution
near the meniscus to some extent. EMS can improve this nonuniform distribution of
velocity and temperature to some extent.
When EMS is applied (shown in Figure 5b,c), for plane 1, a horizonal recirculating
flow has been generated due to the effect of electromagnetic force, this trend is similar
to Yin’ work in [31]. Meanwhile, it can be seen that the velocity at both side of SEN
increases significantly compared with that without EMS. For plane 2, the upper and lower
recirculation flow disappears gradually, instead, molten steel at the left side tends to move
towards the right side of mold. In fact, the flow pattern in the mold is the interaction
of inertial flow caused by jet flow and driving flow induced by electromagnetic force;
Figure 5d shows the schematic of flow pattern with EMS, two different accelerated and
decelerated regions exist both at the longitudinal plane and transverse plane which are
related to the direction of electromagnetic force and initial flow pattern without EMS. With
the increase of EMS current, it can be seen that an integrated recirculation flow has been
generated at plane 1, and the velocity magnitude also increases, while for plane 2, the jet
flow has obviously been inhibited at the left side of the SEN, meanwhile, the jet angle at
the right side of the SEN decreases due to large electromagnetic force.
73
Materials 2021, 14, 3681
(a) (b)
(c) (d)
Figure 5. Velocity distribution at transverse plane (plane 1) and a quarter longitudinal plane (plane 2): (a) without-EMS, (b)
400 A-EMS, (c) 700 A-EMS (d) schematic of flow pattern with EMS.
Further justification of velocity under the influence of EMS is plotted in Figure 6. As

shown in Figure 6a, when EMS is off, the X-velocity has the same magnitude at both sides
of the SEN but with opposite direction; the maximum value is 0.18 m/s which corresponds
to the impinging region of upper recirculation flow of Figure 5a. When EMS is on, with
the increase of EMS current, the value of X-velocity at the left side of the SEN increases
due to the accelerated region of Figure 5d, while for the right side of SEN, the direction of
X-velocity changed firstly because of the horizonal swirling flow, and then the value of X-
velocity increases with the increase of EMS current. For the small EMS current (400–600 A),
it can be observed that the maximum value of X-velocity is less than that without EMS
74
Materials 2021, 14, 3681
(0.18 m/s), it means that small parts of upper recirculation flow still exist which hinders the
horizonal recirculation flow, while when EMS current is greater than 600 A, the maximum
value of X-velocity is 0.21 m/s which is larger than 0.18 m/s, it indicates that the upper
recirculation flow has been broken completely in Figure 6b, with the increase of EMS
current, the velocity in the mold decreases, as the EMS current increases to 700 A, the
Z-velocity all changes to negative, which indicates that a plug flow has generated near the
mold narrow face. This flow pattern may not benefit to the removal of inclusion.

/LQH
$ $
= PP
$ $
$

'LVWDQFHIURPWKHPHQLVFXVP
;YHORFLW\PV

$ $
$ $
$
6(1

:LGHIDFHP =YHORFLW\PV
(a) (b)
Figure 6. Variation of velocity under different EMS currents: (a) X-velocity at line 1, (b) Z-velocity at
line 2.
3.3. Effect of EMS Current on Initial Solidification

The 3D variation of solidified shell thickness along the casting direction under different
EMS currents at the casting speed of 1.2 m/min are shown in Figure 7.
(a) (b) (c)
(d)
Figure 7. 3D variation of solidified shell thickness along the casting direction under different EMS currents: (a) 0 A, (b) 400 A,
(c) 700 A, (d) schematic of transverse flow near jet flow with EMS.
As shown in Figure 7, the initial solidified shell was generated at −0.042 m distance
from the meniscus, and its thickness increases along the casting direction. When EMS is
off, the jet flow pouring from the SEN impinges to the solidified shell at the narrow face of
the mold, which may result in the thin and uneven distribution of solidified shell at the
impact region. When EMS is applied, the solidified shell thickness at the mold narrow
face increases because that the direct impact of jet flow has been weakened under the
influence of electromagnetic force. While for the mold wide face, due to the washing effect
75
Materials 2021, 14, 3681
of transverse flow, the uniformity of solidified shell thickness changes even for a suitable
EMS current (Figure 7b), however, for the higher EMS current (700 A), the solidified shell
thickness at the right side of wide face decreases The main reason is that as shown in
Figure 7d, an electromagnetic force gradient exists from the edge to the center of the slab,
the jet flow may deviate towards the mold wide face and lead to the remelting of solidified
shell; this phenomenon may not benefit the uniformity of the solidified shell to any extent.
Figure 8 shows the variation of solidified shell thickness at the center of the mold
wide face and narrow face under different EMS currents, respectively. It can be seen that
the solidified shell becomes thinner as the EMS current increases for the mold wide face.
This phenomenon can be attributed to the superheat dissipation effect of EMS. While for
the mold narrow face, when the EMS is not applied, it can be observed that the solidified
shell stops growing due to the remelting phenomenon of high temperature liquid steel,
which may increase the risk of break out. With the increase of EMS, the solidified shell
thickness at the mold exit increases. When the EMS current increases from 0 to 700 A,
the shell thickness of wide face and narrow face at the mold exit changes from 19.51 and
10.2 mm to 18.85 and 18.5 mm, respectively.

$
$
$
$
$
6ROLGLILHGVKHOOWKLFNQHVVPP
$
$
$
$
$

'LVWDQFHIURPWKHPHQLVFXVP 'LVWDQFHIURPWKHPHQLVFXVP
(a) (b)
Figure 8. Variation of solidified shell at the center of: (a) wide face, (b) narrow face under different
EMS currents.
In the continuous casting process, the initial solidified shell near the meniscus around
the perimeter of the mold is of great importance because it is associated with crack for-
mation [1]. Therefore, in order to describe the uniformity of solidified shell thickness
quantitatively, the uniform index of solidified shell thickness is set up, Ui , and the relevant
definitions are depicted in Equations (12)–(14):
σi
Ui = 1 − (12)
hi− ave

m
hi− ave = ∑ hi − j /mi (13)
j =1
#
$ m
$ 2
$ ∑ (hi− j − hi− ave )
% j =1
σi = (14)
mi
where, i represents the wide face (WF) and narrow face (NF) at mold transverse plane,
respectively. hi − j , hi − ave and mi are solidified shell thickness in mesh element j, averaged
solidified shell thickness and total mesh element number, respectively, and σi is the stan-
dard deviation of the solidified shell. The larger the uniform index is, the more uniformly
solidified the shell will be. What is more, because the influence of the electromagnetic
force is mainly concentrated at −0.4 m distance from the meniscus for the current stirrer
position, therefore, in the current study, we use the z = −0.4 m transverse plane to analyze
76
Materials 2021, 14, 3681
the uniformity. Figure 9 and Table 3 show the uniform index at the mold wide face and
narrow face under different EMS currents.

:LGHIDFH
1DUURZIDFH
8QLIRUPLQGH[ı

(06FXUUHQW$
Figure 9. Uniform index at z = −0.4 m under different EMS currents.
Table 3. Uniform index at z = −0.4 m under different EMS currents.
Uniform Index
EMS Current (A)
Wide Face Narrow Face
0 0.948 0.828
400 0.961 0.971
500 0.965 0.97
600 0.923 0.972
700 0.9 0.973
As shown in Figure 9, compared with that without EMS, EMS can effectively improve
the uniformity of the solidified shell, especially for the mold narrow face. For the mold
narrow face, the uniform index changes little as the EMS current increases from 400 to
700 A, the main reason is that the electromagnetic driving flow can suppress the direct
impact of jet flow on the mold narrow face, and benefit to the uniform growth of solidified
shell. For the mold wide face, it can be observed that the uniform index increases at first
and then decreases accompanied with the increase of EMS current, this can be attributed
that for the high EMS current, the remelt phenomenon at region II intensified, so the
uniform index decreased.
3.4. Effect of Casting Speeds on Fluid Flow in the Mold

Considering that the flow velocity uniformity on the left and right sides of the mold is
better at 600 A, as shown in Figure 6a, and the thickness of the shell with narrow surface is
thicker at 600 A, as shown in Figure 8b, the thickness of the shell with narrow surface is
conducive to improving the speed. According to Figures 6a and 8b, it is better at 700 A, but
considering the uniform index of the solid shell as shown in Figure 9, the uniformity index
of the wide face shell is worse at 700 A, therefore, 600 A was selected for further study.
The velocity distribution at plane 1 and plane 2 under different casting speeds when
the EMS current is 600 A and frequency is 4 Hz are shown in Figure 10.
77
Materials 2021, 14, 3681
(a) (b)
Figure 10. Velocity distribution at transverse plane (plane 1) and a quarter longitudinal plane (plane 2) at (600 A, 4 Hz):
(a) 0.8 m/min, (b) 1.4 m/min.
As shown in Figure 10, for plane 2, molten steel tends to flow from the left side of
the SEN to the right side of the SEN due to the effect of electromagnetic force. When the
casting speed is 0.8 m/min, it can be seen that the jet flow at the left side of SEN has nearly
disappeared by comparing it with that of 1.4 m/min. While at the right side of the SEN, the
jet angle increases significantly as the casting speed increase from 0.8 to 1.4 m/min. The
main reason is that under the same EMS current, for the low casting speed (0.8 m/min), the
jet velocity poured from the SEN is relatively small. Therefore, under this condition, the jet
flow at the left side of the SEN is constrained more significantly, while for the right side of
the SEN, the jet flow can also be easily dragged under the low casting speed, and results in
a small jet angle. For plane 1, although a horizonal recirculating flow has been generated
for these two cases, the velocity at the decelerated region is small when the casting speed
is 1.4 m/min due to the hindering effect of stronger upper recirculation flow. Figure 11a
shows the variation of velocity under different casting speeds at line 1, the velocity at the
left side of SEN is greater than that of the right side; because of the accelerated region,
with the increase of casting speed, although a transverse flow was generated for all casting
speeds, the difference of maximum velocity at both sides of the SEN increases, as casting
speed increases from 0.8 to 1.4 m/min, the difference increases from 0.0188 to 0.072 m/s.
For the longitudinal line (Figure 11b), for the low casting speed, a plug flow has generated
near the mold wall, it means that a critical casting speed exists to avoid the plug flow.
78
Materials 2021, 14, 3681
/LQH
PPLQ
= PP PPLQ
PPLQ
PPLQ

'LVWDQFHIURPWKHPHQLVFXVP

;YHORFLW\PV

PPLQ

PPLQ
6(1

PPLQ
PPLQ

=YHORFLW\PV
:LGHIDFHP
(a) (b)
Figure 11. Variation of velocity under different casting speeds at (600 A, 4 Hz): (a) X-velocity at line 1, (b) Z-velocity at
line 2.
3.5. Effect of Casting Speed on Solidification
The variation of solidified shell thicknesses along the casting direction and center of
the mold wide face and the narrow face under different casting speeds when the EMS
current is 600 A are shown in Figure 12.
(a) (b) (c)
Figure 12. 3D variation of solidified shell thickness along the casting direction under different casting speeds at (600 A,
4 Hz): (a) 0.8 m/min, (b) 1.2 m/min, (c) 1.4 m/min.
As shown in Figures 12 and 13, the solidified shell thickness decreases with the
increase of casting speed. For the mold wide face, a remelting concave has generated at
region II because of the deviated jet flow induced by electromagnetic force gradient, and
it is aggravated with the increase of casting speed. For the mold narrow face, it can be
seen that with the increase of casting speed, the remelt phenomenon of solidified shell
thickness aggravates due to the strong jet flow. When the casting speed increases from 0.8
to 1.4 m/min, the shell thickness of the wide face and narrow face at the mold exit changes
from 20.72 and 20.4 mm to 15.95 and 17.31 mm, respectively.
79
Materials 2021, 14, 3681

PPLQ PPLQ

PPLQ PPLQ

PPLQ
PPLQ
PPLQ PPLQ

'LVWDQFHIURPWKHPHQLVFXVP 'LVWDQFHIURPWKHPHQLVFXVP
(a) (b)
Figure 13. Variation of the solidified shell at the center of, (a) wide face, (b) narrow face under
different casting speeds at (600 A, 4 Hz).
The uniform index at z = −0.4 m under different casting speeds when the EMS current
is 600 A is shown in Table 4. The uniform index at the mold narrow face and the wide face
is shown in Figure 14. As shown in Figure 14, the uniform index at the mold narrow face is
larger than that of the mold wide face, the main reason is that when EMS is applied, the jet
flow which directly impacts the narrow face has been suppressed, and this may benefit
uniform growth of the solidified shell, while for the mold wide face, the deviated jet flow
can break the uniformity of the solidified shell. With the increase of casting speed, the
uniform index at the mold wide face decreases firstly and then increases, while at the mold
narrow face it increases but changes are not significant.
Table 4. Uniform index at z = −0.4 m under different casting speeds.
Uniform Index
Casting Speed (m/min)
Wide Face Narrow Face
0.8 0.944 0.956
1.0 0.895 0.973
1.2 0.923 0.972
1.4 0.919 0.982

:LGHIDFH
1DUURZIDFH

8QLIRUPLQGH[ı

&DVWLQJVSHHGPPLQ
Figure 14. Uniform index at z = −0.4 m under different casting speeds.
4. Application Effects
According to the optimal parameters of the mathematical simulation, the industrial
test including level fluctuation and number density of inclusion was carried out with and
without EMS when the casting speed was 1.2 m/min. The relevant results are shown in
Figure 15.
80
Materials 2021, 14, 3681

$
$(06 $
$(06
'HIHFWLQGH[RIVODEQXPP

)OXFWXDWLRQKHLJKWPP

'LVWDQFHIURPWKHVODEVXUIDFHPP
7LPHV
(a) (b)
Figure 15. Industrial test: (a) transient level fluctuation at quarter of mold, (b) number density
of inclusion.
As shown in Figure 15a, the largest level fluctuation is less than ±5 mm whether
the EMS is on or off. However, when EMS is applied, the amplitude of fluctuation level
is relatively small compared to that without EMS, it indicates that EMS can decrease
the fluctuation level for a suitable EMS current. In addition, the inclusion has also been
detected by magnifying glass and the size which is larger than 100 μm can be observed by
this method. It can be seen that the number density within 15 mm from the slab surface
have a decreasing trend when the EMS is applied, which demonstrates the benefits of the
optimized flow control parameters.
5. Conclusions
A three-dimensional mathematical model is conducted to investigate the fluid flow,
solidification and level fluctuations under different EMS currents and casting speeds. The
uniform index of the solidified shell thickness has been introduced to judge the uniformity
of the solidified shell. Conclusions of this study are summarized as follows:
(1) When EMS is applied, a horizonal recirculating flow has been generated, accelerated
and decelerated regions exist in the mold. With the increase of EMS current, the
difference of velocity near free surfaces decreases; large EMS current may generate
plug flow near the mold wall.
(2) EMS can obviously improve the uniformity of the solidified shell, with the increase of
EMS current, the uniform index at the mold narrow face increases while at the mold
wide face it first increases and then decreases.
(3) Under the same EMS current, with the increase of casting speed, the difference of
velocity near the free surface increases, the uniform index at the mold narrow face
changes little, while at the mold wide face it first decreases and then increases.
(4) EMS can weaken the level fluctuation and reduce the number density of inclusion for
a suitable EMS current through industry test.
(5) A rational EMS current range exists to obtain optimal steel quality. In the current
study, when the casting speed is 1.2 m/min, the rational EMS current is 500–600 A.
Author Contributions: Methodology, Q.L. and J.Z.; formal analysis, Z.L. and H.L.; data curation, B.L.
and C.J.; writing—original draft preparation, G.L. All authors have read and agreed to the published
Funding: This project was financially supported by the National Science Foundation of China (No.
U1860107 and 52074181).
81
Materials 2021, 14, 3681
Data Availability Statement: Data sharing not applicable. No new data were created or analyzed in
this study. Data sharing is not applicable to this article.
Acknowledgments: The authors express sincere thanks to Xiaoxuan Deng in Shougang Group
Co., Ltd., Research Institute of Technology for his helpful assistance with structure and format of
this paper.
References
1. Thomas, B.G. Review on Modeling and Simulation of Continuous Casting. Steel Res. Int. 2018, 89. [CrossRef]
2. Cho, S.-M.; Thomas, B.G. Electromagnetic Forces in Continuous Casting of Steel Slabs. Metals 2019, 9, 471. [CrossRef]
3. Jin, K.; Thomas, B.G.; Ruan, X. Modeling and Measurements of Multiphase Flow and Bubble Entrapment in Steel Continuous
Casting. Metall. Mater. Trans. B 2016, 47, 548–565. [CrossRef]
4. Li, Z.; Zhang, L.; Ma, D.; Wang, E. Numerical Simulation on Flow Characteristic of Molten Steel in the Mold with Freestanding
Adjustable Combination Electromagnetic Brake. Met. Mater. Trans. A 2020, 51, 2609–2627. [CrossRef]
5. Zhang, L.-S.; Zhang, X.-F.; Wang, B.; Liu, Q.; Hu, Z.-G. Numerical Analysis of the Influences of Operational Parameters on the
Braking Effect of EMBr in a CSP Funnel-Type Mold. Met. Mater. Trans. A 2014, 45, 295–306. [CrossRef]
6. Dubke, M.; Tacke, K.-H.; Spitzer, K.-H.; Schwerdtfeger, K. Flow fields in electromagnetic stirring of rectangular strands with
linear inductors: Part II. Computation of flow fields in billets, blooms, and slabs of steel. Met. Mater. Trans. A 1988, 19, 595–602.
[CrossRef]
7. Yamada, T.; Satoh, S.; Fujisaki, K.; Shouji, S.; Takahiro, Y. Consideration of heat transfer and solidification in 3-D MHD calculation.
IEEE Trans. Magn. 2000, 36, 1300–1304. [CrossRef]
8. Okazawa, K.; Toh, T.; Fukuda, J.; Kawase, T.; Toki, M. Fluid Flow in a Continuous Casting Mold Driven by Linear Induction
Motors. ISIJ Int. 2001, 41, 851–858. [CrossRef]
9. Li, B.; Lu, H.; Zhong, Y.; Ren, Z.; Lei, A.Z. Influence of EMS on Asymmetric Flow with Different SEN Clogging Rates in a Slab
Continuous Casting Mold. Metals 2019, 9, 1288. [CrossRef]
10. Wang, B.; Yang, Z.G.; Zhang, X.F.; Wang, Y.T.; Nie, C.P.; Liu, Q.; Dong, H.B. Analysis of the effects of electromagnetic stirring on
solidification structure of bearing steel. Metalurgija 2015, 54, 327–330.
11. Yang, Z.-G.; Wang, B.; Zhang, X.-F.; Wang, Y.-T.; Dong, H.; Liu, Q. Effect of Electromagnetic Stirring on Molten Steel Flow and
Solidification in Bloom Mold. J. Iron Steel Res. Int. 2014, 21, 1095–1103. [CrossRef]
12. Yang, Z.G.; Wang, B.; Zhang, X.F.; Wang, Y.T.; Dong, H.B.; Liu, Q. Effects of electromagnetic stirring on the flow field and level
fluctuation in bloom molds. J. Univ. Sci. Technol. Beijing 2014, 36, 1354–1360.
13. Fang, Q.; Ni, H.; Wang, B.; Zhang, H.; Ye, F. Effects of EMS Induced Flow on Solidification and Solute Transport in Bloom Mold.
14. Ren, B.; Chen, D.-F.; Wang, H.-D.; Long, M.-J.; Han, Z.-W. Numerical simulation of fluid flow and solidification in bloom
continuous casting mould with electromagnetic stirring. Ironmak. Steelmak. 2015, 42, 401–408. [CrossRef]
15. Wang, X.; Zheng, S.-G.; Liu, Z.-H.; Zhu, M.-Y. Numerical Simulation on Multiple Physical Fields Behaviors in Billet Continuous
Casting with Different Stirrer Positions. Steel Res. Int. 2019, 91, 1900415. [CrossRef]
16. Maurya, A.; Jha, P.K. Influence of electromagnetic stirrer position on fluid flow and solidification in continuous casting mold.
Appl. Math. Model. 2017, 48, 736–748. [CrossRef]
17. Maurya, A.; Jha, P.K. Numerical investigation of in-mold electromagnetic stirring process for fluid flow and solidification.
COMPEL-Int. J. Comput. Math. Electr. Electron. Eng. 2017, 36, 1106–1119. [CrossRef]
18. Maurya, A.; Jha, P. Two-phase analysis of interface level fluctuation in continuous casting mold with electromagnetic stirring. Int.
J. Numer. Methods Heat Fluid Flow 2018, 28, 2036–2051. [CrossRef]
19. FFujisaki, K.; Ueyama, T. Magnetohydrodynamic calculation for free surfaces. J. Appl. Phys. 1998, 83, 6356–6358. [CrossRef]
20. Fujisaki, K.; Ueyama, T.; Toh, T.; Uehara, M.; Kobayashi, S. Magnetohydrodynamic calculation for electromagnetic stirring of
molten metal. IEEE Trans. Magn. 1998, 34, 2120–2122. [CrossRef]
21. Li, B.; Lu, H.; Zhong, Y.; Ren, Z.; Lei, Z. Numerical Simulation for the Influence of EMS Position on Fluid Flow and Inclusion
Removal in a Slab Continuous Casting Mold. ISIJ Int. 2020, 60, 1204–1212. [CrossRef]
22. Moffatt, H.K. Electromagnetic stirring. Phys. Fluids A Fluid Dyn. 1991, 3, 1336–1343. [CrossRef]
23. Zhang, W.; Luo, S.; Chen, Y.; Wang, W.; Zhu, M. Numerical Simulation of Fluid Flow, Heat Transfer, Species Transfer, and
Solidification in Billet Continuous Casting Mold with M-EMS. Metals 2019, 9, 66. [CrossRef]
24. Chen, W.; Ren, Y.; Zhang, L. Large Eddy Simulation on the Fluid Flow, Solidification and Entrapment of Inclusions in the Steel
along the Full Continuous Casting Slab Strand. JOM 2018, 70, 2968–2979. [CrossRef]
25. Liu, G.; Liu, Q.; Ji, C.; Chen, B.; Li, H.; Liu, K. Application of a Novel Chamfered Mold to Suppress Corner Transverse Cracking
of Micro-Alloyed Steel Slabs. Metals 2020, 10, 1289. [CrossRef]
26. Li, B.; Okane, T.; Umeda, T. Modeling of biased flow phenomena associated with the effects of static magnetic-field application
and argon gas injection in slab continuous casting of steel. Metall. Mater. Trans. B 2001, 32, 1053–1066. [CrossRef]
82
Materials 2021, 14, 3681
27. Vives, C.; Ricou, R. Experimental study of continuous electromagnetic casting of aluminum alloys. Metall. Mater. Trans. B 1985,
16, 377–384. [CrossRef]
28. Qiu, S.; Liu, H.; Gan, Y. Numerical Simulation of Solidification Process of Slab Casting Based on Continuum Model. J. Iron Steel
Res. 2003, 15, 16–20.
29. Lu, H.; Li, B.; Li, J.; Zhong, Y.; Ren, Z.; Lei, Z. Numerical Simulation of In-mold Electromagnetic Stirring on Slide Gate Caused
Bias Flow and Solidification in Slab Continuous Casting. ISIJ Int. 2021, 61, 1860–1871. [CrossRef]
30. Li, B.; Lu, H.; Shen, Z.; Sun, X.; Zhong, Y.; Ren, Z.; Lei, Z. Physical Modeling of Asymmetrical Flow in Slab Continuous Casting
Mold due to Submerged Entry Nozzle Clogging with the Effect of Electromagnetic Stirring. ISIJ Int. 2019, 59, 2264–2271.
[CrossRef]
31. Yin, Y.; Zhang, J.; Wang, B.; Dong, Q. Effect of in-mould electromagnetic stirring on the flow, initial solidification and level
fluctuation in a slab mould: A numerical simulation study. Ironmak. Steelmak. 2018, 1–10. [CrossRef]
83
materials
Article
Multi-Parameter Characteristics of Electric Arc Furnace Melting
Michał Moskal 1, *, Piotr Migas 2 and Mirosław Karbowniczek 2
1 COGNOR SA Branch HSJ in Stalowa Wola, Kwiatkowskiego 1, 37-450 Stalowa Wola, Poland
2 Department of Ferrous Metallurgy, Faculty of Metals Engineering and Industrial Computer Science,
AGH-University of Science and Technology, 30 A. Mickiewicza A., 30-059 Krakow, Poland;
[email protected] (P.M.); [email protected] (M.K.)
Abstract: The article presents the results of analyses of numerical modelling of selected factors in
electric arc furnace melts. The aim of the study was to optimise the melting process in an electric arc
furnace using statistical-thermodynamic modelling based on, among other things, multiple linear
regression (MLR). The article presents tools and methods which make it possible to identify the
most significant indicators of the process carried out on the analysed unit from the point of view of
improvement. The article presents the characteristics of the process and creation of the MLR model
and, by applying its numerical analyses and results of calculations and simulations for selected
variables and indicator, identifying the operation of a selected furnace. Developed model to demand
of electric energy identification was used for calculations of energy balances, the distribution of the
energy used in the furnace was presented.
Keywords: electric arc furnace; energy efficiency modelling; process optimization; mass and en-
ergy balances
Citation: Moskal, M.; Migas, P.;

1. Introduction
Karbowniczek, M. Multi-Parameter
Characteristics of Electric Arc Observed climatic changes in the environment are causing recycling technologies for
Furnace Melting. Materials 2022, 15, steel production to increase their share in the total balance of crude steel production. This is
1601. https://doi.org/10.3390/ particularly noticeable in the European Union. Therefore, the electric arc furnace is becom-
ma15041601 ing one of the leading processing units for steel scrap used in crude steel production [1].
Electricity consumption in this process still accounts for about 50–80% of the total energy
Jiangshan Zhang
consumption during melting. Technological progress and the development of information
technology create opportunities to reduce the unit demand for electrical or chemical energy.
Received: 31 January 2022 Mathematical descriptions and characterisations of the arc furnace smelting process
Accepted: 17 February 2022 are complicated due to its nature resulting from, among others, high temperature and
Published: 21 February 2022 chemical composition gradients, turbulent flows, multiphase and heterogeneity of the
Publisher’s Note: MDPI stays neutral system, significant mass flow of substrates, high speed of phase transformations and non-
with regard to jurisdictional claims in linearity of phenomena. Therefore, the development of modelling of selected stages or the
published maps and institutional affil- entire arc furnace smelting process remains justified [2].
iations. The models used to describe the melting technology can be physical models—
reproducing the processes on a smaller scale (e.g., laboratory and/or using other me-
dia than liquid metal and slag with the use of criterion numbers). An example of the
application of a high-temperature model can be the studies on the performance of burners
Copyright: © 2022 by the authors. in an electric arc furnace carried out by Yonmo Sung [3]. They presented modification of
Licensee MDPI, Basel, Switzerland. oxy–gas burners, which was validated on a test bench.
Studies of scrap melting phenomena were also carried out based on water models [4]
(using ice, three heating rods, and steam nozzles, which reflected scrap, electric arc, and
oxy–gas burners, respectively).
Numerical methods can also be used to model the process. Mombeni [5] used a
CFD model calculated in the commercial software ANSYS FLUENT to address the crack
4.0/).

Materials 2022, 15, 1601
formation of water-cooled panels installed in the furnace vault. CFD modelling can also
be used to determine the temperature distribution in a metal bath in the context of burner
placement and media flow rates [6].
However, the mathematical models are the most universal method, which take into
account physicochemical processes, thermodynamic and theoretical equations, mass and
energy balances, or only technological data from the process [7]. They can be divided into
two categories: linear and non-linear.
Linear models are created as a linear combination of predictor variables [8] based on
average statistical input and output data from the process, or assumed theoretical values.
They describe the relationship between individual parameters and their importance for
the outcome of the calculation. Linear models are static and refer to only two points in
time; they are the result of input data and, as mentioned above, they can only be used with
reference to average values of a given process [9].
The most popular is the Kohle model, which describes the influence of process param-
eters on electricity demand.
Another model based on a multivariate linear regression equation is the one developed
by Haupt [10]. He determined the influence of scrap quality on the melting process in an
electric arc furnace.
However, non-linear models have the ability to describe the process more accurately. R.
Morales created one of the first models, which included heat and mass transfer calculations
and chemical reaction kinetics in terms of use the direct reduced iron [11]. The author
created a dynamic model, which took into account the relationships between individual
phenomena in terms of their physical course and not just the result, which was a statistical
outcome of many variables. Thus, it makes it possible to analyse single melts and to follow
the course of change of the result in time.
Attempts have been made to simplify the models and organise their construction.
Clear boundaries of individual zones in the furnace were set. In terms of both: furnace
construction, melting stages, and division into individual phases of materials/substances
present in the furnace space. Bekker [12] specified two groups of solids and liquids in
the constructed model. On the other hand, Logar [13,14] created a system of dependent
modules representing particular groups of phenomena (including solid scrap zone, solid
slag zone, liquid slag zone, liquid metal zone, and wall and roof zone). Models using neural
networks are becoming increasingly popular. The first versions were developed in the
1990s. Ledoux and Bonnard, in their work, built a model based on a multi-layer perceptron.
The model described the dynamics of an electric arc during the melting process.
Due to the process complexity, it is necessary to select models in such way that
sufficient representation of the processes is maintained with the best possible ability to
interpret and understand the simulation results [15]. The balance between these two
characteristics depends largely on the type of models and the modelling methods [16].
2. Materials and Methods

2.1. Theoretical Mathematical Models
Mathematical models are the most commonly used in EAF simulation. They can be
both linear and non-linear in nature. This classification results from taking into account the
above-described variability of process efficiency as well as the variability of the parameters
themselves during the melt [17]. Taking into account the kinetics of the reaction or the vari-
ation in the efficiency of media use or their control transforms the model into a non-linear
form. Non-linearity is understood as effects ranging from media consumption dependent
on melting programmes to the dependence of heat losses on the number and height of
scrap heaps and thus arc exposure. When creating mathematical models, theoretical data or
available technological and statistical data are used—which, depends on their availability
and quality.
86
Materials 2022, 15, 1601
2.2. Linear Modelling of Steel Melting in Electric Arc Furnace

Regression linear models are created as a linear combination of predictor variables
based on statistical mean input and output data from a process, or assumed theoretical
values. This means that they refer to (only) two points in time, they are the result of input
data and, e.g., output data, and as mentioned above, and they can only be used with respect
to the average values of a given process. Static models often do not take into account some
process disturbances and micro-scale phenomena (of low weights) and, therefore, their
use for the analysis of single melts becomes difficult or, even impossible, due to the high
variability of the real process conditions [18].
The most widely used statistical model in the literature is the Köhle model [18,19].
Köhle’s Equation (1) describing the demand for electricity:

Wr G G T t + tN M M
= 300 + 900 E − 1 + 1600 Z + 0.7 ◦ A − 1600 + 0.85 S − 8 3G − 4.3 3 L (1)
kWh/t GA GA C min m /t m /t
where Wr is the electric energy demand, kWh/Mg; GE is the weight of ferrous materials,
kg; GA is the tap weight (mass of the bath), kg; GZ is the weight of slag components, kg; TA
is the temperature before taping, ◦ C; tS is the power on time, min; tN is the power off time,
min; MG is the natural gas consumption, Nm3 ; and ML is the oxygen consumption, Nm3 .
This model characterises the electric energy demand, which in this case (of the units
analysed by Köhle) is between 380 and 600 kWh/Mg with an accuracy equal to the standard
deviation of 5 kWh/Mg.
The disadvantage of this solution is the low sensitivity of the model to changes in
process conditions (occurrence of disturbances, change of the melting programme for the
arc control system, or burners) which translates into a high error for individual melts.
However, it is also an advantage in terms of long-term prediction. The Köhle model, due
to its hybrid empirical–theoretical character, remains easy to interpret while maintaining
sufficient accuracy.
In industrial practice, problems are encountered due to, e.g., insufficient process
metering and difficult definition of boundary conditions. The amount of electrical energy
supplied is easy to determine, whereas the efficiency of heat transfer through the arc
depends on a large number of factors, e.g., the melting stage, the phenomena associated
with arc burning (arc length, arc stability), or the plasma temperature associated with the
gas atmosphere of the furnace, as well as the parameters characterising the steel scrap.
Identifying the efficiency of the use of chemical energy—dependent on the reactions taking
place in the reactor, which is the volume of the furnace bowl—is also a challenge. Its
determination requires the identification of the reactions taking place and the distribution
of the heat obtained between the phases, as well as the dynamics of the process [11–13,20].
When constructing models, it is necessary to perform mass and energy balances.
Carrying out correctly defined balance calculations (of mass and energy) is essential from
the point of view of the overall characteristics of the energy efficiency of the process and
the determination of process boundary conditions [11].
In the course of the analyses, a database of indicators and technological parameters
was created, which included 1953 records from one specific EAF. Due to a wide range of
steel grades, the database was based on the most frequently manufactured group of steel
grades and data was selected by rejecting melts:
• from furnace start-up periods;
• with furnace operation disturbances;
• switching from or to grades with large differences in chemical composition, when
there was a justified suspicion of wet pouring;
• with process parameter values that are outliers.
The parameters included in the database are batch, material, and energy parameters.
87
Materials 2022, 15, 1601
2.3. Mass Balances

The mass balance is based on the assumption that the sum of the masses of the inputs
min equals the sum of the outputs mout .
mscr + msf + mc + mfa + mox + mCH4 + mla = mhm + msl + mf + mwg (2)
where mscr is the mass of the scrap, kg; msf is the mass of the slag formers, kg; mox is the
mass of the oxygen, kg; mCH4 is the mass of the natural gas, kg; mc is the mass of the
carburizer, kg; mfa is the mass of the foaming agent, kg; mla is the mass of the leaked air,
kg; mhm is the mass of the hot metal, kg; msl is the mass of the slag, kg; mwg is the mass of
the waste gases, kg; and mf is the mass of the fumes, kg.
For the purposes of consideration, the assumption has been made that the mass of
the hot heel does not change between heats. In reality, the mass of hot heel changes and is
difficult to estimate due to the relatively small volume it occupies and the constant irregular
changes in the scale changing the volume of the furnace hearth.
2.3.1. Slag Weight Calculations

Slag mass was calculated on the basis of the chemical composition of the scraps used,
assuming that the elements contained in them: Si, Mn, Cr, C, P, and Al are oxidised to the
average of the level characterising the composition of the metal bath, and the amount of
oxidised Fe constitutes 5.5% of the charge mass. These elements and the oxygen needed to
burn them together with the addition of CaO and MgO make up the slag mass. The slag
mass also includes ash from carbon carriers.
Chemical compositions of the slags were identified via energy dispersive X-ray flu-
orescence (XRF Oxford X-MET 3000TX +). In order to identify the chemical composition
of the steel, a spectrophotometric method of analysis was used on a FOUNDRY-MASTER
spark spectrometer.
The correctness of assumptions and calculations was checked by calculating the slag
mass on the basis of CaO introduced into the process, which, as a component not dissolving
in the metal bath, theoretically remains entirely in the slag.

mCaO met. ·0.958 + mCaO inj. ·0.864 ·100%
msl = (3)
%CaOslag
where mCaO met. is the mass of the metallurgical calcium, kg; mCaO inj . is the mass of the
injected pulverized calcium, kg; and %CaOslag is the percentage of mass weight of CaO
in slag.
A calculation was then made of the mass of each element that was oxidised during
bath melting and slag formation, thus reducing the metal yield.
MAe ·%tl · msl

mae = . (4)
MAo
where mae is the mass of the alloying element melted in slag, kg; %tl is the percentage of
mass weight of given element in slag; MAe is the atomic mass of element, u; and MAo is the
atomic mass of oxide of given element, u.
2.3.2. Calculation of Waste Gas Masses

In order to calculate the mass of waste gases taking into account the course of the
reaction of combustion of coal and methane, Reactions (5)–(7) were written:
1
C + O2 → CO (5)
2
C + O2 → CO2 (6)
88
Materials 2022, 15, 1601
CH4 + 2O2 → CO2 + 2H2 O (7)

The mass of carbon monoxide and carbon dioxide for each reaction for one kilogram
of product burnt was calculated. The values obtained were, respectively, 2.33 kg, 3.67 kg,
and 2.75 kg (from Reactions (5)–(7)). Due to the lack of measurement of the flue gas
composition, it was assumed that all the substrates burned completely, and left entirely
with the flue gases. Therefore, in the mass of waste gasses mwg , the components according
to the Equation (8) are included:
mwg = mN2 + mCO2 + mH2 O (8)
2.3.3. Calculation of Metal Bath Weight

The theoretical weight of the metal bath was calculated as the difference between the
weight of the scrap, the alloying elements contained in the bath (including iron), and the
estimated amounts of fumes.
mhm = mscr − Σmae − mf (9)
where mhm is the mass of hot metal, kg; mscr is the mass of the scrap, kg; mae is the sum of
masses of alloying elements that went into slag, kg; and mf is the mass of the fumes, kg.
2.4. Energy Balances

The energy balance assumes that the sum of the energy input Ein into the process is
equal to the sum of the energy output from the process in the form of product heat Qout
and thermal losses Qloss . The energy input to the system consists of electrical energy and
chemical energy. In the other direction, the energy is transferred out of the process in the
form of metal baths, liquid slag, waste gases, dusts, and heat losses.
Eel. + Echem. = Qhm + Qsl + Qwg + Qf (10)
where Eel. is the electrical energy, MJ; Echem. is the chemical energy, MJ; Qhm is the heat of
hot metal, MJ; Qsl is the heat of the slag, MJ; Qwg is the heat of waste gases, MJ; and Qf is
the heat of the fumes, MJ.
The chemical energy was calculated on the basis of enthalpies of exothermic reactions—
oxidation of bath components and combustion of methane. The Perry Nist Janaf method
was used to calculate the enthalpies of the reactions.
2.4.1. Calculation of the Physical Heat of Hot Metal

Calculation of the physical heat of the hot metal is based upon the values of the specific
heat of the hot metal, its mass, and the temperature rise. It also takes into account the
phase transition by using the values of latent heat, liquid temperature, and differentiation
of specific heat into that occurring in the liquid and solid state.
For the calculation of the physical heat of the metal bath, the Formula (11) is used:
Qhm = mhm ·(css ·(Tm − Ta ) + Qlst + cls (T − Tm ))/1000 (11)
where, Qhm is the physical heat of the metal bath at temperature T, MJ; mhm is the mass of
hot metal, kg; css is the average specific heat of steel in the solid state, kJ/kg K; cls is the
average specific heat of steel in the liquid state, kJ/kg·K; Qlst is the latent heat of fusion of
steel, kJ/kg; T is the steel tap temperature, K; Tm is the steel melting temperature, and K;
Ta is the ambient temperature, K.
The specific heat of the liquid state was taken as 0.836 MJ/Mg·K and that of the solid
state as 0.698 MJ/Mg·K. The latent heat of fusion for the steel was taken as 271.7 kJ/kg.
89
Materials 2022, 15, 1601
The melting point of steel with a hot metal composition was calculated according to
the Formula (12):
n
Tm = TFe − ∑ pi ki (12)
i=1
where Tm is the melting point, K; TFe is the melting point of pure iron (1812 K); pi is the
percentage content of the element in the metal bath, %; and ki is the temperature reduction
factor (according to the Table 1) [21].
Table 1. Temperature reduction factor for individual elements [21].
Element C Mg Si P S Cu Ni V Mo Cr
ki 65 5 8 30 25 5 4 2 2 1.5
2.4.2. Calculation of the Physical Heat of Slag

The calculation of the physical heat of slag was based on the values of the slag’s
specific heat, mass, temperature rise, and latent heat. The following Formula (15) was used
to calculate the physical heat of slag:
msl (csl (T − Ta ) + Qlsl )

Qsl = (13)
1000
where Qsl is the physical heat of slag at temperature T, MJ; msl is the slag mass, kg; csl is
the specific heat of slag at temperature T, kJ/kg·K; Qlsl is the latent heat of melting of slag,
kJ/kg; T is the slag tap temperature, K; and Ta is the ambient temperature, K.
The specific heat of slag was calculated from the Formula (14):
csl = 0.736 + 2.93·10−4 T (14)
The assumed latent heat of fusion is 209 kJ/kg.
2.4.3. Calculation of the Physical Heat of Waste Gases

The calculation of the physical heat of the waste gases is based on the specific heat
values of the gases of their mass and temperature rise. The formula was used to calculate
the physical heat of the waste gases:
Qwg = mwg cwg (T − Ta )/1000 (15)
where Qwg is the physical heat of the waste gases at temperature T, MJ; mwg is the mass of
the waste gases, kg; cwg is the specific heat of the waste gases at temperature T, kJ/kg·K; T
is the temperature of the gases, K; and Ta is the ambient temperature, K.
The average specific heat of the waste gases is 1.33 kJ/Nm3 , whereas 1450 ◦ C was
taken as the value of gas temperature.
3. Results
In the field of research, considering modelling (MLR) as a first step, an attempt was
made to apply and subsequently modify the Köhle equation [22]. Based on the analyses
carried out on a furnace operating under the industrial conditions, one should state the
lack of universality for the form of the model proposed by Köhle. The obtained values
differ from the measured ones by an average of 50% for the 1992 version and by an average
of 10% for the 2002 version of the equation [23]. As part of the work performed, the model
was modified, and one of the changes made was the replacement of the ML coefficient
relating to oxygen consumption by an equation taking into account the heat gain from the
oxidation of individual elements. This is due to the previously mentioned large number of
scrap classes, as well as the large share of the iron oxidation reaction as a source of chemical
90
Materials 2022, 15, 1601
energy. Therefore, the replacement of the coefficient that represents the average heat gain
by a separate function is justified. This function is written in the form:
⎛ ⎞
kg
MAM ·avg%MO.sl ·1000 Mg
⎝
EM = ∑ Msl · ·ΔHM ⎠ (16)
M
100%·MAMO
where Msl is the mass of slag, Mg; MAM is the atomic weight of considered metal, u;
avg%MO.sl is the average considered metal oxide content of slag, %; MAMO is the atomic
weight of considered metal oxide, u; and ΔHM is the the enthalpy of the oxidation reaction
of a considered element, converted to kWh/kg; (where M = Cr, Mn, Fe, Al, Si).
Figure 1 shows the comparison of the real values of electricity consumption with the
model values obtained based on two versions of the Köhle model: the 2002 version and its
modification. It can be noted that for the modified version a much higher model fit was
obtained (average difference 1%, maximum 20%).
450
M. Kohl 2002
Theoretical energy consumption,
Modyfied
400
[kWh/Mg]
Average Real
350
Average M.
Kohl 2002
300 Average
Modyfied
250
250 300 350 400 450
Real energy consumption on the furnace, [kWh/Mg]
Figure 1. Comparison of Köhle model results with real electricity consumption [23].
In the realities of the furnace characterised, the model did not adequately provide
quantitative answers as to what steps and extent of change should be taken to optimise the
process.
Therefore, mass and energy balances were calculated and a linear multiple regression
(MLR) model was undertaken.
For the calculation of mass and energy balances, it was necessary to identify the
chemical composition of the liquid products (slag and steel). For this purpose, slag and
metal samples were taken. A study of the chemical composition of slags was carried out for
200 melts in the same group of steel grades. The obtained average chemical compositions
of slag and steel are presented in Tables 2 and 3.
Table 2. Average chemical composition of EAF slags before tap.
Oxide CaO MgO MnO Cr2 O3 SiO2 P2 O5 Al2 O3 TiO2 FeO

% of weight 29.54 5.71 9.09 7.52 9.55 0.41 1.19 0.42 36.57
91
Materials 2022, 15, 1601
Table 3. Average chemical composition of liquid bath before tap.
Element C Mn Cr Ni Si P S
% of mass 0.071 0.083 0.107 0.109 0.005 0.008 0.107
In the case of the unit under consideration, the insufficient metering, the high diver-
sification of the steel grades, and the high variability of the charge (based on more than
40 scrap grades) led to some simplifications. For example, assuming a constant hot heel
level; ignoring the change in furnace geometry resulting from erosion of the lining and,
thus, assuming that the oxygen stream always reaches the bath and is fully consumed; and
assuming that the chemical composition of the scrap corresponds to the average of the
ranges described in PN 85/H 15000.
The results of the mass and energy balances obtained, and their analysis, clearly
indicate the areas which should be investigated in order to improve the scrap melting
process, e.g., appropriate use of natural gas, problems with furnace leak-tightness, and
use of oxygen. However, the predominance of chemical energy in the melting process
suggests that its optimisation will have a significant impact on the overall energy balance
of the melt.
The parameters used to create the MLR were time, which is related to the amount of
heat loss in the process; oxygen and gas consumption, as sources of chemical energy; the
amount of slag-forming materials used, whose melting requires a significant amount of
energy; and the scrap yield, which gives information on the amount of iron oxidised.
Calculations and multiple regression analyses were carried out to obtain the equation
identifying the demand for electric energy. Table 4 presents the results for the selected back-
ward MLR analysis, which gave the selected statistical model—the form of Equation (17).
Statistical calculations were performed for the total sample size, which was N = 1959; the
significance level of α = 0.05 was assumed. Statistical correlation between the considered
independent variables was checked; the relationships between the variables present in the
equation were statistically insignificant.
Table 4. Statistical parameters of calculated MLR model.
Variable b* Std. Err. of b* B Std. Err. of b t(1953) p-Value

Intercept 169.942 8.689 19.558 0.000
Tap-to-tap time 0.248 0.020 0.263 0.021 12.324 0.000
Oxygen to charge 0.403 0.020 3271 0.166 19.670 0.000
Natural gas to charge −0.055 0.020 −2633 0.952 −2767 0.006
Slag formers to charge 0.064 0.020 0.113 0.036 3166 0.002
Scrap yield 0.123 0.021 48.787 8.377 5.82 0.000
b* is a structural parameters of the regression function; B is an estimators of obtained regression equation; t is a
t-Student test probability distribution; and p-value is a probability level of variable.
Statistical results were obtained: the regression coefficient at the level of R = 0.497;
comparing for Rtable = 0.1946 (for N > 1000), one can state the significance of the calculated
linear correlation. Fisher–Snedecor’s F-value, at the α = 0.05 level, F(5.1953) = 128; for
Ho’s hypothesis, comparing it with the table critical value of Ftable = 2.108, Ho ’s hypothesis
of statistical significance R2 of the equation—can be accepted at the significance level of
α = 0.05, which implies the possibility of rejecting the alternative hypothesis. It can also be
seen that for all selected variables, the probability p is less than the significance level of 5%
(p < α), and the independent variables present in the equation are individually statistically
significant.
Following the above, a linear regression model was created (3), resulting in:
mox mCH4 m m
Eel = 169.94 + 0.26·tTTT + 3.27· − 1.98· + 0.11· sf + 48.79· hmł (17)
mscr mscr mscr mscr
92
Materials 2022, 15, 1601
where tttt is the tap to tap time, min.; mscr is the mass of scrap in charge, kg; mox is the mass
of oxygen, kg; mCH4 is the mass of natural gas, kg; msf is the mass of slag formers, kg; and
mhm is the mass of hot metal, kg
In Figure 2, the normality analysis of the distribution of the residuals was shown. It
seems, Shapiro–Wilk test: SW–W = 0.9971, p = 0.0009. Variance constancy: fulfilled, the
uniformity of the scatter of the residuals is constant across the width of the interval.
Figure 2. Normal distribution of residuals.
Simulations using the created MLR model confirmed the conclusions of the analysis
of the balances performed, i.e., the predominance of chemical energy and the inverse
proportionality of methane consumption to electrical energy consumption.
Taking into account the analysis results, which confirmed the conclusions of the mass
and energy balance, the description of the process efficiency (defined as the ratio of the
supplied energy to the amount of physical heat of the metal bath) as a function of the share
of chemical energy was undertaken.
In Figure 3, the analysis of the dependence of process efficiency on the share of chemical
energy was shown. A decrease in process efficiency (ηEAF ) is visible with an increase in the
share of chemical energy in the process. This trend in melt efficiency prompted an analysis
of the efficiency of chemical energy use in the process under study.
0.75
R² = 0.6498
0.70
0.65
ηEAF [%]
0.60
0.55
0.50
0.40 0.45 0.50 0.55
Share of chemical energy [%]
Figure 3. Dependence of process efficiency on the share of chemical energy.
93
Materials 2022, 15, 1601
As an example, the amount and timing of CH4 addition was analysed as a function of
the efficiency of the heat generated by its combustion.
It was assumed that the efficiency will change with the progress of scrap melting.
This is due to phenomena such as a change in the scrap surface with the progress of
melting, but most importantly due to a decrease in the temperature difference between the
flame and the charge, which heats up during the process. The change in burner efficiency
was re-calculated on the basis of the approach presented in [21,24,25] and is presented in
Figure 4.
Figure 4. Burner efficiency as a function of melting time progression [%] (own study based on data
from [21]).
The energy efficiency of the burners was considered first. This varies with the progress
of melting. This efficiency is, on average, between 70 and 40%, and can drop as low as
20%. To identify the energy transfer efficiency, the Gottardi methodology was used [25],
which made the energy efficiency of the burner dependent on, among other things, the
temperature difference between the flame and the charge, which decreases with the progress
of melting and, additionally, with a decrease in the scrap surface area. This dependence is
described by the function (18):
ηCH4 = 70.272 − 8·10−3 ·p2t + 5·10−5 ·p3t (18)
where ηCH4 is the energy efficiency of the burner [%]; and pt is the melting progression [%].
Based on the mathematical description of the change in burner efficiency as a function
of melting progress, three patterns of energy gain as a function of the average gas flow
possible under the conditions of the unit analysed were simulated. Boundary conditions
were established: minimum gas flow in burner mode, 90 Nm3 /h; the currently used average
gas flow rate, 110 Nm3 /h; and near-maximum flow rate, 200 Nm3 /h. The maximum burner
operation time—resulting from the melting stage characterised by the presence of scrap in
the “cold regions” of the walls—is 30 min in total for all the baskets.
The change in energy contributed by the burners over time (Figure 4), the cumulative
energy per 30 min (Figure 5a), and the cumulative energy as a function of the maximum
possible amount of gas supplied (Figure 5b) were considered.
94
Materials 2022, 15, 1601
90 2000
90 110 200 90 110 200
Cumulative energy input from burners

80 1800
Energy input of the burner [kWh]
70 1600
1400
60
1200
50
[kWh]
1000
40
800
30
600
20 400
10 200
0 0
0 10 20 30 0 10 20 30
Burner operating time [min] Burner operating time [min]
(a) (b)
Figure 5. (a) Variation of burner energy input as a function of burner operating time (b) cumulative
energy as a function of burner operating time
From the simulations obtained, it can be concluded that it is beneficial to intensify

the gas flow in the initial stage of the melt, as this allows the burner to use its moment of
greatest efficiency. The calculated maximum achievable gas consumption is, respectively,
135, 165, and 300 Nm3 , and the energy contributed is 826 kWh, 1009 kWh, and 1835 kWh.
Finally, the equation was created to simulate production costs (in the sense of direct
manufacturing costs), CO2 emissions, and process efficiency (Mg of hot metal/h)—it is
assumed that such a construction of the model will allow to control the technology in a
manner depending also on economic objectives.
The attempt to consolidate the results in terms of obtaining the unit melt cost function
contributed to the creation of Equation (19).
44
Pc = Eel ·X1 + mCH4 ·X2 + (mscr − mhm )·X4 + mox ·X3 + mC + mCH4 · ·X5 + msf ·X6 (19)
16
where Pc is the production cost; X1 is the cost per unit of electrical energy, PLN/kWh; X2
is the cost per unit of natural gas, PLN/Nm3 ; X3 is the cos per unit of technical oxygen,
PLN/Nm3 ; X4 is the average unit cost of scrap purchase, PLN/Mg; X5 is the cost per
emission of CO2 unit PLN/Mg; and X6 is the cost per unit of slag formers, PLN/kg.
Equation (5) was used to calculate production costs (in terms of direct production costs
of liquid steel), CO2 emissions, and process efficiency (Mg of hot metal/h)—assuming that
such a model construction would allow to control the technology in a manner depending
also on economic objectives. After calculating the production costs for the smelters under
consideration (forming the database), a correlation analysis was carried out between the
production cost (less the price of scrap metal forming the metal bath) and the amount of
chemical energy input; the results are presented in Figure 6.
95
Materials 2022, 15, 1601
250
Cost of producing [PLN/tonne of hot metal]

240
230
220
210
200 y = 708.63x-0,208
R² = 0.261
190
200 250 300 350 400
Amount of chemical energy [kWh/tonne of hot metal]
Figure 6. The amount of chemical energy in relation to the direct cost of producing a tonne of
metal bath.
The correlation coefficient is R = 0.513 and the t-Student’s t-value reaches t = 41.51 with
ttabular = 1.960. From the existence of this relationship (Figure 6) it should be concluded
that it is advantageous to increase the share of chemical energy in the process, but taking
into account the efficiency of its transfer by the oxygen treating bath stage.
4. Discussion
Modelling of the melting process in an electric arc furnace is possible under industrial
conditions, even with incomplete metering. However, the individual characteristics of the
unit under consideration have to be taken into account as existing models do not show
universality. The use of models developed for other units requires their modification and
taking into account the characteristics resulting from the proportion and share of chemical
and electric energy as well as the input and technological conditions of the furnace.
Increasing the chemical energy input into the melt lowers the required total amount
of electrical energy—methane consumption is inversely proportional to electrical energy
consumption. However, due to the technological effect of the fact that oxygen is fed most
intensively at the time of charge reheating—throughout this period—its correlation with
electrical energy is positive.
Based on model calculations for the analysed furnace unit using a modified Köhle
model from 2002, the model and actual data converged at the level of R = 0.36. The
calculated electricity demand differed from the actual demand at the maximum level of
20%. However, it did not give sufficient guidance for process control.
Thanks to the mass and energy balance calculations performed, results were obtained
from which it can be concluded that there is a predominance of chemical energy (52%) in
the process compared to electrical energy.
It can be concluded that it is advantageous to intensify the gas flow at the initial
stage of melting, as this makes it possible to use the moment of maximum efficiency of the
burners (presented on Figure 5a,b). These conclusions require a complementary analysis;
the real conditions of the process should be taken into account, which limit the maximum
flow on the oxy–gas burners in the initial stages of the process (risk of burner damage by
flame reflection from the scrap metal near the walls) and the fact that a higher energy input
will lead to a faster melting of the deposited scrap metal
The statistical calculations carried out and the analyses of the statistical indices ob-
tained made it possible to select the multiple linear regression model Equation (17)—as a
method giving the possibility of an easy interpretation with an appropriate representation
96
Materials 2022, 15, 1601
of the process in these conditions. The regression coefficient for the calculated model is
R = 0.497, the difference between the calculated value and the real one is, at maximum, 17%
(with mean 0). The obtained model was extended by the analysis of natural gas efficiency
and consolidated with the cost calculation equation.
Author Contributions: Conceptualization, P.M. and M.K.; methodology, M.M.; resources, M.M.; data
curation, M.M.; writing—original draft preparation, M.M.; writing—review and editing, P.M. and
M.K. All authors have read and agreed to the published version of the manuscript.
Acknowledgments: The present work was write as part of the second edition of “Implementation
Doctorate” programme. The authors gratefully acknowledge their support.
References
1. World Steel Association. Steel’s Contribution to a Low Carbon Future and Climate Resilient Societies—World Steel Position Paper; World
Steel Association: Brussels, Belgium, 2020; ISBN 978-2-930069-83-8. Available online: https://www.worldsteel.org/en/dam/jcr:
7ec64bc1-c51c-439b-84b8-94496686b8c6/Position_paper_climate_2020_vfinal.pdf (accessed on 10 November 2020).
2. Carlsson, L.S.; Samuelsson, P.B.; Jönsson, P.G. Using Statistical Modeling to Predict the Electrical Energy Consumption of an
Electric Arc Furnace Producing Stainless Steel. Metals 2020, 10, 36. [CrossRef]
3. Sung, Y.; Lee, S.; Han, K.; Koo, J.; Lee, S.; Jang, D.; Oh, C.; Jang, B. Improvement of Energy Efficiency and Productivity in an
Electric Arc Furnace through the Modification of Side-Wall Injector Systems. Processes 2020, 8, 1202. [CrossRef]
4. Xiaojun, X.; Shufeng, Y.; Jingshe, L.; Jinqiang, W.; Mengjing, Z.; Maolin, Y. Physical model experiment and theoretical analysis of
scrap melting process in electric arc furnace combined blowing system. Ironmak. Steelmak. 2020, 47, 748–756.
5. Gharib Mombeni, A.; Hajidavalloo, E.; Behbahani-Nejad, M. Transient simulation of conjugate heat transfer in the roof cooling
panel of an electric arc furnace. Appl. Therm. Eng. 2016, 98, 80–87. [CrossRef]
6. Coskun, G.; Yiğit, C.; Buyukkaya, E. CFD modelling of a complete electric arc furnace energy sources. Innovations 2016, 4, 22–24.
7. He, K.; Wang, L.; Li, X. Review of the Energy Consumption and Production Structure of China’s Steel Industry: Current Situation
and Future Development. Metals 2020, 10, 302. [CrossRef]
8. Faraway, J.J. Generalized linear models. In International Encyclopedia of Education, 3rd ed.; Peterson, P., Baker, E., McGaw, B., Eds.;
Elsevier: Amsterdam, The Netherlands, 2010; pp. 178–183. [CrossRef]
9. Cameron, L.T.; Hangos, K.M. Process Modelling and Model Analysis, 1st ed.; Academic Press: San Diego, CA, USA, 2001; Volume 4,
pp. 10–17; 253–275.
10. Haupt, M.; Vadenbo, C.; Zeltner, C.; Hellweg, S. Influence of Input-Scrap Quality on the Environmental Impact of Secondary
Steel Production. J. Ind. Ecol. 2017, 21, 391–401. [CrossRef]
11. Morales, R.D.; Rodriguez, H.; Conejo, A.N. A Mathematical Simulator for the EAF Steelmaking Process Using Direct Reduced
Iron. ISIJ Int. 2001, 41, 426–436. [CrossRef]
12. Bekker, J.G.; Craig, I.K.; Pistorius, P.C. Modeling and Simulation of an Electric Arc Furnace Process. ISIJ Int. 1999, 39, 23–32.
[CrossRef]
13. Logar, V.; Dovžan, D.; Škrjanc, I. Modeling and Validation of an Electric Arc Furnace Part 1, Heat and Mass Transfer. ISIJ Int. 2012,
52, 402–412. [CrossRef]
14. Logar, V.; Dovžan, D.; Škrjanc, I. Modeling and Validation of an Electric Arc Furnace Part 2, Thermo-chemistry. ISIJ Int. 2021, 52,
15. Štrumbelj, E.; Kononenko, I. A General Method for Visualizing and Explaining Black-Box Regression Models. In Adaptive and
Natural Computing Algorithms; ICANNGA 2011; Lecture Notes in Computer Science; Dobnikar, A., Lotrič, U., Šter, B., Eds.;
Springer: Berlin, Germany, 2011; Volume 6594, pp. 21–30.
16. Banks, H.T.; Tran, H.T. Mathematical and Experimental Modeling of Physical and Biological Processes, 1st ed.; Chapman and Hall/CRC:
New York, NY, USA, 2009; pp. 3–5.
17. Carlsson, L.S.; Samuelsson, P.B.; Jönsson, P.G. Predicting the Electrical Energy Consumption of Electric Arc Furnaces Using
Statistical Modeling. Metals 2019, 9, 959. [CrossRef]
18. Köhle, S. Effects on the electric energy and electrode consumption of arc furnaces. Stahl Eisen 1992, 112, 59–67.
19. Köhle, S.; Lichterbeck, R.; Paura, G. Verbesserung der Energetischen Betriebsführung von Drehstrom-Lichtbogenöfen; European
Commission: Brussels, Belgium, 1996.
97
Materials 2022, 15, 1601
20. Meier, T.; Gandt, K.; Echterhof, T.; Pfeifer, H. Modeling and Simulation of the Off-gas in an Electric Arc Furnace. Metall. Mater.
Trans. B 2017, 48, 3329–3344. [CrossRef]
21. Karbowniczek, M. Electric Arc Furnace Steelmaking, 1st ed.; CRC Press Taylor & Francis Group: Boca Raton, FL, USA, 2021; p. 239.
22. Köhle, S. Recent improvements in modelling energy consumption of electric arc furnaces. In Proceedings of the 7th European
Electric Steelmaking Conference, Venice, Italy, 26–29 May 2002.
23. Migas, P.; Moskal, M.; Karbowniczek, M.; Czyż, J. Analysis of the energy consumption during the electric arc process. In Proceed-
ings of the Iron and Steelmaking 2019, Modern Metallurgy the XXVIII International Scientific Conference, Liptovský Mikuláš,
Slovakia, 23–25 October 2019.
24. Pretorius, E.; Oltmann, H.; Jones, J. EAF Fundamentals: Charging, Melting and Refining; LWB Refractories: New York, NY, USA,
2005; pp. 16–21.
25. Bergman, K.; Gottardi, R. Design criteria for the modern UHP electric arc furnace with auxiliaries. Ironmak. Steelmak. 1990, 17,
282–287.
98
materials
Article
Numerical Simulation Study of Gas-Solid Heat Transfer and
Decomposition Processes of Limestone Calcined with Blast
Furnace Gas in a Parallel Flow Regenerative Lime Kiln
Shaopei Duan 1,2 , Baokuan Li 1,2, * and Wenjie Rong 1,2
1 School of Metallurgy, Northeastern University, No. 3-11, Wenhua Road, Heping District,
Shenyang 110819, China; [email protected] (S.D.); [email protected] (W.R.)
2 Key Laboratory of Data Analytics and Optimization for Smart Industry, Ministry of Education,
Northeastern University, Shenyang 110819, China
Abstract: Quicklime is an essential reducing agent in the steel smelting process and its calcination
from limestone is accompanied by considerable energy consumption. As a relatively economical lime
kiln, the Parallel Flow Regenerative (PFR) lime kiln is used as the main equipment for the production
of quicklime by various steel industries. PFR lime kilns generally use natural gas as the fuel gas.
Although natural gas has a high calorific value and is effective in calcination, with the increasing
price of natural gas and the pressure saves energy and protect the environment, it makes sense of
exploring the use of cleaner energy sources or other sub-products as fuel gas. The use of blast furnace
gas (BFG) as a low calorific value fuel gas produced in the steel smelting process has been of interest.
This paper therefore develops a set of mathematical models for gas-solid heat transfer and limestone
decomposition based on a Porous Medium Model (PMM) and a Shrinking Core Model (SCM) to
numerically simulate a PFR lime kiln using BFG in order to investigate the feasibility of calcining
Citation: Duan, S.; Li, B.; Rong, W.
limestone with low calorific fuel gas and to provide a valuable reference for future development of
Numerical Simulation Study of
such processes and the kiln structure improvement.
Gas-Solid Heat Transfer and
Decomposition Processes of
Limestone Calcined with Blast
Keywords: PFR lime kiln; lime calcination; BFG; numerical simulation; PMM; SCM
Furnace Gas in a Parallel Flow
Regenerative Lime Kiln. Materials
2022, 15, 4024. https://doi.org/
10.3390/ma15114024 1. Introduction
Academic Editor: Irina Hussainova
The main component of quicklime is calcium oxide (CaO), which is widely used in
the metallurgical, construction, agricultural and chemical industries because of its ease of
Received: 7 May 2022 preparation, low price and good reducing properties. Nearly half of the world’s quicklime is
Accepted: 4 June 2022 used in metal smelting, so the metallurgical industry has the greatest demand for quicklime.
Published: 6 June 2022 Quicklime is obtained by calcining limestone at high temperatures, the main component
Publisher’s Note: MDPI stays neutral of which is calcium carbonate (CaCO3 ) with small amounts of silicon dioxide (SiO2 ) and
with regard to jurisdictional claims in magnesium oxide (MgO).
published maps and institutional affil- Among all types of quicklime products, the quantity of quicklime used in metal
iations. smelting is not only huge, but there are also strict requirements for its high reducibility. At
temperatures less than 1173 K the calcination does not occur and is called “under-burning”;
at temperatures above 1423 K the surface porosity becomes smaller and thus less chemically
active and is called “hard-burnt” or “over-burnt” At temperatures between 1173 K and
Copyright: © 2022 by the authors. 1423 K, calcination results in large surface porosity and strong chemical activity, making it
suitable as an excellent reducing agent for metal smelting, which is known as “soft-burnt”.
Many researches have been carried out on the association between calcination and
the chemical activity of quicklime.Suju Hao finds the lime activity degree was determined
by acid-base titration, and the lime pore distribution was measured by mercury intrusion
porosimetry [1]. Vola deal with thermal analyses, burning trials and reactivity tests on
15 carbonate rocks, i.e., pure and impure carbonates, mud-supported and grain-supported
4.0/).

Materials 2022, 15, 4024
limestones, crystalline marbles, and dolomites, used for the production of different lime
products in industrial vertical shaft kilns worldwide [2]. Wang does many studies on
the relationship between microstructure and physicochemical properties of limestone
including, porosity, bulk density, pore size distribution, specific surface area and activity
were studied under the condition of rapid heating from room temperature to 1350–1550 ◦ C,
the results showed that complete decomposition of limestone (diameter 12.5–15 mm) at
1350 ◦ C, 1450 ◦ C and 1550 ◦ C took 11.7 min, 9.2 min and 6.9 min, respectively [3]. The
subject of was to evaluate the influence of lime particle size and lime calcination level
as well as slag properties (basicity, viscosity, CaO activity and density) on lime particle
dissolution at temperature in the early stage of process blowing time of the basic oxygen
furnace (1400 ◦ C) [4].
The Parallel Flow Regenerative (PFR) lime kiln is widely used in the metallurgical
industry due to its high daily production capacity (600 t/d–800 t/d), good chemical activity
of the product and high thermal efficiency. At the same time, PFR lime kilns are constantly
developing and evolving into a variety of improved models.
A three-dimensional steady-state model to predict the flow and heat transfer in a
rotary lime kiln is presented [5]. Senegačnik et al. present a case of a remodelled annular
shaft kiln, in which the air excess ratio can be reduced to its optimal level by recirculation
of recuperator waste gas and its injection into the combustion chambers [6]. The objective
of Gutiérrez is to analyze the energy and energy consumption of the calcination process in
vertical shaft kilns, in order to identify the factors affecting fuel consumption [7]. The PFR
lime kiln has established itself worldwide for the type of product. Piringer explains the
functional principle of such lime kilns [8]. Krause present DEM-CFD simulations of the
transient processes occurring in an industrial scale PFR-kiln [9].
There are also many CFD numerical simulations for lime kilns. The proposed pre-
dictive control method can make the output value of the calcination zone temperature of
the lime rotary kiln fast and stable to track the change of the reference value [10]. Suárez
builds a regression model for each one of the three prime variables (gas consumption,
SO2 , and NOx emissions) of a lime kiln employed in the paper manufacturing process
using the multivariate adaptive regression splines (MARS) method in combination with
the artificial bee colony (ABC) technique [11]. Metz evaluates the influence of electric arc
furnace dust (EAFD) and lime kiln waste (LKW) [12]. Juneja covers the modeling and
control aspects of Lime Kiln process [13]. CFD simulations of an industrial scale rotary
kiln for cement clinker production are also conducted [14]. The impact of these coating
layers on the clinker production process within a rotary kiln is investigated with CFD
simulations [15]. Sagastume and Wenjie analyzed the thermal performance of the lime kiln
using the worm-efficiency, pointing out that the performance of a new direction for the
analysis of the performance of lime kilns [16,17].
Cheng, et al. [18,19] used the shrinking core model as the basis to find the gas-solid
temperature profile inside the ASK by numerical simulation and analyzed the limestone
decomposition rate related to the limestone type. Bluhm-Drenhaus [20] and Krause [21]
combined the finite volume method (FVM) with the finite element method(FEM) to simulate
the temperature field of the gas-solid flow inside the lime kiln and calculated the trajectory
of the particles in the kiln with the flow process of the fuel gas. Ning Jingtao [22] studied the
calcination of fine limestone using a shrinking core model and confirmed that 1200–1473 K
is a more desirable calcination temperature for limestone. The coupled simulation of DEM
(discrete element method) and CFD (finite volume method) is a method to solve the stacked
bed problem in lime kilns, where the DEM model solves the limestone decomposition
behavior while the CFD model simulates the gas-solid two-phase heat exchange and the
pressure drop in the stacked bed [23]. Mikulčić, et al. [24] used this coupled approach to
study the cement industry for CO2 emissions. Meanwhile, a two-energy equation model
based on a porous media model is developed by Long Huilong, et al. [25] to simulate the
gas-solid chemical reaction in a non-stationary manner, which is in agreement with the
experimental results. Shanshan Bu, et al. [26] studied the effect of three particle contact
100
Materials 2022, 15, 4024
models on the heat transfer in a stacked bed and found that the three particle contact
models have a large influence on the calculation results in turbulent flow. Liu, et al. [27]
used a gas-burnt lime shaft kiln as a research object and solved the temperature field inside
the lime kiln using a local non-thermal equilibrium model.
Previously, in order to ensure the calcination effect, natural gas was generally used
as the fuel gas, mixed with air and then burned to generate heat and heat as the heat
source for the calcination of limestone. While this traditional process has ensured a good
quality of limestone calcination, it has also created significant problems. Firstly, the main
component of natural gas is methane, and with the progressively stricter restrictions on
greenhouse gas emissions in countries around the world, strict regulations have been
made on the relevant equipment for metallurgical enterprises, and PFR lime kilns, as the
main equipment consuming natural gas, have to consider energy saving improvements;
secondly, the price of natural gas continues to rise due to fluctuations in global political
and economic factors, bringing huge cost pressures on the relevant enterprises; finally, in
some metallurgical enterprises, blast furnace gas is difficult to be widely used in various
production processes because the combustion produces a temperature of only 1673 K to
1773 K, and the utilisation rate is low. In addition, the PFR lime kiln supports 40 mm
to 120 mm particle size for calcination when using natural gas as the fuel gas, and there
is still a lack of research into the range of particle sizes supported by the change of fuel.
In summary, the consideration of low calorific value sub-products, represented by blast
furnace gas, as an alternative to natural gas for limestone calcination is gaining interest in
the industry. Therefore, this paper makes an attempt in this direction by replacing natural
gas with blast furnace gas and establishing a reliable mathematical model of gas-solid
heat exchange and calcium carbonate decomposition reaction for the study of its effect,
providing a reference basis for future equipment design and the process improvement.
2. Mathematical models
The simulation of the PFR lime kiln is based on the Finite Volume Method (FVM), and
for any scalar quantity such as temperature, the solution equation as follows:
∂ρφ ∂ → ∂φ
+ ( ρ ui φ − Γ ) = Sφ (1)
∂t ∂xi ∂xi
The mass conservation equation in fluid mechanics theory, also known as the con-
tinuity equation, is a fundamental equation of computational fluid dynamics based on
a continuous model. It assumes that density, coordinates, time and other variables are
continuous in space. Its expression is in the form of Equation (2).
∂ρ →
+ ∇ · ( ρ v ) = Sm (2)
∂t
Equation (2) is the basic form of the mass conservation equation. The model is used in
PFR lime kilns to calculate the mass transfer between gas and solid, for example, the mixing
of heating and cooling gases, and the emission of carbon dioxide during decomposition.
The Navier-Stokes equations are equations describing the conservation of momentum
in the motion of incompressible fluid, i.e., the sum of the external forces acting on the
control body is equal to the rate of change of momentum on the control body with time,
and have the following basic form.
→
∂(ρ v ) →→ = → →
+ ∇ · (ρ v v ) = −∇ p + ∇ · ( τ ) + ρ g + F (3)
∂t
In PFR lime kilns, the momentum equation is used to simulate the gas flow and wind
resistance variations.
The space inside the lime kiln is filled with limestone particles and the particles move
slowly downward, so it can be treated as a moving bed problem inside the lime kiln. The
101
Materials 2022, 15, 4024
porous media model allows accurate calculation of the pressure drop problem for the
moving bed/packed bed, which assumes that the space is occupied by a mixture of solids
and fluids.
The introduction of the porous media model in the PFR lime kiln simulation is essen-
tially the introduction of a source term for the momentum loss in the momentum equation
to describe the pressure drop (flow wind resistance) generated by the fluid flowing through
fluid zone. And the losses originate from the flow viscous resistance and inertial resistance,
which is calculated as below.
Ergun equation
|∇ p| 150μ(1 − γ)2 1.75ρ(1 − γ) 2

= v∞ + v∞ (4)
L Dp2 γ3 Dp γ3
1 150(1 − γ)2
= (5)
α Dp2 γ3
3.5(1 − γ)
C2 = (6)
Dp γ3
The shrinking core model assumes that the reactants are spheres, that heat is supplied
uniformly from all directions, and that the chemical composition of the raw materials is uniform.
∂rCaCO3 MCaCO3
= −k · · RD (7)
∂t ρCaCO3

RD = kD peq − pco2 (8)

kD = 0.0001Tp exp −4026/Tp · YT.C (9)
peq = 101325 exp[17.74 − 0.00108Ti + 0.332 log( Ti ) − 22020/Ti ] (10)

480
TP > 1150K
YT.C = TP −958 (11)
2.5 TP ≤ 1150K
4πλ1 λ2
λ= (12)
λ1 1
rc1 − rc2.m + λ2 r
1 1
c1.m
− 1
rc1
Limestone decomposition rate is the volume fraction of decomposed calcium carbonate:

3
rc2
XS = 1 − 3
(13)
rc1
The temperature equations for gas-solid convective heat transfer, internal and external
particle heat conduction, and between solid particles
∂* + →
ϕCaCO3 ρi c pi Ti + ∇ v down ϕCaCO3 ρi c pi Ti = λ(1 − γ)( To − Ti )/Vs − k · QD ΔHR (14)
∂t
2
4πrc1
QD = (1 − γ ) × RD (15)
Vp
Energy equation of the external particle (calcium oxide)
∂ →
∂t ( ϕCaO ρo c po To ) + ∇· ( v down ϕCaO ρo c po To ) = (16)
∇ · ((kCaO + eb )∇ To ) + av hv Tg − To − λ(1 − γ)( To − Ti )/Vs
102
Materials 2022, 15, 4024
The radiation equivalent thermal conductivity treatment of porous media used for
radiation between solids [28]
eb = 16σTo3 /(3β) (17)
The convective heat transfer coefficient hv is approximately calculated by flow-across tube
bundle model in the heat exchanger model [29], which is calculated as following equations
Nu · λ g
hv = (18)
lz
1.6274Re−0.575
Nu = Pr1/3 · (19)
γ
lz = 0.0178ρ0.596 (20)
ρg · Dp · u
Re = (21)
μ
Pr = ν/a (22)
λg
a= (23)
ρg c pg
Sp
a v = (1 − γ ) × (24)
Vp
The gas energy equation
∂ →
γρg cpg Tg + ∇ γ v ρg cpg Tg = ∇ · kg ∇ Tg + av hv To − Tg (25)
∂t
! "
MCaO /ρCaO
ε = 1− × XS (26)
MCaCO3 /ρCaCO3
ϕg = (1 − γ · ε + γ (27)
ϕCaCO3 = (1 − γ) × (1 − XS ) (28)
ϕCaO = 1 − ϕg − ϕCaCO3 (29)
3. Physical Models
3.1. Geometric Model
The PFR lime kiln has a total height of 20 m and is divided into two chambers, which is
connected by a cross channel. Each chamber is divided into a preheating zone, a calcination
zone and a cooling zone from top to bottom. The preheating zone is 6 m high, The upper
part is the limestone inlet and also the exhaust gas outlet, and the lower part has 8 fuel gas
jets. The height of the calcination zone is 9 m. Below the calcination zone is the cooling
zone, which is 5 m high and is surrounded by a channel in the upper outer ring, which
gradually contracts downwards in the middle to the plane of the cooling air inlet and then
gradually expands and extends to the CaO outlet (Figure 1). Chamber 2 has the same
components as chamber 1.
A structural grid is used to discretize the entire computational domain. In the re-
gion of structurally complex calcination and preheating zones, the grid is sufficiently
refined. The grid resolution affects not only the computational time and resources, but
also the reliability of the simulation results. The sensitivity for grid numbers of 4.6 million,
5.2 million and 5.5 million was examined in this PFR lime kiln model to obtain grid inde-
pendent results. The comparison results show that a denser grid has a minor effect on the
computational results.
103
Materials 2022, 15, 4024
3.2. Operating Procedures

The PFR lime kiln consists of two chambers, with the fuel gas in Chamber 1 injected
into the calcination zone to calcine the limestone while the fuel gas jet in Chamber 2 does
not operate. The main component of limestone is the calcium carbonate, which is heated at
high temperatures to decompose into calcium oxide and carbon dioxide, a process known
as limestone calcination. When calcination begins, the high temperature gas heats the
calcium carbonate, which starts to decompose into calcium oxide and carbon dioxide from
the surface to the inside, and as the decomposition process proceeds, the material gradually
moves down and from the calcination zone into the cooling zone. Cooling air is blown into
the cooling zone from the bottom up by the cooling air inlet to quickly cool the material
that has just entered the cooling zone so that its chemical activity is not reduced by the
high temperature. The material is cooled as it gradually moves down and is eventually
discharged through the CaO outlet. The high temperature gas is involved in the calcination
and the temperature is reduced, while the cooling air is involved in the cooling and the
temperature is increased. The two are mixed in the cross channel to form the exhaust gas.
The heat carried by the exhaust gas has a preheating effect on the unreacted material in
Chamber 2, while the cooling air is continuously blown into the Chamber 2 cooling gas
inlet to enhance the exhaust gas convection and accelerate the efficiency of the exhaust gas
upwards. Finally, after passing through the material in the calcining and preheating zones
in Chamber 2, the exhaust gas is discharged by the upper exhaust gas outlet, while the
limestone inlet in Chamber 1 is only feeding the material at this time and is not operating as
an exhaust gas outlet. After 15 min of operation of Chamber 1, fuel gas is instead injected
by the fuel gas jet in Chamber 2, while Chamber 1 functions as an exhaust gas outlet for
15 min. The PFR lime kiln completes one cycle every 30 min and continues to operate
(Figure 2).
Figure 1. Schematic diagram of the PFR lime kiln components.
104
Materials 2022, 15, 4024
3.3. Boundary Conditions and Calculation Methods

As the outer walls of the PFR lime kiln are well insulated, the inner surface of the kiln
is used as an adiabatic boundary. The interior of the kiln is divided into two fluid domains,
the porous media area filled with the calcium carbonate and the gas area where only the
exhaust gas is present. The boundary below the fuel gas nozzle is used as the fuel gas
inlet boundary, while the rest of the nozzle is used as the exterior. Also, the temperature of
the blast furnace gas after combustion is used as the fuel gas inlet temperature and other
calculation conditions are shown in Table 1.
Figure 2. Schematic diagram of the operating procedure of the PFR lime kiln.
Table 1. Calculation conditions.
Calculation Conditions Value Unit

Fuel gas inlet velocity 20 m/s
Fuel gas inlet temperature 1673 K
Fuel gas nozzle diameter (8) 70 mm
Cooling air inlet velocity 10 m/s
Cooling air inlet temperature 300 K
Cooling air inlet diameter 1000 mm
Material movement speed 1.54 m/h
Initial temperature of material (limestone) 300 K
Average diameter of material 40, 80, 120 mm
Calcium carbonate density 3310 kg/m3
Calcium oxide density 2810 kg/m3
Calcium carbonate thermal conductivity 2.26 W/m·k
Calcium oxide thermal conductivity 0.07 W/m·k
Calcium carbonate decomposition
1073 K
temperature
Void fraction 0.36, 0.41, 0.46 -
Density, thermal conductivity and void fraction is determined in the laboratory.

The semi-implicit method of coupled pressure equations (referred to as the SIMPLE
algorithm) was used to solve the coupled velocity and the pressure problem in the process
of solving the equations; two equations in the turbulence equations were solved in the
105
Materials 2022, 15, 4024
first-order windward difference format, and the remaining equations were treated by the
second-order windward difference; the convergence factor of each equation was set to
10−6 .
The following hypotheses will also be established in this study in order to exclude any
interference to the results.
i. The limestone is composed exclusively of calcium carbonate, disregarding the in-
fluence of impurities that are in fact present in the limestone, such as MgO, on the
calcination results.
ii. The limestone raw material is spherical and the particle size refers to the diameter.
iii. Ignoring the effect of airflow on the movement of the material.
4. Results and Analysis

A number of locations within the PFR lime kiln were taken for analysis, as shown
in Figure 3, where the two intermediate lines are located in the geometric centre of the
chambers called mid lines and the near wall lines are located 40 mm from the chamber
side wall, with at least one material particle accommodated between the line and the side
wall, which is not only farther away from the fuel gas inlet and poorly heated, but is also
affected by the side wall effect, with differences in particle movement and decomposition
downwards from mid lines. The exhaust gas temperature was analysed by taking the
exhaust gas outlet above chamber 2. The central profile Plane 1 is taken to show the overall
distribution of all fields. Plane 2 is taken in the centre of the intermediate channel to
analyse the mixture of exhaust gas and cooling air after calcination. Plane 3 is taken at the
intersection of the calcining and cooling zones and analyses the surface temperature of the
solids after calcination, the core temperature and the decomposition of the limestone.
Figure 3. Schematic diagram of the analysis position.
106
Materials 2022, 15, 4024
4.1. Gas Velocity Field

The gas velocity distributions on the mid lines and near wall lines at 40 mm, 80 mm
and 120 mm particle sizes are shown in Figure 4. At a distance of 14 m from the CaO
outlet, fuel gas enters the calcination zone from the boundary to heat the limestone, and
as the distance gradually decreases, the mid lines gas velocity increases, but is blocked
by the solid material and the increase is small. Until about 5 m from the CaO outlet, the
gas velocity decreases and then sharply increase to about 7 m/s at about 2 m from the
CaO outlet due to the effect of cooling air blowing in. Due to the high temperature and
low density of the fuel gas, some of it rises into the preheating zone after being blown
into the calcination zone of the lime kiln. In addition, after the cooling air has cooled the
high temperature material, some of the exhaust gas does not flow through the intermediate
passage towards chamber 2, but enters the preheating zone via the calcination zone of
chamber 1.
(a) (b) (c)
Figure 4. Gas velocity variation on the mid lines and near wall lines for 40 mm, 80 mm and 120 mm
particle sizes. (a) Gas velocity variation at 40 mm; (b) Gas velocity variation at 80 mm; (c) Gas velocity
variation at 120 mm.
The gas velocity distribution of the near wall line in the calcination zone shows an
opposite trend to that of the mid line. As one approaches the CaO outlet, the near wall
line gas velocity decreases and the gas reaches the cooling zone with a velocity close to
that of mid lines gas velocity. However, at closer distances from the CaO outlet, the near
wall line gas velocity does not show a sharp increase in velocity similar to mid line gas
velocity, but increases slowly to below 2 m/s. Similarly, in the preheating zone, the near
wall line gas velocity is less affected by the rising high temperature fuel gas and cooling
air convergence, and the gas velocity is smaller, approaching 0 near the top limestone
inlet. It should be noted that the velocity at 14 m from the CaO outlet is not the fuel gas
inlet velocity, as neither mid line nor the near wall line passes through the fuel gas inlet
boundary. (Figure 4a–c)
Similarly, the variation of gas velocity tends to be the same for the three particle sizes
of 40 mm, 80 mm and 120 mm (Figure 5a–d), indicating that the effect of different void
fractions on gas velocity is small.
Combined Figures 4 and 5 show that the gas velocity varies less with particle size.
Whereas the velocity variations in the centre of the chamber (mid lines) and the kiln side
wall (near wall lines) converge in the calcination zone, they show different variations in
the preheating and cooling zones. The centre of the cooling air inlet is on the same axis as
the centre of the chamber, which inevitably leads to higher velocities in the centre of the
chamber. The side walls of the chamber, on the other hand, are less affected by the fuel gas
and cooling air, and their velocity variation over the sections is also less.
107
Materials 2022, 15, 4024
(a) (b)
(d)
(c)
Figure 5. Comparison of gas velocities at mid lines and near wall lines in chamber 1 and 2. (a) Change
in gas velocity at the mid line of chamber 1; (b) Change in gas velocity at the mid line of chamber 2;
(c) Change in gas velocity at the near wall line of chamber 1; (d) Change in gas velocity at the near
wall line of chamber 2.
4.2. Gas Temperature Field

As the gas temperature distribution is directly influenced by the gas velocity distri-
bution, there is little difference in the gas temperature distribution at 40 mm, 80 mm and
120 mm particle sizes (Figure 6a–c). At 40 mm particle size, the temperature distribution in
the calcination zone of chamber 2 (where calcination does not occur) is uniform (Figure 6a),
while at 80 mm particle size, the temperature in the centre of the calcination zone of cham-
ber 2 is significantly higher than that in the side walls (Figure 6b), and as the particle size
increases to 120 mm, the temperature in the centre of both the calcination and preheating
zones of chamber 2 is higher, with less difference between the temperature in the near wall
zone and the center (Figure 6c).
As shown in Figure 7a, the temperature of the near wall line of chamber 1 at 40 mm
particle size varies little from the preheating zone to the lower part of the calcination zone,
until the junction of the cooling zone, where the gas temperature is influenced by the
cooling air and decreases to around 350 K. The temperature of the mid line of chamber 1
remains similar to the near wall line in the preheating zone, but increases rapidly up to
1200 K upon entering the calcination zone. The mid line of chamber 1 remains at a similar
temperature to the near wall line in the preheating zone, however, the temperature increases
rapidly in the calcination zone due to the high temperature of the fuel gas up to 1200 K.
After entering the cooling zone, the mid line gas temperature is also affected by the cooling
air and decreases rapidly to around 350 K. The gas temperature at the mid line of chamber
2 varies more gently with the height of the chamber and only changes significantly between
the cooling zone and the cross channel. The change in gas temperature in chamber 1 at
80 mm and 120 mm particle size is more or less the same as the change in gas temperature
at 40 mm (Figure 7b,c).
108
Materials 2022, 15, 4024
(a) (b) (c)
Figure 6. Gas temperature distribution at 40 mm, 80 mm and 120 mm particle sizes. (a) Gas
temperature distribution at 40 mm (b) Gas temperature distribution at 80 mm (c) Gas temperature
distribution at 120 mm.
(a) (b) (c)
Figure 7. Gas temperature variation on the mid and near wall lines for 40 mm, 80 mm and 120 mm
particle sizes. (a) Gas temperature variation at 40 mm (b) Gas temperature variation at 80 mm (c) Gas
temperature variation at 120 mm.
The variation in mid line and near wall line temperatures in chamber 1 is essentially
the same for all three particle sizes (Figure 8a,c). In chamber 2, the variation in mid line
and near wall line gas temperatures is more pronounced with lime kiln height for all three
particle sizes. 120 mm particle size chamber 2 mid line gas temperatures is higher than
80 mm and 40 mm chamber 2 mid line gas temperatures (Figure 8b). Similarly, the near
wall gas temperature of chamber 2 for 120 mm particle size is higher than the near wall gas
temperature of chambers 2 for 80 mm and 40 mm (Figure 8d).
The cross channel has only high temperature exhaust gas and no solid material, so
the temperature variation is more pronounced than in the chambers, and is characterised
by uneven gas mixing and temperature distribution in the outer ring of chamber 1, while
entering the centre of the channel, the mixing is uniform and the average temperature is
highest at this point, and gradually decreases as the gas enters the outer ring of chamber 2
(Figure 9).
109
Materials 2022, 15, 4024
Figure 8. Comparison of gas temperature variations in the mid line and near wall line of chamber 1
and 2. (a) Change in gas temperature at the mid line of chamber 1; (b) Change in gas temperature
at the mid line of chamber 2; (c) Change in gas temperature at the near wall line of chamber 1; (d)
Change in gas temperature at the near wall line of chamber 2.
Figure 9. Gas temperature distribution and variation in the cross channel at 40 mm, 80 mm and
120 mm particle sizes.
The gas temperature distribution at the exhaust gas outlet in Figure 10 matches the gas
temperature in Figure 6. 120 mm particle size has the highest gas temperature at the exhaust
gas outlet, followed by 80 mm, while 40 mm particle size has the lowest gas temperature.
110
Materials 2022, 15, 4024
The gas temperature at the centre of the exhaust gas outlet is also higher than near wall,
which is consistent with the results in Figure 7.
(a) (b) (c)

Figure 10. Gas temperature distribution and variation in exhaust gas outlet at 40 mm, 80 mm and
120 mm particle sizes. (a) Gas temperature in exhaust gas outlet of 40 mm; (b) Gas temperature in
exhaust gas outlet of 80 mm; (c) Gas temperature in exhaust gas outlet of 120 mm.
Combined with the gas temperature variations and distributions in Figures 6–10, the
gas temperature distributions in chamber 1 are similar for 40 mm, 80 mm and 120 mm
particle sizes, while the gas temperature in chamber 2 increases as the particle size gradually
increases, as reflected in the cross channel and the exhaust gas outlet cross section. This
indicates that the gas temperature distribution varies with the particle size (void fraction)
for a given gas velocity. The effect of particle size (void fraction) on gas temperature
decreases as the gas velocity increases.
4.3. Solid Surface Temperature

For CaO used in reduction reactions in metallurgical processes, the chemical activity
is the most important indicator, and maintaining a solid surface temperature of no more
than 1423 K during calcination is a decisive parameter in maintaining chemical activity. If
the solid surface pore size is reduced over a long period of time above 1423 K, the chemical
activity decreases and is referred to as “hard-burnt” or “over-burnt”, so it is important to
maintain a reasonable solid surface temperature.
The 40 mm solid material briefly reaches about 1600 K near the fuel gas inlet in the
calcination zone of chamber 1, then the temperature drops rapidly as the material moves
down into the cooling zone, where the solid surface temperature is rapidly cooled to room
temperature by the cooling air. In contrast, the solid surface temperature in the preheating
zone does not exceed 1200 K (Figure 11a). The solid material surface temperatures at 80 mm
and 120 mm particle size also show similar temperature distributions, but in chamber 2 the
solid material surface temperature distribution at 80 mm and 120 mm particle size differs
from that at 40 mm, which is more in line with the gas temperature distribution in Figure 6.
As shown in Figure 12a–c, the solid surface temperatures at the mid and near wall
lines of chamber 1 are similar in the preheating zone and gradually increase as they move
from the calcination zone towards the CaO outlet, with a greater temperature gradient
at the mid line than at the near wall line. The mid line and near wall line solid surface
temperatures are rapidly cooled by the cooling air until they reach the cooling zone, where
the temperature rapidly decreases to around 300 K to 350 K. Chamber 1 shows similar
variations for all three particle sizes: 40 mm, 80 mm and 120 mm. In chamber 2, however,
the temperature of the near wall line is characterised by a larger particle size and a higher
surface temperature due to the influence of the high temperature exhaust gas brought in by
the cross channel at different temperatures.
111
Materials 2022, 15, 4024
(a) (b) (c)
Figure 11. Solid surface temperature distribution at 40 mm, 80 mm and 120 mm particle
sizes. (a) 40 mm solid surface temperature distribution (b) 80 mm solid surface temperature
distribution (c) 120 mm solid surface temperature distribution.
(a) (b) (c)
Figure 12. Variation of solid surface temperature on the mid lines and near wall lines for 40 mm,
80 mm and 120 mm particle sizes. (a) 40 mm solid surface temperature variation (b) 80 mm solid
surface temperature variation (c) 120 mm solid surface temperature variation.
The mid line and near wall line solid surface temperatures of chamber 1 converge
for each particle size (Figure 13a,c). In the preheating zone, the solid surface temperature
remains around 1050 K. After calcinate in the calcining zone, the mid line solid surface
temperature rises close to 1200 K and the near wall line solid surface temperature reaches
1100 K. Neither exceeds 1423 K, which does not lead to hard-burnt and reduced chemical
activity. The trend in the mid line and near wall line solid temperatures in chamber 2
correlates with the change in gas temperature, with the largest solid surface temperature
gradient at 120 mm particle size reaching 700 K in the mid line and 750 K in the near wall
line, and up to 650 K in the mid line and near wall line line at 40 mm.
Plane 3 is the interface between the calcination and cooling zones, as shown in
Figure 14 which reflects the temperature field of the solid material for 40 mm, 80 mm
and 120 mm particle sizes before entering the cooling zone. At particle sizes 40 mm
and 80 mm, the solid surface temperature distribution in chamber 1 is basically the same,
with a maximum value of 1170 K, while the 80 mm particle size has a higher solid surface
temperature gradient in chamber 2 than at 40 mm particle size, with a maximum tempera-
ture of 690 K, while the 40 mm particle size is only about 650 K. The maximum temperature
112
Materials 2022, 15, 4024
in chamber 1 is still around 1170 K when the particle size is 120 mm, but the distribution is
wider than at 40 mm and 80 mm, and the maximum solid surface temperature in chamber
2 reaches 740 K. In addition, the lowest surface temperature for all three particle sizes in
chamber 2 is around 510 K.
Figure 13. Comparison of solid surface temperature variations in the mid line and near wall line of
chamber 1 and 2. (a) Change in solid surface temperature at the mid line of chamber 1; (b) Change in
solid surface temperature at the mid line of chamber 2; (c) Change in solid surface temperature at
near wall line of chamber 1; (d) Change in solid surface temperature at near wall line of chamber 2.
Figure 14. Distribution and variation of solid surface temperatures in double chambers on Plane 3.
The solid surface temperature did not reach the upper limit of 1423 K hard-burnt
temperature, which is also related to the fact that the heat of combustion of BFG is lower
than the heat of combustion of natural gas. In addition, the near wall line has a significant
difference between the downward velocity and the temperature gradient of the material
113
Materials 2022, 15, 4024
and the mid line material due to the side wall effect. Although the gas temperature fields
converge for all three particle sizes, the solid surface temperature still shows a greater
temperature gradient at 120 mm in chamber 2 as it gradually approaches the cooling zone.
In contrast, the solid surface temperature in chamber 1 does not change significantly with
particle size.
4.4. Solid Core Temperature

During the calcination of limestone, the temperature must be greater than 1173 K
before decomposition reactions begin. As the decomposition reaction progresses from the
outside to the inside, the core temperature rises at a slower rate and eventually there may
be a part of the core with a temperature less than 1173 K where no decomposition reaction
takes place, a situation generally referred to as “under burning”. The “under burning”
means that the limestone has not decomposed completely.
As shown in Figure 15a, the core temperatures in both the calcination and preheating
zones at 40 mm particle size reach above 1000 K, while the core temperature in the central
part of the cooling zone reaches above 1000 K. The core temperature near wall drops rapidly.
On the one hand, this is due to the fact that the combustion heat of the fuel gas cannot reach
a long distance, and on the other hand, it is also due to the fact that the cooling air cools the
solid in the cooling zone so quickly that not only the surface temperature of the material
drops, but also the core temperature of part of the material. The core temperature in the
cooling zone at 120 mm particle size is significantly smaller than that at 1000 K (Figure 15c).
The core temperature distribution in chamber 2 for the three particle sizes increases with
increasing particle size, resulting in a more pronounced increase in the area with higher
core temperatures.
(a) (b) (c)
Figure 15. Solid surface temperature distribution at 40 mm, 80 mm and 120 mm particle sizes.
(a) 40 mm solid core temperature distribution (b) 80 mm solid core temperature distribution
(c) 120 mm solid core temperature distribution.
In combination with Figures 15a,b and 16a,b, the variation of the mid line and near
wall line temperatures in chamber 1 are essentially the same for 40 mm and 80 mm grain
sizes, while for 120 mm grain size, the mid line solid core temperature remains the same as
the former, while the near wall line shows a greater drop as it enters the cooling zone (as in
Figure 16c). For the solid core temperature in chamber 2, the changes in the mid line and
near wall line are essentially the same, remaining at around 650 K and 550 K respectively.
114
Materials 2022, 15, 4024
(a) (b) (c)
Figure 16. Solid core temperature variation on the mid line and near wall line for 40 mm, 80 mm and
120 mm particle sizes. (a) Variation in solid core temperature at 40 mm (b) Variation in solid core
temperature at 80 mm (c) Variation in solid core temperature at 120 mm.
As shown in Figure 17a,b, the core temperatures in the mid line of chamber 1 are
essentially the same, with minor differences in the near wall line only in the cooling zone.
In contrast, the core temperatures for the 40 mm, 80 mm and 120 mm particle sizes at the
mid line and near wall lines of chamber 2 show significant differences. During the upward
movement away from the cooling zone of chamber 2, the solid core temperature reaches
700 K at the mid line of the 120 mm grain size and even 750 K at the near wall line at one
point and stabilises around 650 K during the upward movement, while the temperature
gradient decreases at the mid line and near wall line of the 80 mm and 40 mm grain sizes,
with the maximum temperature gradient reaching over 50 K.
(a) (b)
Figure 17. Comparison of mid and near wall line solid core temperature variations. (a) Variation in
solid core temperature at the mid line (b) Variation in solid core temperature at the near wall line.
Combining the temperature variations of chamber 1 in Figure 17a,b, it can be seen

that the core temperature variations of Plane 3 at the interface between the calcination and
cooling zones are mainly manifested in chamber 2, therefore the core temperature analysis
of Plane 3 is also for chamber 2, as shown in Figure 18.
The maximum temperature of Plane 3 at the interface of the calcination and cooling
zones in chamber 2 at 40 mm particle size is 650 K and the minimum temperature is 550 K.
As the particle size increases to 80 mm, the core temperature reaches 740 K and 560 K
respectively, and as the particle size continues to increase to 120 mm, the maximum core
temperature basically remains unchanged at 740 K, but the minimum temperature rises to
around 575 K. There is a significant reduction in the temperature difference.
115
Materials 2022, 15, 4024
Figure 18. Solid core temperature distribution on double chambers and solid core temperature
variation in chamber 2 on Plane 3.
The core temperature distribution and changes in the solids at 40 mm, 80 mm and
120 mm particle sizes shows that the core temperature in chamber 1 varies little with
particle size, especially at 40 mm and 80 mm, and does not change much until 120 mm. The
main reason for this is that as the decomposition reaction continues, the core temperature
rises more slowly in the larger particle sizes, and by the time it reaches the cooling zone, the
core of some of larger solids has not yet risen to a sufficient temperature and is affected by
the cooling air and no longer continues to rise. The core temperature contrast in chamber 2
is even more pronounced, i.e., the temperature gradient is smaller for the smaller particle
sizes and larger for the larger particle sizes, mainly because the smaller particle sizes warm
up faster and the heat is more evenly distributed, while the larger particle sizes show an
external hot and internal cold distribution of core temperature at the same time.
4.5. Limestone Decomposition Field

The calcination of limestone is completely dependent on the remaining calcium carbon-
ate thickness and therefore the limestone decomposition is represented by the distribution
of undecomposed thickness (e.g., Figure 19a–c). As the particle size gradually increases, the
decomposition reaction becomes more difficult to carry out and the remaining limestone
thickness no more decreases, indicating that the limestone reaction gradually increases in
unreacted thickness as the particle size increases.
At 40 mm particle size, the undecomposed thickness of the chamber 1 calcination zone
gradually decreases as the mid line moves down towards the CaO outlet, and the undecom-
posed thickness remains the same in the upper part of the preheating and calcination zones,
remaining at around 2.5 mm after reaching the cooling zone. The undecomposed thickness
near the wall line is above 3.5 mm in the preheating zone until it rapidly decreases and
remains around 3 mm after entering the calcining zone. A similar trend was observed for
the near wall line at 80 mm particle size, where the undecomposed thickness was close to
12 mm, while the undecomposed thickness of the mid line remained around 7 mm and
did not change significantly with height (Figure 20b). At 120 mm grain size, the mid line
undecomposed thickness is the same as at 80 mm, while the near wall line shows a large
difference. The undecomposed thickness of the near wall line was as high as 30 mm in
the preheating zone, and only after entering the calcination zone did the undecomposed
thickness decrease significantly and was similar to the decomposition of the 40 mm and
80 mm grain sizes (Figure 20c).
116
Materials 2022, 15, 4024
(a) (b) (c)
Figure 19. Distribution of undecomposed thickness of CaCO3 at 40 mm, 80 mm and 120 mm particle
sizes. (a) undecomposed thickness of CaCO3 at 40 mm (b) undecomposed thickness of CaCO3 at
80 mm (c) undecomposed thickness of CaCO3 at 120 mm.
(a) (b) (c)
Figure 20. Variation of undecomposed thickness of CaCO3 at 40 mm, 80 mm and 120 mm particle
sizes. (a) Variation of undecomposed thickness of CaCO3 at 40 mm (b) Variation of undecomposed
thickness of CaCO3 at 80 mm (c) Variation of undecomposed thickness of CaCO3 at 120 mm.
At the CaO outlet, the undecomposed thickness distribution is shown in Figure 21.
At a particle size of 40 mm, the undecomposed thickness near the inner side is small, with
an average undecomposed thickness of about 3 mm, and increases to 7 mm as it moves
towards the outer side, while at a particle size of 80 mm, the undecomposed thickness
reaches 7 mm on the inner side and more than 13 mm on the outer side. The decomposition
distribution at all grain sizes indicates that the side wall effect has a significant influence
on the final decomposition, especially at larger grain sizes where the side wall effect has a
greater influence on the decomposition process.
The variation of limestone decomposition rates for the three sizes is shown in Figure 22.
At 40 mm particle size, the mid line limestone decomposition rate is close to 90%, while the
near wall decomposition rate is between 75% and 85%. The mid line decomposition rate
reaches about 83% with 80 mm particle size, while the near wall decomposition rate is the
same as at 40 mm particle size, between 75% and 85%. When the particle size increases to
120 mm, the mid line decomposition rate remains around 20% and the decomposition effect
is poor, while the near wall line can only reach 50% decomposition rate in the position
between the preheating and calcination zones, and in other positions, the decomposition
rate is generally below 30%. The main reason of its close its close to 50% decomposition
rate is that the limestone inlet is closed when calcination takes place in chamber 1, i.e., a
large amount of high temperature exhaust gas accumulates at the top, which promotes
117
Materials 2022, 15, 4024
limestone decomposition to a certain extent, but this promotion is not stable and the final
decomposition effect is far from the decomposition demand.
Figure 21. Variation of 40 mm, 80 mm and 120 mm undecomposed thickness of CaO outlet.
Figure 22. Variation in mid lines and near wall lines decomposition rates at 40 mm, 80 mm and
120 mm particle sizes.
The undecomposed thickness varies significantly from particle size. The larger the
grain size, the more difficult the decomposition and the greater the effect on the final
decomposition rate. The limestone decomposition rate is also strongly influenced by the
side wall effect, with limestone decomposition in the mid line outperforming that in the
near wall line. At the same time, the side wall effect has a greater influence on the small
particle size than on the large particle size. At 120 mm particle size, the effect of not reaching
the decomposition temperature at the core is much greater than the effect of the side wall
effect on decomposition.
5. Conclusions
This paper presents a comparative analysis of the gas velocity field, gas temperature
field, solid surface and core temperature field, and limestone decomposition field for
limestone decomposition processes at 40 mm, 80 mm and 120 mm particle sizes. The results
show that.
(i) Changing the fuel gas from natural gas to BFG while keeping the PFR lime kiln
structure unchanged will reduce the range of supported calcined grain sizes from
118
Materials 2022, 15, 4024
40 mm to 120 mm to below 80 mm, and that limestone of 120 mm grain size is difficult
to calcine and decompose under current conditions.
(ii) Even at 40 mm grain size, the decomposition rate is over 90% after calcination by
natural gas, whereas with BFG calcination, the decomposition rate is 75% to 85% and
the under burning rate is even higher.
(iii) By switching to calcination by BFG, the problem of hard burnt is largely avoided as
the solid surface temperature does not exceed 1200 K.
(iv) Analysis of the exhaust gas outlet temperature shows that the exhaust gas is dis-
charged at a temperature of no more than 700 K and even less than 500 K at 40 mm
particle size, which is a further improvement in the utilisation of energy.
(v) The side wall effect has a greater impact on the calcination of small grain size limestone
and is the main reason for the higher under burning rate.
(vi) There is room for further research and exploration in this study as follows. Firstly,
it is clear that the original kiln structure is no longer suitable for BFG calcination
of limestone and needs to be optimised for low calorific value fuel calcination of
limestone; secondly, the particle size range for BFG calcination of limestone must be
less than the range of 40 mm to 120 mm, which is suitable for natural gas calcination,
but the exact particle size support range still needs to be further investigated in
conjunction with experiments after optimising the PFR lime kiln structure. Finally,
experimental validation of low calorific value gas-fired limestone calcination needs to
be established and this part of the work will be necessary.
Author Contributions: S.D., writing—original draft, mathematical model establishing, program-

ming; B.L., conceptualization, methodology, writing—review and editing; W.R., data curation, field
measurements. All authors have read and agreed to the published version of the manuscript.
Funding: This work was supported by the National Natural Science Foundation of China (Grant No.
51934002).
Acknowledgments: Firstly: I would like to thank Baokuan Li for his excellent contribution to this
paper. He was keenly aware of the importance of PFR lime kilns for energy efficiency and environ-
mental protection, and has done a great deal of preliminary research in this area, as well as providing
highly creative guidance on mathematical model building and development. Secondly, thanks Wenjie
Rong’s preliminary research on rotary lime kilns, which has been very informative for this paper, her
research on sidewall effects and discrete elements has been very thorough and pioneering and has
led to significant progress. Finally, thanks Cai Zhang for her significant contribution to the geometric
modelling, her skilful BIM skills helped to convert the 2D drawings into a 3D model, which is the
best PFR model by far. I have also benefited from her constant companionship in writing, which has
been a constant source of inspiration and a relationship that I will continue to cherish for the rest of
my life.
Conflicts of Interest: The authors declare that they have no conflict of interest.
Nomenclature
ρg Fluid density; Dp Diameter of the particles; α Permeability in porous media; C2 Inertial drag
coefficient; MCaCO3 Relative molecular mass of calcium carbonate; dCaCO3 Diameter of the internal
core calcium carbonate; ρCaCO3 density of the calcium carbonate; RD Limestone decomposition rate;
kD Reaction constants; peq Equilibrium partial pressure of carbon dioxide at the front of the reaction
zone; pco2 CO2 partial pressure in the environment; Tp Average particle temperature; Ti Temperature
of internal core calcium carbonate; YT.C Reaction rate correction factor; λ1 Thermal conductivity of
internal core calcium carbonate, λ2 Thermal conductivity of external core calcium oxide, rc1 Calcium
carbonate radius rc1.m 1/2 calcium carbonate radius rc2.m calcium carbonate radius plus 1/2 calcium
oxide layer radius; ϕCaCO3 Volume fraction of calcium carbonate, ρi density of calcium carbonate, ci
specific heat capacity of calcium carbonate; QD Particle reaction rate; ΔHR Heat of decomposition of
calcium carbonate(183,000 J/mol); Vp Volume of individual particle; ϕCaO Volume fraction of calcium
oxide, ρo density of calcium oxide, co specific heat capacity of calcium oxide; av Specific surface area
of limestone; hv heat transfer coefficient of gas solid two phase flow; σ Stephen Boltzmanns constant;
119
Materials 2022, 15, 4024
5.6697 × 10−8 W/(K4 ·m2 ); β Radiation attenuation coefficient of porous media; Nu Nussel number,
lz Stacked bed feature length, Pr Prandtl number, Re Reynolds number, a Thermal diffusion coeffi-
cient; ν Kinematic viscosity of gas, μ dynamic viscosity of gas, u Velocity of gas movement.
References
1. Hao, S.; Hwang, J.Y.; Jiang, W.; Zhang, Y.Z. Preparation for High Activity Degree Lime and Effect on Desulfurization of Hot
Metal Pretreatment. In 9th International Symposium on High-Temperature Metallurgical Processing; The Minerals, Metals & Materials
Series; Springer: Cham, Switzerland, 2018.
2. Vola, G.; Bresciani, P.; Rodeghero, E.; Sarandrea, L.; Cruciani, G. Impact of Rock Fabric, Thermal Behavior, and Carbonate
Decomposition Kinetics on Quicklime Industrial Production and Slaking Reactivity. J. Therm. Anal. Calorim. 2018, 136, 967–993.
[CrossRef]
3. Wang, G.; Ge, K.; Sun, T.; Shui, Z.; Hu, T.; Jiang, H.; Wang, Z. Comparison of Hydration of Thermally Activated Water-Washed
Kaolins in Cement Mortar. J. Test. Eval. 2019, 48, 1837–1849. [CrossRef]
4. Vieira, L.M.; de Oliveira, H.C.C.; Telles, V.B.; Junca, E.; Vieira, E.A.; de Oliveira, J.R. Influence of Lime Particle and Slag Properties
on Lime Dissolution in BOF Converter. J. Mater. Res. Technol. 2020, 9, 14878–14886. [CrossRef]
5. Georgallis, M.; Nowak, P.; Salcudean, M.; Gartshore, I.S. Modelling the Rotary Lime Kiln. Can. J. Chem. Eng. 2008, 83, 212–223.
[CrossRef]
6. Senegačnik, A.; Oman, J.; Širok, B. Annular Shaft Kiln for Lime Burning with Kiln Gas Recirculation. Appl. Therm. Eng. 2008, 28,
7. Gutiérrez, A.S.; Martínez, J.B.C.; Vandecasteele, C. Energy and Exergy Assessments of A Lime Shaft Kiln. Appl. Therm. Eng. 2013,
51, 273–280. [CrossRef]
8. Piringer, H. Lime Shaft Kilns. Energy Procedia 2017, 120, 75–95. [CrossRef]
9. Krause, B.; Liedmann, B.; Wiese, J.; Bucher, P.; Wirtz, S.; Piringer, H.; Scherer, V. 3D-DEM-CFD Simulation of Heat and Mass
Transfer, Gas Combustion and Calcination in An Intermittent Operating Lime Shaft Kiln. Int. J. Therm. Sci. 2017, 117, 121–135.
[CrossRef]
10. Zhongda, T.; Shujiang, L.; Yanhong, W.; Xiangdong, W. SVM Predictive Control for Calcination Zone Temperature in Lime Rotary
Kiln with Improved PSO Algorithm. Trans. Inst. Meas. Control. 2018, 40, 3134–3146. [CrossRef]
11. Suárez, V.M.G.; García-Gonzalo, E.; Bayón, R.M.; Nieto, P.J.G.; Antón, J.C.Á. Predictive Model of Gas Consumption and Air
Emissions of A Lime Kiln in A Kraft Process Using The ABC/MARS-based Technique. Int. J. Adv. Manuf. Technol. 2018, 100,
1549–1562. [CrossRef]
12. Metz, J.C.; Maciel, E.F.; Garbin, M.; Modolo, R.C.E.; Moraes, C.A.M.; Gomes, L.B.; Brehm, F.A. Influence of Electric Arc Furnace
Dust and Lime Kiln Waste in Portland Cement Hydration. Ambiente Construído 2020, 20, 225–241. [CrossRef]
13. Juneja, P.J.; Sunori, S.; Sharma, A.; Sharma, A.; Joshi, V. Modeling, Control and Instrumentation of Lime Kiln Process: A Review.
In Proceedings of the 2020 International Conference on Advances In Computing, Communication & Materials (ICACCM),
Dehradun, India, 21–22 August 2020.
14. Wirtz, S.; Pieper, C.; Buss, F.; Schiemann, M.; Schaefer, S.; Scherer, V. Impact of Coating Layers in Rotary Cement Kilns: Numerical
Investigation with A Blocked-off Region Approach for Radiation and Momentum. Therm. Sci. Eng. 2020, 15, 100429. [CrossRef]
15. Pieper, C.; Wirtz, S.; Schaefer, S.; Scherer, V. Numerical Investigation of the Impact of Coating Layers on RDF Combustion and
Clinker Properties in Rotary Cement Kilns. FUEL 2021, 283, 118951. [CrossRef]
16. Sagastume, A.; Vandecasteele, C. Exergy-based indicators to evaluate the possibilities to reduce fuel consumption in lime
production. Energy 2011, 36, 2820–2827.
17. Rong, W.; Li, B. Energy and exergy analysis of an annular shaft kiln with opposite burners. Appl. Therm. Eng. 2017, 119, 629–638.
[CrossRef]
18. Cheng, C.; Specht, E. Reaction rate coefficients in decomposition of lumpy limestone of different origin. Thermochim. Acta 2006,
449, 8–15. [CrossRef]
19. Duc, H.D.; Specht, E. Determination of reaction coefficient, thermal conductivity and pore diffusivity in decomposition of
limestone of different origin. In Proceedings of the World Congress on Engineering and Computer Science, San Francisco, CA,
USA, 19–21 October 2011; Volume 11, pp. 978–984.
20. Bluhm-Drenhaus, T.; Simsek, E.; Wirtz, S. A couple fluid dynamic-discrete element simulation of heat and mass transfer in a lime
shaft kiln. Chem. Eng. Sci. 2010, 65, 2821–2834. [CrossRef]
21. Krause, B.; Liedmann, B.; Wiese, J. Coupled three dimensional DEM-CFD simulation of a lime shaft kiln Calcination, particle
movement and gas phase flow field. Chem. Eng. Sci. 2015, 134, 834–849. [CrossRef]
22. Ning, J.-T.; Zhong, B.-J.; Fu, W.-B. Study on the Calcination of Fine Limestone Powder at High Temperature. J. Combust. Sci.
Technol. 2003, 9, 205–208.
23. Wu, H.; Gui, N.; Yang, X.; Tu, J.; Jiang, S. Parameter analysis and wall effect of radiative heat transfer for CFD-DEM simulation in
nuclear packed pebble bed. Exp. Comput. Multiph. Flow 2021, 3, 250–257. [CrossRef]
24. Mikulčić, H.; Vujanović, M.; Fidaros, D.K. The application of CFD modelling to support the reduction of CO2 emissions in cement
industry. Energy 2012, 45, 464–473. [CrossRef]
120
Materials 2022, 15, 4024
25. Long, H.; Xu, M.; Yu, D.; Ding, Y.; Wang, K. Two-temperature model of water gas shift reaction in porous media based on fluent.
Comput. Appl. Chem. 2012, 29, 981–985.
26. LI, B.S.S.Y.J.; Qiu-Wang, S.Y.W. Effects of Point Contact Treatment Methods on Flow and Heat Transfer of Structured Packed Beds.
J. Eng. Thermophys. 2013, 34, 534–537.
27. Liu, G.; Cui, G.; Ma, S.; Han, N. Simulation of the Temperature Field in a Lime Kiln by Using a Local Non-heat-equilibrium
Model and Its Optimized Analysis. J. Eng. Therm. Energy Power 2018, 33, 96–105.
28. Lu, Z. Calculation of Effective Thermal Conductivity of Foam Porous Media. J. Nanjing Univ. Sci. Technol. 2001, 25, 257–261.
29. Zhu, L.; Bai, J.; Qiao, B. Calculation of Convective Heat Transfer for Active Limes Pre-heater. Metall. Equip. 2012, 6–8.
121
materials
Article
Effect of Preheating Temperature on Thermal–Mechanical
Properties of Dry Vibrating MgO-Based Material Lining in
the Tundish
Xiaodong Deng 1 , Jianli Li 1,2 and Xiao Xie 3, *
1 The State Key Laboratory of Refractories and Metallurgy, Wuhan University of Science and Technology,
Wuhan 430081, China
2 Hubei Provincial Key Laboratory for New Processes of Ironmaking and Steelmaking,
Wuhan University of Science and Technology, Wuhan 430081, China
3 School of Computer Science and Technology, Wuhan University of Science and Technology,
Wuhan 430081, China
Abstract: For the collapse of the working layer of dry vibrating material during preheating, the
four-strand tundish of a steel plant was taken as a prototype for numerical simulation. The software
ANSYS was used to calculate the temperature field and stress and strain field on the working layer
under three preheating stages through the indirect coupling method. The results show that during
the preheating process, the temperature field distribution on the hot surface of the working layer
gradually develops toward uniformity with the increase in preheating temperature. However, the
temperature gradient between the cold and hot surfaces increases subsequently, and the highest
temperature between the cold and hot surfaces reaches 145.31 ◦ C in the big fire stage. The stress on
the top of the working layer is much larger than in other areas, and the maximum tensile stress on the
top reaches 39.06 MPa in the third stage of preheating. Therefore, the damage to the working layer
Citation: Deng, X.; Li, J.; Xie, X. Effect starts from the top of the tundish. In addition, the strain of the area near the sidewall burner nozzle
of Preheating Temperature on in the casting area is much larger than that in the middle burner area with the increase in preheating
Thermal–Mechanical Properties of temperature. Thus, the working layer near the sidewall burner nozzle is more prone to damage and
Dry Vibrating MgO-Based Material collapse compared with the middle burner nozzle.
Lining in the Tundish. Materials 2022,
15, 7699. https://doi.org/10.3390/ Keywords: dry vibrating material; collapsed ladle; mathematical modeling; temperature field; stress
ma15217699 field; strain field
Academic Editor: Scott M.
Thompson
Received: 26 September 2022 1. Introduction

Accepted: 31 October 2022
Continuous casting tundish, working as a buffer and distributor of liquid steel be-
tween the ladle and continuous casting molds, plays a key role in affecting the performance
Publisher’s Note: MDPI stays neutral of casting and solidification, as well as the quality of final products [1–3]. Therefore, the
with regard to jurisdictional claims in quality of the refractory material of the tundish is important for the full development and
published maps and institutional affil- improvement of the metallurgical performance of the tundish. The development of the
iations. tundish working layer can be divided into four stages: (i) no working layer stage, (ii) insu-
lation board stage, (iii) gunning material stage, and (iv) dry vibrating material stage [4].
Generally, tundish refractory lining mainly includes an insulation layer, a permanent layer,
and a working layer. The working layer is in contact with molten steel. Thus, the working
lining material should have good slag resistance, thermal shock resistance, volume stability,
excellent thermal insulation performance, and high-temperature strength [5–7]. The dry
vibrating material as the fourth generation of tundish working layer material, not only has
conditions of the Creative Commons the advantages of both insulation board and gunning material but also has the advantages
Attribution (CC BY) license (https:// of easy construction, high thermal efficiency, fast tundish turnover, long service life, and
creativecommons.org/licenses/by/ low energy consumption, which is widely used in the metallurgical industry of continuous
4.0/). casting tundish [8,9].

Materials 2022, 15, 7699
With the development of continuous casting technology, it is gradually realized that

the thermal state of the tundish has an important role in controlling the temperature of
the molten steel in the tundish, maintaining the stability of the superheat of the molten
steel, improving the quality of the continuous casting billet, saving energy, and prolonging
the service life of the tundish [10–12]. During the continuous casting process, the molten
steel will be poured from the ladle into the tundish. There is conductive heat loss through
the wall of the tundish and radiative heat loss through the bath surface [13,14]. In order
to reduce the heat loss of the molten steel, the lining can be preheated to absorb a large
amount of heat before pouring to reduce the temperature difference between the lining and
the molten steel.
However, during the use of the dry vibrating material working layer, it was found that
the dry vibrating material with phenolic resin as the low-temperature binding agent will
undergo an oxidation reaction to form a decarbonized layer during the preheating process,
and the original physical properties of the working layer of the tundish will be changed by
the formation of the decarbonized layer [15]. When the preheating regime of the tundish
is imperfect, the working layer is prone to large area collapse and local spalling, which
leads to the production of instable running. In order to solve this problem, Shi et al. [16]
analyzed the current status of refractory temperature and thermal stress research, and
established and analyzed the temperature and thermal stress distribution of two types of
refractories for ladles using APDL language. Li et al. [17] carried out a large number of
follow-up analyses on the collapsed tundish of magnesium dry vibrating material tundish
and optimized and improved the preheating process by testing and analyzing the strength
of the dry vibrating material preheating process and the change law of the preheating
temperature rise.
In the present study, a 1:1 model with a four-strand tundish of a steel mill as the
prototype was established. Finite element simulation of the temperature field and the stress
field during preheating in a tundish body was carried out by ANSYS software. According
to the distribution and changes in temperature, stress, and strain fields on the working
layer during the preheating process, the most vulnerable areas of the working layer were
analyzed. This research aims to identify the stress distribution at the critical point and
analyze the cause of the damage. The obtained results are of great significance to the
normal operation of the continuous casting process.
2. Model Description
2.1. Fluid Flow and Heat Transfer
In the preheating process of tundish, the mixture of gas and air enters the tundish after
ignition. Then, the combusted gas transfers heat to the tundish lining through convection
and radiation in the flow process. During the preheating process fluid, flow, and heat
transfer are modeled using conservation equations of mass Equation (1), momentum
Equation (2), and energy Equation (3):

∂ρ ∂ ρu j
+ =0 (1)
∂t ∂x j

∂ui ∂ui ∂P ∂ ∂ui ∂u j
ρ + ρu j =− + (μ + μt ) + + gi ( ρ − ρ 0 ) (2)
∂t ∂xi ∂xi ∂x j ∂x j ∂xi
! "
∂T ∂ uj T ∂ C p μt ∂T
ρC p + ρC p = k0 + + ST (3)
∂t ∂x j ∂x j Prt ∂x j
where ρ is the density; CP is the heat capacity; μt is the turbulent viscosity; Prt is the
turbulent Prandtl number; ST is the source term of the energy equation; u is the velocity; x
is the Cartesian space coordinates, and subscripts i, j are for the coordinate directions.
124
Materials 2022, 15, 7699
The standard k-ε models [18] are shown in Equations (4) and (5):
! "
∂ ∂ ∂ μt ∂k
(ρk) + (ρkui ) = μ+ + Gk + Gb − ρε − γ M + SK (4)
∂t ∂xi ∂x j σk ∂x j
! "
∂ ∂ ∂ μt ∂k ε ε2
(ρε) + (ρεui ) = μ+ + C1 ( Gk + C3 Gb ) − C2 ρ + Sε (5)
∂t ∂xi ∂x j σε ∂x j k k
where k is the kinetic energy of turbulence per unit mass; εF is the turbulent energy
dissipation rate; μ is the molecular viscosity; μt is the turbulent viscosity; Gk is the turbulent
kinetic energy generated by the laminar velocity gradient; Gb is the turbulent kinetic
energy generated by buoyancy [19]; γM is the fluctuation generated by the transition
diffusion in compressible turbulence; σk and σε are the turbulent Prandtl numbers of k and
ε, respectively; C1 , C2 , and C3 are the constants; and SK and Sε are the source term of the
turbulent kinetic energy(k) and its dissipation rate (εF ).
As mentioned by Launder and Spalding [18], the other values for model constants in
this study were C1 = 1.44, C2 = 1.92, σk = 1.0, and σε = 1.3. For buoyant shear layers for
which the main flow direction is aligned with the direction of gravity, C3 becomes 1. For
buoyant shear layers that are perpendicular to the gravitational vector, C3 becomes 0.
2.2. Solid Heat Transfer and Thermal Stress

During the preheating process, the heat transfer of the tundish package can be con-
sidered as a three-dimensional steady-state heat transfer, and the heat-transfer behavior is
expressed by the following Equation (6):
∂2 T ∂2 T ∂2 T
+ + =0 (6)
∂x ∂y ∂z
where T is the temperature; x, y, and z are the x, y, and z coordinate directions.

According to the temperature distribution and thermal expansive coefficient of each
part of the tundish lining, the deformation is calculated in the special constraints. Then,
the strain of each point of the tundish is calculated using the deformation of tundish
displacement with the geometric equation. Finally, the stress in the tundish of each point is
calculated through the strain according to the physical equation of tundish material [20,21].
The thermal stress field geometry equation is used to calculate the relationship between
strain and displacement. It can be expressed by the following Equation (7).
⎡ ∂
⎤
∂x 0 0
⎢0 ∂
0⎥
⎢ ∂y ⎥
⎢ ∂ ⎥
⎢0 0 ⎥
ε=⎢
⎢∂
∂z ⎥ λ (7)
⎢ ∂y
∂
∂x 0⎥ ⎥
⎢ ∂ ∂ ⎥
⎣0 ∂z ∂y ⎦
∂ ∂
∂z 0 ∂x
* +T * +T
where ε = ε x ε y ε z γxy γxz γyz is the strain at any point in tundish; λ = u v w
is the displacement along the directions of x, y, and z.
125
Materials 2022, 15, 7699
The physics equation of the stress field is used to calculate the relationship between
strain and stress. It can be expressed by the following Equation (8).
⎡ v v
⎤
1 1− v 1− v 0 0 0
⎢ v 1 v
0 0 0 ⎥
⎢ 1− v 1− v ⎥
⎢ v v ⎥
E (1 − v ) ⎢ 1− v 1− v 1 0 0 0 ⎥
σ= ⎢ 1−2v ⎥ (8)
(1 + v)(1 + 2v) ⎢
⎢
0 0 0 2(1− v )
0 0 ⎥
⎥
⎢ 0 0 0 0 1−2v
0 ⎥
⎣ 2(1− v ) ⎦
1−2v
0 0 0 0 0 2(1− v )
where E is the elastic modulus; v is Poisson’s radio; σ is the stress; and ε is the strain.
According to the above stress–strain relations, any stress of each point is calculated
by each point of its inner strain obtained from the previous step, and the products are the
object of the force, with the force meeting the balance equation.
2.3. Materials and Methods

In the present study, the characterization of the working and permanent layers re-
quired for the simulation calculations was made in the laboratory. The raw materials of
the working and permanent layers were taken from the industrial raw materials of a steel
mill. Test samples of the dry vibrating material were put into a mold with an inner size
of 40 mm × 40 mm × 160 mm, followed by tamping until the density of raw materials was
about 2.40 g/cm3 . These prepared samples were initially dried at 220 ◦ C for 3 h to obtain
handling strength, and then were fired in an electric furnace at 1000 ◦ C for 2 h under air
conditions (oxidizing atmosphere) and buried carbon conditions (reducing atmosphere), re-
spectively, where the materials were heated at a rate of 5 ◦ C/min, followed by cooling to room
temperature in the furnace. Test samples of the permanent layer materials were put into a
mold with the inner size of 40 mm × 40 mm × 160 mm, followed by vibrating. After heating
for 3 h at 220 ◦ C condition, all the specimens were fired at 1000 ◦ C for 2 h in an electric furnace
under an air atmosphere condition with the heated rate of 5 ◦ C/min and then cooled down to
room temperature. The bulk density of three test samples was evaluated using Archimedes’
principle in kerosene medium according to GB/T 2997-2015 standard.
Thermal conductivity of the disc specimens with Φ180 mm × 20 mm at 350 ◦ C, 600 ◦ C,
800 ◦ C, and 1000 ◦ C was evaluated using the hot-wire method per the standard GB/T 5990-
2006. The characterization of the dynamic elastic modulus at room temperature was evalu-
ated according to GB/T 30758-2014. The cylindrical specimens of Φ8 mm × 50 mm were
prepared, and the thermal expansion rate of the specimen was measured at 100–1400 ◦ C us-
ing a high-temperature thermal (Precondar, Luoyang, China) expansion meter per standard
GB/T 7320-2008.
2.4. Geometry, Mesh, and Boundary Conditions

In this study, a four-strand tundish of a steel mill is modeled as a prototype 1:1. Con-
sidering the symmetry of the tundish structure, half of the model is taken as the calculation
area to reduce the calculation volume. The carbon element in the dry vibrating material
of the working layer will react with the oxidizing substances in the preheating gas during
the preheating process of the tundish, resulting in the delamination of the working layer
and the generation of two layers of decarbonized layer and the original layer. Therefore,
the working tundish lining with two layers was modeled, and the decarburization layer
thickness was 25% of the whole working layer. The specific geometric model is shown
in Figure 1a. The tundish body is composed of the original layer, decarbonized layer,
permanent layer, insulation layer, steel shell, slag retaining wall, flow control device, and
lid. The burner nozzle in the center of the pouring area on the lid is burner No. 1 and the
burner nozzle in the pouring area near the side wall is burner No. 2. In addition, the burner
nozzle in the impact area is burner No. 3. The physical properties of refractory linings used
126
Materials 2022, 15, 7699
in this simulation are shown in Tables 1 and 2. The characterization of the insulation layer
and the steel shell in Table 1 can be found by Zhang [22].
Figure 1. Geometry and mesh: (a) geometry; (b) grid of computational domain.
Table 1. Physical properties of refractory linings.
P Cp k α E
(Kg·m−3 ) (J·kg−1 k−1 ) (w·m−1 k−1 ) (◦ C−1 ) (MPa)
Original layer 3330 988.6 - 1.36 × 10−5 7500
Decarbonized layer 2349 988.6 - 1.36 × 10−5 880
Permanent layer 2600 800 - 6 × 10−7 5700
Insulation layer 850 816.4 0.17 5 × 10−6 1.2 × 105
Steel shell layer 7820 502 46.4 1.17 × 10−5 2 × 105
Table 2. Thermal conductivity of materials at different temperatures.
Temperatures Original Layer Decarbonized Layer Permanent Layer

(◦ C) (W·m−1 k−1 ) (W·m−1 k−1 ) (W·m−1 k−1 )
350 0.377 0.455 0.694
600 0.639 0.686 0.898
800 0.953 1.001 1.077
1000 1.015 1.13 1.27
The model is meshed using ANSYS MESH software. Based on the hybrid mesh
division method, the areas of different complexity are divided into meshes of different
sizes and densities to reduce the number of meshes as much as possible while ensuring
calculation accuracy, so as to save the calculation resources and calculation time. The total
number of meshes in the tundish model is about 8 million, in which the meshes of the
working layer and burner of the tundish are encrypted. The mesh division results are
shown in Figure 1b.
Non-slip conditions were applied on all wall boundaries for the fluid phase. A constant
velocity flow was used at the inlet. At the outlet of the tundish, the outflow was applied.
The heat loss is calculated based on the heat transfer coefficient at the side and bottom
walls and the environment temperature. A summary of input parameters and boundary
conditions used for computational fluid dynamics simulations is provided in Table 3. The
heating process of the tundish is composed of three stages of preheating, namely, the first
stage of preheating, the second stage, and the third stage of preheating. The preheating
127
Materials 2022, 15, 7699
parameters for each stage are shown in Table 4. The data in Table 4 are from the steel
mill site.
Table 3. Input parameters and boundary conditions used for CFD simulations.
Parameter Value
Density 1.7878 kg·m−3
Viscosity 1.37 × 10−5 kg·m−1 s−1
Heat capacity 840.37 J·kg−1 K−1
Thermal conductivity 0.0145 W·m−1 k−1
Wall no-slip
Side wall (heat loss coefficient) 15 W·m−2
environment temperature 25 ◦ C
Table 4. The conditions of the burner nozzle.
Flow Rate (m3 ·h−1 ) Velocity (m·s−1 ) Temperature (◦ C)

The first stage 355 1.3951 400
The second stage 651 2.5583 800
The third stage 1071 4.2088 1150
3. Results
3.1. Temperature Field Distribution
Figure 2 shows the temperature field distribution of the hot surface of the wall in
lengthwise direction of the working layer. The result shows that with the increase in
preheating temperature, the temperature distribution on the hot surface of the working
layer gradually tends to be uniform. As shown in Figure 2a,b, the temperature field
distribution on the hot surface of the working layer is less uniform at the first and second
stages of preheating. The red area is mainly concentrated in the area of the No. 1 burner
nozzle and the bottom area of the No. 2 burner nozzle. The distribution area is small, and
the temperature difference between zones is 16 ◦ C and 20 ◦ C. Figure 2c displays that the
distribution of the temperature field of the hot surface tends to be uniform during the third
stage of preheating, and the temperature difference between the red zones is only 11.6 ◦ C.
Figure 3 shows the temperature field distribution of the cold surface. The result shows
that the trend of temperature field change on the cold side of the working layer is slightly
different from that on the hot side. The temperature field on the cold side does not develop
toward uniformity with the increase in preheating temperature. The high-temperature area
is mainly concentrated in the middle and lower regions of the wall, and in the region near
the top of the tundish, the working layer temperature is lower. There is a larger temperature
gradient compared with the middle and lower regions. In addition, the temperature on
the cold side has a large temperature difference compared with the temperature on the hot
side. The temperature gradient between the cold and hot sides further expands with the
increase in preheating temperature. In the third stage of preheating, the difference between
the highest temperature on the cold side in Figure 3c and the highest temperature on the
hot side in Figure 2c is 145.31 ◦ C, and the temperature gradient further increases in the
orange temperature zone of both figures. The results of this calculation indicate that there
is a large temperature gradient in the longitudinal direction in the working layer of the dry
vibrating material during the preheating process, which increases the possibility of thermal
shock damage to the material and is harmful to the pouring of the tundish [20].
128
Materials 2022, 15, 7699
ȱ
Figure 2. Temperature field distribution of the hot surface of the working layer. (a) The first stage of
preheating; (b) the second stage of preheating; and (c) the third stage of preheating.
129
Materials 2022, 15, 7699
ȱ
Figure 3. Temperature field distribution of the cold surface of the working layer. (a) The first stage of
preheating; (b) the second stage of preheating; and (c) the third stage of preheating.
3.2. Stress and Strain Field Distribution

Figure 4 shows the distribution of the maximum principal stress applied to the hot
surface of the wall in lengthwise direction at different preheating stages, which is oriented
perpendicular to the wall of the working layer. When the maximum principal stress is
positive, the stress is tensile stress and the direction is toward the fluid domain; when the
maximum principal stress is negative, the stress is compressive stress and the direction is
toward the permanent layer. As shown in Figure 4, the boundary region of the working
layer belongs to the stress concentration region and the stress increases significantly with the
increase in the preheating temperature. However, the maximum principal stress suffered
130
Materials 2022, 15, 7699
by the working layer in the middle region has less change in stress value with the increase
in the preheating temperature. The tensile stress at the top of the wall in the boundary
region is significantly greater than that at the sides and bottom, which indicates that the
possibility of cracks from the top of the working layer during preheating is much greater
than that in other regions.
Figure 4. Distribution of the maximum principal equivalent stress field of the working layer. (a) The
first stage of preheating; (b) the second stage of preheating; and (c) the third stage of preheating.
Figure 5 shows the maximum principal strain field on the cold surface of the working
layer. The direction of the strain field is consistent with the direction of the maximum
principal stress. As shown in Figure 5, the area where positive deformation occurred in
the working layer was mainly concentrated in the middle of the wall, while the area where
negative deformation occurred was distributed in the boundary area of the working layer.
This indicates that during the preheating process, the working layer material deformed
toward the fluid domain in the middle area, while the boundary area deformed toward
the permanent layer, which exacerbated the stress concentration in the boundary area to a
certain extent. The maximum positive deformation area on the cold surface of the working
layer gradually shifts from the bottom to the side with the increase in preheating tempera-
131
Materials 2022, 15, 7699
ture. In the second and third stages of preheating, the maximum positive deformation area
was concentrated in the area near the side wall. Moreover, the larger the value of positive
deformation, the greater the possibility of detachment between the working layer and the
permanent layer. When the deformation volume reaches a certain value, the working layer
will be separated from the permanent layer, which is harmful to the pouring of the tundish.
Figure 5. Distribution of the maximum principal elastic strain field of the working layer. (a) The first
stage of preheating; (b) the second stage of preheating; and (c) the third stage of preheating.
4. Analysis and Discussion

In order to judge more intuitively the stress changes suffered by the working layer,
combined with the simulation results of the temperature field and stress field of the dry
vibrating material of the tundish in the third stage of preheating, a straight line called
path 1 and path 2 was made in the area corresponding to the center of the burner nozzles 1
and 2 of the working layer, respectively, to calculate the temperature field and stress field
distribution on the paths. The two paths are located on the hot surface of the working
layer, as shown in Figure 6, while the distribution curves of temperature and maximum
principal stress on the paths are shown in Figures 7 and 8. The horizontal coordinate 0 mm
is the bottom of the middle pack, and L is the length on the path, and the larger its value,
132
Materials 2022, 15, 7699
the closer the path is to the top of the pack. The red curve is the variation curve of the
maximum principal stress with the path, and the black curve is the variation curve of
temperature with the path.
Figure 6. Path 1 and path 2.
Figure 7. Temperature and maximum principal stress distribution on path 1.
Figure 8. Temperature and maximum principal stress distribution on path 2.
As shown in Figure 7, the magnitude of the maximum principal stress in the region of
80–980 mm on path 1 varies extremely, and its stress value is close to 0 MPa, which cannot
cause damage to the working layer. While in the bottom region of path 1, the maximum
tensile stress is 1.336 MPa and the maximum compressive stress is 7.2 MPa. However,
in the top region, the maximum tensile stress on path 1 is 3.03 MPa and the maximum
compressive stress reaches 57.98 MPa. Figure 9 shows the strength characteristics of the dry
vibrating material. As can be seen from the figure, the resin in the dry vibrating material
is gradually cured with the increase in preheating temperature so that the working layer
has a certain strength at low temperatures. However, the curing resin starts to oxidize
133
Materials 2022, 15, 7699
and decompose at 200 ◦ C, resulting in a loss in the strength of the working layer. After
1000 ◦ C, the strength of the working layer gradually increases with the sintering of the dry
vibrating material [23]. According to the temperature curve in Figure 7, it is known that
the temperature at the top of the working layer is lower than in other areas. Thus, in the
case where the dry vibrating material has lower strength and greater stress, the top of the
working layer is more prone to be damaged, cracks are formed, and spalling occurs [17].
Figure 9. Strength characteristics of the dry vibrating material of the working layer [24].
Figure 8 shows that the maximum tensile stress is 1.14 MPa and the maximum com-
pressive stress is 9.50 MPa in the bottom area of path 2, while the maximum tensile stress
is 5.34 MPa and the maximum compressive stress is 64.6 MPa in the top area of path 2.
Compared with path 1, the stress in the top area of path 2 has increased significantly, while
the temperature has decreased slightly. Combined with the strength characteristics of the
dry vibrating material in Figure 9, it is known that the working layer is more prone to
be damaged and cracked at path 2. On the other hand, according to Figure 5c, there is a
possibility of separation of the working layer from the permanent layer. When cracks are
generated at the top of path 2 and the cracks extend to the bottom of the working layer, the
steel will penetrate the working layer along the cracks during the pouring process, which
will lead to large spalling and tundish collapse.
5. Conclusions
(1) The temperature field distribution on the hot surface of the working layer gradually
develops toward uniformity with the increase in preheating temperature, while the
temperature gradient between the cold and hot surfaces increases. The highest tem-
perature difference between the cold and hot surfaces reaches 201.31 ◦ C during the
third stage of preheating, which increases the possibility of thermal shock damage to
the refractory of the working layer.
(2) During the preheating process, the stress on the working layer is mainly concentrated
in the boundary area. The maximum tensile stress at the top of the working layer
reaches 39.06 MPa during the third stage of preheating, which is significantly greater
than that in other areas. It can be seen that the possibility of damage from the top of
the working layer to forming cracks is much greater than in other areas.
(3) In the high-temperature preheating stage, the maximum principal strain in the area
of the No. 2 burner nozzle is 0.0076198 mm, which is much larger than in other
areas. When cracks are generated at the top of the working layer and the cracks
extend downward, the working layer will collapse and spall in a large area due to the
penetration of steel.
Author Contributions: Methodology, X.D.; Supervision, J.L. and X.X.; Writing—original draft, X.D.;
Writing—review & editing, X.X. and J.L. All authors have read and agreed to the published version
of the manuscript.
134
Materials 2022, 15, 7699
Funding: This work was financially supported by the National Natural Science Foundation of China
(51974210, 52274341), Hubei Provincial Natural Science Foundation (2019CFB697), and State Key
Laboratory of Refractories and Metallurgy, Wuhan University of Science and Technology.
Conflicts of Interest: The authors declare that they have no conflict of interest.
Nomenclature
Cp Heat capacity
E Elastic modulus
Gb Generation term of turbulent kinetic energy
Gk Generation term of turbulent kinetic energy
k Turbulent kinetic energy
L Length
Prt Turbulent Prandtl number
ST Source term of energy equation
Sε Source term of turbulent energy dissipation rate
SK Source term of turbulent kinetic energy
T Temperature
t Time
u Velocity
v Poisson’s radio
x Cartesian coordinates
γM Dilatation dissipation term
ε Strain
εF Turbulent energy dissipation rate
μ Molecular viscosity
μt Turbulent viscosity
ρ Density
σ Stress
σk Turbulent Prandtl number for k
σε Turbulent Prandtl number foe ε
References
1. Sahai, Y. Tundish Technology for Casting Clean Steel: A Review. Metall. Mater. Trans. B 2016, 47, 2095–2106. [CrossRef]
2. Mazumdar, D. Review, Analysis, and Modeling of Continuous Casting Tundish Systems. Steel Res. Int. 2019, 90, 1800279.
[CrossRef]
3. Sheng, D.Y. Design Optimization of a Single-Strand Tundish Based on CFD-Taguchi-Grey Relational Analysis Combined Method.
4. Xu, Y. Research progress of dry vibrating mixes for continuous casting tundish. Refractory 2018, 52, 389–395.
5. Zhang, J.H.; Chen, Y.X.; Yang, D.Z. Process Design and Life Analysis of Tundish Lining. Metalurgija 2021, 60, 191–194.
6. Vdovin, K.N.; Tochilkin, V.V.; Yachikov, I.M. Designing Refractories for the Tundish of a Continuous Caster. Refract. Ind. Ceram.
2016, 56, 569–573. [CrossRef]
7. Biswas, S.; Sarkar, D. Modern refractory practice for clean steel. In Introduction to Refractories for Iron- and Steelmaking; Springer:
Cham, Switzerland, 2020; pp. 409–425.
8. Musmeci, M.; Rendtorff, N.M.; Musante, L.; Martorello, L.; Galliano, P.; Aglietti, E.F. Characterization of MgO-Based Tundish
Working Lining Materials, Microstructure and Properties. Ceram. Int. 2014, 40, 14091–14098. [CrossRef]
9. Wei, J.C.; Tu, J.B.; Wang, Y.L.; Zhang, L.J. Environment-Friendly Dry Vibration Refractories Used for Continuous Casting Tundish.
Adv. Mater. Res. 2013, 750–752, 1335–1338.
10. Botnikov, S.A.; Khlybov, O.S.; Kostychev, A.N. Development of the Metal Temperature Prediction Model for Steel-Pouring and
Tundish Ladles Used at the Casting and Rolling Complex. Metallurgist 2019, 63, 792–803. [CrossRef]
11. Sheng, D.Y.; Windisch, C. A Simulation-Based Digital Design Methodology for Studying Conjugate Heat Transfer in Tundish.
12. Yang, B.; Lei, H.; Bi, Q.; Jiang, J.; Zhang, H.; Zhao, Y.; Zhou, J.A. Fluid Flow and Heat Transfer in a Tundish with Channel Type
Induction Heating. Steel Res. Int. 2018, 89, 1800173. [CrossRef]
13. Gastón, A.; Sarmiento, G.S.; Begnis, J.S.S. Thermal Analysis of a Continuous Casting Tundish by an Integrated Fem Code. Lat. Am.
Appl. Res. 2008, 38, 259–266.
14. Sheng, D.Y.; Jönsson, P.G. Effect of Thermal Buoyancy on Fluid Flow and Residence-Time Distribution in a Single-Strand Tundish.
135
Materials 2022, 15, 7699
15. Xu, L.; Gao, S.; Chen, M.; Wu, Y.; Shinozaki, K. Improvement in Microstructure and Thermo-Mechanical Properties of MgO-Based
Dry Vibratable Material by Addition of Fe. Mater. Chem. Phys. 2020, 253, 123368. [CrossRef]
16. Shi, S.; Li, G.; Jiang, G.; Xie, L.; Liu, J. Temperature and Thermal Stress Analysis of Refractory Products. Sens. Transducers J. 2013,
21, 53.
17. Li, S.B.; Chen, Y.Q.; Huang, J.J.; Geng, H.L.; Li, Z.C.; Wang, D.W. Improvement of tundish baking process for magnesia dry
material. Steelmaking 2021, 37, 46–49.
18. Jones, W.P.; Launder, B.E. The Prediction of Laminarization with a Two-Equation Model of Turbulence. Int. J. Heat Mass Transf.
1972, 15, 301–314. [CrossRef]
19. Chattopadhyay, K.; Isac, M.; Guthrie, R.I.L. Modelling of Non-Isothermal Melt Flows in a Four Strand Delta Shaped Billet Caster
Tundish Validated by Water Model Experiments. ISIJ Int. 2012, 52, 2026–2035. [CrossRef]
20. Vasil’ev, D.V.; Grigor’ev, V.P. On the Problem of Thermomechanical Stresses in the Lining and Shell of Steel-Pouring Ladles.
Refract. Ind. Ceramc. 2012, 53, 118–122. [CrossRef]
21. Li, G.; Liu, J.; Jiang, G.; Liu, H. Numerical Simulation of Temperature Field and Thermal Stress Field in the New Type of Ladle
with the Nanometer Adiabatic Material. Adv. Mech. Eng. 2015, 7, 168781401557598. [CrossRef]
22. Zhang, L.L. Study of Flow and Heat Transfer Phenomena and its Influence on Deformation of Blank and Mold for Beam Blank
Continuous. Ph.D. Thesis, Chongqing University, Chongqing, China, 2017.
23. Behera, S.; Sarkar, R. Sintering of Magnesia: Effect of Additives. Bull. Mater. Sci. 2015, 38, 1499–1505. [CrossRef]
24. Yan, P.Z.; Weng, X.Y.; Wu, K.D.; Tang, L.; Ye, X.X.; Jiang, X.X. Solution and control of collapse of dry vibration mix in tundish
during heating. World Iron Steel 2011, 11, 60–62+67.
136
materials
Article
New Insights into the Mechanism of Nucleation of ZrO2
Inclusions at High Temperature
Yutang Li 1,2 , Linzhu Wang 1,2, *, Chaoyi Chen 1 , Shufeng Yang 3 and Xiang Li 4
1 School of Materials and Metallurgy, Guizhou University, Guiyang 550025, China

2 State Key Laboratory of Public Big Data, Guizhou University, Guiyang 550025, China
3 School of Metallurgical and Ecological Engineering, University of Science and Technology Beijing,
Beijing 100083, China
4 Department of Materials and Metallurgy Engineering, Guizhou Institute of Technology,
Guiyang 550003, China
Abstract: It is difficult to observe the nucleation mechanism of inclusions in real-time. In this study,
the nucleation process of zirconium oxide inclusions was systematically studied by classical nucle-
ation theory and first principles. Zr deoxidized steel with 100 ppm Zr addition was processed into
metallographic samples for scanning electron microscopy energy-dispersive spectroscopy observa-
tion. The electrolytic sample was analyzed by micro X-ray diffraction and transmission electron
microscopy, and the zirconium oxide in the sample was determined to be ZrO2 . The nucleation
rate and radius of the ZrO2 inclusions were calculated by classical nucleation theory, and they were
compared with the experimental values. There was a considerable difference between the experimen-
tal and theoretical values of the nucleation rate. The effect of the nucleation size was analyzed by
first-principles calculation, and the thermodynamic properties of ZrO2 clusters and nanoparticles
were analyzed by constructing (ZrO2 )n (n = 1–6) clusters. The thermodynamic properties of ZrO2
calculated by first principles were consistent with the values in the literature. Based on two-step
nucleation theory, the nucleation pathway of ZrO2 is as follows: Zratom + Oatom → (ZrO2 )n → (ZrO2 )2
Citation: Li, Y.; Wang, L.; Chen, C.; → core (ZrO2 particle)–shell ((ZrO2 )2 cluster) nanoparticle → (ZrO2 )bulk .
Yang, S.; Li, X. New Insights into the
Mechanism of Nucleation of ZrO2 Keywords: inclusion; ZrO2 ; nucleation mechanism; first principles
Inclusions at High Temperature.
Materials 2022, 15, 7960. https://
doi.org/10.3390/ma15227960
Academic Editor: Daniela Kovacheva

1. Introduction
The mechanical properties and service life of steel are seriously affected by inclusions.
Received: 17 October 2022
After the proposal of the concept of oxide metallurgy [1], researchers have realized that
Accepted: 3 November 2022
it is important to control the size of inclusions [2–4] rather than increase the smelting
cost [5–8]. Formation of inclusions begins with nucleation, and exploring the nucleation
Publisher’s Note: MDPI stays neutral mechanism and properties of inclusions in steel is important to control the inclusion size.
with regard to jurisdictional claims in However, owing to the high speed of inclusion nucleation and high smelting temperature,
published maps and institutional affil- the inclusion-nucleation process is difficult to detect and observe directly. Therefore,
iations. research on the inclusion of nucleation in steel is a challenge.
Researchers have found that Zr-containing inclusions can promote acicular ferrite
transformation in Zr-deoxidized steel under certain conditions [9–13]. Zr-containing inclu-
sions in Zr-deoxidized steel also play an important role in oxide metallurgy. On the one
hand, they can induce nucleation of intragranular ferrite. There are several theories about
the nucleation mechanism of ferrite. The change in the chemical composition of austenite
around inclusions promotes nucleation. Inclusions and precipitates are coherent with ferrite
conditions of the Creative Commons to reduce the potential nucleation barrier and promote nucleation. In addition, the strain
Attribution (CC BY) license (https:// energy caused by the difference in thermal shrinkage between inclusions and austenite,
creativecommons.org/licenses/by/ as well as inclusions acting as an inert interface, promotes nucleation [14]. Among the
4.0/). proposed theories, the Mn-depleted-zone mechanism formed by the precipitation of MnS

Materials 2022, 15, 7960
on ZrO2 is considered to be one of the most effective mechanisms for intragranular ferrite
nucleation [15,16]. On the other hand, Zr-containing inclusions play an important role in
controlling the distribution of MnS inclusions in the steel. The density of ZrO2 is close to
that of liquid steel, and the volume is small. Therefore, it is not easy for ZrO2 to float up
in liquid steel, and it is easier for ZrO2 to disperse and distribute in steel to improve the
distribution of sulfide through heterogeneous nucleation. Theoretical calculation shows
that the lattice mismatch degree between MnS and ZrO2 is only 5.2%, and ZrO2 is the
most effective nucleation core to promote MnS nucleation [17]. The thermodynamics from
first-principles calculation show that Mn will diffuse into ZrO2 because ZrO2 has cationic
vacancies and can absorb Mn [18]. Whether it induces intragranular ferrite nucleation
or acts as a heterogeneous nucleation core, the key is the size control of ZrO2 . However,
there has been limited research on the nucleation mechanism and size control of ZrO2 . In
addition, the inclusion-nucleation speed is fast and the smelting temperature is high, so the
inclusion-nucleation process is difficult to detect and observe directly.
To study the evolution of alumina inclusions at the atom scale, Wang et al. [19–21]
investigated the cluster structure from experimental and theoretical aspects, and they
proposed a two-step nucleation mechanism. Using quenching and three-dimensional
atomic-probe detection technology, Zhao et al. [22] captured the intermediate structure
of titanium oxide, and they proposed the cluster-assisted nucleation mechanism. Yang
et al. [23] simulated the growth process of clusters by molecular dynamics, and they found
that clusters grow through collision. However, there has been limited research on the
nucleation of zirconium oxide. Thus, it is necessary to study the nucleation process of
zirconium oxide.
In this study, a high-temperature Zr deoxidation experiment was carried out, and the
characteristics of the zirconium oxide inclusions in liquid steel, such as the composition,
morphology, size, quantity, and area density, were statistically analyzed. According to
classical thermodynamic nucleation theory, the relationships between the solute element
activities and the nucleation radius and nucleation rate of inclusions were obtained, and
the theoretical nucleation rate was compared with the experimental nucleation rate. Cluster
models of zirconium oxide were constructed by Materials Studio software. The cluster
structure and thermodynamic properties of nanoparticles after geometric optimization
were calculated, and the accuracy of the first-principles calculation was verified. By
combining the high-temperature experimental results, classical nucleation calculations,
and first-principles analysis, the nucleation mechanism of zirconium oxide inclusions is
proposed.

2.1. Sample Preparation
Pure iron was used as the raw material, and it was heated in a Si–Mo heating electric
resistance furnace (Braveman Special Testing Furnace CO. LTD., Luoyang, Henan, China).
The chemical composition of the pure iron sample is shown in Table 1. The pure iron
sample was heated to 1873 K (1600 ◦ C) in an alumina crucible in the Si–Mo heating electric
resistance furnace. After the temperature was maintained at 1873 K for 30 min, the Zr–Fe
alloy (60% Zr) wrapped in a high-purity iron belt was added to the melted pure iron,
followed by stirring for 10 s to ensure uniform distribution of the Zr–Fe alloy. Finally, 120 s
after adding the Zr–Fe alloy, samples were removed in quartz tubes, followed by quenching
in water. The whole experimental process was protected by high-purity argon gas.
Table 1. Chemical contents of the impurities in the pure iron sample (wt%).
CompositionC Si Mn P S Cr Al Cu Ni Ti N Fe
Amount 0.016 0.0033 0.01 0.0053 0.0017 0.0107 0.003 0.0037 0.0038 0.001 0.002 Bal.
138
Materials 2022, 15, 7960
The inclusions were extracted by electrolysis. A copper plate was used as the cathode,
and the sample was the anode. After electrolysis, the anode was placed in anhydrous
ethanol. By ultrasonic cleaning, the inclusions attached to the anode were dispersed in
anhydrous ethanol. Finally, the inclusions extracted by electrolysis were analyzed by micro
X-ray diffraction (μXRD, Bruker D8 Advance, Bruker, Berlin, Germany) and transmission
electron microscopy (TEM, JEOL JEM-F200, JEOL, Tokyo, Japan).
2.2. Microstructure and Composition Characterization

To measure the composition and morphologies, the samples were processed into
∅5 mm × 15 mm metallographic samples. Scanning electron microscopy with energy-
dispersive X-ray spectroscopy (SEM-EDS) (EM-30PLUS, COXEM, Daejeon, Korea) was
then performed. In addition, the samples were processed into ∅5 mm × 10 mm bars for
total oxygen and nitrogen content detection by the fusion-infrared absorption method. The
oxygen and nitrogen contents were measured three times to investigate the uniformity of
the total oxygen and nitrogen contents in the molten iron. In addition, the total Zr content
was measured by inductively coupled plasma–atomic emission spectroscopy. The chemical
contents of oxygen, nitrogen, and zirconium are given in Table 2.
Table 2. Chemical contents of O, N, and Zr in the steel sample after Zr addition.
Zr Addition Holding Time/s [O]/ppm [N]/ppm [Zr]/ppm

125 21 24
100 ppm 120 89 28 31
106 34 22
To measure the composition of the inclusions in the steel after zirconium addition,
the inclusions extracted by electrolysis were observed by μXRD and TEM. The results
provided an experimental reference for subsequent first-principles calculation of inclusion
crystal-type selection.
2.3. Nucleation Calculation

According to classical nucleation theory, the critical nucleation size and nucleation
rate I [cm−3 ·s−1 ] can be calculated by [24]

16πγSL 3 V2
1 1
ln I = O
− (1)
3k B R2 T 3 ln S∗
2
(ln SO )2
O
2γ 2rSL VO
rC = − SL = (2)
ΔGV RT ln SO
where kB is the Boltzmann constant (1.38 × 10−23 J/K), R is the gas constant (8.314 J/(mol K)),
T [K] is the absolute temperature, and VO [m3 /mol] is the molar volume of oxide. γSL [J/m2 ]
is the interfacial energy between the oxide and liquid steel, and it can be expressed by
Young’s equation:
γSL = γSV − γLV cos θ (3)
γLV = 1.75 − 0.279 ln(1 + 140· aO ) (4)
where γSV is the surface energy of the solid inclusion, γLV is the surface energy of the
liquid steel, which has been calculated in previous studies [25,26], and θ is the contact angle
between liquid steel and the inclusion, as illustrated at Table 3.
139
Materials 2022, 15, 7960
Table 3. Data related to the calculation of the critical nucleation size and nucleation rate of oxide in-
clusions.
Inclusion Θ (deg) γSV (J/m2 ) VO (m3 /mol) logKeq

ZrO2 111 [27] 1.395 [27] 10.5× 10−6 −57,000/T + 21.8 [28]
The experimental values of the nucleation rate I can be obtained by [24]
fv
I= (5)
3 πr · t
4 3
where t is the nucleation time (generally taken to be 0.2 s [29]) and r is the critical nucleation
radius obtained by Equation (2). fv is the volume fraction of oxide particles: [24]
ρ Fe MZrO2
fv = · ·[ppm insol. Zr] × 10−6 (6)
ρ ZrO2 xMZr
where ρFe and ρZrO2 are the densities of Fe and ZrO2 , respectively (ρFe = 7.8 g/cm3 ,
ρZrO2 = 5.85 g/cm3 ) [30], MZrO2 is the molecular weight of ZrO2 , Mzr is the atomic weight
of Zr, and [ppm insol. Zr] represents the Zr content.
2.4. First-Principles Calculation

The DMol [3] module based on density functional theory in the Materials Studio
software package (Materials Studio8.0, Accelrys, California, America) was used for cluster
optimization and thermodynamic property calculation. The Broyden–Fletcher–Goldfarb–
Shanno mechanism was used for geometric optimization. The Perdew–Burke–Ernzerhof
functional with the generalized gradient approximation was selected as the electron
exchange–correlation potential function [31].
3. Results
3.1. Inclusion Characterization
To obtain the size distribution of the ZrO2 inclusions, 100 SEM images were continu-
ously taken at 5000× magnification, and the total observed area was 0.11 mm2 . The average
size of the ZrO2 inclusions in the sample was 0.56 μm, and the size distribution of ZrO2
inclusions approximately approached a normal distribution (Figure 1).
100
Average size: 0.56 μm
Number density (mm_2)
80 100ppm Zr
60
40
20
0
0.2 0.3 0.4 0.5 0.6 0.7 0.8
Size distribution (μm)
Figure 1. Size distribution of ZrO2 in Zr deoxidized steel.
140
Materials 2022, 15, 7960
The inclusions extracted by electrolysis were characterized by μXRD, and the mor-
phology and composition were analyzed by SEM-EDS (Figure 2). From the mapping image,
the inclusions in the Zr deoxidized steel were zirconium oxide. A few Al inclusions were
also detected because of the trace amount of Al in the raw materials. The results of μXRD
suggested that the zirconium oxide was ZrO2 . In addition, both monoclinic and tetragonal
ZrO2 were detected. This may be because of the transformation of tetragonal ZrO2 to
monoclinic ZrO2 during rapid cooling.
Figure 2. Morphology, composition, and X-ray diffraction pattern of the inclusions in Zr deoxi-
dized steel.
3.2. Classical Nucleation Calculation

According to classical nucleation theory, the relationship between the critical nu-
cleation radius of ZrO2 and the activities of the solute elements at 1873 K is shown in
Figure 3a. When the Zr activity is in the range 0.0001–1, and the oxygen activity is in the
range 0.0001–0.1, the critical nucleation radius of ZrO2 is 0.3–1.2 nm. The relationship
between the nucleation rate and solute element activities is shown in Figure 3b. When the
Zr activity is in the range 0.0001–1, and the oxygen activity is in the range 0.001–0.1, the
critical nucleation rate of ZrO2 is in the range 100–560 cm−3 ·s−1 . The points in Figure 3 are
the experimentally measured Zr and oxygen activities.
(a) (b)
_ _
Zirconium Activity (aZr)
Zirconium Activity (aZr)
-40 cm 3·s 1
_ _
10-1 10-1 100 cm 3·s 1
_ _
_ _ 200 cm 3·s 1
-40 cm 3·s 1
10-2 10-2 _ _
0.3 nm 300 cm 3·s 1
0.4 nm _ _
400 cm 3·s 1
0.6 nm
10-3 10-3 _ _
0.8 nm 500 cm 3·s 1
1.0 nm _ _
1.2 nm 560 cm 3·s 1
10-4 10-4
10-3 10-2 10-1 10-3 10-2 10-1
Oxygen Activity (ao) Oxygen Activity (ao)
Figure 3. Relationships between the activities of the solute elements and the (a) critical nucleation
radius and (b) nucleation rate of ZrO2 .
141
Materials 2022, 15, 7960
To obtain the activities of zirconium and oxygen, the composition of the zirconium
deoxidized steel and corresponding thermodynamic data were substituted into
ai = f i [mass% i ] (7)
∑ ei [mass% i]
j
log f i = (8)
where ai , f i , and [mass% i ] are the activity, activity coefficient, and concentration of element
j
i, respectively, and ei is the first-order interaction coefficient (Table 4).
Table 4. Interaction coefficients of O and Zr at 1873 K [26].
j
ei Zr O
O −23 −0.17
Zr 0.032 −4
The experimentally estimated value of ln I can be calculated by Equations (5) and (6).
The experimental value of ln I was 57 cm−3 ·s−1 . From Figure 3, the theoretical value of ln I
is approximately −40 cm−3 ·s−1 . Therefore, the experimental value of the nucleation rate I
was approximately 40 orders of magnitude larger than the theoretical value.
3.3. First-Principles Calculations

According to the two-step nucleation mechanism, the nucleation process of inclusions
in liquid steel should include the multiphase deoxidation reaction, which can be expressed
by the following two steps [32–35]. In the first step, the deoxidized elements in the molten
steel melt and dissolve, and the deoxidized element atoms react with the dissolved oxygen
in the molten steel to form oxide clusters. In the second step, the clusters combine to form
cluster aggregates. The cluster aggregates then form critical crystal nuclei.
3.3.1. Structures of (ZrO2 )n Clusters

With increasing n value, the average bond length of the (ZrO2 )n cluster initially slightly
increases, and it finally fluctuates at approximately 2.0 Å (Table 5). The nucleon binding
ability in the nucleus is stronger and more stable for larger average binding energy (Ebin ).
The average binding energies of the (ZrO2 )n (n = 1–6) clusters are negative (Table 5),
indicating that the binding between nuclei is relatively stable.
142
Materials 2022, 15, 7960
Table 5. Structures, average bond lengths, sizes, and average binding energies of (ZrO2 )n
(n = 1–6) clusters.
Clusters Bond Length(Zr-O) (nm) Size (Dmax ) nm Ebin eV
0.189 0.38 −5.55

(ZrO2 )1
0.192 0.48 −7.05

(ZrO2 )2
0.204 0.60 −7.38

(ZrO2 )3
0.20 0.66 −7.52

(ZrO2 )4
0.206 0.70 −7.68

(ZrO2 )5
0.203 0.93 −7.75
(ZrO2 )6
With an increasing number of atoms (n = 1–6), the average binding energy of the
(ZrO2 )n cluster decreases, especially between the (ZrO2 )1 and (ZrO2 )2 clusters. This may be
because of the energy error caused by the different numbers of atoms in different clusters.
Therefore, the energy gap was used to compare the stabilities of the clusters. The energy
gap is the difference between the lowest unoccupied molecular orbital (LUMO) and the
highest occupied molecular orbital (HOMO). The energy gap reflects the ability of electrons
to transition from an occupied orbital to an empty orbital, and it represents the ability of
molecules to participate in chemical reactions. The system is more stable for a larger energy
gap [36].
The energy gaps of the (ZrO2 )n clusters are given in Table 6. The (ZrO2 )2 cluster
has the largest energy gap, indicating that the (ZrO2 )2 cluster is the most stable of the
(ZrO2 )n clusters (n = 1–6). The LUMOs and HOMOs of the (ZrO2 )n clusters are shown in
Figure 4. The blue and yellow area denote the orbitals of electron cloud, where the color is
to distinguish the plus or minus of orbital wave function.
Table 6. LUMO–HUMO energy gaps of the (ZrO2 )n (n = 1–6) clusters.
Clusters (ZrO2 )1 (ZrO2 )2 (ZrO2 )3 (ZrO2 )4 (ZrO2 )5 (ZrO2 )6

LUMO −0.15927 −0.12413 −0.15856 −0.15835 −0.11532 −0.13762
HOMO −0.18301 −0.2138 −0.21179 −0.19151 −0.20248 −0.22429
LUMO-HUMO 0.023742 0.089665 0.05323 0.033158 0.087153 0.086679
143
Materials 2022, 15, 7960
Figure 4. HOMOs and LUMOs of the (ZrO2 )n clusters.
3.3.2. Thermodynamic Properties of the (ZrO2 )n Clusters

The thermodynamic properties of ZrO2 are shown in Figure 5. Where S is the entropy,
and Cp is the heat capacity. In this study, the ZrO2 crystal structure was tetragonal. The
lines and points represent the calculated thermodynamic properties and values in the
literature, respectively [37]. In the temperature range of 0–1000 K, there is a certain degree
of deviation between the calculated and the values in the literature, but the variation
trend of the thermodynamic properties with temperature is consistent. To the best of our
knowledge, the reason that caused deviation is mainly the machine error derived from the
thermochemical software package, and it was not possible to meet strict consistency criteria
when combining data from various sources to form a data set for a substance. Another
reason is that phase transformation occurred between monoclinic and tetragonal ZrO2 . In
general, the calculated value is in good agreement with the value from the literature.
144
Materials 2022, 15, 7960
80
Thermodynamic Properties
Calculated values
_1 _1
S (J·mol ·K )
60
_1 _1
CP (J·mol ·K )
40
Experimental values
20
_1 _1
S (J·mol ·K )
0 _1 _1
CP (J·mol ·K )
0 200 400 600 800 1000

Temperature (K)
Figure 5. Thermodynamics properties of ZrO2 .
3.3.3. Gibbs Free Energy Changes of (ZrO2 )n Clusters and Nanoparticles

The formation Gibbs free energy (ΔG) curves of the (ZrO2 )n clusters and ZrO2(s) are
shown in Figure 6a,b, respectively. The formation Gibbs free energies of (ZrO2 )n (n = 1–6)
are negative, suggesting the (ZrO2 )n (n = 1–6) clusters form. However, the formation Gibbs
free energy change from (ZrO2 )1 to ZrO2(s) is positive, so the reaction from (ZrO2 )1 to
ZrO2(s) does not occur when the temperature is greater than 1000 K.
(b)
(a)
-200 -200
'G (kJ/mol)
'G (kJ/mol)
-400 -400
(ZrO2)1
(ZrO2)1
(ZrO2)2 (ZrO2)2
-600 -600 (ZrO2)3
(ZrO2)3
(ZrO2)4 (ZrO2)4
(ZrO2)5 (ZrO2)5
-800 -800 (ZrO2)6
(ZrO2)6
0 500 1000 1500 2000 0 500 1000 1500 2000

Temperature (K) Temperature (K)
Figure 6. (a) Formation Gibbs free energy curves of (ZrO2 )n . (b) Formation Gibbs free energy curve
of ZrO2(s) .
The Gibbs free energy changes from nanoscale ZrO2 to ZrO2(bulk) and from the Zr
and O atoms to nanoscale ZrO2 are shown in Figure 7a,b, respectively. The Gibbs free
energy of the macroscale ZrO2 crystal is less than zero, indicating that nanoscale ZrO2 will
spontaneously transform to the macroscale ZrO2 crystal.
145
Materials 2022, 15, 7960
(a) 0 2 nm (b) 0 2 nm
5 nm 5 nm
10 nm 10 nm
-200 15 nm -200 15 nm
20 nm 20 nm
'G (kJ/mol)
'G (kJ/mol)
25 nm 25 nm
-400 30 nm -400 30 nm
40 nm 40 nm
50 nm 50 nm
-600 60 nm
-600
60 nm
70 nm 70 nm
80 nm -800 80 nm
-800 90 nm 90 nm
100 nm 100 nm
200 nm -1000 200 nm
-1000 500 nm 500 nm
0 500 1000 1500 2000 0 500 1000 1500 2000
Temperature (K) Temperature (K)
Figure 7. (a) Gibbs free energy change from nanoscale ZrO2 to ZrO2(bulk) . (b) Gibbs free energy
changes from the Zr and O atoms to nanoscale ZrO2 .
4. Discussion
The formation of Gibbs free energy changes of the (ZrO2 )n clusters and various ZrO2
nanoparticles are shown in Figure 8. Most of the Gibbs free energy change curves of the
(ZrO2 )n clusters are higher than those of the ZrO2 nanoparticles, which is in agreement with
the law of the step-by-step change of the thermodynamic stability., from the perspective of
the gradual decrease of the energy, this law indicates that the formation process of ZrO2
occurs from the atoms to clusters to nanoscale crystal particles to the macroscale crystal.
(ZrO2)1
(ZrO2)2
200
(ZrO2)3
(ZrO2)4
(ZrO2)5
100 (ZrO2)6
2 nm
'G (kJ/mol)
5 nm
0 10 nm
15 nm
20 nm
25 nm
-100 30 nm
40 nm
50 nm
60 nm
-200 70 nm
80 nm
90 nm
100 nm
-300 200 nm
500 nm
ZrO2(bulk)
1800 1850 1900 1950 2000
Temperature (K)
Figure 8. Formation Gibbs free energy changes of (ZrO2 )n clusters and various ZrO2 nanoparticles.
146
Materials 2022, 15, 7960
The surfaces of nanoparticles usually contain two or three atomic layers [38,39], and
the average length of the Zr–O bond is approximately 0.214 nm. Therefore, the surface
atomic layer of zirconium oxide nanoparticles is approximately 0.428–0.642 nm thick. From
Table 4, the sizes of the (ZrO2 )n clusters are also within this range. In addition, from Table 5,
the energy gap of the (ZrO2 )2 cluster is the largest, indicating that this cluster is the most
stable. Therefore, the (ZrO2 )2 cluster may be the atomic layer on the surface of the ZrO2
nanoparticle. Based on the two-step nucleation theory, it is speculated that the nucleation
pathway of ZrO2 is Zratom + Oatom → (ZrO2 )n → (ZrO2 )2 → core (ZrO2 particle)–shell
((ZrO2 )2 cluster) nanoparticle → (ZrO2 )bulk . The nucleation process of ZrO2 at 1873 K is
shown in Figure 9.
Figure 9. Nucleation process of zirconium oxide at 1873 K.
5. Conclusions
High-temperature deoxidation experiments and inclusion-extraction experiments
have been performed. The nucleation process of ZrO2 inclusions in Zr deoxidized steel
was investigated by classical nucleation theory and first-principles calculation. The main
conclusions are as follows.
When the Zr content was 100 ppm, SEM-EDS showed that the main inclusions in
the steel were ZrO2 . μXRD analysis confirmed the existence of ZrO2 , and monoclinic and
tetragonal ZrO2 were simultaneously detected, which may be because tetragonal ZrO2
transformed to monoclinic ZrO2 during the rapid cooling process. The average size of
the ZrO2 inclusions was 0.56 μm. Through classical nucleation theory, the relationships
between the solute element activities and the nucleation radius and nucleation rate of ZrO2
were obtained. The theoretical value of the nucleation rate was compared with the values
in the literature, and the experimental value of I was approximately 40 orders of magnitude
larger than the theoretical value.
The thermodynamic properties of macroscale ZrO2 were calculated by first principles,
and the results were consistent with the experimental values. (ZrO2 )n (n = 1–6) cluster
models were constructed, and the thermodynamic properties of the geometrically opti-
mized cluster structures and nanoscale ZrO2 particles were calculated, which verified the
rationality of the existence of the pre-nucleation phase in terms of the thermodynamics.
Based on two-step nucleation theory, the nucleation pathway of ZrO2 is proposed to be
Zratom + Oatom → (ZrO2 )n → (ZrO2 )2 → core (ZrO2 particle)–shell ((ZrO2 )2 cluster) nanopar-
ticle → (ZrO2 )bulk .
147
Materials 2022, 15, 7960
Author Contributions: L.W., C.C. and S.Y. conceived the idea and designed the research. Y.L.
performed the theoretical calculations. Y.L. and X.L. performed the μXRD analyses. X.L. performed
the high-temperature experiments. Y.L. and L.W. wrote the original draft of the paper, and all authors
contributed to the review and editing of the paper. L.W. acquired the funding and supervised the
study. All authors have read and agreed to the published version of the manuscript.
Funding: This research was supported by the National Science Foundation of China (Grant Nos.
52064011 and 52274331) and the Science and Technology Planning Project of Guizhou (Grant Nos.
Qian Ke He Ji Chu ZK [2021]258 and Qian Ke He Zhi Cheng [2021]342). This study was also supported
by the Research Program for Talented Scholars of the Guizhou Institute of Technology (Grant No.
XJGC20190962).
Data Availability Statement: Part of the data presented in this study are available on request from
the corresponding author. The data are not publicly available due to intellectual property.
Acknowledgments: Thanks for the computing support of the State Key Laboratory of Public Big
Data, Guizhou University.
Conflicts of Interest: The authors state that there are no conflict of interest to disclose.
References
1. Takamura, J.I.; Mizoguchi, S. Roles of oxides in steel performance. Proceedings of The Sixth International Iron and Steel Congress,
ISIJ International, Nagoya, Japan, 21–26 October 1990; pp. 591–597.
2. Wang, L.; Yang, S.; Li, J.; Chen, C.; Li, C.; Li, X. Study on the Capillary Interaction Between Particles on the Surface of High-
Temperature Melts. Steel Res. Int. 2021, 92, 2100013. [CrossRef]
3. Li, Y.; Wang, L.; Li, J.; Yang, S.; Chen, C.; Li, C.; Li, X. In-Situ Observation on the Agglomeration and Dispersion of Particles at the
Interface of High-temperature Melts. ISIJ Int. 2021, 61, 753–762. [CrossRef]
4. Abraham, S.; Bodnar, R.; Raines, J.; Wang, Y. Inclusion engineering and metallurgy of calcium treatment. J. Iron Steel Res. Int.
2018, 25, 133–145. [CrossRef]
5. Liu, W.; Yang, S.F.; Li, J.S.; Wang, F.; Yang, H.B. Numerical model of inclusion separation from liquid metal with consideration of
dissolution in slag. J. Iron Steel Res. Int. 2019, 26, 1147–1153. [CrossRef]
6. Peng, B.W.; Li, F.J.; Zheng, S.B.; Li, H.G.; Zhai, Q.J. Inclusion control study in high aluminum steel. J. Iron Steel Res. Int. 2018, 25,
7. Zhao, K.W.; Zeng, J.H.; Wang, X.H. Nonmetallic inclusion control of 350 km/h high speed rail steel. J. Iron Steel Res. Int. 2009, 16,
20–26. [CrossRef]
8. Wang, L.; Yang, S.; Li, J.; Wu, T.; Liu, W.; Xiong, J. Improving Cleanliness of 95CrMo Drill Rod Steel by Slag Refining. Met. Mater.
Trans. A 2016, 47, 99–107. [CrossRef]
9. Yang, Y.; Zhan, D.; Lei, H.; Qiu, G.; Li, Y.; Jiang, Z.; Zhang, H. In situ observation of acicular ferrite nucleation and growth at
different cooling rate in Ti-Zr deoxidized steel. Metall. Mater. Trans. B 2019, 50, 2536–2546. [CrossRef]
10. Chai, F.; Yang, C.F.; Su, H.; Zhang, Y.Q.; Xu, Z.; Yang, Y.H. Effect of Zr addition to Ti-killed steel on inclusion formation and
microstructural evolution in welding induced coarse-grained heat affected zone. Acta Metall. Sin. Engl. Lett. 2008, 21, 220–226.
[CrossRef]
11. Guo, A.M.; Li, S.R.; Guo, J.; Li, P.H.; Ding, Q.F.; Wu, K.M.; He, X.L. Effect of zirconium addition on the impact toughness of the
heat affected zone in a high strength low alloy pipeline steel. Mater. Charact. 2008, 59, 134–139. [CrossRef]
12. Wu, H.B.; Hou, M.; Liang, G.L.; Tang, D. Effect of zirconium on the low-temperature toughness of CGHAZ in F40 ship plates
containing titanium. J. Univ. Sci. Technol. Beijing 2012, 34, 137–142.
13. Min, J.; Hu, Z.; Wang, X.; Pak, J.J. Characterization of microstructure and non-metallic inclusions in Zr-Al deoxidized low carbon
steel. ISIJ Int. 2013, 53, 1386–1391.
14. Bhadeshia, H.K.D.H.; Honeycombe, R.W.K. Steels Micro-Structure and Properties, 3rd ed.; Elsevier Ltd.: Oxford, UK, 2006; Volume
7, pp. 1–360.
15. He, K.; Baker, T.N. Zr-contaning precipitates in a Ti-Nb micro-alloyed HSLA steel containing 0.016% Zr addition. Mater. Sci. Eng.
1996, 215, 57–66.
16. Sarma, D.S.; Karasev, A.V.; Jonsson, P.G. On the role of nonmetallic inclusions in the nucleation of acicular ferrite in steels. ISIJ Int.
2009, 49, 1063–1074. [CrossRef]
17. Guo, Q.Y.; Song, B.; Song, M.M. Effects of Ti, Al and Zr deoxidation on morphology of sulfides in medium-sulfur non-quenched
and tempered steel. Trans. Mater. Heat Treat. 2019, 40, 133–139.
18. Li, Y.; Wan, X.L.; Cheng, L.; Wu, K.M. First-principles calculation of the interaction of Mn with ZrO2 and its effect on the formation
of ferrite in high-strength low-alloy steels. Scr. Mater. 2014, 75, 78–81. [CrossRef]
19. Wang, G.; Zhang, L. Thermodynamics dependent on size in nucleation process of inclusions and calculation of critical nucleus
radius in clean molten steel. Iron Steel 2012, 47, 22–26. [CrossRef]
148
Materials 2022, 15, 7960
20. Wang, G.C.; Wang, Q.; Li, S.L.; Ai, X.G.; Fan, C.G. Evidence of Multi-step Nucleation Leading to Various Crystallization Pathways
from an Fe-O-Al Melt. Sci. Rep. 2014, 4, 5082. [CrossRef]
21. Lei, H.; Xiao, Y.; Wang, G.; Zhang, H.; Jin, W.; Zhang, L. Thermodynamic insight into the growth of nanoscale particle of
Al-deoxidation in Fe–O–Al melt. Sci. Rep. 2020, 10, 16909. [CrossRef]
22. Zhao, D.; Bao, W.; Li, H.; Zheng, S.; Chou, K.-C. Cluster-assisted nucleation mechanism of titanium oxides in Fe-Ti supercooled
alloys. J. Alloys Compd. 2018, 744, 797–800. [CrossRef]
23. Yang, L.; Zhang, W.; He, L.; Li, H.; Zheng, S. Study on the growth and morphology evolution of titanium oxide clusters in molten
iron with molecular dynamics simulation. RSC Adv. 2019, 9, 32620–32627. [CrossRef] [PubMed]
24. Suito, H.; Ohta, H. Characteristics of Particle Size Distribution in Early Stage of Deoxidation. ISIJ Int. 2006, 46, 33–41. [CrossRef]
25. Wang, L.; Li, J.; Yang, S.; Chen, C.; Jin, H.; Li, X. Nucleation and Ostwald Growth of Particles in Fe-O-Al-Ca Melt. Sci. Rep. 2018, 8,
1135. [CrossRef] [PubMed]
26. Li, Y.; Wang, L.; Chen, C.; Li, J.; Li, X. Effect of Mg Treatment on the Nucleation and Ostwald Growth of Inclusions in Fe-O-Al-Mg
Melt. Materials 2020, 13, 3355. [CrossRef] [PubMed]
27. Humenik, M.; Kingery, W.D. Metal-Ceramic Interactions: III, Surface Tension and Wettability of Metal-Ceramic Systems. J. Am.
Ceram. Soc. 1954, 37, 18–23. [CrossRef]
28. Hino, M.; Ito, K. Thermodynamic Data for Steelmaking, 2nd ed.; Tohoku University Press: Tohoku, Japan, 2010; pp. 247–264.
29. Ohta, H.; Suito, H. Efects of Dissolved Oxygen and Size Distribution on Particle Coarsening of Deoxidation Product. ISIJ Int.
2006, 46, 42–49. [CrossRef]
30. Sakata, K.; Suito, H. Dispersion of fine primary inclusions of MgO and ZrO2 in Fe-10 mass pct Ni alloy and the solidification
structure. Met. Mater. Trans. B 1999, 30, 1053–1063. [CrossRef]
31. Perdew, J.P.; Burke, K.; Ernzerhof, M. Generalized Gradient Approximation Made Simple. Phys. Rev. Lett. 1997, 77, 3865.
[CrossRef]
32. Wang, G.; Xiao, Y.; Zhao, C.; Jing, L.; Shang, D. Atomic Cluster Aggregates in Nucleation of Solid Alumina Inclusion in the
Aluminum Deoxidation for Liquid Iron. Metall. Mater. Trans. B 2017, 49, 282–290. [CrossRef]
33. Wang, G.; Xiao, Y.; Song, Y.; Zhou, H.; Tian, Q.; Li, F. A density functional study on the aggregation of alumina clusters. Res.
Chem. Intermed. 2017, 43, 1447–1463. [CrossRef]
34. Xiao, Y.; Lei, H.; Yang, B.; Wang, G.; Wang, Q.; Jin, W. Nucleation and growth for magnesia inclusion in Fe-O-Mg melt. RSC Adv.
2018, 8, 38336–38345. [CrossRef]
35. Xiao, Y.; Lei, H.; Yang, B.; Zhao, Y.; Wang, Q.; Wang, G. Thermodynamic insight into the growth of Calcia inclusions at the
nanoscale: The case of Fe-O-Ca melt. RSC Adv. 2019, 9, 11135–11141. [CrossRef]
36. Aihara, J.-I. Reduced HOMOLUMO Gap as an Index of Kinetic Stability for Polycyclic Aromatic Hydrocarbons. J. Phys. Chem. A
1999, 103, 7487–7495. [CrossRef]
37. Chase, M.W. NIST-JANAF Thermochemical Tables; American Chemical Society: Washington, DC, USA, 1998; Volume 2,
pp. 1880–1881.
38. Byun, J.S.; Shim, J.H.; Cho, Y.W.; Lee, D.N. Non-metallic particle and intragranular nucleation of ferrite in Ti-killed C-Mn steel.
Acta Mater. 2003, 51, 1593–1606. [CrossRef]
39. Phillpot, S.R.; Wolf, D.; Gleiter, H. Molecular dynamics study of the synthesis and characterization of a fully dense, three
dimensional nanocrystalline material. J. Appl. Phys. 1995, 78, 847–861. [CrossRef]
149
materials
Article
Unveiling the Effect of CaF2 on the Microstructure and
Transport Properties of Phosphosilicate Systems
Yizhe Du 1,2 , Zhidan Huang 1,2 , Mujun Long 1,2 , Huamei Duan 1,2 and Dengfu Chen 1,2, *
1 Laboratory of Materials and Metallurgy, College of Materials Science and Engineering, Chongqing University,
Chongqing 400044, China
2 Chongqing Key Laboratory of Vanadium-Titanium Metallurgy and New Materials, Chongqing University,
Chongqing 400044, China
Abstract: As an effective flux, CaF2 is beneficial in improving the fluidity of slag in the steel-making
process, which is crucial for dephosphorization. To reveal the existence form and functional mech-
anism of CaF2 in phosphosilicate systems, the microstructures and transport properties of CaO-
SiO2 -CaF2 -P2 O5 quaternary slag systems are investigated by molecular dynamics simulations (MD)
combined with experiments. The results demonstrate that the Si-O coordination number does not
vary significantly with the increasing CaF2 content, but the P-O coordination number dramatically
decreases. CaF2 has a minor effect on the single [SiO4 ] but makes the structure of the silicate sys-
tem simple. On the contrary, F− ions could reduce the stability of P-O bonds and promoted the
transformation of [PO4 ] to [PO3 F], which is beneficial for making the P element-enriched phosphate
network structure more aggregated. However, the introduction of CaF2 does not alter the tetrahedral
character of the original fundamental structural unit. In addition, the results of the investigation of
the transport properties show that the self-diffusion coefficients of each ion are positively correlated
with CaF2 content and arranged in the order of F− > Ca2+ > O2− ≈ P5+ > Si4+ . Due to CaF2 reducing
Citation: Du, Y.; Huang, Z.; Long, M.; the degree of polymerization of the whole melts, the viscosity decreases from 0.39 to 0.13 Pa·s as the
Duan, H.; Chen, D. Unveiling the CaF2 content increases from 0% to 20%. Moreover, the viscosity of the melt shows an excellent linear
Effect of CaF2 on the Microstructure dependence on the structural parameters.
and Transport Properties of
Phosphosilicate Systems. Materials
Keywords: molecular dynamics simulation; CaF2 ; melts’ structure; transport properties; viscosity
2022, 15, 7916. https://doi.org/
10.3390/ma15227916
Academic Editor: Philippe

Colomban 1. Introduction
Received: 20 October 2022 The physical and chemical properties of slag are crucial for mass transfer and chemical
Accepted: 8 November 2022 reactions between liquid steel and slag. It is well-known that the physical and chemical
Published: 9 November 2022 properties of slag are determined by its structural characteristics [1,2], and it is of extraordi-
nary interest to study the structural information of slag to understand its performance.
There is a large quantity of experimental approaches that have been applied to study
the structure information of slag, and they mainly include nuclear magnetic resonance,
X-ray diffraction, neutron diffraction, Raman spectroscopy, etc. [3–5]. These methods help
iations.
one to effectively understand the microstructure and unique properties of slag, which
is a significant breakthrough in this research direction. In recent years, with the rapid
development of computer technology, a large number of simulation techniques have
Copyright: © 2022 by the authors. gradually entered the field of vision of scholars. In particular, MD simulations are expected
Licensee MDPI, Basel, Switzerland. to provide an effective way to understand the slag structure from a microscopic point of
This article is an open access article view with its advantages. Specifically, it is not affected by experimental conditions, such
distributed under the terms and as high temperature and pressure. At present, a large number of scholars have used the
conditions of the Creative Commons molecular dynamics (MD) simulation method to conduct studies on the microstructure and
Attribution (CC BY) license (https:// properties of metallurgical slag and achieved remarkable results [6–8].
4.0/).

Materials 2022, 15, 7916
At present, structural information of binary and ternary silicate and aluminate melts
has been extensively and carefully studied. As for phosphate melt, it has not been system-
atically studied due to the complexity and diversity of its structure. However, phosphorus
is one of the harmful elements in steel. The excessive amount of phosphorus in steel is
detrimental to its quality and properties. Therefore, dephosphorization is one of the key
tasks in steel-making. In addition, phosphorus removal relies on the reaction between
steel and slag, so a comprehensive study of the microstructure and transport properties of
phosphorus removal slag could help clarify the underlying causes of the alteration in its
macroscopic properties. In previous studies, Diao et al. [9,10] studied the microstructure
of the ternary slag system of CaO-SiO2 -P2 O5 through MD simulation, and the results
showed that silicon and phosphorus mainly formed a tetrahedral structure. Additionally,
the concentration of free oxygen decreases significantly with increasing P2 O5 content and
the degree of polymerization of the melt increases. Fan et al. [11] reported the existence
form of Si and P in CaO-SiO2 -P2 O5 melt under high basicity, and the results showed that
both Si and P tended to form complex anions, and P ions are more inclined to form a
single tetrahedron structure. Moreover, Jiang et al. [12] used MD simulation to study the
structure and properties of the molten CaO-SiO2 -P2 O5 -FeO slag system and concluded that
the polymerization degree of the system decreased with the increase of basicity.
In our previous studies, a relatively deep understanding of the structural properties
of binary phosphate systems has been obtained [1]. In actual production, additional
components will be used to adjust the comprehensive physicochemical properties of the
slag to meet the requirements of the metallurgical production process. CaF2 is widely used
as a conventional flux to reduce the viscosity of slag and improve its mobility. At present,
some scholars have carried out studies on the structural properties of CaF2 -containing
glasses. For example, Kansal et al. [13] studied the effect of the CaO/MgO ratio on
the structure and thermal properties of CaO-MgO-SiO2 -P2 O5 -CaF2 , and found that CaF2
always tends to combine with [PO4 ] to form composite structures. Pedone et al. [14]
investigated the influence of halides on the structure of phosphosilica bioactive glasses by
MD simulation. The results show that in the mixed-fluoride/chloride-containing glasses,
fluorine tends to surround phosphate, whereas chloride moves toward the silicate network.
Furthermore, interestingly, according to relevant reports, F− can also directly participate in
the dephosphorization reaction, thus directly affecting the structure of dephosphorization
slag [15,16]. However, no reports have been found regarding the presence of CaF2 in
dephosphorized slag and its effects on structure and properties.
Therefore, to shed light on the morphological and functional mechanisms responsible
for the presence of CaF2 in dephosphorized slag, in this paper we focus on the quaternary
slag system CaO-SiO2 -CaF2 -P2 O5 and perform a comprehensive analysis of its microstruc-
ture using molecular dynamics simulations. Under the conditions of certain contents of
(xCaO)/(xSiO2 ) and P2 O5 in the slag, the influence of CaF2 on the microstructure and
transport performance of the slag system under high temperature is investigated combined
with experiments. The results of this investigation can provide some valuable information
to understand the microstructure of dephosphorized slag and clarify the intrinsic link
between melt flow properties and the evolution of structural units.
2. Computational Methodology
2.1. Interatomic Potential
For molecular dynamics simulation, selecting appropriate potential function and
corresponding parameters is the basis of accurate calculation. All molecular dynamics
simulations were carried out using the Born-Mayer-Huggins (BMH) model [1,7,9–11] in
this study, and the potential can be expressed as Equation (1):
Zi Zj e2 Cij
Uij (r ) = + Aij exp − Bij r − 6 (1)
4πε 0 rij rij
152
Materials 2022, 15, 7916
where, Zi and Zj are the effect charges of ions i and j, respectively, e represents the charge of
a single electron, ε0 represents the vacuum permittivity, rij is the distance between atoms i
and j, and Aij , Bij , and Cij are the adjustable parameters for BMH potentials. The three items
on the right side of the above formula represent the coulombic interaction, short-range
repulsion interaction, and the van der Waals interactions. Various interaction potential
parameters between the selected particles are listed in Table 1 [11,17,18].
Table 1. Born–Mayer–Higgins potential parameters for atomic pairs in CaO-SiO2 -CaF2 -P2 O5 systems.
Atom 1 Atom 2 Aij (eV) Bij (1/Å) Cij (eV·Å6 )

Ca Ca 329,051.60 6.25 4.33
Ca Si 26,674.68 6.25 0
Ca P 164,585.80 12.5 0
Ca O 717,827.00 6.06 8.67
Ca F 496,191.60 6.06 8.67
Si Si 2162.39 6.25 0
Si P 1081.60 12.5 0
Si F 43,406.00 6.06 0
Si O 62,794.37 6.06 0
P P 0 0.0 0
P F 365,232.88 11.8 0
P O 1847.70 3.45 0
F F 730,722.80 5.88 17.34
F O 1,046,135.40 5.88 17.34
O O 1,497,049.00 5.88 17.34
2.2. Simulation Approach

The target sample composition was discussed in the full melting component range
at the corresponding temperatures. The slag samples of CaO-SiO2 -CaF2 -P2 O5 were di-
vided into five groups, where the first group with 0% CaF2 was used as a comparison
with the other four groups. As Figure 1 shows, according to the liquid phase range of the
CaO-SiO2 -CaF2 -P2 O5 quaternary slag system at 1600 ◦ C obtained through Factsage 8.0 ther-
modynamic calculation software, the chemical composition of each group in this study
was determined. The varying number of atoms was then calculated based on the mole
fraction of each group. Referring to the empirical formula in relevant literature and research
results [19], the density of various groups at 1600 ◦ C was calculated, respectively. The
chemical composition, atomic number, and other information about each group of samples
are listed in Table 2.
The computational methods used and the choice of parameters are critical factors
in achieving efficient and accurate simulations. In this study, about 6000 atoms were
randomly placed in a model box. Since the number of calculated atoms is always finite,
periodic boundary conditions were performed on all faces of the model box to obtain an
infinite system of atoms without boundaries. The results obtained with periodic boundary
conditions are sufficient to reflect the actual situation. All MD simulations use a canonical
ensemble (NVT), which means that the calculations were performed in a system with a
constant atomic number (N), sample volume (V), and temperature (T). Additionally, the
sum method of Ewald was used for the long-distance coulomb force, and the motion
equation of atoms was explained by the jump integral method of a 1 fs timestep. The
potential cutoff radius was set to 10 Å in the calculation of the repulsive force. Besides,
the total time length of each group of simulations was determined to be 60 ps, equivalent
to 60,000 steps. After the beginning of the simulation, the initial temperature was set
at 5000 K (4727 ◦ C) for 15,000 timesteps to agitate the atoms and eliminate the effect
of the intentional distribution. Secondly, the temperature was cooled down to 1873 K
(1600 ◦ C) with 30,000 timesteps. Subsequently, the system was relaxed at 1600 ◦ C for
another 15,000 timesteps in the equilibrium calculation. The temperature, volume, and
153
Materials 2022, 15, 7916
enthalpy remained nearly constant for 15,000 timesteps, demonstrating that the system has
reached equilibrium.
Figure 1. Liquid phase interval of the CaO-SiO2 -CaF2 -P2 O5 system.
Table 2. Composition, number of atoms, and density of CaO-SiO2 -CaF2 -P2 O5 melts at 1600 ◦ C.
Mole Fraction (%) Number of Atoms Total Density (g/cm3 )

Groups
CaO SiO2 CaF2 P2 O5 Ca Si P F O
G1 49 41 0 10 1010 845 412 0 3732 5999 2.5159
G2 46 39 5 10 1041 796 408 204 3551 6000 2.5109
G3 44 36 10 10 1095 730 405 405 3365 6000 2.5083
G4 41 34 15 10 1124 682 401 602 3191 6000 2.5018
G5 38 32 20 10 1152 636 397 795 3020 6000 2.4947
2.3. Statistics of Structural Information

The radial distribution function (RDF) is commonly used to investigate the character
of the short-range order of melts. Equation (2) lists the mathematical expression of the
RDF [20]: 2
V nij (r − Δr/2, r + Δr/2)
Ni Nj ∑
gij (r ) = (2)
j
4πr2 Δr
where, V is the volume of MD-simulated cells, N is the number of particles, and nij is the
average number of atom j surrounding the atom i within a distance r ± Δr/2. The abscissa
of the first peak and the first trough in the RDF curve represent the average bond length
and cutoff radius of the corresponding atoms, respectively.
In addition, integration of the corresponding partial RDF of the particles generates the
average CN function, which represents the number of atoms j around atoms i within the
cutoff radius. The CN function is expressed as Equation (3):
4πNj r 2
Nij (r ) = r gij (r )dr (3)
V 0
Finally, as far as the concentration of oxygen species and the distribution of structural
units, Qn , were concerned, this structural information will be counted by the Matlab
program, based on the spatial atomic coordinates derived from the MD simulation.
154
Materials 2022, 15, 7916
2.4. Viscosity Calculation

Viscosity is one of the most significant physical parameters of slag. The viscosity
of melts can reflect the degree of polymerization. Through statistical analysis of atomic
coordination by MD simulation, a function of MSD would be generated, displayed as
Equation (4):
1 N
MSD = Δr (t)2
= ∑ |ri (t) − ri (0)|2
(4)
N i =1
where, N is the number of particles, ri (t) represents the position coordinate of atom i at
time t, and angular brackets denote a statistical average of many function values. The
self-diffusion coefficient could be obtained from MSD as shown below [21]:
* +
1 d Δr (t)2
D = lim (5)
t→∞ 6 dt
Then, the shear viscosity information of the melts can be obtained by combining the
self-diffusion coefficient, D, with the Stokes–Einstein equation [22,23]:
KB T
η= (6)
Dλ
where, K B is the Boltzmann constant, which is 1.38 × 10−23 J/K, T is the system temperature,
and λ is the particle transition step size, which is commonly considered λ = 2RO = 2.8 Å [24–26].
Based on the above calculation method, the partial transport performance of the melts can
be obtained, and the relationship between the structural information and performance can
be established.
3. Experimental Method
Based on the mole fraction of each sample in Table 2, the composition of the exper-
imental slag was obtained by mass conversion. The results are shown in Table 3. The
reagents used in our experiments are all from a specialist chemical reagent company in
Chongqing, China. The purity of the reagents (CaO, SiO2 , CaF2 , and P2 O5 ) used was above
99.5 wt.%. The weighted sample powder was well-mixed and placed in a graphite crucible
before the viscosity was measured.
Table 3. The chemical compositions of experimental slags.
Slag Composition (Mass Percent)

Groups
CaO SiO2 CaF2 P2 O 5
G1 41.5 37.1 0 21.4
G2 38.3 34.8 5.8 21.1
G3 36.1 31.7 11.4 20.8
G4 33.1 29.5 16.9 20.5
G5 30.3 27.3 22.2 20.2
The viscosity was measured using the rotating cylinder method. The viscometer was
calibrated at room temperature using an oil with known viscosity prior to the experiment.
Approximately 250 g of each slag sample was placed into a graphite crucible to melt,
and the average heating rate was 5 ◦ C/min. Since P2 O5 has a low boiling point, it is
prone to volatilization and produces white smoke at high temperatures. Consequently,
the other three components were firstly added into the crucible and heated to 1500 ◦ C
for 20 min. After the sample was fully melted, P2 O5 was added, and the crucible was
covered to prevent volatilization. Then, the crucible was opened after keeping it for 5 min.
If there was no obvious white smoke, the crucible was reheated to 1600 ◦ C and kept for
20 min to homogenize the chemical composition. Finally, the viscosity of each sample was
determined from an average of 60 consecutive measurements.
155
Materials 2022, 15, 7916

4.1. Local Structural Characteristics
The local structure information of melts can be preliminarily obtained by RDF and CN.
Taking G2 as an example, Figure 2a,b show the distribution of RDF and CN in the system
of CaO-SiO2 -CaF2 -P2 O5 at 1600 ◦ C when CaF2 content was 5%, respectively. According to
the RDF curve, the average bond length of each atom pair in the melts can be concluded.
As can be obtained from Figure 2a, the average bond lengths of Ca-O, Si-O, P-O, and Ca-F
were 2.31, 1.62, 1.50, and 2.30 Å, respectively. The results are in good accordance with
previous research obtained from MD simulations and experiments [9–11,27]. Table 4 shows
the variation of the bond lengths for various pairs of atoms from G1 to G5.
Figure 2. (a) RDF and (b) CN for different atom pairs in CaO-SiO2 -CaF2 -P2 O5 systems and changes
of P-O and Si-O coordination numbers: (c) Si-O and (d) P-O.
Table 4. Variation of average bond lengths of different atom pairs.
Pair Ca-O Si-O P-O P-F Ca-F P-Ca

Rij /Å (G1) 2.31 1.61 1.50 — — 3.60
Rij /Å (G2) 2.31 1.61 1.52 1.01 2.31 3.60
Rij /Å (G3) 2.31 1.61 1.54 1.01 2.31 3.60
Rij /Å (G4) 2.32 1.61 1.55 1.01 2.32 3.61
Rij /Å (G5) 2.32 1.61 1.56 1.01 2.32 3.61
In general, a strong and sharp peak in the RDF curve indicates a steep stabilization
of the corresponding bond. Similarly, for CN curves, a broad flat plateau implies a large
stability of the corresponding polyhedron. It can be observed in Figure 2a that both Si-O
and P-O curves had a sharp peak, meaning Si and P tended to combine with O atoms and
form stable structures. From Table 4, the average Si-O bond length remained constant with
the increasing CaF2 content in the melt, demonstrating that the Si-O bond was particularly
stable and was not subject to CaF2 . However, the length of the P-O bond became longer,
confirming the character of the P-O structure affected by CaF2 , while indicating a decrease
156
Materials 2022, 15, 7916
in the strength of the P-O bond, which may lead to the evolution of the phosphate melt
structure. Furthermore, the P-F bond appeared in the system due to the addition of CaF2 .
Interestingly, the RDF curve for the P-F bond had an unusually sharp peak and the P-F
bond length did not change significantly with CaF2 content, indicating that the P-F bond
was considerably more stable than the P-O bond, which is unprecedented. In addition,
the Ca-O and Ca-F bonds had slightly increased lengths, indicating that they were more
loosely bound with the addition of CaF2 .
As can be seen from Figure 2b, CNSi-O , CNP-O , and CNCa-O were 4.04, 3.87, and 5.51, re-
spectively. The plateau on the CNSi-O curve was smoother than that on the CNP-O curve,
indicating that the stability of the Si-O structure was higher than that of the P-O in the
CaO-SiO2 -CaF2 -P2 O5 systems. Since Ca2+ is typically present as a network modifier,
CNCa-O exhibited a sloping plateau, meaning that no stable structure was formed be-
tween Ca-O, which is consistent with previous studies on slag or glassy structures con-
taining CaO [10,28–30]. Additionally, it is worth noting that CNP-F had an extremely flat
plateau between 0 and 1, which means that F− and P5+ had a strong coordination tendency.
Wang et al. [31] introduced CaF2 into CaO-SiO2 -Al2 O3 slag systems and found that F−
has a strong tendency to replace an O in the [AlO4 ] structure to form an Al-F bond. They
attributed the phenomenon to the difference between the electronegativity of F− and O2− .
Therefore, the addition of CaF2 causes a shift in the original structure of the melts, especially
for phosphate systems. The tendency of F− to coordinate with P5+ is so strong that it may
form a competitive relationship with O2− , leading to a large-scale transformation of the
P-O structure. This trend may be more significant in high-temperature conditions. Besides,
the CN curves for all pairs of atoms except Si-O, P-O, and P-F did not have a clear plateau,
suggesting that they do not typically form stable structures, and they are therefore not
discussed in detail here.
Figure 2c,d show the alters of Si-O and P-O coordination numbers as CaF2 content in
slag from 0% to 20%, respectively. At 0% CaF2 content, the coordination numbers of Si-O
and P-O were close to 4.0, indicating that most of them exist as 4-coordinates and conform
to the tetrahedral form. From Figure 2c, the coordination number changes of Si-O were not
obvious in the range of the mole fraction of (CaF2 ) = 0~20%, which were all around 4.0. Due
to the high stability of [SiO4 ], it is difficult for F− to break through the bond energy barrier
between Si-O to coordinate with Si4+ , which was also discussed in previous studies [32,33].
However, when increasing the CaF2 content from 0% to 20%, the coordination plateau of
P-O became increasingly tortuous and the average coordination number decreased, with
values of 4.05, 3.87, 3.74, 3.66, and 3.57. Moreover, it can be seen from the variation law of
the bond length that P-O kept increasing, indicating that its stability decreased. In contrast,
the P-F bond length was much smaller than the P-O bond length. All indications show that
the affinity between P5+ and F− is greater than that of O2− , which confirms that CaF2 will
affect the coordination of P-O and alter the original [PO4 ] structure.
Figure 3 shows the coordination distributions for Si-O and P-O, with superscripts
indicating coordination numbers. The content of SiIV was always above 95%, which shows
that [SiO4 ] is the main structural unit in silicate systems and the content of [SiO4 ] did not
alter significantly with the increase of CaF2 content, which is consistent with the findings
of Fan et al. [34]. From Figure 3b, when the CaF2 content was 0, virtually all P-O in G1
appeared in a 4-coordination structure, indicating that the majority of P exists in slag
in the form of a [PO4 ] structure and serves as the basic structural unit of the phosphate
systems. However, as the CaF2 content increased, the PIV content decreased and the
PV gradually disappeared, while the PIII content continued to increase. Therefore, in
contrast to silicate systems, the structural units of phosphate systems absolutely change
with increasing CaF2 content, and new structures may emerge as the coordination number
of P-O gradually evolves from high to low. With the gradual increase of tri-coordinated P
content, combined with the coordination of P-F in Figure 2b, it indicated that the addition
of CaF2 prompted F− to replace O2− , and a [PO4 ] to [PO3 F] structural transition occurred.
A similar phenomenon also appeared in the study of the phosphate glass structure by
157
Materials 2022, 15, 7916
Rao et al. [35] and Touré et al. [36]. However, in Pedone et al.’s [14] work, no P-F/Cl bonds
were found at room temperature. It may be that the particles become more active and their
diffusion ability is enhanced at high temperatures compared with normal temperatures,
which provides favorable thermodynamic and kinetic conditions for the bonding between
P5+ and F− .
Figure 3. Coordination of Si-O and P-O under different CaF2 contents: (a) Si-O and (b) P-O.
4.2. Distribution of Bond Angles

The distribution of bond angles is also a critical parameter to characterize the structure
of the melt. Figure 4 shows the statistics of the O-Si-O and O-P-O bond angles’ information
with varying CaF2 content. CaF2 had a negligible effect on the distribution of O-Si-O and
O-P-O bond angles in the CaO-SiO2 -CaF2 -P2 O5 system. When CaF2 content increased
from 0% to 20%, the average bond angles of O-Si-O and O-P-O were 109.2◦ and 108.7◦ ,
respectively, which are very close to the theoretical value of 109.5◦ of standard tetrahedron.
It indicates that although F− tended to replace the position of an O2− in [PO4 ] tetrahedron,
it did not affect some structural characteristics of the original P-O bond, and the network
structure with Si4+ and P5+ as the core still maintained the tetrahedral structure. Moreover,
CaF2 did not appear to cause large-scale rearrangements of the atoms in the whole systems,
which consisted of a polymeric tetrahedral structure of Si4+ , P5+ , F− , and O2− , as well as
network modifiers such as Ca2+ dispersed.
Figure 4. Distribution of bond angles in the system of varying CaF2 content: (a) O-Si-O and (b) O-P-O.
4.3. Structural Unit Evolution

The silicate and phosphorene systems mainly consist of a network structure with
O atoms connected to Si and P atoms. There are three types of distinct oxygen, which
are divided into free oxygen (Of ), non-bridging oxygen (Onb ), and bridging oxygen (Ob ).
158
Materials 2022, 15, 7916
Additionally, a unique tri-coordinated oxygen structure has been found in the aluminate
system according to the literature [7]. Bridging oxygen with two tetrahedra, including Si-O-
Si, Si-O-P, and P-O-P, improved the degree of polymerization of the system. Non-bridged
oxygen was attached to only one tetrahedron, namely O-Si and O-P, while the other end
was attached to a metallic cation. They function in the opposite way to bridging oxygen.
The free oxygen is not connected to any tetrahedron. The cutoff radii of Si-O and P-O were
selected to be 2.3 and 2.5 Å, respectively, and the distribution of various oxygen types in
the melts was collected in Figure 5a. With the increase of CaF2 content, the amount of free
oxygen in the melts slightly increased, while the shift of the number of bridging oxygen and
non-bridging oxygen had no obvious rule and was approximately in dynamic equilibrium.
Figure 5. Structural unit evolution. (a) Types of oxygen atoms and (b) types of oxygen atoms after
subdivision. (c,d) Distribution of Qn with different CaF2 contents: (a) Qn Si and (b) Qn P.
Ob and Onb were further subdivided in the melts, as shown in Figure 5b. There
are three types of Ob : Si-O-Si, Si-O-P, and P-O-P. As the CaF2 content increased, Si-O-Si
decreased from 24.1% to 20.0%, Si-O-P increased from 13.0% to 16.7%, and P-O-P increased
from 0.7% to 4.8%. It has been shown that CaF2 is beneficial in disrupting Si-O-Si and
losing the initially polymeric silicate network structure. The Si-O-P structure in the system
increased; that is, the addition of CaF2 promoted the connection between [SiO4 ] and [PO4 ] or
[PO3 F], resulting in a silicophosphate composite structure that was more easily established
in the systems. Moreover, the increase of P-O-P also indicates that the connectivity of the
phosphate network structure became higher, which makes the phosphate melt structure
more complex.
To further quantitatively analyze the influence of CaF2 on the network structure of
the systems, Qn was introduced to characterize the polymerization degree of silicate and
phosphate systems respectively, where n represents the number of bridging oxygen (Ob ) in a
single tetrahedral unit. The current results show that Qn can be classified into five types: Q0 ,
Q1 , Q2 , Q3 , and Q4 , indicating that 0, 1, 2, 3, and 4 Ob are connected in a tetrahedral element.
Figure 5c,d show the distribution of Qn in silicate and phosphate systems, respectively. As
CaF2 content increased, the Q0 and Q1 in the silicate system increased, while the Q2 , Q3 ,
159
Materials 2022, 15, 7916
and Q4 decreased, again confirming that CaF2 breaks the high connectivity between [SiO4 ]
tetrahedral structures, simplifying the structure of the silicate systems. Besides, only Q0 and
Q1 structures originally existed in the phosphate system, indicating that [PO4 ] normally
exists in the form of a single tetrahedron or pairings, which is consistent with the research
results of Fan et al. [11]. However, as the CaF2 content increased, the Q0 rapidly decreased
and the Q1 increased. In addition, several Q2 and Q3 structures appeared and continued to
increase. The results indicate that the original phosphate structure was not complicated,
and the connectivity between the [PO4 ] tetrahedra was low. However, the addition of CaF2
reduced the number of single [PO4 ] tetrahedral elements and the current [PO3 F] structure
tended to form a chain or network composite structure, which increased the connectivity of
the phosphate network. Macroscopically, higher connectivity is beneficial to the enrichment
of P elements. In other words, CaF2 can enrich the phosphate network, which is favorable
for dephosphorization.
4.4. Transport Properties and Viscosity

The above results indicate that increasing the CaF2 content simplified the structure
of the silicate system in CaO-SiO2 -CaF2 -P2 O5 melts but complicated the structure of the
phosphate system. Therefore, to further understand the effect of CaF2 on the degree of
polymerization of the whole melt and to assess the changes in macroscopic properties,
it is necessary to quantitatively analyze the transport properties of the system. Liquid
molecules do not stay in a fixed position but are constantly moving [37]. The self-diffusion
coefficient is a momentous parameter that reflects the diffusivity of the particles in the
melt. As shown in Figure 6a, based on the MSD function and the Einstein relation, the
self-diffusion coefficients of distinct ions can be obtained. It can be seen that the order of
the self-diffusion coefficients of different ions was F− > Ca2+ > O2− ≈ P5+ > Si4− , and they
were all in direct proportion to the content of CaF2 , indicating that the addition of CaF2
can make each ion become more active. It mainly results from the depolymerization of the
network structure in the melt by CaF2 , which lowers the energy barrier for the migration of
ions in the melt and enhances the mobility of each particle. In addition, these phenomena
can also lead to changes in the macroscopic properties of the melts. It is worth noting that
the diffusion capacity of F− in the melts was most prominent and much larger than that
of O2− , indicating that the substitution of F− for O2− improved the overall mobility of
the phosphate structural units. Moreover, the diffusion coefficients of P5+ and O2− were
equivalent, which means that P and O always maintained the stable structure of [PO4 ] or
[PO3 F] and diffused cooperatively throughout the melts.
Figure 6. (a) The self-diffusion coefficient of each atom with different CaF2 contents. (b) Viscosity
comparison of MD simulation, models, and experimental measurement.
The melts’ viscosity was calculated from the self-diffusion coefficient of each ion and
compared with the experimental measurements. The results of the MD simulation and
experiment in Figure 6b both show that with the increase of CaF2 content, the viscosity of
160
Materials 2022, 15, 7916
the CaO-SiO2 -CaF2 -P2 O5 systems decreased and led to an improvement of melt liquidity.
Clearly, the viscosity, which reflects the viscous resistance of the melt during the flow and
depends prominently on the degree of polymerization of the melt, would be reduced in a
melt with simple structural units. Besides, the NPL model [38] and Pal model [39] were also
used to compare the calculation results. It can be observed that although there were some
errors between the calculated viscosity consequences and the experimentally measured
ones, the trends were in perfect agreement, which indicates that the MD simulations
were able to predict the viscosity of the system accurately to some extent and reflects the
reliability of the MD simulations. The predictions of both models differed significantly from
the experimental data due to discrepancies in some of the components. Consequently, the
MD viscosity calculations are in better agreement with the experimental results compared
to both models.
In the process of steel-making dephosphorization, P is usually enriched in 2CaO·SiO2 -
3CaO·P2 O5 (C2 S-C3 P) solid solution [40–42]. Dephosphorization depends on the concen-
tration of phosphorus in the solid solution, and the flow properties of the dephosphorized
product in the slag also determine whether phosphorus can be efficiently removed from
the slag. As can be seen from the above analysis, the introduction of CaF2 directly changed
the basic structural units of the phosphate melt, making the phosphate network units more
easily enriched. On the other hand, CaF2 reduced the viscosity and improved the fluidity of
the slag, so that the dephosphorized products enriched in P could be better transported to
and removed from the slag layer. CaF2 is thus favorable for dephosphorization both from
the microscopic reaction point of view in slag and from the macroscopic flow properties.
Our study links the microscopic to the macroscopic and essentially defines the critical role
of CaF2 in the dephosphorization of slag.
4.5. Correlation between Viscosity and Structural Properties

The viscosity of the slag depends on its degree of polymerization. Researchers have
proposed two common approaches to describe the complexity of melts. The first one
amounts to counting the number of non-bridged oxygen atoms, denoted as NBO, based
on the results of molecular dynamics simulations. The parameter NBO/T, which re-
flects the degree of melts’ polymerization, can be obtained by combining the number
of network formers, T (Si or P), in the system [43]. The larger the NBO/T, the higher
the ratio of non-bridging oxygen in the melts, that is, the simpler the structure of the
melts is, then the viscosity and other parameters of the melts will also change accordingly.
The second is to judge the melts’ complexity according to the evolution of total Qn . It
is usually expressed by the ratio of high-complexity Qn to low-complexity Qn , such as,
DOP = (Q3 + Q4 + Q5 )/(Q0 + Q1 + Q2 ). [44]. The higher the DOP, the more complex the
systems. Comparing the calculated melts’ viscosity with the above two parameters, we
observed a correspondence between the viscosity and the two parameters, as shown in
Figure 7.
It can be observed in Figure 7a that NBO/T increased with the increase of CaF2
content, while DOP was the opposite. The results show that CaF2 can effectively reduce the
complexity of the system, and the variation of both quantities has a good correspondence
with the trend of the viscosity value, suggesting that the melting viscosity is directly related
to the complexity of the system. Specifically, the introduction of CaF2 simplified some
complex network units formed by interweaving [SiO4 ], [PO4 ], and [PO3 F] structures in
the whole melt, and formed simple structures such as single or chain, greatly reducing the
connectivity of the whole melt. Furthermore, the complexity of the slag structure depends
on the competing effects of silicates and phosphates on the polymerization of the molten
slag. In the CaO-SiO2 -CaF2 -P2 O5 systems, CaF2 promoted the disaggregation of complex
network units into small units, which made the diffusion of micro-particles easier. The
macroscopic manifestation of this phenomenon is a reduction of the total viscosity.
161
Materials 2022, 15, 7916
Figure 7. (a) Change of structural parameters and viscosity and (b) the relationship between viscosity
and structural parameters.
In Figure 7b, the relationship between the viscosity and the above two parameters was
obtained by linear fitting. For viscosity and DOP, y = 0.5728x − 0.0167, R2 = 0.9821, and
for viscosity and NBO/T, y = −0.3130x + 1.1755, R2 = 0.9712. The correlation coefficients
of the above two fitting results were high enough, so the relationship between viscosity
and microstructure of CaO-SiO2 -CaF2 -P2 O5 melts could be accurately described, and at the
same time, the viscosity could also be predicted by the microstructure of the systems.
5. Conclusions
We have presented the microstructure information of the CaO-SiO2 -CaF2 -P2 O5 melts
at 1600 ◦ C by MD simulation and explored the evolution of each structural unit with
the increase of CaF2 content. Combined with the analysis of microscopic particle trans-
port and macroscopic flow properties, it is clear that the crucial role played by CaF2 in
phosphosilicate melts has been investigated.
By analyzing the distributional properties of the coordination and bond angles between
different atoms, we found that both S4+ and P5+ were present in tetrahedral form in the
molten CaO-SiO2 -CaF2 -P2 O5 system. The coordination number of Si-O was maintained
at around 4.0 when increasing the CaF2 content from 0% to 20%, while the coordination
number of P-O decreased from 4.05 to 3.57. Therefore, CaF2 had little effect on the structure
of [SiO4 ] but decreased the stability of the [PO4 ] structure. Specifically, F− tended to
replace O2− and promote the transformation of [PO4 ] to a [PO3 F] structure, and at the same
time, it is beneficial to make the P element-enriched phosphate network structure more
aggregated. However, the addition of CaF2 did not lead to a large-scale rearrangement
of the atoms in the whole system, and the network structure with Si4+ and P5+ as cores
remained tetrahedral.
The results of the MD simulation and experiment showed that CaF2 is beneficial for
reducing the degree of polymerization of the melt and thereby reducing the melt viscosity,
which decreased from 0.39 to 0.13 Pa·s as the CaF2 content increased from 0% to 20%,
and it had a good linear relationship with the structural parameters. In summary, CaF2
is beneficial for dephosphorization both from the microscopic reaction point of view and
from the macroscopic flow properties in slag.
Author Contributions: Formal analysis, methodology H.D.; funding acquisition, conceptualization

D.C.; methodology, data curation Z.H.; resources, supervision M.L.; software, writing—original draft,
Y.D. All authors have read and agreed to the published version of the manuscript.
Funding: The authors gratefully acknowledge the financial support provided by the National Natural
Science Foundation of China, project No. 51874060 and No. 52274320.
Institutional Review Board Statement: The study did not require ethical approval.
162
Materials 2022, 15, 7916
Informed Consent Statement: Informed consent was obtained from all subjects involved in the study.
References
1. Du, Y.; Yuan, Y.; Li, L.; Long, M.; Duan, H.; Chen, D. Insights into structure and properties of P2 O5 -based binary systems through
molecular dynamics simulations. J. Mol. Liq. 2021, 339, 116818. [CrossRef]
2. Liu, J.; Liu, Z.; Feng, J.; Li, B.; Chen, J.; Ren, B.; Jia, Y.; Yin, S. Reaction Mechanism of CA6 , Al2 O3 and CA6 -Al2 O3 Refractories
with Refining Slag. Materials 2022, 15, 6779. [CrossRef] [PubMed]
3. Lee, S.; Min, D.J. Investigation of sulfide capacity of aluminosilicate slag based on ionic structure considerations. J. Am. Ceram.
Soc. 2018, 101, 634–643. [CrossRef]
4. Yang, J.; Kim, Y.; Sohn, I. Gaining insights on high-temperature thermal conductivity and structure of oxide melts through
experimental and molecular dynamics simulation study. J. Mater. Res. Technol. 2021, 10, 268–281. [CrossRef]
5. Yao, Y.; Wang, Y.; Wei, Q.; Cui, S.; Hao, L. Effect of the Formation of Amorphous Networks on the Structure and Hydration
Characteristics of Granulated Blast Furnace Slag. Materials 2020, 13, 1462. [CrossRef]
6. Stoch, P.; Goj, P.; Wajda, A.; Stoch, A. Alternative insight into aluminium-phosphate glass network from ab initio molecular
dynamics simulations. Ceram. Int. 2021, 47, 1891–1902. [CrossRef]
7. Bi, Z.; Li, K.; Jiang, C.; Zhang, J.; Ma, S.; Sun, M.; Wang, Z.; Li, H. Performance and transition mechanism from acidity to basicity
of amphoteric oxides (Al2 O3 and B2 O3 ) in SiO2 –CaO–Al2 O3 –B2 O3 system: A molecular dynamics study. Ceram. Int. 2021, 47,
12252–12260. [CrossRef]
8. Li, K.; Khanna, R.; Bouhadja, M.; Zhang, J.; Liu, Z.; Su, B.; Yang, T.; Sahajwalla, V.; Singh, C.V.; Barati, M. A molecular dynamic
simulation on the factors influencing the fluidity of molten coke ash during alkalization with K2 O and Na2 O. Chem. Eng. J. 2017,
313, 1184–1193. [CrossRef]
9. Diao, J.; Fan, G.; Liu, X.; Xie, B. Computer simulation of anionic structures of molten CaO-SiO2 -P2 O5 system. Metall. Mater. Trans.
B 2014, 45, 1942–1947. [CrossRef]
10. Wang, Z.; Cai, S.; Zhang, M.; Guo, M.; Zhang, Z. Structural investigation of phosphorus in CaO-SiO2 -P2 O5 ternary glass. Metall.
Mater. Trans. B 2017, 48, 1139–1148. [CrossRef]
11. Fan, G.; Diao, J.; Jiang, L.; Zhang, Z.; Xie, B. Molecular dynamics analysis of the microstructure of the CaO-P2 O5 -SiO2 slag system
with varying P2 O5 /SiO2 ratios. Mater. Trans. 2015, 56, 655–660. [CrossRef]
12. Diao, J.; Ke, Z.; Jiang, L.; Zhang, Z.; Zhang, T.; Xie, B. Structural Properties of Molten CaO–SiO2 –P2 O5 –FeO System. High Temp.
Mater. Process. 2017, 36, 871–876. [CrossRef]
13. Kansal, I.; Goel, A.; Tulyaganov, D.U.; Rajagopal, R.R.; Ferreira, J.M. Structural and thermal characterization of CaO–MgO–SiO2 –
P2 O5 –CaF2 glasses. J. Eur. Ceram. Soc. 2012, 32, 2739–2746. [CrossRef]
14. Pedone, A.; Chen, X.; Hill, R.G.; Karpukhina, N. Molecular dynamics investigation of halide-containing phospho-silicate bioactive
glasses. J. Phys. Chem. B 2018, 122, 2940–2948. [CrossRef]
15. Diao, J.; Xie, B.; Wang, Y.; Guo, X. Effect of fluorine on the minerals phase and citric acid solubility of CaO–SiO2 –FetO–P2 O5 –CaF2
system. ISIJ Int. 2010, 50, 768–770. [CrossRef]
16. Ueda, S.; Watanabe, Y.; Ikeda, T.; Maeda, M. Phase equilibria in the CaF2 -CaO-P2 O5 system at 1623 K. Metall. Mater. Trans. B
Process Metall. Mater. Process. Sci. 2001, 32, 967–969. [CrossRef]
17. Belashchenko, D.; Ostrovskii, O. Computer Simulation of Noncrystalline Ionic–Covalent Oxides CaO–P2 O. Inorg. Mater. 2002, 38,
18. Hirao, K.; Kawamura, K. Material design using personal computer. Shokabo Tokyo 1994, 52, 52–54.
19. Burylev, B.P.; Laptev, D.M.; Moisov, L.P. Calculation of the viscosity of slag melts. Svarochnoe Proizv. 1999, 10, 8–11.
20. Ziman, J.M. Principles of the Theory of Solids; Cambridge University Press: Cambridge, UK, 1972.
21. Kubo, R. The fluctuation-dissipation theorem. Rep. Prog. Phys. 1966, 29, 255. [CrossRef]
22. Shimizu, N.; Kushiro, I. Diffusivity of oxygen in jadeite and diopside melts at high pressures. Geochim. Cosmochim. Acta 1984, 48,
1295–1303. [CrossRef]
23. Mungall, J.E. Empirical models relating viscosity and tracer diffusion in magmatic silicate melts. Geochim. Cosmochim. Acta 2002,
66, 125–143. [CrossRef]
24. Tinker, D.; Lesher, C.E.; Baxter, G.M.; Uchida, T.; Wang, Y. High-pressure viscometry of polymerized silicate melts and limitations
of the Eyring equation. Am. Mineral. 2004, 89, 1701–1708. [CrossRef]
25. Reid, J.E.; Poe, B.T.; Rubie, D.C.; Zotov, N.; Wiedenbeck, M. The self-diffusion of silicon and oxygen in diopside (CaMgSi2 O6 )
liquid up to 15 GPa. Chem. Geol. 2001, 174, 77–86. [CrossRef]
26. Lv, X.; Xu, Z.; Li, J.; Chen, J.; Liu, Q. Molecular dynamics investigation on structural and transport properties of Na3 AlF6 –Al2 O3
molten salt. J. Mol. Liq. 2016, 221, 26–32. [CrossRef]
27. Goj, P.; Stoch, P. Influence of CaO on structural features of polyphosphate P2 O5 -Fe2 O3 -FeO glasses by molecular dynamics
simulations. J. Non-Cryst. Solids 2020, 537, 120014. [CrossRef]
163
Materials 2022, 15, 7916
28. Sasaki, Y.; Iguchi, M.; Hino, M. The role of Ca and Na ions in the effect of F ion on silicate polymerization in molten silicate
system. ISIJ Int. 2007, 47, 638–642. [CrossRef]
29. Christie, J.K.; Ainsworth, R.I.; de Leeuw, N.H. Ab initio molecular dynamics simulations of structural changes associated with
the incorporation of fluorine in bioactive phosphate glasses. Biomaterials 2014, 35, 6164–6171. [CrossRef]
30. Park, J.H.; Min, D.J.; Song, H.S. The effect of CaF2 on the viscosities and structures of CaO-SiO2 (-MgO)-CaF2 slags. Metall. Mater.
Trans. B 2002, 33, 723–729. [CrossRef]
31. Wang, Z.; Tang, P.; Wen, G.; Liu, Q. Effect of F− Replacing O2− on Crystallization Behavior of CaO–SiO2 –Al2 O3 Continuous
Casting Mold Flux. ISIJ Int. 2019, 59, 367–374. [CrossRef]
32. Takeda, O.; Okawara, T.; Sato, Y. Development of a new viscometer based on rotating crucible method and viscosity measurement
of SiO2 –CaO–CaF2 system. ISIJ Int. 2012, 52, 1544–1549. [CrossRef]
33. Sasaki, Y.; Iguchi, M. The Coordination of F Ions around Mg and Ca Ions in Molten CaO–CaF2 –MgO–SiO2 System at 1873K. ISIJ
Int. 2009, 49, 602–604. [CrossRef]
34. He, S.; Wang, S.; Jia, B.; Li, M.; Wang, Q.; Wang, Q. Molecular dynamics simulation of the structure and properties of CaO-SiO2 -
CaF2 slag systems. Metall. Mater. Trans. B 2019, 50, 1503–1513. [CrossRef]
35. Venkateswara Rao, G.; Shashikala, H. Structural, optical and mechanical properties of ternary CaO-CaF2 -P2 O5 glasses. J. Adv.
Ceram. 2014, 3, 109–116. [CrossRef]
36. Touré, A.B.; Mele, E.; Christie, J.K. Atomic-scale clustering inhibits the bioactivity of fluoridated phosphate glasses. Biomed. Glas.
2019, 5, 76–84. [CrossRef]
37. Jiang, C.; Zhang, H.; Xiong, Z.; Chen, S.; Li, K.; Zhang, J.; Liang, W.; Sun, M.; Wang, Z.; Wang, L. Molecular dynamics
investigations on the effect of Na2 O on the structure and properties of blast furnace slag under different basicity conditions. J.
Mol. Liq. 2020, 299, 112195. [CrossRef]
38. Mills, K.; Sridhar, S. Viscosities of ironmaking and steelmaking slags. Ironmak. Steelmak. 1999, 26, 262–268. [CrossRef]
39. Pal, S.K.; Ray, H.S.; Choudhury, P.; Das, S.; Majumdar, M. An approach towards theoretical prediction of viscosity of glasses over
a wide range of temperature. Trans. Indian Ceram. Soc. 2005, 64, 81–86. [CrossRef]
40. Song, Y.; Hu, X.; Chou, K. The Reaction Behavior of 2CaO· SiO2 with CaO–SiO2 –FeO–P2 O5 Slag. Materials 2022, 15, 6594.
[CrossRef]
41. Xia, Y.; Li, J.; Fan, D.; Hou, G. Effects of interfacial oxygen potential and slag phase changing during slag formation process on
dephosphorization behavior. ISIJ Int. 2019, 59, 1519–1526. [CrossRef]
42. Suito, H.; Inoue, R. Behavior of phosphorous transfer from CaO–FetO–P2 O5 (–SiO2 ) slag to CaO particles. ISIJ Int. 2006, 46,
43. Wu, T.; He, S.; Liang, Y.; Wang, Q. Molecular dynamics simulation of the structure and properties for the CaO–SiO2 and
CaO–Al2 O3 systems. J. Non-Cryst. Solids 2015, 411, 145–151. [CrossRef]
44. Jiang, C.; Li, K.; Zhang, J.; Qin, Q.; Liu, Z.; Liang, W.; Sun, M.; Wang, Z. Molecular dynamics simulation on the effect of
MgO/Al2 O3 ratio on structure and properties of blast furnace slag under different basicity conditions. Metall. Mater. Trans. B
2019, 50, 367–375. [CrossRef]
164
materials
Article
Characterization of Hot Deformation of near Alpha Titanium
Alloy Prepared by TiH2-Based Powder Metallurgy
Rongxun Piao 1 , Wenjin Zhu 1 , Lan Ma 2 , Peng Zhao 3, * and Biao Hu 4
1 School of Mechanical Engineering, Anhui University of Science and Technology, Huainan 232001, China
2 Department of Vanadium and Titanium, Panzhihua University, Panzhihua 617000, China
3 School of Materials Science and Engineering, Shandong Jianzhu University, Jinan 250101, China
4 Anhui International Joint Research Center for Nano Carbon-Based Materials and Environmental Health,
Huainan 232001, China
Abstract: TiH2 -basd powder metallurgy (PM) is one of the effective ways to prepared high tempera-
ture titanium alloy. To study the thermomechanical behavior of near-α titanium alloy and proper
design of hot forming, isothermal compression test of TiH2 -based PM near-α type Ti-5.05Al-3.69Zr-
1.96Sn-0.32Mo-0.29Si (Ti-1100) alloy was performed at temperatures of 1123–1323 K, strain rates
of 0.01-1 s−1 , and maximum deformation degree of 60%. The hot deformation characteristics of
alloy were analyzed by strain hardening exponent (n), strain rate sensitivity (m), and processing
map, along with microstructure observation. The flow stress revealed that the difference in soften-
ing/hardening behavior at temperature of 1273–1323 K and the strain rate of 1 s−1 compared to the
lower deformation temperature and strain rate. The strain hardening exponents at temperatures
of 1123 K are all negative under all strain rates, and the most severe flow softening with minimum
value of n was observed at 1123 K and 1 s−1 . The strain rate sensitives showed that the peak region
with m value greater than 0.5 generally appeared in the high temperature range of 1273–1323 K,
while strain rate sensitivity at low temperature behaved differently with strain rates. The processing
Citation: Piao, R.; Zhu, W.; Ma, L.; map developed for strain of 0.6 exhibited high power dissipation efficiency at high temperatures
Zhao, P.; Hu, B. Characterization of of 1273–1323 K and a low strain rate of 0.01 s−1 , due to microstructure evolution of β phase. The
Hot Deformation of near Alpha decrease of strain rate at 1323 K resulted in the formation of globularization of α lamellae. The
Titanium Alloy Prepared by instability domain of flow behavior was identified in the temperature range of 1123–1173 K and at
TiH2 -Based Powder Metallurgy.
the strain rate higher than 0.01 s−1 reflecting the localized plastic flow and adiabatic shear banding,
Materials 2022, 15, 5932. https://
and inhomogenous microstructure. The variation of power dissipation energy (η) slope with strain
doi.org/10.3390/ma15175932
demonstrated that the power dissipation mechanism during hot deformation has been changed
Academic Editor: Jae Wung Bae from temperature-dependent to microstructure-dependent with the increase of temperature for the
alloy deformed at 0.1 s−1 . Eventually, the optimum processing range to deform the material is at
Received: 23 July 2022 .
1273–1323 K and a strain rate range of 0.01–0.165 s−1 (lnε = −4.6–−1.8).
Published: 27 August 2022
Keywords: near-α titanium alloy; TiH2 -based powder metallurgy; hot compression; strain hardening
exponents; strain rate sensitivity; processing map
iations.
1. Introduction
Near-α titanium alloys are widely used in compressor discs and blades of jet engines
Copyright: © 2022 by the authors. because of their excellent high-temperature fatigue and creep properties, high strength-
Licensee MDPI, Basel, Switzerland. to-weight ratio, and good corrosion resistance [1]. Ti-1100 (Ti-6Al-2.75Sn-4Zr-0.4Mo-0.4Si,
This article is an open access article wt%) alloy has the highest capability to tolerate temperature up to 873 K without any
distributed under the terms and degradation in mechanical properties among near-α titanium alloys [2]. Like most titanium
conditions of the Creative Commons alloys, near-α titanium alloy has poor formability and high cost, which has always been
Attribution (CC BY) license (https:// two key problems in a wide range of commercial applications [3]. Powder metallurgy (PM)
is one of the most promising methods for production of titanium alloy components at low
4.0/).

Materials 2022, 15, 5932
cost [4]. However, compared with ingot metallurgy (IM) titanium alloys, challenge of PM
titanium alloy are high porosity and high oxygen contamination, which usually lead to
detrimental effect on the final product by lowering ductility or other defects [5].
One of the effective ways to solve the two problems of insufficiently high density
and oxygen contamination is to use TiH2 powder as the main raw material. It has
been demonstrated previously that PM parts developed using TiH2 powder feedstock
had better control over lower oxygen content, the microstructures, and good chemical
homogeneity [6–8]. Using TiH2 powder, density larger than 95% and good performance
with fine grain microstructure can be easily obtained [6,8,9]. Previously, Hagiwara et al. [10]
have prepared Ti-1100 alloy with ultra low chlorine hydride-dehydride (ELCL-HDH) ti-
tanium powder and master alloy powder as main raw materials through the means of
blended elemental (BE) PM synthesis of cold pressing-vacuum sintering-heat treatment-hot
isostatic pressing, and the prepared alloys have high density of 95% and better perfor-
mance in tensile strength and high cycle fatigue strength compared to the conventional
PM process. Zhang et al. [3] investigated the mechanical behavior of TiH2 -based near-α
Ti–3Al–2Zr–2Mo titanium alloy prepared by a series process of cold pressing-sintering-
hot extrusion-vacuum annealing-common annealing, and the prepared PM alloy samples
showed a satisfying combination of superior tensile strength and excellent ductility. Using
similar methods, recently, Zhang et al. [11] and Wu et al. [12] in the same research group
studied the tensile properties of a hot-extruded TiH2 -based near-α Ti-3Al-2Zr-2Mo-0.36O
and Ti-6Al-2Sn-titanium alloy, and the prepared material has shown high performance of
tensile ductility.
For the proper design of thermal deformation of metal, it is necessary to study the
thermomechanical behavior. The mechanism of flow stress softening as well as strain hard-
ening phenomena determines the thermomechanical behavior of a material, depending on
processing parameters such as strain, strain rate, and temperature. For the description of the
hot deformation behavior of titanium alloys, the dynamic material model (DMM) [12,13]
based processing map technology has been widely used to provide the optimization of
processing parameters and processing windows with instability regions [14]. Likewise,
the strain hardening, strain rate sensitivity, and temperature sensitivity of the alloys dur-
ing deformation can provide the reference and guidance for better thermal deformation
of the alloy. Previously, many studies have focused on the hot deformation of as-cast
near-α titanium alloys. For example, Krishna et al. [15] investigated the hot deformation
mechanism in near α titanium alloy 685 by constructing a processing map, obtaining the
optimal condition of 1248 K and 0.001 s−1 , and proposing that dynamic recrystalization
(DRX) occurred due to low oxygen level present in the alloy. Balasundar et al. [16] dis-
cussed geometric dynamic recrystallization behavior of a near-α titanium alloy TTTAN
29A (equivalent to IMI834) with acicular microstructure through hot compression tests at
temperatures of 1123–1333 K and a strain rate of 3 × 10−4 –1 s−1 , and proposed the optimum
condition is at 1193–1303 K along with strain rates of 3 × 10−4 –10−3 s−1 and dynamic
recrystallization (DRX) or globularization of lamellar α phase is dominant mechanism of
microstructure evolution. Zhou et al. [17] studied hot workability of near-α titanium alloy
Ti–6Al–3Nb–2Zr–1Mo with an initial duplex microstructure by means of isothermal com-
pressions. By integrating process maps and constitutive relationship, they proposed that
the optimal hot working domain appeared at condition of 1198-1248 K/0.01–0.1 s−1 , and
the main mechanism of microstructure evolution in α + β phase field is dynamic recrystal-
lization of β phase and super-plasticity deformation (SPD). Su et al. [18] have study the hot
deformation behavior of near-α DsTi700 titanium alloy and revealed that the deformation
mechanisms are mainly kinking of α lamellas or flow localization at high strain rate of
0.5–1 s−1 in α + β two phase field, while DRV is the dominated deformation mechanism
in β single phase. Morakabati and Hajari [19] studied the high temperature deformation
behavior of the Ti−5.7Al−2.1Sn−3.9Zr−2Mo−0.1Si (Ti-6242S) alloy with an initial acicular
microstructure by using processing map and revealed that the dynamic globularization
occurred in lower temperature conditions of α/β two-phase region (T ≤ 1223 K) and DRV
166
Materials 2022, 15, 5932
was observed at high temperature (T = 1273 K). All above studies have been focused on as-
cast alloy. However, the study of hot deformation behavior for PM alloy is rarely reported
at present. Ding et al. [20] investigated the PM Ti600 alloy prepared by a HDH titanium
powder-based PM-extrusion-annealing method and revealed that the microstructure was
predominant by equiaxed grains after annealing below β transition temperature, while the
bi-model structure was dominant after annealing at 1323 K in the β single phase. In our
previous study [21], hot deformation of TiH2 -based PM Ti-1100 at a relatively low temper-
ature range of 973–1173 K and strain rate of 0.01–1 s−1 was investigated. The results of
strain rate sensitivity and temperature sensitivity analysis showed that the flow instability
mainly occurred at strain rates of 0.01 and 1 s−1 under all temperature range and the safe
working window was limited at a narrow zone of 1073–1173 K with strain rate of 0.1 s−1 .
This paper mainly focused on the hot deformation behavior of TiH2 -based PM near-α
titanium alloy Ti-1100 in the upper α or α/β two phase region of 1123–1323 K and strain
rates of 0.01–1.0 s−1 . For the characterization of hot deformation behavior of Ti-1100 alloy,
the strain hardening exponents, strain rate sensitivity, microstructure, and processing map
were investigated.

Titanium hydride powder (Ti-4.5H-0.12O-0.015C-0.023N-0.016Fe in mass%) with
a particle size of 45 μm and high grade of Al, Sn, Zr, Mo, and Si powders with a par-
ticle size of 75 μm were well mixed by the ball-mill machine. The mixture was subjected
to cold isostatic pressing (CIP) at molding pressure of 200 MPa and holding time of
180 s to obtain the cylindrical lump alloy, and then the vacuum sintering was conducted at
1423 K for 240 min. under Ar atmosphere of 80~120 Pa. Finally, the CIP sintered lump was
subjected to hot isostatic pressing (HIP) at 1223 K and 200 MPa for 180 min.
In the hot compression test, a cylindrical sample with dimensions of ϕ 8 mm × 12 mm
(in accordance to ASTM E209, preserving the aspect ratio) was machined along the axis
of the HIP cylinder with a wire-electrode cutting machine. The hot compression tests
of 15 samples were conducted by THERMECMASTER–Zthermos–simulation machine at
deformation temperatures ranging from 1123 K to 1323 K with an interval of 50 K and
strain rates of 0.01, 0.1, and 1 s−1 . The whole isothermal compression test process was
carried out under a vacuum degree lower than 10 Pa to prevent the samples oxidation. The
specimens were heated to the target temperature at a rate of 10 K/s and held for 180 s. The
fixed height of each sample was reduced by 60%, and the samples were quickly cooled
to room temperature in order to retain high temperature deformation microstructure. To
minimize friction, 0.1 mm thick mica wafer was placed at both end faces between the speci-
mens and die. The true stress–strain curves were acquired automatically by a computer-
equipped monitor.
The isothermal compression specimen was cut along the radial direction, and the
cutting section of specimen was grinded with SiC paper followed by mechanical polishing
with a solution of water and diamond polishing paste. Then, the specimens were subjected
to the etching in Kroll’s reagent (1 mL HF, 5 mL HNO3 , and 100 mL water). Lastly, the
microstructures of the central area of the cut specimens were characterized by optical
microscopy (OM, DMM-490C, Caikang Optical Instrument Co., LTD, Shanghai, China)
and scanning electron microscope (SEM, TESCAN-MIRA 4, Tescan Co., LTD, Shanghai,
China branch).
The microstructure of the as-sintered alloys is shown in Figure 1. It shows a typical
two-phase microstructure, mainly consisting of lath α phase with a width of about 20 μm,
and β phase in the α phase boundary. From preliminary density measurements, the density
of as-sintered alloy was determined to be 97.1% of the theoretical density, and the chemical
composition of sintered alloy is shown in Table 1. The β-transus temperature of the alloy
was preliminarily determined as 1293 K by differential scanning calorimeter (DSC), and
the details of DSC analysis can be found in Appendix A.
167
Materials 2022, 15, 5932
Figure 1. Microstructure of as-prepared Ti-1100 alloy.
Table 1. Chemical composition of sintered near-α Ti-1100 alloy (wt%).
Ti Al Zr Sn Mo Si C N H O
88.3 5.05 3.689 1.96 0.32 0.29 0.0071 0.0013 0.0016 0.12

3.1. True Stress–Strain Curves
3.1.1. General Behavior
Shown in Figure 2 is the macro-appearance of deformed samples at different processing
conditions. It can be seen from the top and side views in Figure 2 that severe cracks were
observed in the specimens deformed at 1123 K/0.1 s−1 and 1173 K/1 s−1 , indicating that
the sample under these conditions may not be suitable for thermal deformation. Likewise,
slight cracks were observed on the side of the drum-shaped samples deformed at high
temperature of 1273–1323 K and strain rate of 0.01 s−1 . It should be noted that slight cracks
on the side of the specimen are usually caused by compressive stress and additional tensile
stress, which is different from the deformation behavior of the central part of the specimen
subjected to a single pressure, so the hot workability in this case cannot be judged by these
macroscopic appearances alone [22]. Under other experimental conditions, there were no
obvious macro-cracks and micro-cracks, and the deformation of the sample is uniform.
Figure 2. Macro-appearance (top and side views) of deformed alloys at different processing conditions.
168
Materials 2022, 15, 5932
Based on the hot compression experimental data of Ti-1100 alloy, the flow stress–strain
curve was obtained, as shown in Figure 3. As seen at the beginning of the hot deformation
process, the flow stress increases rapidly to the peak stress (σP ). This hardening effect is
mainly attributed to the dislocation proliferation and intersection [23].
(a) (b)
(c)
Figure 3. The true strain–stress curves of Ti-1100 alloy at different deformation conditions:
(a) 0.01 s−1 , (b) 0.1 s−1 , and (c) 1 s−1 .
Beyond the peak stress, the flow stress exhibits the different behaviors. At a strain rate
of 0.01 s−1 , the flow stress exhibits softening response under all deformation temperatures,
and this softening effect is more obvious at low temperature range. At a strain rate of
0.1 s−1 , the flow stress exhibits the general behaviors of softening response in the low
part of α/β region (T ≤ 1223 K). By comparison, the flow curves reached to steady-state
condition at upper part of α/β region (T = 1323 K). This behavior can be explained by the
fact that at a high temperature of 1323 K, the volume fraction of β-BCC phase with low
mechanical property and high diffusivity is increased [24]. As the strain rate increases to
1 s−1 , the flow stress curves show increasing tendency in the high temperature range of
1273–1323 K, implying work hardening effect was dominant at this condition. Obviously,
169
Materials 2022, 15, 5932
these hardening effects are more significant at high strain rate (1 s−1 ) compared to that
at 1273 K and 0.1 s−1 . These behaviors indicated that dynamic softening by temperature
and the work-hardening by strain rate interact each other, resulting the different behaviors
under the different temperature and different strain rate conditions. As consequence, it
shows the softening effects at low strain rate of 0.01 s−1 , while both softening and hardening
effects were observed at relatively high strain rate of 0.1–1 s−1 depending on temperature
and strain rate.
3.1.2. Effect of Temperature on Flow Stress

The true stress of the alloy is sensitive to the deformation temperature and strain rate.
Figure 4 shows a typical example of temperature dependence on flow stress at a strain of
0.1 under the different strain rates. As seen in Figure 4, at constant strain rate, the true flow
stress decreases continuously with the increase of temperature. Obviously, the decreasing
rate of the flow stress decreases with increasing temperature, and this behavior can be
attributed to the increasing the volume fraction of β phase with higher diffusivity and high
stacking fault energy against to α phase [20,24]. Another possible cause of temperature
dependence of flow stress is the high kinetics of β to α transformation. Thus, as β phase
has lower strength and higher active slip system than α phase [25], the true flow stress
decreases by an increase in deformation temperature. Similar behavior can be found in the
deformation of IMI834 alloy [26] and Ti-6242S alloy [19].
Figure 4. Temperature dependence of the flow stress at a strain of 0.1 under various strain rates.
3.2. Strain Hardening Exponents

To analyze the characteristics of hot work hardening of alloy, the concept of strain
hardening was introduced based on the following functional model [27]:
σ = σ0 + Kεn (1)
where σ0 is the yield strength (MPa); K is the strength coefficient; ε is the strain; n is the
strain hardening exponent, which can be express by Equation (1) [28]:

∂(lgσ)
n= (2)
∂(lgε) ε,. T
170
Materials 2022, 15, 5932
The parameter n reflects the comprehensive effect of strain hardening and flow stress
softening process [28,29]. It can be known from Equation (2) that the negative value of n
represents the trend of stress softening, while the positive value of n represents the work
hardening behavior. By the data processing and substitution into Equation (2), the strain
hardening exponent n was obtained. Figure 5 shows the influence of temperature on the
strain hardening exponent n under different strains. At a relatively low strain rate of
0.01 s−1 , the values of n under a strain of 0.1 shown in Figure 5a fluctuates near the balance
state line of n = 0 with temperature. However, the value of n at a strain rate of 0.01 s−1 are
mostly negative as the strain rate increases to 0.3 and 0.5, and it becomes more negative
at higher temperature. When the strain rate is 0.1 s−1 and 1 s−1 , the values of n shown
in Figure 5b,c are all negative in the temperature range of 1123–1223 K, and it becomes
positive at temperature above 1223 K, meaning that the softening effect has been changed
to the hardening effect near at 1223 K. Where the temperature is 1323 K and strain rate is
1 s−1 , the value of n increases from the minimum value of −0.8 to its maximum value of
0.8. Since the strain hardening exponent n is a parameter of comprehensive effects of strain
hardening and flow stress caused by thermal effects, the increase of the n value is basically
caused by the decreases of softening effect as the deformation temperature increases [28].
Similar behavior of the increasing tendency with the temperature was reported at SP700
superplastic titanium alloy under the high strain rate of 0.1 s−1 [30].
(a) (b)
(c)
Figure 5. Influence of deformation temperature on the strain hardening exponent n of Ti-1100

titanium alloy under different strains: (a) ε = 0.1, (b) ε = 0.3, and (c) ε = 0.5.
171
Materials 2022, 15, 5932
Figure 6 shows the variation of the strain hardening exponent of the alloy with the
change of strain under the different strain rates. It can be found that at temperatures of
1123 K under all strain rates, the curve of n value generally decreases during strain of 0.6
and approaches to stable state, indicating that the flow softening was gradually increased
due to the temperature rise effect, and then decreased as temperature rise effect disappears
at the high strain [30]. Moreover, under all strain conditions, the value of n decreases
more negatively with the increase of strain rate. Similarly, at temperature of 1173 K, the
n-value curve under all strain conditions moves to a more negative region with the increase
of strain rate. As seen in Figure 6a at the temperature of 1273 and 1323 K, the thermal
softening effect is more significant when the strain rate is 0.01 s−1 . However, when the
strain rate increases to 0.1 s−1 , the strain hardening effect is dominant, and it becomes more
significant as the strain rate further increases to 1 s−1 as shown in Figure 6c. It was revealed
from Figure 6b,c that, the higher temperature the more prone to strain hardening and the
lower temperature the more prone to thermal softening. At the temperature of 1223 K, the
value of n is negative under all strain conditions and strain rates, and slightly increases
with increase of strain rate. No significant change of n value with strain was observed at
this temperature, showing better uniform plastic deformation ability.
(a) (b)
(c)
Figure 6. Variation of the strain hardening exponent as a function of strain under different strain
. . .
rates: (a) ε = 0.01 s−1 , (b) ε = 0.1 s−1 , and (c) ε = 1 s−1 .
172
Materials 2022, 15, 5932
It is noted that the softening effect in term of negative n-value only considers the
temperature rise effect, but not the microstructural evolution related softening effects such
as dynamic recrystallization (DRX) or dynamic recovery (DRV). In general, the softening
caused by DRX or DRV is favorable to hot deformation. However, the softening caused
by temperature rise is not conducive to hot deformation. It is generally known that the
temperature rise during alloy deformation will generate deformation heat, leading to
instability phenomena such as local flow or adiabatic shear [31]. In this work, the most
severe flow softening with minimum value of n occurred at low temperature of 1123 K
with a high strain rate of 1 s−1 , so the deformation in this condition may be dangerous and
should be avoided. In addition, obvious flow softening effect were also observed at high
temperature of 1323 K with low strain rate of 0.01 s−1 . This is mainly due to the formation
of the β phase as discussed in detail in the following section.
3.3. Microstructure Observation

As seen in Figure 3, at low temperature range of α/β region (T ≤ 1273K), the strain–
stress curves show a peak stress at relatively low strain followed by flow softening, while
flow curves show steady-state behavior at high temperature (T = 1323 K). In single-phase
alloys, these two behaviors are often attributed to occurrence of DRV, i.e., the occurrence
of dislocation climb and DRX, i.e., the nucleation and growth of new grains, respectively.
However, different microstructural evolution is known to be responsible for the observed
flow softening in two-phase Ti-alloys [32].
The microstructures of Ti-1100 alloy compressed at 1123 K, 1273 K, and 1323 K un-
der the strain rates of 1~0.01 s−1 are presented in Figure 7(a1–a3), (b1–b3), and (c1–c3),
respectively. According to Figure 7, the evolution of randomly distributed α lamellar phase
with a length of 10~20 μm and a thickness less than 10 μm, during isothermal deforma-
tion, was a prevailing microstructure feature at deformation temperature below 1273 K,
while at 1323 K, the various form of transformed β phase with non-lamellar α structure
is its main feature, mainly due to an increase in volume fraction of β-BCC phase. During
deformation at 1123–1273 K, the α colony, characterized by the fine parallel array structure,
was observed, and tends to rotate perpendicular to the compressive axis. This, α colony
formation is caused by local overheating. Due to the poor thermal conductivity of Ti-1100
alloy, the distortion energy caused by deformation cannot be released in the form of heat
immediately, which may further lead to local overheating and the formation of fine parallel
structure [33]. There is no specific grain size change with the change of strain rate. In
addition, the bending/kinked lamella and sub-grain formation are the general character-
istics of microstructural evolution. It is generally known that the thick initial lamellar
microstructure undergoes buckling and elongates, while thin microstructure is more eas-
ily fragmented, so the bending/kinked lamella are possibly formed under the different
deformation conditions. The sub-grain formation may be caused by boundary-splitting
mechanism through intense localized shear during hot deformation [20]. Moreover, the
globular shape or equiaxed grains with a size less than 5 μm are usually observed near the
α lath boundary, indicating that dynamic recrystallization (DRX) of α phase occurs. This is
because the fractured lamellar grains during hot deformation promote the formation of
refined equiaxed microstructure through recrystallization of α grains. [4].
Differently, as observed in Figure 7(c1–c3), the microstructure of the alloy deformed at
1323 K has the characteristics of transformed β phase with different structures. As seen, the
microstructure of the specimen deformed at the strain rate of 1 s−1 contained transformed
β phase with basketweave (widmanstätten) structure, and discontinuous α layers at the
transformed β grain boundary, and also within grain. This morphorlogy of the transformed
microstructure is usually formed near β-transus temperature [34]. Clearly, compared to
lower temperature, the volume of β phase at this condition was increased and the boundary
of α phase becomes more irregular. As the strain rate decreases to 0.1 s−1 , the proportion of
transformed β increased and the α/β grain boundary became blurred. When the strain rate
is further decreased to 0.01 s−1 , the enrichment of transformed β near the α phase boundary
173
Materials 2022, 15, 5932
was observed. In addition, the globularized α phase was also observed. In this case, the
globularization of α phase is mainly controlled by termination migration mechanism. It
is believed that the β phase diffuses into the boundary as an interlayer phase, so it can
reduce the dihedral angle and further leads to grooves in the α-layers. In other words, the
mass transfer from the curved surfaces of the lamellar grain terminations to the vicinal flat
surface of the lamellae drives the termination migration [35]. More clear observations and
phase identifications of this condition can be seen from the SEM micrographs in Figure 8.
Figure 8a,c shows the secondary electron image (SE) of Ti-1100 alloy deformed at
1323 K/ 0.01 s−1 and 1323 K/1 s−1 , respectively, in which both α and β phases can be easily
recognized. Figure 8b presents back-scattering electron (BSE) image of alloy deformed at
1323 K/1 s−1 . As shown by the white arrow in Figure 8b, due to the shear strain applied
during deformation, there exists sub-boundary in “A” layer. As an unstable interface, this
sub-boundary provides favorable conditions for β phase penetration. As a consequence,
the α lamella is fractured and the α/β interface is formed. Similar observations have been
reported previously for globularization of Ti-6242S alloy [20]. For phase identification, EDS
(energy dispersive spectroscopy) analysis was conducted for matrix α phase (point 1 in
Figure 8b) and transformed β phase (point 2 in Figure 8b), and the results are shown in
Figure 8d. It can be seen from the EDS analysis that the α-stable element of Al is higher
in the matrix, while the β-stable element of Mo or Si is rich in phase with fine parallel
structure, thus confirming that the matrix is α phase and the other is the transformed
β phase.
In general, DRX or DRV is the main mechanism of the flow stress softening in hot
deformation. Usually, DRX occurs more obviously in the sample deformed at lower tem-
perature and lower strain rate [22]. In present study, the flow stress softening is mainly
controlled by DRX and microstructural evolution. As seen in Figure 7, dynamic recrystal-
lization occurred in the alloys deformed in lower part of two-phase α/β region (T ≤ 1273 K),
while microstructural evolution of transformed β grain and dynamic globalization of α
by diffusional control are the dominant mechanism of flow softening at high temperature
of 1323 K.
(a1) (b1) (c1)
Sub-grain
Transformed
1 sƺ1
Bending/Kinked
20μm 20μm 10μm
(a2) (b2) (c2)

0.1 sƺ1
΅ colony
Transformed Ά
DRX
DRX
20μm 20μm 10μm
(a3) (b3) (c3)

Compression axis
Transformed Ά
0.01 sƺ1
DRX
Globularised ΅
20μm 20μm 20μm
1123 K 1273 K 1323 K
Figure 7. Microstructure of Ti-1100 alloy with a strain of 0.6 at (a1) 1123 K/1 s−1 , (a2) 1123 K/0.1 s−1 ,
(a3) 1123 K/0.01 s−1 ; (b1) 1273 K/1 s−1 , (b2) 1273 K/0.1 s−1 , (b3) 1273 K/0.01 s−1 ; and (c1) 1323 K/
1 s−1 , (c2) 1323 K/0.1 s−1 , (c3) 1323 K/0.01 s−1 .
174
Materials 2022, 15, 5932
Figure 8. SEM micrographs of Ti-1100 alloy compressed at 1323 K: (a) SE image at 0.01 s−1 , (b) BSE
image at 0.1 s−1 , (c) SE image at 1 s−1 , and (d) EDS analysis of point 1 and 2 in Figure 8b.
3.4. Strain Rate Sensitivity

It is widely known that the strain rate sensitivity (m) is a parameter of great importance
to determine the deformation behavior of the material. The high strain rate sensitivity can
either induce the occurrence work hardening or flow stress softening depending on the
temperature and strain [21]. In general, the positive m value is favorable to the plasticity
of the material, while the negative m value is adverse to hot deformation because the
material is prone to the defects [30]. In addition, when the m value is greater than 0.5, the
superplasticity occurs [36].
In terms of the high temperature deformation theory, the strain rate sensitivity is given
in the following [37]:
∂lnσ Δlnσ
m= . ≈ . (3)
∂lnε T, ε Δlnε T, ε
After data processing, the values of m were obtained. Figure 9 shows the variation of
strain rate sensitivity under different deformation conditions. As seen in Figure 9a at the
strain rate of 0.01 s−1 , the m values are all positive and greater than 0.18. It generally shows
increasing tendency with the increase of strain under most of temperature conditions,
indicating the plasticity of material is enhanced as the strain increases. When the strain rate
is increased to 0.1 s−1 and 1 s−1 , the strain rate sensitivity behaves differently depending
on deformation temperature. As seen in Figure 9b,c under low temperature of 1123–1173 K,
the m value decreases with the increase of strain, and this reduction of m is more significant
at higher strain rates. These phenomena implied that the deformation mechanism has been
changed as increasing the strain rate. When the high strain rate is 1s−1 , the value of m
becomes negative as the strain approaches above 0.3, showing poor plasticity of material at
this domain.
In all cases, the peak zone of m value mostly appeared in the high temperature range
of 1273–1323 K and the strain of 0.6, indicating good plasticity of material. The maximum
175
Materials 2022, 15, 5932
m-value of 0.64 was obtained at 1273 K and strain rate of 0.6, and the maximum m-values of
0.52 and 0.56 under strain rate of 0.1 s−1 and 1 s−1 were obtained at 1273 K and strain rate of
0.6. These values are all greater than 0.5, indicating the material has superplasticity. This is
mainly due to the increase in volume fraction of β phase transformed near α/β temperature
region, resulting in higher m value and easy deformation. According to superplastic theory,
the strain rate sensitivity value can be affected by the deformation parameters such as T, ε,
.
and ε, alloy composition, and phase fraction. Earlier studies on Ti60 [38], and Ti-6242S [20]
alloys have shown that the maximum m-value was obtained at 0.93 Tβ . Similarly, in the
present study, the maximum m-value was obtained in the temperature range of 1273–1323 K,
near the β-transus temperature.
(a) (b)
(c)
.
Figure 9. Contour maps of strain rate sensitivity under different strain rates: (a) ε = 0.01 s−1 ,
. .
(b) ε = 0.1 s−1 , and (c) ε = 1 s−1 .
The strain rate sensitivity near 1223 K is relatively stable with strain under a fixed
strain rate, but it increases with the increase of strain rate. When the strain rate is
0.01 s−1 and 0.1 s−1 , the m values are in the range of 0.18-0.30, which is a typical de-
formation controlled by climb-limited glide processes characteristic of power-low creep or
176
Materials 2022, 15, 5932
dislocation glide [34]. When the strain rate is 1 s−1 , the m values are higher than 0.3, which
is probably related to grain-boundary sliding (GBS) [28].
3.5. DMM Processing Map

In order to characterize the hot working behavior of materials under plastic deforma-
tion without losing stability at high temperature, processing map of a material based on
the dynamic materials model (DMM) [39] was developed. This theory assumes that the
workpiece is regarded as power dissipation in high temperature deformation and the power
observed by workpiece essentially dissipates during the hot deformation process. In this
map, the microstructure evolution such as DRV, DRX, and superplasticity are considered
as “safe” deformation conditions, while the “unsafe” deformation conditions are various
defects such as shear band and flow localization. To develop the processing map of the
alloy, the efficiency of power dissipation (η) representing the power dissipation capacity is
defined as Equation (4):
. 2m
η (T, ε ) = (4)
m+1
where m is the strain rate sensitivity coefficient for a given flow stress and expressed by
.
m = ∂(ln σ )/∂ ln ε . Variation of power dissipater efficiency with temperature and strain
rate constitutes the power dissipater map, which presents various power domains corre-
lated to specific microstructural mechanisms.
To identify the microstructural instability, Ziegler flow instability [40] was used as
Equation (5):
. ∂ ln[m/(m + 1)]
ξ ε = . +m (5)
∂lnε
.
where ξ ε is a dimensionless instability parameter. This model assumes that when the
instability parameter is negative, the flow instability of material occurs. The variation of
instability criteria with strain rates and temperature constitutes the instability map. By
superimposing power dissipater map and instability map, the processing map of a material
is developed. Figure 10a,b show the processing maps of the alloy at the strain of 0.3 and
0.6, respectively. As seen, the value of η generally increased with increasing temperature,
and there are multiple peak domains in high strain rate region and low strain region.
As shown in Figure 10a, the instability mainly occurred in the area of temperature below
1173 K and higher strain rate, and this instability areas coincided with low efficiency regions.
This instability region gradually disappears as the temperature increases and the strain rate
decreases. This reduction of instability may be attributed to the reduction of deformation
heating effects, which is considered as main factor of an adiabatic shear banding [31]. With
the increase of strain, the instability regions expanded to the temperature of 1173 K and
.
to the strain rate of 0.023 s−1 (lnε = −3.78) as show in Figure 10b. In addition, another
instability zone was appeared at a high strain rate of 1s−1 and a temperature of 1273 K. The
optimum processing range of material deformation is 1273–1323 K along with strain rate
.
range of 0.01–0.165 s−1 (lnε = −4.6~−1.8) since the power dissipation efficiency generally
shows high values at this region and the instability criteria has a positive value.
Figure 11 shows the microstructures of specimens deformed under various deforma-
tion conditions. It can be known from microstructural observation that the shear deforma-
tion is the dominant instability mechanism. As seen in Figure 11a,b, the microstructure
exhibited a severe flow localization at near 45◦ to the compression axis and large amount
of shear band. In general, the localized plastic flow or shear deformation are easy to form
at low temperatures and high strain rates [41], and similar results have been observed in
Ti-6242S alloy [19]. As described in Section 3.2, the minimum value of strain hardening
exponents n at 1123 K also demonstrated severe flow softening due to temperature rise.
Another instability is inhomogenous microstructure as shown in Figure 11c at 1273 K/
1 s−1 . This may be due to the insufficient phase transformation behavior occurring at high
strain rate and high temperature, which leads to the inhomogeneity of microstructure.
177
Materials 2022, 15, 5932
These microstructures are undesirable in achieving appropriate mechanical properties, so it

should be avoided during hot deformation.
(a) (b)
Figure 10. Processing maps of Ti-1100 alloy at strains of (a) 0.3 and (b) 0.6 (shaded zone represents
instability domain; numbers represent power dissipation efficiency (%)).
Figure 11. Optical micrographs of specimens deformed under different conditions: (a) 1123 K, 1 s−1 ,
(b) 1123 K, 0.1 s−1 , (c) 1273 K, 1 s−1 , and (d) 1273 K, 0.1 s−1 .
178
Materials 2022, 15, 5932
According to DMM, η is related to the microstructural mechanism which occurs dy-

namically to dissipate power, such as DRV, phase transformation, and grain refinement [42].
The higher η leads to a better dissipative behavior from the workpiece associated with
microstructure evolution, resulting better deformation capacity. On the other hand, the
lower η leads to worse dissipative behavior due to temperature rise effect. Therefore, it
was worth to investigate the variations of η under different processing conditions. Shown
in Figure 12a is the variation of η with strain at different temperatures. At 1123 K shown
in Figure 12a, the values of η are all lower than 0.5, and it decreased more with the in-
crease of strain. This behavior implies that the deformation power was mainly dissipated
into viscoplastic heat due to temperature rise effect, indicating low formability at warm
deformation region. So, this is the reason why the alloy is prone to defects when it is
deformed at low temperature as discussed above. As shown in Figure 12a at moderate
temperature of 1223 K, the value of η remains constant with strain, but it shows increasing
tendency with strain at higher temperatures of 1273 K and 1323 K. Obviously, the slope of
η with respect to strain gradually increases with increase of temperature, and this means
that the power dissipation mechanism during hot deformation has been changed from
temperature-dependent to microstructure-dependent with the increase of temperature. The
variations of power dissipation efficiency with temperature under different strains and
a fixed strain rate of 0.1 s−1 is shown in Figure 12b. Clearly, the power dissipation ef-
ficiencies increased with increasing temperature and the increase in η values is more
significant with increase of strain. As seen in Figure 12a,b at fixed strain of 0.6, the value
of η increases with increase of temperature, implying microstructure optimization were
more likely to occur at higher temperature. These findings exactly explain the different
temperature-dependent behavior of true stress–strain curves as discussed in preceding
sections. Figure 12c shows the variations of power dissipation efficiency with strain under
different strain rates and the fixed temperature of 1223 K. The η values remained constant
with the strain at each strain rate. The η values at a strain rate of 0.01 s−1 remained about
0.6, implying microstructural changes such as DRX or DRV are more likely to occur in this
case. Metallographic observation under this condition shown in Figure 7 also confirmed
the DRXed microstructures. On the other hand, the η values at a strain rate of 1 s−1 are
in the range of 0.3–0.4 and lower than 0.5, implying the microstructural optimization is
difficult to occur in this case, but mostly consumed by viscoplastic heat or local overheating
which may induce the instability such as shear deformation or localized plastic flow.
(a) (b)
Figure 12. Cont.
179
Materials 2022, 15, 5932
(c)
Figure 12. Variations of power dissipation efficiency with (a) strain at 0.1 s−1 , (b) temperature at
strain rate of 0.1 s−1 , and (c) strain at 1223 K.
4. Conclusions
1. The true stress–strain curves showed that at a strain rate of 0.01 s−1 , the flow stress
exhibited softening response under all deformation temperatures. At relatively high
strain rate of 0.1 s−1 and 1 s−1 , the flow stress exhibits the general behaviors of
softening response in the low part of α/β region (T ≤ 1223 K), while both softening
and hardening effects were observed at upper part (T > 1273 K).
2. The strain hardening exponent (n) analysis revealed that the hardening/softening
effects become more significant as the strain increases. In addition, the higher tem-
perature the more prone to strain hardening and the lower temperature the more
prone to thermal softening. When the strain rate is 0.1 s−1 and 1 s−1 , the values of n
are all negative in the temperature range of 1123–1223 K and it becomes positive at
temperature above 1223 K, implying that the softening effect has been changed to the
hardening effect near at 1223 K. The most severe flow softening with minimum value
of n occurred at low temperature of 1123 K with high strain rate of 1 s−1 , and high
temperature of 1323 K with low strain rate of 0.01 s−1 .
3. Lamellar phase with a length of 10~20 μm and a thickness less than 10 μm was
the main structure of alloy deformed at temperature below 1273 K, and the various
form of transformed β phase with irregular shape of α structure is the main feature
at temperature of 1323 K. The dynamic recrystallization of α is the main softening
mechanism in lower part of temperature (T ≤ 1273 K), while microstructural evolution
of transformed β grain and dynamic globalization of α by diffusional mainly control
the softening effect of alloy.
4. The developed processing map demonstrated that the deformation in the temperature
range of 1273–1323 K and strain rates of 0.01–0.165 s−1 was desirable and led to high
efficiencies. At low temperature of 1123 K, increasing strain rate led to the increase of
flow instability, which was primarily manifested as localized plastic flow, adiabatic
shear bands and inhomogenous microstructure. The variation of power dissipation
energy (η) with strain demonstrated that the power dissipation mechanism during
hot deformation has been changed from temperature-dependent to microstructure-
dependent with the increase of temperature for the alloy deformed at 0.1 s−1 .
180
Materials 2022, 15, 5932
Author Contributions: Conceptualization, R.P. and L.M.; methodology, L.M. and B.H.; software, R.P.
and W.Z.; validation, R.P.; investigation, R.P. and W.Z.; writing—original draft preparation, R.P.;
writing—review and editing, P.Z.; funding acquisition, P.Z. All authors have read and agreed to the
published version of the manuscript.
Funding: This research was funded by the National Natural Science Foundation of China (NSFC)
(No. 51804007), Natural Science Research Project of Anhui Universities (Grant No. KJ2019A0127),
and Anhui University of Science and Technology introduction of talent research start-up fund project
of (No. 13200456).
Acknowledgments: School of Materials Science and Technology, Anhui University of Science and
Technology for the supports.
Appendix A
The β-transus temperature (Tβ , the temperature at which the α + β-phases transform
to the β-phase) of the as-sintered Ti-1100 titanium alloy was determined by differential
scanning calorimeter (DSC, NETZSCH STA 449F5, Germany). The main parameters of DSC
analysis are as follows:
(1) Sample mass (mg): 25.18
(2) Type of crucible: Al2 O3 (85 μL) with lid
(3) Heating rate (K/min): 10
As seen in Figure A1, Tβ was determined as 1293 K (or 1019.7 ◦ C).
Figure A1. DSC curve of the as-sintered Ti-1100 titanium alloy.
181
Materials 2022, 15, 5932
References
1. Williams, J.C.; Starke, E.A., Jr. Progress in structural materials for aerospace systems. Acta Mater. 2003, 51, 5775–5799. [CrossRef]
2. Bania, P. Metallurgy and Technology of Practical Titanium Alloys; TMS: Warrendale, PA, USA, 1994; pp. 9–18.
3. Zhang, H.R.; Niu, H.Z.; Zang, M.C.; Yue, J.K.; Zhang, D.L. Microstructures and mechanical behavior of a near α titanium alloy
prepared by TiH2-based powder metallurgy. Mater. Sci. Eng. A 2020, 770, 138570. [CrossRef]
4. Fang, Z.Z.; Paramore, J.D.; Sun, P.; Ravi Chandran, K.S.; Zhang, Y.; Xia, Y.; Cao, F.; Koopman, M.; Free, M. Powder metallurgy of
titanium–past, present, and future. Int. Mater. Rev. 2018, 63, 407–459. [CrossRef]
5. Fang, Z.Z.; Sun, P.; Wang, H.T. Hydrogen sintering of titanium to produce high density fine grain titanium alloys. Adv. Eng. Mater.
2012, 14, 383–387. [CrossRef]
6. Joshi, V.V.; Lavender, C.; Moxon, V.; Duz, V.; Nyberg, E.; Weil, K.S. Development of Ti-6Al-4V and Ti-1Al-8V-5Fe alloys using
low-cost TiH2 powder feedstock. J. Mater. Eng. Perform. 2013, 22, 995–1003. [CrossRef]
7. Azevedo, C.R.F.; Rodrigues, D.; Neto, F.B. Ti–Al–V powder metallurgy (PM) via the hydrogenation–dehydrogenation (HDH)
process. J. Alloy. Compd. 2003, 353, 217–227. [CrossRef]
8. Ivasishin, O.M.; Savvakin, D.G. The impact of diffusion on synthesis of high-strength titanium alloys from elemental powder
blends. Key Eng. Mater. 2010, 436, 113–121. [CrossRef]
9. Zhang, S.Z.; Liu, J.W.; Zhao, Q.Y.; Zhang, C.J.; Bolzoni, L.; Yang, F. Microstructure characterization of a high strength Ti–6Al–4V
alloy prepared from a powder mixture of TiH2 and 60Al40V masteralloy powders. J. Alloy. Compd. 2020, 818, 152815. [CrossRef]
10. Hagiwara, M.; Emura, S. Blended elemental P/M synthesis and property evaluation of Ti-1100. Mater. Sci. Eng. A 2003, 352, 85.
[CrossRef]
11. Zhang, H.R.; Niu, H.Z.; Liu, S.; Zang, M.C.; Zhang, D.L. Significantly enhanced tensile ductility and its origin of a<0001>
micro-textured extrusion bar of a powder metallurgy near alpha titanium alloy. Scr. Mater. 2022, 213, 114633.
12. Prasad, Y.V.R.K.; Seshacharyulu, T. Processing maps for hot working of titanium alloys. Mater. Sci. Eng. A 1998, 243, 82–88.
[CrossRef]
13. Seshacharyulu, T.; Medeiros, S.C.; Frazier, W.G.; Prasad, Y.V.R.K. Hot working of commercial Ti-6Al-4V with an equiaxed α-β
microstructure: Materials modeling considerations. Mater. Sci. Eng. A 2000, 284, 184–194. [CrossRef]
14. Babu, K.A.; Mandal, S.; Kumar, A.; Athreya, C.N.; Boer, B.; Sarma, V.S. Characterization of hot deformation behavior of alloy
617 through kinetic analysis, dynamic material modeling and microstructural studies. Mater. Sci. Eng. A 2016, 664, 177–187.
[CrossRef]
15. Krishna, V.G.; Prasad, Y.V.R.K.; Birla, N.C.; Rao, G.S. Hot-deformation mechanisms in near-alpha titanium alloy 685. JOM 1996,
48, 56–59. [CrossRef]
16. Balasundar, I.; Raghu, T.; Kashyap, B.P. Hot working and geometric dynamic recrystallisation behaviour of a near-α titanium
alloy with acicular microstructure. Mater. Sci. Eng. A 2014, 600, 135–144. [CrossRef]
17. Zhou, D.; Zeng, W.; Xu, J.; Chen, W.; Wang, S. Characterization of Hot Workability for a Near Alpha Titanium Alloy by Integrating
Processing Maps and Constitutive Relationship. Adv. Eng. Mater. 2019, 21, 1801232. [CrossRef]
18. Su, Y.; Kong, F.; You, F.H.; Wang, X.; Chen, Y. The high-temperature deformation behavior of a novel near-α titanium alloy and
hot-forging based on the processing map. Vacuum 2020, 173, 109135. [CrossRef]
19. Morakabati, M.; Hajari, A. Hot working behavior of near alpha titanium alloy analyzed by mechanical testing and processing
map. Trans. Nonferrous Met. Soc. China 2020, 30, 1560–1573. [CrossRef]
20. Ding, C.; Shi, Q.; Liu, X.; Zheng, L.; Li, R.; Hang, Z.; Yu, B.; Wu, W. Microstructure and Mechanical Properties of PM Ti600 Alloy
after Hot Extrusion and subsequent Annealing Treatment. Mater. Sci. Eng. A 2019, 748, 434–440. [CrossRef]
21. Piao, R.; Yang, S.L.; Zhu, Y.L.; Ma, L. Hot Deformation Behavior of Near-α Titanium Alloy Ti-1100 Prepared by TiH2 -based
Powder Metallurgy. Rare Met. Mater. Eng. 2020, 49, 3314–3324.
22. Yang, C.; Liu, X.; Si, J.Y.; Liu, F.; Jiang, L. Processing map and cracking behavior of powder metallurgy superalloy FGH96 during
hot compression. Chin. J. Nonferrous Met. 2015, 25, 2707–2719.
23. Lei, J.; Zhu, W.; Chen, L.; Sun, Q.; Xiao, L.; Sun, J. Deformation behaviour and microstructural evolution during the hot
compression of Ti-5Al4Zr8Mo7V alloy. Mater. Today Commun. 2020, 23, 100873. [CrossRef]
24. Li, A.; Huang, L.; Meng, Q.; Geng, L.; Cui, X. Hot working of Ti−6Al−3Mo−2Zr−0.3Si alloy with lamellar α+β starting structure
using processing map. Mater. Des. 2009, 30, 1625–1631. [CrossRef]
25. Hossenin, R.; Morakabati, M.; Abbasi, S.M.; Hajari, A. Development of a trimodal microstructure with superior combined
strength, ductility and creep-rupture properties in a near alpha titanium alloy. Mater. Sci. Eng. A 2017, 696, 155–165. [CrossRef]
26. Wanjar, P.; Jahazi, M.; Monajati, H.; Yue, S.; Immarigeon, J.P. Hot working behavior of near-α alloy IMI834. Mater. Sci. Eng. A
2005, 396, 50–60. [CrossRef]
27. Holloman, J.H. Tensile deformation. Met. Technol. 1945, 12, 268–290.
28. Luo, J.; Li, M.; Yu, W.; Li, H. The variation of strain rate sensitivity exponent and strain hardening exponent in isothermal
compression of Ti-6Al-4V alloy. Mater. Des. 2010, 31, 741–748. [CrossRef]
29. Luo, J.; Li, M.Q. Strain rate sensitivity and strain hardening exponent during the isothermal compression of Ti60 alloy. Mater. Sci.
Eng. A 2012, 538, 156–163. [CrossRef]
30. Qiu, Q.; Wang, K.; Li, X.; Wang, J.; Gao, X.; Zhang, K. Hot deformation behavior and processing parameters optimization of SP700
titanium alloy. J. Mater. Res. Technol. 2021, 15, 3078–3087. [CrossRef]
182
Materials 2022, 15, 5932
31. Wang, J.; Wang, K.L.; Lu, S.Q.; Li, X.; OuYang, D.L.; Qiu, Q. Softening mechanism and process parameters optimization of
Ti-4.2Al-0.005B titanium alloy during hot deformation. J. Mater. Res. Technol. 2022, 17, 1842–1851. [CrossRef]
32. Ghavam, M.H.; Morakabati, M.; Abbasi, S.M.; Badri, H. Flow behavior modelling of IMI834 titanium alloy during hot tensile
deformation. Trans. Nonferrous Met. Soc. China 2015, 25, 748–758. [CrossRef]
33. Cui, X. Study on Microstructure Evolution and Properties of TC11 Titanium Alloy during High Temperature Deformation Process.
Master’s Thesis, Harbin Institute of Technology, Harbin, China, 2008.
34. Semiatin, S.L.; Bieler, T.R. The effect of alpha platelet thickness on plastic flow during hot working of TI−6Al−4V with
a transformed microstructure. Acta Mater. 2001, 49, 3565–3573. [CrossRef]
35. Stefansson, N.; Semiatin, S. Mechanisms of globularization of Ti−6Al−4V during static heat treatment. Metall. Mater. Trans. A
2003, 34, 691–698. [CrossRef]
36. Ma, L.; Wan, M.; Li, W.; Shao, J.; Bai, X.; Zhang, J. Superplastic deformation mechanical behavior and constitutive modelling of
a near-a titanium alloy TNW700 sheet. Mater. Sci. Eng. A 2021, 817, 141419. [CrossRef]
37. Narayana Murty, S.V.S.; Nageswara Rao, B.; Kashyap, B.P. Instability criteria for hot deformation of materials. Int. Mater. Rev.
2014, 45, 15–26. [CrossRef]
38. Jia, W.J.; Zeng, W.D.; Zhou, Y.G.; Liu, J.R.; Wang, Q.J. High-temperature deformation behavior of Ti60 titanium alloy. Mater. Sci.
Eng. A 2011, 528, 4068–4074. [CrossRef]
39. Prasad, Y.V.R.K.; Gegel, H.L.; Doraivelu, S.M.; Malas, J.C.; Morgan, J.T.; Lark, K.A.; Barker, D.R. Modeling of dynamic material
behavior in hot deformation: Forging of Ti-6242. Metall. Trans. 1984, 15A, 1883–1892. [CrossRef]
40. Ziegler, H. Progress in Solid Mechanics; John Willey and Sons: New York, NY, USA, 1963; p. 93.
41. Yang, J.; Wang, G.; Jiao, X.; Li, Y.; Liu, Q. High-temperature deformation behavior of the extruded Ti-22Al-25Nb alloy fabricated
by powder metallurgy. Mater. Charact. 2018, 137, 170–179. [CrossRef]
42. Qiang, L.; Song, H.; Ke, T.; Yu, Y.; Ye, W.; Song, S.Y. Investigation of high temperature behavior and processing map of
Ti-6Al-4V-0.11Ru titanium alloy. J. Alloy. Compd. 2019, 787, 527–536.
183
materials
Article
In Situ Observation and Phase-Field Simulation Framework of
Duplex Stainless-Steel Slab during Solidification
Tong Wang 1 , David Wexler 1 , Liangliang Guo 2 , Yangfan Wang 1 and Huijun Li 1, *
1 School of Mechanical, Materials, Mechatronics and Biomedical Engineering, University of Wollongong,

Northfields Ave., Wollongong, NSW 2522, Australia
2 BaoWu Steel Group, Baoshan, Shanghai 201900, China
Abstract: The melting and solidification process of S32101 duplex stainless steel (DSS) was inves-
tigated using high-temperature confocal microscopy (HTCM). The method of concentric HTCM
was employed to study microstructure evolution during the solidification process of S32101 DSS.
This method could artificially create a meniscus-shaped solid–liquid interface, which dramatically
improved the quality of in situ observations. During the heating stage, γ-austenite transformed to
δ-ferrite, and this transformation manifested itself in the form of grain boundaries (GBs) moving.
The effects of cooling rate on the solidification pattern and microstructure were revealed in the
present research. An enhanced cooling rate led to a finer microstructure, and the solidification pattern
changed from cellular to dendritic growth. As the temperature decreased, the commencement and
growth of precipitates were observed. In this paper, the experimental data, including parameters such
as temperature, cooling rate, and growth mode, were used as the benchmark for the simulation. A
simulation framework using Micress linked to a 1D heat transfer model enabling consistent analysis
of solidification dynamics in DSS across the whole cast slab was established. Simulating the dendrite
growth and elemental segregation of DSS at specific cooling rates shows that this framework can be a
Citation: Wang, T.; Wexler, D.; Guo, powerful tool for solving practical production problems.
L.; Wang, Y.; Li, H. In Situ
Observation and Phase-Field Keywords: high-temperature confocal microscope; phase-field simulation; duplex stainless steel;
Simulation Framework of Duplex cooling rate
Stainless-Steel Slab during
Solidification. Materials 2022, 15, 5517.
https://doi.org/10.3390/ma15165517

1. Introduction
Jiangshan Zhang S32101 (US designation) is a duplex stainless steel (DSS) with low content of Ni and
high content of N. High contents of Mn and N were added to replace expansive Ni to reduce
Received: 14 July 2022
the cost and improve mechanical properties and corrosion resistance. It is widely used
in the chemical industry or offshore technologies, where a combination of high corrosion
Published: 11 August 2022
resistance and good tensile strength is required [1,2].
Publisher’s Note: MDPI stays neutral Many factors, such as cooling rate, can affect the microstructure of steel during the
with regard to jurisdictional claims in solidification process. Importantly, the microstructure can affect the mechanical properties
published maps and institutional affil- of S32101 DSS and finally influence the quality of the slab. The dual-phase structure is
iations. the root of the excellent properties of DSS. Mazumdar and Kary [3], from various per-
spectives, such as heat transfer during crystallization and solid–liquid interface tribology,
expounded on the control of the continuous casting process and pointed out that control-
ling the solidification conditions controlled the ratio of δ and γ to obtain products with
excellent quality.
High-temperature confocal microscopy (HTCM) was used to observe the solidification
process of S32101 DSS in the present research, and this technique can provide a so-called
conditions of the Creative Commons in situ observation. Zhao and Sun [4] applied laser scanning confocal microscopy (LSCM)
Attribution (CC BY) license (https:// in their research on DSS. They observed and discussed the effect of different N contents
creativecommons.org/licenses/by/ on the solid-state δ ༸ γ phase transformations and concluded that N could hinder the
4.0/). migration of δ/γ interface boundaries (IBs).

Materials 2022, 15, 5517
Computational simulation can also be used predict the solidification process from
experimental observation. The Microstructure Evolution Simulation Software (Micress® ,
Version 7.1) is a software package that calculates the growth of newly formed phases during
solidification or phase transformation of a metallurgical system on a temporal or spatial
scale. The software is based on the multiphase-field concept, developed by ACCESS e.V.
(affiliates of Aachen University, Aachen, Germany) scientists since 1995. It is widely used
in solidification, solid-state transformations, grain growth, recrystallization, heat treatment,
etc. [5,6]. The solidification process and subsequent solid-state phase transformation
always start with nucleation, which is the first step of forming a new phase. This process
is dominated by thermodynamic driving forces, diffusion, and interfacial curvature. The
backbone of Micress is the multiphase field method for multicomponent alloys, which
enables the treatment of multiphase, multigrain, and multicomponent problems. As a
well-established commercial phase-field simulation software, Micress provides a simulation
platform for materials researchers without a computer science background.
In continuous casting, liquid steel is poured into a water-cooled copper mold, with
the high heat extraction rate chilling a solid steel shell that contains the yet-to-be solidified
liquid in the center. This shell is continuously extracted from the copper mold, the rate
of which is an important operating parameter. Secondary cooling through water mist is
applied to the shell as it exits the mold, thus inducing sufficient cooling to ensure complete
solidification. The solidification process occurs with a liquid-to-solid phase transformation,
and segregation of alloying elements such as Cr and Ni, as well as N and Mo for DSS
in particular, could lead to defects. For the continuous casting of S32101 DSS, a deeper
understanding of the effect of thermodynamic and kinetic parameters on the rate and
mechanism is the key to improving the final product quality.
Many studies have been published about DSS with respect to the influence of the
different alloying element contents. Scarce research has been conducted to investigate the
thermodynamic process from melting to solidification. This paper aimed to reveal the
melting process of S32101 DSS and the solidification process regarding different cooling
rates through ‘in situ’ observation; furthermore, on the basis of the experiments, a phase-
field simulation framework was established to determine whether concentric solidification
can be used to provide a basis for benchmarking the relationship between cooling rate and
solidification structure and to what extent this can be used to provide data for developing
the Micress phase-field simulation framework.
2. Experimental and Modeling Approaches

2.1. Experimental Approaches
As an important tool widely used to observe solidification and phase transformation,
HTCM sometimes reduces the observation surface because the sample forms semicircu-
lar droplets in the crucible. If there is further obstruction by oxides or precipitates, the
availability of observational data will be significantly reduced. The design of a concentric
solidification method can tackle this problem.
2.1.1. Concentric HTCM

In 1961, Minski [7] established the principles of LSCM, which are widely applied in
biological sciences. This equipment has been applied to investigate many fields of research,
including the morphology of solidification and an analysis of the δ/γ interphase boundary
dynamic behavior in low-carbon steels [8,9], inclusion agglomeration and engulfment in
steel melt/solid interface [9,10], and crystallization of oxide melts [11]. Since the concentric
HTCM was designed, significant developments in peritectic reaction and transformation
have been achieved. Stephan [12,13] employed the concentric HTCM method to investigate
the peritectic transformation of Fe–C steel and cast new light on the reason for massive-type
phase transformation with thermodynamic and kinetic arguments. Lasertec Corporation
manufactures the HTCM equipment in Yokohama, Japan. A 1.0 kW halogen lamp is located
at a focal point in a gold-plated ellipsoid cavity in the lower half of the infrared furnace.
186
Materials 2022, 15, 5517
The sample was heated by radiation in this chamber. A gold-coated ellipsoidal chamber
concentrated the heat onto the surface of the sample located at the other focal point of the
furnace chamber. Figure 1 shows the components of the 1LM21H type HTCM with an
oxygen detector, monitors, and proportional–integral–derivative (PID) controller.
Figure 1. Picture of HTCM 1LM21H.
Reid et al. [14,15] developed and described the concentric solidification technique to
improve the quality of in situ observation of HTCM. A thin sample is partially melted in
the center, forming a pool of melt surrounded by a solid rim. A platinum sample holder
holds the crucible in place, and a type B thermocouple is attached to the crucible. Since
the in situ observation of HTCM is limited to the sample surface, the thickness becomes an
important parameter. After several attempts with different sample thicknesses, ~250 μm
was determined to be the minimum thickness for concentric solidification experiments on
S32101 DSS. On the one hand, if the thickness is less than 250 μm, the liquid pool is easily
ruptured. On the other hand, if the thickness exceeds 300 μm, the output power required
to melt the sample will be too high to be achieved.
In DSS, initial melting is uneven due to segregation during the casting of the slab,
which was not previously homogenized. Care needs to be taken for the rate of further
heating to achieve a uniform liquid pool in the central region of the sample. With care,
a pool of diameter around 5 mm can be achieved with a planar interface between liquid
and solid. This then becomes the starting condition for all subsequent imposed cooling
rates. The range of potential solidification microstructures possible includes planar at low
cooling rates, and then cellular as the cooling rate increases. The subsequent transition is a
dendritic microstructure which reverts to cellular and planar if a sufficiently high cooling
rate is achieved. In the concentric solidification technique, the maximum controlled cooling
rate that can be achieved is 500 to 1000 ◦ C/min, far below the rate required for reverting to
cellular/planar structures.
The sample and the holder were inserted into the top half of the furnace chamber, and
high-purity argon gas (99.9999 vol.%) was flowed through the chamber. The gas was passed
through a stainless-steel tube filled with titanium turnings to ensure a high-integrity inert
gas, held at a temperature of 850 ◦ C, before entering the furnace chamber. The oxygen level
was reduced to 10–14 ppm before running the experiments. Moisture was also removed
when the gas flowed through the Ti furnace.
The Omron ES100P digital PID controller panel was used to set the program profile
before the experiments, as well as change the temperature during the experiments. The
temperature program is shown in Figure 2a. All HTCM experiments were based on this
procedure. The whole procedure can be divided into five stages. The first stage is to
187
Materials 2022, 15, 5517
stabilize the equipment. The second stage is the heating stage, which involves heating to
1300 ◦ C at a rate of 150 ◦ C/min. The target temperature is below the melting point; one
reason is to avoid overshooting which could leading to sample penetration, and another
reason is to reserve enough operating range for the third step. Stage 3 is the most important
because the liquid pool is formed at this stage, and it is necessary to artificially slowly adjust
the temperature until a stable liquid pool of sufficient size is formed. The fourth stage is the
heat preservation stage to form a stable solid–liquid interface. The last stage is the cooling
stage with five different cooling rates, 5 ◦ C/min, 10 ◦ C/min, 50 ◦ C/min, 80 ◦ C/min, and
150 ◦ C/min, to investigate the effect of cooling rate on solidification microstructures.
Figure 2. (a) Temperature profile of concentric HTCM experiments. I—Stabilizing stage, II—Fast
heating stage, III—Slow heating stage, IV—Holding stage, and V—Cooling stage. Different dot lines
represent different cooling rates; (b) schematic diagram and pictures of S32101 DSS samples.
2.1.2. Materials
The samples used in present research were provided by Baosteel (as shown in Figure 2b),
named 1.4162 (X2CrMnNiN21-5-1) according to the EN standard. This grade of steel is widely
used in extreme environments which require both excellent mechanical properties and high
resistance to intergranular and pitting corrosion. The chemical composition of S32101 DSS is
given in Table 1. All the samples were cut into slices. The diameter of a single slice was 9.7 mm,
and the thickness was 0.25 mm. To prepare samples, a TechCut 5TM precision high-speed saw
with water cooling was used to cut the sample into specific thicknesses before grinding on 800,
1200, 2400, and 4000 SiC papers. The samples were polished with standard metallographic
techniques using a 6 μm diamond paste and 1 μm diamond paste. A fine sample surface can
provide better images, especially in the heating stage or for solid-state studies.
Table 1. The chemical composition of S32101 (wt.%).
Grade C Mn Cr Ni Mo Cu N
S32101 <0.04 5 21.5 1.5 0.5 0.5 0.2
2.2. Modeling Approaches

2.2.1. Thermodynamics: Thermo-Calc
Released over 30 years ago, Thermo-Calc has become one of the most popular thermo-
dynamic calculation software packages in the world. Thermo-Calc simulations can improve
design during manufacturing and aid in selecting heat treatment temperatures. Calcula-
tions can predict experimental outcomes and minimize the number of experiments. In this
study, the 2021b version of Thermo-Calc with the TCFE9 database was used for the calcula-
tion of phasors and their compositions, transition temperatures, and phase diagrams.
Another use of Thermo-Calc was the generation of optimized thermodynamic and
kinetic datasets encapsulated in a GES5 file. This was applied in the subsequent phase-field
simulation as a basis for microstructure development.
188
Materials 2022, 15, 5517
2.2.2. Phase-Field Simulation: Micress

At the interfaces, the phase-field variables change continuously over an interface
thickness η, which could be significantly large compared with the atomic interface thickness
and tiny compared to the microstructure length scale. Their time evolution is calculated by a
set of phase-field equations derived from the minimization of the free energy functional [16].
! "
. v σαβ π2
n
∅α = ∑ Mαβ v
1
2 ∇2 ∅α − ∇2 ∅ β + 2
2ηαβ
∅α − ∅ β + ∑ Jαβγ + |∇∅|Δαβ , (1)
β=α γ=α= β
! "
1 π2
Jαβγ = σ − σαγ ∅γ + ∇2 ∅γ , (2)
2 βγ η 2
Mαβ = M 3 αβ 8η , (3)
π2
where Jαβγ denotes the higher-order terms accounting for the interaction with additional
phases in triple- or multiphase regions.
For the sake of simplicity, only the isotropic formulation is shown here. In Equation (3),
Mαβ is the mobility of the interface as a function of the interface orientation, Kαβ is related
to the local curvature of the interface, and σαβ ∗ is the anisotropic surface stiffness. The
thermodynamic driving force and the solute partitioning are calculated separately using
the quasi-equilibrium approach [16] and are introduced into the equation for the multiphase
fields (Equation (3)). If concentration coupling is activated, an explicit 1D temperature
field in the z-direction can be optionally defined by the keyword ‘1d_temp’. With the 1D
temperature field, heat flow and latent heat release are solved explicitly. For numerical
reasons, the temperature in the 1D temperature field is calculated using a direct, explicit
solver with a default kinetic coefficient [17]. The default value is preconfigured for typical
casting processes and needs not to be changed in most cases.
The temperature field can be evaluated in a linear (cartesian), cylindrical (cylindrical),
or polar (polar) coordinate system; however, in the present research, a linear setup was
used. Furthermore, we ran simulations at 0 mm, 10 mm, and 50 mm in the cast slab.
As shown in Table 2, the thermodynamic data were provided by online coupling
to the database TCFE9 using the elements given in Table 1 and the phases in Table 2.
Similarly, diffusion data for γ and δ, including cross-terms, were taken from the MOBFE4
mobility database.
Table 2. Phases and phase-related parameters.
No. Database ID ID Diffusion Data μ1/x (cm4 /Js) μ2/x (cm4 /Js) σ1/x (J/cm2 ) σ2/x (J/cm2 )
0.8 × 10−4 to
1 FCC_A1 γ MOBFE4 Diff Limit
1.2 × 10−4
2 BCC_A2 δ MOBFE4 Diff Limit 0.8 × 10−4
1 × 10−5 to 0.8 × 10−4 to 0.3 ×10−4 to
3 Liquid L Fixed Diff Limit
1 × 10−10 1.2 × 10−4 1.2 × 10−4

DSSs have a so-called dual-phase structure, which, for S32101, contains about 50%
volume face-centered cubic (FCC) γ islands and 50% volume body-centered cubic (BCC) δ
matrix. After the samples were ground and finally polished, the surface of S32101 DSSs
was observed with an optical microscope using a 0.04 μm particle size silica gel cloth. For
the S32101 DSS, as shown in Figure 3, even without etching, the dual-phase structure was
prominent, with almost equal proportions for each phase.
189
Materials 2022, 15, 5517
Figure 3. Optical microscopy of S32101 DSS cast with δ matrix and γ island structures.
A property diagram, which reveals the response of all phases to temperature change,
was calculated using Thermo-Calc under equilibrium conditions in the present research.
The phase equilibria were calculated with a temperature range from 500 ◦ C to 1500 ◦ C. As
shown in Figure 4, as temperature decreased, the liquid phase first solidified to δ, and, after
solidification was finished, δ remained for a while and then partially transformed to γ. As
temperature decreased, under equilibrium solidification conditions, the precipitation of
secondary phases, such as chromium nitrides, σ phase, and χ phase, proceeded.
Figure 4. Property diagram of S32101 DSS calculated using Thermo-Calc showing the changes in all
phases with temperature under equilibrium conditions.
According to the thermodynamic calculation of S32101 DSS, the δ-to-γ phase trans-
formation occurred around 1270 ◦ C. As shown in the figure, the gas phase formed after
the steel fully solidified and before the commencement of δ-to-γ solid-state phase transfor-
mation, approximately in the range of 1280–1430 ◦ C. For S32101 DSS, nitrogen is used as
an alloying element, and its content is very high. The solubility of nitrogen was the most
significant factor in the gas pore formation. The solubility of nitrogen in the solid phase is
much smaller than in liquid. During the solidification process, the nitrogen was rejected
from the liquid phase and formed nitrogen gas pores. The nitrogen gas pore formation
was also influenced by nitrogen content, cooling rate, partial pressure, etc. Hence, the
temperature range of gas-phase formation is not fixed.
190
Materials 2022, 15, 5517
3.1. Concentric HTCM Experimental Results

The HTCM results are discussed according to the timeframe from heating to cooling,
and the time is counted from the start of the experiment. The temperature values shown
in the concentric HTCM images are those of the sample edges. Since the diameter of the
sample was 9.7 mm, the temperature of the center was higher than that of the edge.
3.1.1. Heating Stage

In Figure 5, a red circular area P is used as a reference marker. As shown in frames (a)
and (b), the color of δ darkened, making it straightforward to distinguish γ and δ in this
period. Frame (c) shows that the morphology of γ changed with the temperature rising.
The γ and δ interface boundaries were apparent. The γ → δ phase transformation was not
clear during the heating stage except for the morphology change. Frames (d) and (e) show
the growth of δ/δ grain boundaries (GBs). The schematic diagram of frame (f) shows the
morphology of γ. The γ accumulated along the grain boundaries of δ with a finger- or
needle-like pattern. The intragranular γ-cells maintained an island pattern in the δ matrix.
The grain boundary motion was in the form of a γ → δ phase transformation.
Figure 5. Concentric HTCM images revealing (a–c) the morphology changes of the γ phase and
(d–f) the growth of δ/δ GBs; (f) a schematic diagram of the γ morphology of S32101 DSS.
As the temperature increased, γ fully transformed to δ, and δ GBs showed up. In

Figure 6, the red dotted line represents the δ/δ GBs, while the white dotted line represents
the solid/liquid IBs. In frame (b), a new δ GB showed up; after the growth of new δ GBs,
the sample started melting, as shown in frame (c). The liquid phase in frame (d) was located
along the δ GBs.
191
Materials 2022, 15, 5517
Figure 6. Concentric HTCM images revealing (a–c) the growth of δ GBs before melting and (d) the
location of the liquid phase.
As the temperature increased, γ transformed to δ. The GBs enlarged and slipped

before melting, and thermal etching appeared, as shown in Figure 7. These phenomena
indicated that the steel would melt soon.
Figure 7. Concentric HTCM images revealing the δ/δ GBs enlarging and slipping (a–c) and S32101
DSS melting (d–f).
192
Materials 2022, 15, 5517
When polished samples are heated to high temperatures and exposed to an inert
atmosphere (e.g., argon gas in present experiments), the GBs will slip. Figure 7 shows
the thermal etching in the solid rim of concentric experiments. As indicated in Figure 7b,
grooves formed at elevated temperatures and remained intact after cooling. Figure 7 also
reveals mobile grain boundaries and stationary grain boundaries. In frame (a), the GBs
slipped in one direction and later toward the opposite direction (frame (b)). Moreover,
the stationary δ/δ GBs enlarged, which is called thermal grooving. The slipping of GBs
is called thermal etching, leading to the formation of a new δ-cell, as shown in frame (c).
After the temperature reached the melting point, the liquid phase firstly showed up inside
the stationary δ GB. Frames (e) and (f) show the commencement and growth of the liquid
phase. Areas I, II, and III represent the liquid phase in different places. I indicates the
liquid phase inside GBs, II indicates the liquid within two δ GBs, and III is the inside of a
δ-cell. Later, all the liquid phases merged into a liquid pool. The concentric solidification
experiment improved the quality of in situ observations due to the minimization of the
meniscus and the increase in the length of the visual interface between liquid and solid [15].
3.1.2. Solidification Process

In the continuous casting process, the quality of products is highly related to the
grain structure during solidification, including grain size, morphology, and orientation [18].
The cooling rate can influence the solidification behavior and the solid shapes during the
solidification process.
As shown in Figure 8, with a cooling rate of 5 ◦ C/min, the sample solidified initially
with a planar solidification front due to the required time for solute gradients to form as
the solid grew and for instability to commence. After this initial planar solidification, the
structure destabilized into a cellular solidification structure. As shown in frame (a), the
sample started by solidifying from outside the liquid pool to inside gradually, and the
solid/liquid interface boundary (S/L IB) moved slowly. The distance of S/L IB motion
(D) under 5 ◦ C/min was around 588 μm. Frame (b) shows that the δ solid grew with a
columnar structure. The GBs were easy to observe, and the grain size was large. As shown
in frame (c), the cellular structure could be observed as dendrites without secondary arms.
Because of the prolonged cooling rate, the growth of primary arms was even toward the
center of the liquid pool, with no space for secondary arms to grow. As shown in frame (d),
the cellular structures were compact, and the content of alloy elements and gas precipitates
was low; hence, gas pores were unlikely to form in this area. The temperature gradients in
the liquid promoted the growth of cellular grains. The cooling rate also played an essential
role in the growth of a columnar structure. As the cooling rate increased, the length of the
columnar structure decreased.
Primary dendritic arms grew beneath the liquid surface, but secondary dendritic arms
(SDAs) growing perpendicular to the growth direction appeared as a series of aligned
islands at the liquid surface. When the cooling rate increased to 10 ◦ C/min after the initial
periods of planar and cellular solidification, there was a transition to dendritic solidification,
as shown in frames (a–c). Solidification started in the S/L IB. As shown in frame (a), the D
of S/L IB was 380.25 μm; right after the growth of the columnar δ, the secondary dendritic
arms of δ commenced. In frames (b) and (c), the growth of primary and secondary dendritic
arms can be observed. The appearance of SDAs followed the growth of a primary dendrite,
and the growth of secondary arms was orientated. The arrows in frame (b) point to the
preferred growth direction of primary and secondary arms. The microstructure under a
10 °C/min cooling rate showed large graininess and dendritic shapes.
For the sample cooled at 50 ◦ C/min, as shown in Figure 9d, the D was 87.75 μm. As
shown in frame (e), the dendrites grew in the center of the liquid pool and were surrounded
by several δ GBs, which can be observed clearly in frame (f). Under the cooling rate of
50 ◦ C/min, the grain size of dendritic and cellular δ was still significant.
193
Materials 2022, 15, 5517
Figure 8. Concentric HTCM revealing the solidification process of S32101 under a 5 ◦ C/min cooling
rate: (a) the movement of S/L IB, (b) the growth of solidifying arms, (c) the residual liquid pool, and
(d) the morphology of δ solids (cellular growth).
For the sample cooled at 80 ◦ C/min, as shown in Figure 9g–i, solidification started in
the S/L IB, and SDAs emerged from outside the liquid pool to the inside within 10 s. The
microstructure comprised cellular structure domains with a small number of needle shape
particles and no more dendritic microstructure. In frame (g), the D was 93.6 μm; right after
the commencement of dendritic δ, the δ solids grew along the S/L IB. In frame (h), within
the liquid pool of the sample, the solidification proceeded and then grew into a long and
slender needle-like or dendritic pattern (frame (i)).
For the sample with a 150 ◦ C/min cooling rate, as shown in Figure 9j–l, the region of
cellular solidification was no longer observed. Instead, a small band of planar solidification
destabilized directly to dendritic solidification with a smaller primary and secondary
arm spacing, and solidification first started along the S/L IB. The solids simultaneously
emerged both at the center of the liquid pool and at the edge of the S/L IB. The morphology
comprised grains of finer cellular shape without evidence of a dendritic microstructure. The
grain size became finer and smaller than outside of the liquid pool. In frame (a), the D was
76.05 μm, and the columnar δ stopped growing shortly after. The nucleation started inside
the liquid pool and along the S/L IB shortly after. In frame (k), a number of secondary
dendritic arms commenced inside the center liquid pool, whereas, in frame (l), all solid δ
showed up with a cellular structure.
These results allowed studying the characteristics of the solidification microstruc-
ture response to thermal conditions. They demonstrated that concentric solidification
194
Materials 2022, 15, 5517
offers a good platform for studying duplex stainless steels, forming a good testbed for
subsequent simulations.
Figure 9. Concentric HTCM revealing the solidification process of S32101 under the cooling rates of
(a–c) 10 ◦ C/min (d–f) 50 ◦ C/min (g–i) 80 ◦ C/min, and (j–l) 150 ◦ C/min.
3.2. Micress Simulation

The most significant development of the present research was establishing a frame-
work to utilize the combination of experiments and simulation to study the solidifica-
tion process of S32101 DSS. This simulation framework can reduce the experimental
workload dramatically. Concentric HTCM was also used to authenticate the accuracy
of the simulation.
Micress simulations followed the same conventions, with the bottom of the domain
oriented in the direction of the slab surface and the top of the domain pointing to the
center of the slab. Figure 10 shows the selected outputs from the region of the chill surface
where (a) SDAs formed and started to grow, (b) and (c) dendrite competition and primary
dendrite arm spacing selection occurred, and (d) the dendrite tips reached the end of the
195
Materials 2022, 15, 5517
domain; from this point until the end of the simulation run, secondary arm coarsening and
final solidification of the liquid were achieved.
Figure 10. Micress simulation of S32101 chill surface with 100 ◦ C superheat. (a) SDAs formation,
(b) dendrite competition, (c) primary dendrite arm selection, and (d) dendrite tip across domain.
For each alloy, composition simulations were linked to a 1D heat transfer model allow-
ing simulations to be run at a different position in the slab. In Figure 11, three such locations
are used to illustrate this: (a) the chill surface, (b) 10 mm inside the slab, and (c) 50 mm
inside the slab. The pertinent features are the coarsening of the microstructure; further into
the slab, the simulation indicated that reducing heat transfer conditions led to a thicker
solidified shell. Here, the primary difference was in the primary arm spacing increasing as
the position moved into the cast slab. These simulations were further contrasted at a later
point in the solidification progression in Figure 12 to highlight the impact on secondary
arm spacing development in response to position in the slab ((a) chill surface, (b) 10 mm,
and (c) 50 mm).
These results implied segregation into the liquid as solidification proceeds. As high-
lighted in the Thermo-Calc calculations, the segregation of elements is critical in issues such
as nitrogen gas bubble formation. The results from a simulation of S32101 showing the
segregation of each of the seven elements nominated in the project are shown in Figure 13.
Micress provides the opportunity to probe how heat transfer conditions, microstructural
development, and segregation interrelate. In the figure, the information is presented graph-
ically with differences in concentration displayed by changing color levels. The underlying
data at each of the 400,000 grid points is also accessible in Micress and can be manipulated
using the Micress analysis software or extracted to conventional spreadsheets such as Excel
for further analysis. In this way, the composition of the liquid phase during solidification
can be plotted as a function of temperature (Figure 14).
196
Materials 2022, 15, 5517
Figure 11. S32101, 100 ◦ C superheated microstructure when growing dendrites reached the top of
the simulation domain: (a) chill surface, (b) 10 mm in the slab, and (c) 50 mm in the slab.
Figure 12. S32101, 100 ◦ C superheated microstructure showing the development of SDAs: (a) chill
surface, (b) 10 mm in the slab, and (c) 50 mm in the slab.
197
Materials 2022, 15, 5517
Figure 13. S32101, 100 ◦ C superheated at 10 mm in the slab. Segregation of seven elements displayed
from one timestep of a simulation run.
Running Micress simulations enables to determine the effects of changing process

parameters on the subsequent solidification. For example, by testing two superheating
temperatures on the casting of S32101, we could extract relationships between fraction
liquid as a function of temperature and the nitrogen segregation in the liquid (Figure 15).
In this case, increasing the superheat led to a lower nitrogen composition in the liquid at a
given temperature compared to superheating at 50 ◦ C. As the nucleation of nitrogen gas
bubbles is related to the content of dissolved nitrogen in the liquid, such an analysis offers
opportunities for probing defect formation minimization.
198
Materials 2022, 15, 5517
Figure 14. S32101, 100 ◦ C superheated at 10 mm. The liquid phase’s average composition as a
temperature function. Solid lines, left-hand axis; dashed lines, secondary axis.
Figure 15. S32101 comparison of segregation of carbon and nitrogen in the liquid during solidification
as a function of superheat (50 ◦ C and 100 ◦ C).
The simulation domains comprised 400,000 grid points, and Micress outputted data
on the position of the interface (in this case, solid and liquid). The average occupancy of
grid points is directly related to the fineness of a given microstructure. Therefore, we can
compare the influence of melt superheat on the resulting microstructure. In Figure 16, it
can be observed that a higher superheat led to a coarser microstructure. Alternatively, the
simulations can address heat transfer conditions such as heat flux Q at the slab surface. The
base simulations were conducted with a fixed heat transfer coefficient of 0.25 W/cm2 ·K,
in line with conventional casting practices. A comparative simulation for S32101 with
a superheating temperature of 50 ◦ C but with the heat transfer coefficient increased to
0.35 W/cm2 ·K was also run, and the simulation results at 10 mm in the slab are included
in Figure 16. The critical point is that, through incorporating a 1D heat transfer model into
the Micress simulations of solidification, cooling conditions in the caster can be adjusted
consistently, allowing direct comparison to, for example, the measure of microstructure
refinement that arises from increasing the heat transfer coefficient. However, all the other
199
Materials 2022, 15, 5517
information entailed in the simulations, such as element segregation, secondary phase

nucleation, and growth, can also be assessed simultaneously.
Figure 16. Influence of superheat on microstructural fineness of S32101, 50 ◦ C and 100 ◦ C superheat
at 10 mm in the cast slab, when increasing heat flux at the boundary.
4. Conclusions
The concentric HTCM method was employed to study the heating and cooling stages
of S32101 DSS. During the heating process, the GB presented γ → δ phase transformation.
As the temperature increased, γ fully transformed to δ. The GBs enlarged and slipped
before melting, and thermal etching appeared. These phenomena indicate that the steel
would melt soon. The liquid phase preferably formed inside of or close to the δ GBs. During
the solidification process, the cooling rate affected the microstructure of S32101 DSS; as
the cooling rate increased, the solid’s size was finer, the morphology comprised smaller
grains, and the solidification behavior in the body of the liquid pool commenced more
quickly. The distance (D) of S/L IB motion decreased as the cooling rate increased. The size
of the dendritic structure of δ decreased until it disappeared. The results demonstrate that
concentric solidification offers a good platform for studying DSSs, forming a good testbed
for subsequent simulations. The growth of precipitates was observed.
The experimental approach based on in situ HTCM was found to be helpful in studying
the solidification process of S32101 DSS. Moreover, the data can be used as a benchmark for
simulation. A simulation framework using Micress linked to a 1D heat transfer model that
enables consistent analysis of solidification dynamics in duplex stainless steels across the
whole cast slab was established. The analysis of phase transformations, solute segregation,
and microstructural fineness, among other aspects, was demonstrated.
The establishment of the Micress framework to simulate the solidification process can
dramatically address the limitations of experiments and reduce the experimental workload.
Micress offers the opportunity to explore how heat transfer conditions, microstructural
development, and separations relate. By revealing the elemental segregation of S32101 DSS
under the cooling rates of 5 and 150 ◦ C/min, Micress was proven to be a powerful tool to
tackle the problems encountered in production.
Author Contributions: Conceptualization, T.W. and Y.W.; methodology, T.W.; software, T.W.; vali-
dation, D.W. and H.L.; formal analysis, T.W.; investigation, T.W.; resources, L.G.; writing—original
draft preparation, T.W.; writing—review and editing, H.L. and D.W.; supervision, H.L. and D.W.;
project administration, L.G.; funding acquisition, H.L. and L.G. All authors have read and agreed to
the published version of the manuscript.
Funding: This research was funded by Baosteel-Australia Joint Research and Development Centre
(BAJC), grant number BA17010.
200
Materials 2022, 15, 5517

Data Availability Statement: Data sharing is not applicable. No new data were created or analyzed
in this study.
References
1. Sieurin, H.; Sandström, R.; Westin, E.M. Fracture toughness of the lean duplex stainless steel LDX 2101. Metall. Mater. Trans. A
2006, 37, 2975–2981. [CrossRef]
2. Escriba, D.; Materna-Morris, E.; Plaut, R.L.; Padilha, A.F. Intermetallic phase precipitation in duplex stainless steels during high
temperature exposition. Proc. Mater. Sci. Forum 2010, 636–637, 478–484. [CrossRef]
3. Mazumdar, S.; Ray, S. Solidification control in continuous casting of steel. Sadhana 2001, 26, 179–198. [CrossRef]
4. Zhao, Y.; Sun, Y.; Li, X.; Song, F. In-situ Observation of δ↔γ Phase Transformations in Duplex Stainless Steel Containing Different
Nitrogen Contents. ISIJ Int. 2017, 57, 1637–1644. [CrossRef]
5. Böttger, B.; Apel, M.; Eiken, J.; Schaffnit, P.; Steinbach, I. Phase-Field simulation of solidification and solid-state transformations in
multicomponent steels. Steel Res. Int. 2008, 79, 608–616. [CrossRef]
6. Zhang, L.; Steinbach, I.; Du, Y. Phase-field simulation of diffusion couples in the Ni–Al system. Int. J. Mater. Res. 2011, 102,
7. Minsky, M. Microscopy Apparatus. U.S. Patent 3,013,467, 19 December 1961.
8. Yin, H.; Emi, T.; Shibata, H. Morphological instability ofδ-ferrite/γ-austenite interphase boundary in low carbon steels. Acta
Mater. 1999, 47, 1523–1535. [CrossRef]
9. Chikama, H.; Shibata, H.; Emi, T.; Suzuki, M. “In-situ” real time observation of planar to cellular and cellular to dendritic
transition of crystals growing in Fe–C alloy melts. Mater. Trans. JIM 1996, 37, 620–626. [CrossRef]
10. Shibata, H.; Yin, H.; Yoshinaga, S.; Emi, T.; Suzuki, M. In-situ observation of engulfment and pushing of nonmetallic inclusions in
steel melt by advancing melt/solid interface. ISIJ Int. 1998, 38, 149–156. [CrossRef]
11. Wang, Y.; Sridhar, S. Liquid Al2 O3 -CaO-MgO inclusion entrapment at delta ferrite intercellular boundaries during solidification.
Ironmak. Steelmak. 2003, 30, 223–228. [CrossRef]
12. Griesser, S.; Reid, M.; Bernhard, C.; Dippenaar, R. Diffusional constrained crystal nucleation during peritectic phase transitions.
Acta Mater. 2014, 67, 335–341. [CrossRef]
13. Griesser, S.; Bernhard, C.; Dippenaar, R. Effect of nucleation undercooling on the kinetics and mechanism of the peritectic phase
transition in steel. Acta Mater. 2014, 81, 111–120. [CrossRef]
14. Niknafs, S.; Phelan, D.; Dippenaar, R. High-temperature laser-scanning confocal microscopy as a tool to study the interface
instability during unsteady-state solidification of low-carbon steel. J. Microsc. 2013, 249, 53–61. [CrossRef] [PubMed]
15. Reid, M.; Phelan, D.; Dippenaar, R. Concentric solidification for high temperature laser scanning confocal microscopy. ISIJ Int.
2004, 44, 565–572. [CrossRef]
16. Eiken, J.; Böttger, B.; Steinbach, I. Multiphase-field approach for multicomponent alloys with extrapolation scheme for numerical
application. Phys. Rev. E 2006, 73, 066122. [CrossRef] [PubMed]
17. Böttger, B.; Eiken, J.; Apel, M. Phase-field simulation of microstructure formation in technical castings–A self-consistent homoen-
thalpic approach to the micro–macro problem. J. Comput. Phys. 2009, 228, 6784–6795. [CrossRef]
18. Phelan, D. In-Situ Studies of Phase Transformations in Iron Alloys. Ph.D. Thesis, University of Wollongong, Wollongong,
Australia, 2002.
201
materials
Article
Influence of Mineralogical Structure of Mold Flux Film on Heat
Transfer in Mold during Continuous Casting of Peritectic Steel
Lei Liu, Xiuli Han *, Mingduo Li and Di Zhang
College of Mining Engineering, North China University of Science and Technology, Tangshan 063210, China;
[email protected] (L.L.); [email protected] (M.L.); [email protected] (D.Z.)
Abstract: The mineralogical structure of flux films is a critical factor in controlling heat transfer in
the mold and avoiding the longitudinal cracking of slabs during the continuous casting of peritectic
steel. In this study, the layered structure, crystallization ratio, mineralogical species, and morphology
features of flux films were characterized by optical microscopy, X-ray diffraction, and electron-probe
microanalysis. Microstructural observation revealed that the normal flux films for peritectic steels
present a multilayered structure and high crystallization ratio (60~90 vol%), mainly composed of
well-developed crystalline akermanite and cuspidine. In contrast, the films with outstanding flux
characteristics with abundant longitudinal cracks on the slab surface have a low crystallization ratio
(<50 vol%) or vast crystallite content (>80 vol%). Furthermore, heat transfer analysis showed that
the low crystallization ratio and the vast crystallite content of flux films worsen the heat transfer
rate or uniformity in the mold, whereas the appropriate thickness and cuspidine content of flux
films can improve the heat transfer performance. From the above results, it is concluded that using
strong crystalline flux to obtain the ideal mineral phase structure of flux film is one of the important
measures for reducing longitudinal cracks during continuous casting of peritectic steel slabs.
Keywords: crystallization ratio; microstructure; flux film; heat transfer; longitudinal crack; peritectic
Citation: Liu, L.; Han, X.; Li, M.; steels
Zhang, D. Influence of Mineralogical
Structure of Mold Flux Film on Heat
Transfer in Mold during Continuous
Casting of Peritectic Steel. Materials 1. Introduction
2022, 15, 2980. https://doi.org/
Peritectic steel is widely used in manufacturing high-strength automobile plates and
10.3390/ma15092980
ship boards with excellent properties of thermoplasticity and machinability [1,2]. Because
Academic Editor: Adam Grajcar of the peritectic reaction during continuous casting of a peritectic steel slab, volume con-
traction and thermal stress congestion of the solidification shell always occur, more easily
Received: 17 March 2022
causing nonuniform heat transfer and longitudinal cracks [3–6]. Thus, avoiding longitudi-
Accepted: 13 April 2022
nal cracks during the continuous casting of peritectic steel slabs remains a challenge.
Published: 20 April 2022
In order to improve the slab quality, a large amount of research has been carried out on
Publisher’s Note: MDPI stays neutral steel solidification characteristics, continuous casting process conditions, and continuous
with regard to jurisdictional claims in casting flux [7–10]. Whereas it can be difficult to change the steel solidification characteristic
published maps and institutional affil- and the casting process condition, the mold flux has become a central factor preventing slab
iations. quality defects [11–15]. In-depth knowledge regarding the heat transfer mechanism of mold
flux films is becoming increasingly essential. Shibata et al. [16–18] suggested the average
thickness of mold flux films for various steel grades can be 0.5–1.75 mm and the interfacial
thermal resistance increases with increasing flux film thickness. Tsutsumi et al. [19–22]
measured the surface roughness of solidified flux films using a stylus surface profiler
and analyzed the surface roughness near the mold to correlate and evaluate the air gap
formation. Choi et al. [23–28] considered the crystallization behavior of mold fluxes to be
conditions of the Creative Commons vital in controlling the overall heat transfer in the mold by means of the mold simulator,
Attribution (CC BY) license (https:// differential thermal analysis, hot thermocouple technique, and confocal laser scanning
creativecommons.org/licenses/by/ microscope. Furthermore, many researchers found that the crystallization ratio of mold flux
4.0/). is a key factor in controlling the heat transfer in the mold [29–32]. However, the effect of

Materials 2022, 15, 2980
mineralogical structure on the heat transfer property of flux films has not been investigated
through a systematic approach.
In this work, the microstructure characteristics of flux films matched for peritectic
steels were investigated by optical microscopy, X-ray diffraction (XRD), and electron-probe
microanalysis (EPMA). Then, the effect of mineralogical structure on the heat transfer
property of flux films was analyzed. The aim of the present work is to ravel out the
relationship between the mineralogical structure of flux films and the longitudinal cracks of
slabs, and then to select the suitable mold flux for avoiding the longitudinal cracks during
continuous casting of peritectic steel slabs. The results will provide a theoretical basis for
optimizing mold flux and improving slab quality.
2. Experimental Section
2.1. Materials
Continuous casting experiments were carried out at Hebei Iron and Steel Group Co., Ltd.
in Shijiazhuang, China. Samples of flux films (see Figure 1) were taken from the meniscus
region (see Figure 2).
Figure 1. Samples of solidified flux films.
Figure 2. Position diagram of flux films in mold.
204
Materials 2022, 15, 2980
Combined with the surface quality of the slab, four typical samples of flux films were
selected. Sample No. 1 represents flux films with normal slab quality for peritectic steel A,
and sample No. 2 represents flux films with longitudinal cracks on the surface of the slab
for peritectic steel A. It was the same with sample No. 3 and sample No. 4 for peritectic steel
B. Some parameters of the typical flux films and the corresponding slab quality are listed in
Table 1. Chemical compositions and physical properties of mold fluxes for peritectic steels
A and B are listed in Tables 2 and 3.
Table 1. Parameters of mold flux films for peritectic steels.
Peritectic Steel Flux Film Thickness of Liquid Flux Consumption of Mold Flux Slab
Number Number Number (mm) (kg·t−1 ) Quality
No. 1 10.25 0.596 Normal
A F-A No. 2 10.41 0.568 Longitudinal crack
No. 3 10.33 0.582 Normal
B F-B No. 4 10.32 0.560 Longitudinal crack
Table 2. Chemical compositions of mold fluxes (wt%).
Peritectic Steel Flux

SiO2 Al2 O3 MgO CaO Fe2 O3 K2 O + Na2 O MnO F− C H2 O
Number Number
A F-A 28.99 2.74 4.17 35.83 0.57 7.65 1.73 7.46 4.88 0.33
B F-B 31.77 4.18 3.72 36.29 1.03 7.82 1.80 6.87 6.31 0.34
Table 3. Physical properties of mold fluxes.
Peritectic Steel Molten Heavy

Flux Number Alkalinity Melting Point (◦ C) Melting Speed (s) Viscosity (Pa·s)
Number (g/cm3 )
A F-A 1.24 1050 21 0.108 0.80
B F-B 1.14 1104 21 0.153 0.70
2.2. Methods
In this work, some samples of flux films from the same group were glued together
by epoxy adhesive, and their section along the thickness direction was glued to the slide,
ground to a thickness of 0.03 mm, and eventually made into a polished thin section. In
addition, other samples of flux films and mold fluxes were completely ground to 0.074 mm
before being used.
The thickness, mineralogical composition, crystallization ratio, and microstructure
of flux films were analyzed using the polished thin section with a polarizing microscope
(Axioskop A1 pol, Carl Zeiss Co., Ltd., Bangkok, Thailand). Then, using the polished thin
section with an electron-probe microanalyzer (JXA-8230, JEOL Co., Ltd., Tokyo, Japan), the
chemical compositions of different mineral crystals in flux films were confirmed.
The crystal phase composition of powder-like flux films was detected with an X-ray
diffractometer (BDX-3200, Bruker Co., Ltd., Seongnam-si, Korea) using Cu-Ka radiation
within the scanning range of 10◦ to 80◦ . The melting temperature and melting speed of
powder-like mold fluxes were measured using a melting point and melting speed tester
(RDS-04, Northeastern University in China).
The heat flow density (1673 K) and viscosity (1573 K) of mold fluxes were measured
using a mold heat flux simulation and viscosity tester (HF-201, Chongqing University in
China). The experimental procedure was as follows: First, 350 g prepared powder-like flux
was put in a graphite crucible and heated to 1673 K with the MoSi2 furnace of the tester.
Then, the water-cooled sensor was immersed into the liquid slag and measured 10 data
points of heat flow density within 45 s. Then, the sensor was taken out from the liquid slag,
and the flux film adhered to the sensor was obtained (probe diagram of mold heat flow
simulator can be seen in Figure 3). After a period of time, when the temperature dropped
205
Materials 2022, 15, 2980
to 1573 K, a standard rotating spindle was immersed into liquid slag and measured 10 data
points of viscosity.
Figure 3. The probe diagram of the mold heat flow simulator.
To reduce the error, three or more samples were prepared for each mold flux, and the
physical properties of each flux were determined from the average value of the measured
data of samples; the difference among the measured data was no greater than 3%.

3.1. Mineralogical Composition and Crystallization Ratio of Flux Films
As the formation of flux films in a mold is affected by complex factors, the composition,
content, morphology, and size of the crystalline minerals are obviously different, showing
the variant heat transfer capacity. When the heat transfer capacity of flux films is poor,
it will often lead to longitudinal cracks during casting crack-sensitive steel. In this work,
mineralogical compositions, proportions, and crystallization ratio of flux films for the
peritectic steels were quantitatively analyzed by using a polarizing microscope.
The polarizing microscope analysis results (Table 4 and Figure 4) show that the main
crystalline minerals of flux films for steel A are akermanite, cuspidine, and wollastonite.
The crystallization ratio of flux films for steel A varies widely. The crystallization ratio
of flux films with good slab quality reaches 85~90%, and the content of akermanite is as
high as 75~80%. The glass phase of flux films with longitudinal cracks on the slab surface
increases obviously, and their crystallization ratio is only 45~50%.
Table 4. Mineralogical compositions, proportions, and crystallization ratio of flux films (vol%).
Peritectic Steel Film Slab Crystallization

Cuspidine Akermanite Wollastonite Glass Phase
Number Number Quality Ratio
No. 1 Normal 10~15 75~80 1~5 5~10 85~90
A No. 2 Longitudinal crack 25~30 20~25 1~5 45~50 45~50
No. 3 Normal 40~45 20~25 <1 30~35 60~65
B No. 4 Longitudinal crack 12~17 80~85 <1 1~5 >95
In contrast, the main crystalline minerals of flux films for steel B are akermanite,
cuspidine, and a small account of wollastonite. The crystallization ratio of flux films with
good slab quality is as low as 60~65%, but the content of cuspidine is as high as 40~45%.
Though the crystallization ratio of flux films with longitudinal cracks on the slab surface is
as high as 95%, the content of well-shaped cuspidine is only 12~17%, and the remaining
content is akermanite crystallites.
206
Materials 2022, 15, 2980
D E
F G
H I
Figure 4. Mineralogical structures in flux films: (a) akermanite in film 1; (b) cuspidine in film 1;
(c) akermanite in film 2; (d) akermanite in film 3; (e) cuspidine in film 3; (f) akermanite in film 4.
According to the backscattered electron (BSE) micrographs and energy-dispersive

spectrometer (EDS) images from the electron-probe microanalyzer (Figures 5 and 6), it can
be found that a large number of granular crystals and interlaced crystals are melilite, and a
small number of spearhead-shaped crystals are cuspidine. Melilite has many isomorphisms
such as akermanite (2CaO•MgO•2SiO2 ) and gehlenite (2CaO•Al2 O3 •SiO2 ). Their optical
properties are similar, so it is difficult to distinguish them by microscope. However, it can
be seen from the XRD analysis results (Figure 7) that the melilite in the flux film mainly is
akermanite (2CaO•MgO•2SiO2 ).
The crystallization ability of flux films directly affects the heat transfer in the mold. The
flux film with low crystallization has low roughness, which makes thermal conductivity
become high, and low precipitation of minerals with higher melting points decreases
the interfacial thermal resistance, which may lead to surface longitudinal cracks in the
peritectic steel. The mineralogical structure of the flux film determines the quality and
production efficiency of the slab. The crystallization ratio of the flux film No. 2 for steel A
with longitudinal cracks on the slab surface is obviously lower, worsening the heat transfer
uniformity in mold, which is one of the causes of longitudinal cracks of slabs. However,
207
Materials 2022, 15, 2980
the cuspidine content of the flux film No. 4 for steel B with longitudinal cracks on the slab
surface is only 12~17%, which may bring about an excessive heat transfer rate causing
longitudinal cracks to occur on the surface of the slab. Moreover, the vast akermanite
crystallite content in the flux film No. 4 can cause heat transfer nonuniformity in the mold.
Figure 5. BSE micrographs of flux films by EPMA.
Figure 6. EDS spectra of different mineral crystals in flux films by EPMA.
Figure 7. XRD results of flux films.
208
Materials 2022, 15, 2980
3.2. Microstructure of Flux Films

The layered structures, mineralogical species, and morphology features of flux films
were analyzed and determined using polarizing microscopy. The layered structure of
normal flux film for steel A is obvious, and the crystallizing layer accounts for a large area
of the flux film (Figure 8a). The normal flux film of steel B exhibits a typical two-layered
structure, and part of the flux film has a multilayered structure (Figure 8c). Compared
with the glass layer near the mold side of the two normal flux films, the flux film of steel
A has a higher degree of discontinuity than the flux film of steel B. The layered structure
of flux film for steel A with longitudinal cracks is a typical three-layered structure with
glass–crystalline–glass layers (Figure 8b). The crystallizing layer of flux film for steel B with
longitudinal cracks accounts for all the flux film (Figure 8d).
D E
F G
Figure 8. Layered structures of flux films: (a) film 1; (b) film 2; (c) film 3; (d) film 4; A—the mold side;
B—the shell side; 1—crystallization layer; 2—glass layer.
The research results show that the flux films have differences in microstructure, es-
pecially in the mineral morphology and particle size. The akermanite and wollastonite in
the crystal layer of the flux films are distributed alternately on the shell side, which has an
obvious boundary with the cuspidine. The akermanite of the flux films for steel A is more
prone to be intertexture-shaped near the shell side and to have a radial and chrysanthemum
shape on the mold side. However, the akermanite of flux films for steel B is particle-shaped
crystallites with small grain size, and the coarse granular and spearhead-shaped cuspidine
of the flux films is concentrated on the mold side.
3.3. Heat Transfer Property of Flux Films

The heat flux density can be defined as the amount of heat that passes through the
flux film per unit cross-sectional area and per unit time. In this study, the heat flow density
of mold fluxes was measured by using the mold heat flux simulator when the temperature
was maintained at 1673 K, and the measurements represent maximum and minimum heat
transfer and heat transfer uniformity. The heat flow density of mold fluxes for peritectic
steels was measured at different times by the simulator, as shown in Figure 9. The maximum
heat flow density represents the heat transfer capacity of flux films at the meniscus, and the
209
Materials 2022, 15, 2980
average heat flow density reflects the heat transfer capacity of flux films in the middle and
upper part of the mold. The heat transfer test revealed that the heat flow density decreases
gradually as the cooling water pass time is increased. The maximum and average heat flow
densities of flux films for steel A are higher than those of flux films for steel B, and the same
law applies to the heat transfer property for these films.
Figure 9. The heat flux density of mold flux films at different times.
3.4. Discussion
The relationship between the heat flow density and the mineralogical structure of flux
films is presented in Figure 10. It can be seen that the greater the thickness of the slag film
is, the lower the hindrance to the heat flux is. The greater the crystalline phase of flux film,
the less the obstruction to the heat flow. Furthermore, the high crystallization of the flux
film promotes an increase in its solidification thickness. However, early studies [33,34] on
heat transfer ignored the influence of the mineralogical composition and proportion of
flux films in the casting of peritectic steels. Instead, these studies considered that more
resistance to heat transfer arises from the higher crystallization ratio or thickness of flux
films. Thus, they were unable to explain the phenomenon of heat flow density dropping
and becoming smaller than expected at the lower crystallization ratio and film thickness.
As shown in Figure 10, the thickness and the heat flow density of flux films for steel A
are higher than those of flux films for steel B. According to the thermal conductivity of
minerals, cuspidine has the lowest thermal conductivity among all the mineral phases of
flux films. With respect to the mineralogical structure of flux films for steel B, the content of
cuspidine is as high as 40~45%, which can increase the thermal resistance of the flux film
and maintain the suitable heat transfer rate and uniformity in mold all the time. In addition,
there is an exception in Figure 10: the heat flow density drops and becomes smaller at the
90% crystallization ratio of the flux film; it seems perfectly natural to ascribe this to the
bigger thickness and higher crystallization ratio of the flux film with the lower impact of
cuspidine content. Therefore, considering the effect of the mineralogical structure of flux
films on heat transfer in the mold and slab quality, this study carried out a comprehensive
work and found that the mineralogical compositions and proportions can also be a key
factor in controlling heat transfer suitable for the slab quality, so it is different from previous
investigators’ observations which incorrectly ascribed this effect to the crystallization ratio
or thickness of flux films.
210
Materials 2022, 15, 2980
Figure 10. Relationship between the heat flow density and mineralogical structure of flux films.
Flux films of peritectic steel with longitudinal cracks generally have lower crystalliza-
tion ratios and crystal growth levels than the normal flux films. The direct cause of vast
longitudinal cracks on the peritectic steels is the abnormal microstructure of flux films,
which may be caused by the inappropriate chemical components of flux and the cooling
condition of the mold. The crystallization ratio of the unqualified films for steel A with
longitudinal cracks may be only 45~50%. Compared with the reasonable flux films for steel
B, the unqualified films have a mass of akermanite crystallite and the crystallization can
abnormally reach 80% or more. All of these results suggest that the primary reasons for
longitudinal cracking on the slab surface of the peritectic steels are the low crystallization
ratio and the vast crystallite content, which can decrease the thermal resistance of flux films
and worsen the heat transfer rate and uniformity.
In view of the strong crack sensitivity of peritectic steel, the longitudinal crack ratio
of the slabs can be reduced by improving the mineralogical structure of the flux films.
However, if the heat transfer is controlled only by enhancing the crystallization capacity, the
balance between the heat transfer and lubrication properties of the flux films may be lost.
So, during continuous casting of peritectic steels, the contradiction between heat transfer
and lubrication should be considered comprehensively, and the optimum mineralogical
structure of flux films should be obtained through reasonable composition regulation of the
mold flux. In this work, it is recommended that the content of (CaO/SiO2 ) or (Na2 O + K2 O)
of the mold flux for steel A should be properly increased. While maintaining high alkalinity
(CaO/SiO2 ), the CaF2 content of the mold flux for steel B should be increased appropriately.
4. Conclusions
The influence of the mineralogical structure of mold flux films on heat transfer in the
mold during the continuous casting of peritectic steels for avoiding the longitudinal cracks
of slabs has been investigated. The following conclusions can be drawn:
(1) The mineralogical structure of normal flux films during the continuous casting of peritec-
tic steels presents a multilayered structure with a high crystallization ratio (60~90 vol%),
mainly composed of well-developed crystalline akermanite and cuspidine.
(2) Flux films for peritectic steels with longitudinal cracks have a lower crystallization ra-
tio and crystal growth level than the normal flux films have, showing the characteristic
of low crystallization ratio (<50 vol%) or vast crystallite content (>80 vol%).
(3) The results from heat transfer analysis confirmed that the mineralogical composition
and proportion of flux films is a key factor in controlling heat transfer during the
casting of peritectic steels. The more glass phase and crystallite flux films have, the
worse the heat transfer rate and uniformity in the mold are.
211
Materials 2022, 15, 2980
(4) Using strong crystalline flux to obtain an ideal mineral phase structure of flux films
is one of the important measures for reducing the longitudinal cracking of peritectic
steels, and the ideal mineral phase is characterized by a high crystallization ratio
(>60 vol%), without the vast crystallite content, and as much cuspidine content
as possible.
Author Contributions: Conceptualization, L.L. and X.H.; methodology, formal analysis, L.L. and
M.L.; investigation, resources, L.L. and D.Z.; data curation, M.L. and D.Z.; writing—original draft
preparation, L.L.; writing—review and editing, L.L. and X.H.; visualization, M.L. and D.Z.; supervi-
sion, X.H.; project administration, L.L. and X.H. All authors have read and agreed to the published
51774140 and No. 51574105), the Operation Expenses for Universities’ Basic Scientific Research
of Hebei Province (No. JQN2020038), and the Postgraduate Innovation Funding Project of Hebei
Province (No. CXZZBS2021101).
References
1. Hechu, K.; Slater, C.; Santillana, B.; Sridhar, S. The use of infrared thermography to detect thermal instability during solidification
of peritectic steels. Mater. Charact. 2019, 154, 138–147. [CrossRef]
2. Long, X.; He, S.P.; Xu, J.F.; Huo, X.L.; Wang, Q. Properties of high basicity mold fluxes for peritectic steel slab casting. ISIJ Int.
2012, 19, 39–45. [CrossRef]
3. Suzuki, M.; Yamaoka, Y. Influence of carbon content on solidifying shell growth of carbon steels at the initial stage of solidification.
Mater. Charact. 2003, 44, 836–844. [CrossRef]
4. Liu, T.; Long, M.J.; Chen, D.F.; Wu, S.X.; Tang, P.M.; Liu, S.; Duan, H.M.; Yang, J. Investigations of the peritectic reaction and
transformation in a hypoperitectic steel: Using high-temperature confocal scanning laser microscopy and differential scanning
calorimetry. Mater. Charact. 2019, 156, 109870. [CrossRef]
5. Jeong, H.; Hwang, J.Y.; Cho, J.W. In-depth study of mold heat transfer for the high speed continuous casting process. Met. Mater.
Int. 2016, 22, 295–304. [CrossRef]
6. Hanao, M.; Kawamoto, M.; Yamanaka, A. Influence of mold flux on initial solidification of hypo-peritectic steel in a continuous
casting mold. ISIJ Int. 2012, 52, 1310–1319. [CrossRef]
7. Zhao, H.; Wang, X.; Zhang, J. Research on mold flux for hypo-peritectic steel at high casting speed. Int. J. Miner. Metall. Mater.
2007, 14, 219–224. [CrossRef]
8. Wang, W.L.; Gu, K.; Zhou, L.J. Radiative heat transfer behavior of mold fluxes for casting low and medium carbon steels. ISIJ Int.
2011, 51, 1838–1845. [CrossRef]
9. Park, J.Y.; Ko, E.Y.; Choi, J.; Sohn, I. Characteristics of medium carbon steel solidification and mold flux, crystallization using the
multi-mold simulator. Met. Mater. Int. 2014, 20, 1103–1114. [CrossRef]
10. Ko, E.Y.; Choi, J.; Park, J.Y.; Sohn, I. Simulation of low carbon steel solidification and mold flux crystallization in continuous
casting using a multi-mold simulator. Met. Mater. Int. 2014, 20, 141–151. [CrossRef]
11. Zhang, Y.F.; Han, X.L.; Liu, L.; Liu, L.N. Effect of mineralogical structure of flux film on slab quality for medium carbon steel.
Trans. Indian Inst. Met. 2018, 71, 1803–1807. [CrossRef]
12. Nakada, H.; Susa, M.; Seko, Y.; Hayashi, M.; Nagata, K. Mechanism of heat transfer reduction by crystallization of mold flux for
continuous casting. ISIJ Int. 2008, 48, 446–453. [CrossRef]
13. Wen, G.H.; Tang, P.; Yang, B.; Zhu, X.B. Simulation and characterization on heat transfer through mould slag film. ISIJ Int.
2012, 52, 1179–1185. [CrossRef]
14. Liu, Y.Z.; Wang, W.L.; Ma, F.J.; Zhang, H.H. Study of solidification and heat transfer behavior of mold flux through mold flux heat
transfer simulator technique: Part I. Development of the technique. Metall. Mater. Trans. B 2015, 46, 1419–1430. [CrossRef]
15. Mills, K.C. Structure and properties of slags used in the continuous casting of steel: Part 1. Conventional mould powders. ISIJ Int.
2016, 56, 1–13. [CrossRef]
16. Shibata, H.; Kondo, K.; Suzuki, M.; Emi, T. Thermal resistance between solidifying steel shell and continuous casting mold with
intervening flux film. ISIJ Int. 1996, 36, 79–81. [CrossRef]
17. Mills, K.C.; Fox, A.B. The role of mould fluxes in continuous casting-so simple yet so complex. ISIJ Int. 2003, 43, 1479–1486.
[CrossRef]
212
Materials 2022, 15, 2980
18. Ma, F.J.; Liu, Y.Z.; Wang, W.L.; Zhang, H.H. Study of solidification and heat transfer behavior of mold flux through mold
flux heat transfer simulator technique: Part II. Effect of mold oscillation on heat transfer behaviors. Metall. Mater. Trans. B
2015, 46, 1902–1911. [CrossRef]
19. Tsutsumi, K.; Nagasaka, T.; Hino, M. Surface roughness of solidified mold flux in continuous casting process. ISIJ Int.
1999, 39, 1150–1159. [CrossRef]
20. Wen, G.H.; Zhu, X.B.; Tang, P.; Yang, B.; Yu, X. Influence of raw material type on heat transfer and structure of mould slag. ISIJ Int.
2011, 51, 1028–1032. [CrossRef]
21. Ozawa, S.; Susa, M.; Goto, T.; Endo, R.K.; Mills, K.C. Lattice and radiation conductivities for mould fluxes from the perspective of
degree of crystallinity. ISIJ Int. 2006, 46, 413–419. [CrossRef]
22. Cho, J.W.; Blazek, K.; Frazee, M.; Yin, H.; Park, J.H.; Moon, S.W. Assessment of CaO-Al2 O3 based mold flux system for high
aluminum TRIP casting. ISIJ Int. 2003, 53, 62–70. [CrossRef]
23. Choi, S.Y.; Lee, D.H.; Shin, D.W.; Choi, S.Y.; Cho, J.W.; Park, J.M. Properties of F-free glass system as a mold flux: Viscosity,
thermal conductivity and crystallization behavior. J. Non-Cryst. Solids 2004, 345, 157–160. [CrossRef]
24. Han, X.L.; Zhang, Y.F.; Liu, L.; Liu, L.N. Effects of soda ash on properties of mold flux and mineralogical structures of flux film.
J. Iron Steel Res. Int. 2016, 23, 197–202. [CrossRef]
25. Lei, Y.; Xie, B.; Ma, W.-H. Analysis of crystallization behavior of mold fluxes containing TiO2 using single hot thermocouple
technique. J. Iron Steel Res. Int. 2016, 23, 322–328. [CrossRef]
26. Zhang, L.; Wang, W.L.; Sohn, I.L. Crystallization behavior and structure analysis for molten CaO-SiO2 -B2 O3 based fluorine-free
mold fluxes. J. Non-Cryst. Solids 2019, 511, 41–49. [CrossRef]
27. Zhang, S.D.; Li, M.; Zhu, L.L.; Wang, Q.Q.; He, S.P.; Wang, Q. Effect of substituting Na2 O for SiO2 on the non-isothermal
crystallization behavior of CaO-BaO-Al2 O3 based mold fluxes for casting high Al steels. Ceram. Int. 2019, 45, 11296–11303.
[CrossRef]
28. Kölbl, N. New mold slag compositions for the continuous casting of soft steels. Steel Res. Int. 2021, 93, 2100165. [CrossRef]
29. Wang, W.; Blazek, K.; Cramb, A. A Study of the crystallization behavior of a new mold flux used in the casting of transformation
induced plasticity steels. J. Metall. Mater. Trans. 2008, 39, 66–74. [CrossRef]
30. Kang, Y.; Lee, J.; Morita, K. Thermal conductivity of molten slags: A review of measurement techniques and discussion based on
microstructural analysis. ISIJ Int. 2014, 46, 2008–2016. [CrossRef]
31. Han, X.L.; Zhang, Y.F.; Liu, L.; Liu, L.N. Influence of chemical composition of mold flux on viscosity and texture of slag film.
Environ. Toxicol. Chem. 2016, 98, 511–517. [CrossRef]
32. Costaa, I.T.; Carvalho, C.S.; Oliveira, J.R.; Vieirab, E.A. Physical properties characterization of a peritectic mold flux formed from
the addition of calcitic marble residue in the commercial one. J. Mater. Res. Technol. 2019, 8, 3297–3305. [CrossRef]
33. Cho, J.W.; Emi, T.; Shibata, H.; Suzuki, M. Heat transfer across mold flux film in mold during initial solidification in continuous
casting of steel. ISIJ Int. 1998, 38, 834–842. [CrossRef]
34. Cho, J.W.; Shibata, H.; Emi, T.; Suzuki, M. Thermal resistance at the interface between mold flux film and mold for continuous
casting of steels. ISIJ Int. 1998, 38, 440–446. [CrossRef]
213
materials
Article
The Effect of Heat Source Path on Thermal Evolution during
Electro-Gas Welding of Thick Steel Plates
Jun Fu 1,2 , Qing Tao 1,3 , Xiaoan Yang 1,2 , Bogdan Nenchev 1 , Ming Li 2 , Biao Tao 2 and Hongbiao Dong 1, *
1 School of Engineering, University of Leicester, Leicester LE1 7RH, UK; [email protected] (J.F.);
[email protected] (Q.T.); [email protected] (X.Y.); [email protected] (B.N.)
2 Nanjing Iron & Steel United Co., Ltd., Nanjing 210035, China; [email protected] (M.L.);
[email protected] (B.T.)
3 School of Materials and Physics, China University of Mining and Technology, Xuzhou 221116, China
* Correspondence: [email protected]; Tel.: +44-(0)7718369708
Abstract: In recent years, the shipbuilding industry has experienced a growing demand for tighter
control and higher strength requirements in thick steel plate welding. Electro-gas welding (EGW) is
a high heat input welding method, widely used to improve the welding efficiency of thick plates.
Modelling the EGW process of thick steel plates has been challenging due to difficulties in accurately
depicting the heat source path movement. An EGW experiment on 30 mm thickness E36 steel
plates was conducted in this study. A semi-ellipsoid heat source model was implemented, and
its movement was mathematically expressed using linear, sinusoidal, or oscillate-stop paths. The
geometry of welding joints, process variables, and steel composition are taken from industrial scale
experiments. The resulting thermal evolutions across all heat source-path approaches were verified
against experimental observations. Practical industrial recommendations are provided and discussed
in terms of the fusion quality for E36 steel plates with a heat input of 157 kJ/cm. It was found that
the oscillate-stop heat path predicts thermal profile more accurately than the sinusoidal function and
linear heat path for EGW welding of 30 mm thickness and above. The linear heat path approach is
Citation: Fu, J.; Tao, Q.; Yang, X.; recommended for E36 steel plate thickness up to 20 mm, whereas maximum thickness up to 30 mm is
Nenchev, B.; Li, M.; Tao, B.; Dong, H. appropriate for sinusoidal path, and maximum thickness up to 35 mm is appropriate for oscillate-stop
The Effect of Heat Source Path on path in EGW welding, assuming constant heat input.
Thermal Evolution during
Electro-Gas Welding of Thick Steel Keywords: electro-gas welding; high heat input; heat source movement path; finite element analysis;
Plates. Materials 2022, 15, 2215.
thermal evolution
https://doi.org/10.3390/ma15062215
Academic Editor: Adam Grajcar
Received: 23 January 2022 1. Introduction

In the shipbuilding industry and offshore engineering, the size and structures of ships
Published: 17 March 2022
are increasing rapidly [1,2], thus giving rise to a growing demand for higher control in
Publisher’s Note: MDPI stays neutral welding thick steel plates [3,4]. Electro-gas welding (EGW) has become an indispensable
with regard to jurisdictional claims in welding method for shipbuilding enterprises due to its outstanding advantages, such as
published maps and institutional affil- large heat input and high welding efficiency [5]. As illustrated in Figure 1, EGW is an
iations.
automatic welding process using a special flux-cored wire with CO2 gas protection, and
used for the welding of vertical position of steel plates [6]. During the welding process,
the torch moves along the chosen weld path from bottom to top. A water-cooled copper
slider is placed on the front of the weld, and a ceramic “backing” plate is positioned at the
rear. The welding torch is also used to feed the welding metal into the groove. In thick
plates welding, the welding pool is restrained by the weld pieces, backing plate and sliding
copper shoe, so that single pass welding can be accomplished.
4.0/).

Materials 2022, 15, 2215
Figure 1. The electro-gas welding (EGW). (a) The 3D schematic diagram of EGW equipment, (b) The
2D schematic diagram of EGW equipment, (c) Photographs of EGW equipment, (d) Photographs
of experimentation.
In EGW, heat transfer from the heat source to the plates occurs mainly through
radiation and convection [7,8]. In thick steel plate welding, the efficiency of heat transfer
in the “y” direction (with depth, Figure 1) of single pass EGW is a key factor affecting the
quality of the joint [9,10]. In 2016, Hwang, Kim and Lee [11] introduced double ellipsoidal
moving heat sources to model the temperature profile and residual stress distribution in
EGW of marine steel. However, the ellipsoidal method oversimplifies the oscillate-stop
heat source path of industrial welding process to 1 dimensional linear case, whereby the
movement path in the “y” direction is not considered. Therefore, the heat source paths are
inconsistent with the real EGW process, hence the predicted fusion lines do not agree with
experimental results.
To include an oscillate-stop heat source path within weld modelling, in 2017, Xu, Pan
and Wan [12] used a sinusoidal function to simulate metal active gas arc welding (GAW). In
2018, Yuan et al. [13] applied a piecewise function to model the real oscillate movement path
of gas metal arc welding. In 2019, da Silva Pereira et al. [14] implemented weave patterns
by path parameterization, which improved the accuracy and has been further extended
to predict defects in welding [15]. However, the effect of different heat source paths on
thermal evolution in modelling EGW of thick steel plates and further optimize heat source
paths according to thickness of steel plate has not yet been investigated systematically.
This study investigates the effect of three different heat source paths (linear, sinusoidal
and oscillate-stop) through FEA on the resulting thermal evolution in thick steel plate
welding. The results for EGW welding of E36 marine steel plates with thicknesses ranging
from 20–40 mm and a heat input of 157 kJ/cm are compared based on fusion efficiency. The
FEA models are verified by comparing against experimental observations. Finally, practical
216
Materials 2022, 15, 2215
industrial recommendations are provided for each heat source path’s approach, according
to variation in plate thickness.
2. Materials and Model Description

2.1. Electro-Gas Welding Experiments
This study investigates the effect of heat source path on thermal evolution during
EGW of E36 marine steel plates. For the FEA modelling, experimental parameters, such as
the geometry of the welding joints, thermo-mechanical and EGW process variables, and
steel composition were considered as inputs, listed in Table 1. The temperature variation
of material properties such as density and thermal conductivity of the E36 steel plate was
also taken into account. JMatPro was implemented to calculate the material properties of
the steel [16,17] for the E36 steel composition, shown in Table 2. In addition to the five
main elements: carbon, manganese, phosphorus, and sulphur, microalloying elements such
as niobium, titanium, and aluminium were added to improve the mechanical properties
of steel. The dimensions of the steel plate model are shown in Figure 2. The thickness
of the steel plate is 30 mm, while the length and width of the steel plate are 800 mm and
250 mm, respectively.
Table 1. Parameters of welding experiments.
Heat Input Welding Speed Current Voltage Wire Diameter Wire Feed Rate
(kJ/cm) (cm/min) (A) (V) (mm) (m/min)
157 6.9 420 43 1.6 13
Parameters of the torch movement of horizontal oscillations
Internal stop External stop Amplitude Frequency
(s) (s) (mm) (s−1 )
1 1.5 10 2/11
Figure 2. Geometry of welded steel plate.
217
Materials 2022, 15, 2215
Table 2. Chemical composition of E36 steel (wt %).
C Si Mn S P Nb Ti Al
0.08 0.27 1.45 0.002 0.01 0.014 0.014 0.034
In the electro-gas welding process of 30 mm thick steel E36, flux-cored wire JIS Z3319
DW-S60G with a diameter of 1.6 mm was used. The chemical composition and the proper-
ties of the weld deposit are presented in Tables 3 and 4.
Table 3. Chemical composition of weld deposit.
Grade C Mn Si S P Cr Ni Mo B Ti
DW-S60G 0.07 1.68 0.33 0.006 0.011 0.02 0.77 0.26 / 0.02
Table 4. Mechanical properties of weld deposit.
Yield Strength Rp0.2 Tensile Strength Rm Elongation A Akv at −20 ◦ C

(MPa) (MPa) (%) (J)
534 662 26 124, 139, 120
2.2. Three-Dimensional EGW Model for the Numerical Simulation

Heat transfer during welding is a complex process dependent on multiple thermo-
physical and geometrical factors. In EGW, heat from the heat source is dissipated to the
welding plates mainly by conduction, radiation, and convection. Thus, a balance must
be found between the rate of heat generation (speed and energy of EGW) and the rate of
dissipation, determined by the case specific geometry and chemistry.
The heat transfer equation is given as follows:
∂T
ρC p + ∇(−k∇ T ) = Q − Qr − Qc (1)
∂t
where ρ—density of the material; C p —specific heat capacity; T—absolute temperature;
t—time; k—thermal conductivity; Q—heat source; Qr —radiation heat loss; Qc —convection
heat loss. During the EGW process, the peak temperature of the heat-affected zone can
reach more than 2000 ◦ C, thus the temperature gradient between the steel plate and the
environment is massive, causing significant radiation. The heat dissipation terms for
radiation and convection are:

− Qr = εσ Tamb
4
− T4 (2)
− Qc = h f ( Tamb − T ) (3)
where ε—thermal emissivity; σ—Stephan Postman’s constant: 5.67 × 10−8 [W/(m2 ·K4 )];
Tamb —ambient temperature [K]; T—steel plate surface temperature [K]; qr —radiant heat
transfer flux [W/m2 ]; qc —convective heat transfer flux [W/m2 ]; h f —convection heat
transfer coefficient between welding parts and the environment.
A three-dimensional EGW model was implemented, as shown in Figure 3a,b. This
model uses a free tetrahedral mesh with the following principles designed for improving
the computational efficiency:
• The weld region close to the weld heat source has a larger temperature gradient and,
hence, was divided into a finer grid (minimum mesh tetrahedral edge length 7.67 mm).
• A coarser mesh (maximum mesh tetrahedral edge length 24.24 mm) was applied to
the base steel plate areas far away from the heat source, where a smaller change in
thermal gradients occurred.
218
Materials 2022, 15, 2215
Figure 3. (a) Geometry and mesh of EGW model implemented in this study, (b) the transient
temperature distribution simulated of steel plate at 520 s, (c) schematic diagram welding joint of
pre-welding bevel, (d) welding joint of after-welding showing the fusion line.
Figure 3b illustrates the thermal profile and fusion line of a steel plate during the
EGW process at 520 s at the midsection. The thermal distribution at surface and side of the
steel plate is available from the diagram. Experimental fusion lines are used to verify the
simulation results. A schematic diagram of the pre-welding bevel and micrograph of real
joint after welding is shown in Figure 3c,d, respectively.
The filling of weld metal gradually with the movement of heat source is considered.
The study simulates the ‘activate’ and ‘inactive’ of weld metal. Geometry of weld metal is
pre-drawn in the work piece gap and properties of weld metal are activate point by point
with the heat source movement.
For the welded area of the weld metal, the material properties of the geometry are activated.
For the unwelded area of the weld metal, the material properties of the geometry are
inactivated, as shown in Figure 4.
219
Materials 2022, 15, 2215
Figure 4. Schematic diagram of the activation process of the weld metal.
2.3. Heat Source Movement Path

2.3.1. Heat Source Model
The earliest and simplest heat source model is the point heat source model proposed
by Rosenthal [18] and it has been widely used in welding simulation. Rosenthal’s heat
source model applies a quasi-steady state 3D semi-infinite geometry for point source.
A velocity term was added to Rosenthal’s model to simulate heat source movement by
Lecoanet et al. [19–21]. To describe the heat source distribution, a double ellipsoidal heat
source model [22,23] was proposed by Rouquette et al., combining two different ellipses,
one in the front quadrant and the other in the rear quadrant [24,25]. Laser Welding Processes
have been simulated by a double ellipsoidal heat source model in recent years [26–28]. The
two heat source models, i.e., Gaussian, semi-ellipsoid and double ellipsoidal heat source
model, predict similar temperature distribution and distortion [29,30]. A simplified Gaus-
sian heat source model was to improve calculation efficiency by Cai and Norman [31,32].
In this study, the semi-ellipsoid heat source model was selected. The equation of
semi-ellipsoid heat source model is shown below:

6Q 3 x 2 + y2 + z2
q( x, y, z) = √ exp − (4)
πr3 π r2
where q( x, y, z) represents the heat flow density distribution of ( x, y, z), Q is the effective
power of the arc, r is the radius of the semi-ellipsoid heating source. The schematic diagram
of semi-ellipsoid heat source is shown in Figure 5.
Figure 5. Schematic diagram of heat source (semi-ellipsoid heat source).
220
Materials 2022, 15, 2215
2.3.2. Heat Source Movement Path

The movement of the heat source is expressed using a linear, sinusoidal and oscillate-
stop path. In this study, the movement speed in the z-axis is defined as v1 and in the y-axis
as v2 . As shown in Figure 6, three welding paths are modelled. Their exact coordinate heat
source locations at each time step are expressed in terms of v1 , v2 and then substituted in
Equation (4).
Figure 6. Schematic diagram of heat source path. (a) linear heat source path, (b) sinusoidal path heat
source, (c) oscillate-stop heat source path.
1. Linear path heat source

The linear path heat source considers the welding heat source moving in the welding
direction with a welding speed of v1 ; the speed of the y-axis direction is zero. The expression
of welding heat flux for its movement along the welding direction is shown in Equation (5):
⎛ ⎞
6Q 3 x 2 + y2 + ( z − v1 t )2
q( x, y, z) = √ exp⎝− ⎠ (5)
πr3 π r2
The simulation of the welding heat source moves from bottom to top along the welding
direction with the welding speed during the electro gas welding process. The heat source
moves along the centre of the steel plates for different thicknesses, with a welding speed of
6.9 cm/min, as given in Table 1.
2. Sinusoidal path heat source
The heat source, moving both in the welding direction with speed v1 and in the
thickness direction with speed v2 , is considered in the sinusoidal path heat source model.
The schematic diagram of the path is shown in Figure 6b.
The defined oscillate range is d, the length of the weld beam is L, the welding speed is
v1 , and the oscillate speed of the heat source in the thickness direction is v2 ; the equation of
the welding period is:
2d
T= (6)
v2
Total welding time:
L
t= (7)
v1
The heat source position of the y-axis direction is considered as:
d 2πt
y= sin (8)
4 T
221
Materials 2022, 15, 2215
Steel plates with three different thicknesses of 30 mm, 35 mm and 40 mm were used in
this study to investigate the applicable thickness for different heat source movement paths.
The parameters of the sinusoidal for different thickness steel plates are shown in Table 5.
Table 5. Sinusoidal path parameters of welding heat source.
Thickness of Plate Amplitude (A) Period (T) Oscillate Range Offset (k)
No.
(mm) (mm) (s) (mm) (mm)
1 30 10.0 5.5 20 20
2 35 12.5 5.5 25 20
3 40 15.0 5.5 30 20
According to the parameters of the sinusoidal heat source movement path, the po-
sition of the heat source changes with time during the welding process, as described in
Equation (10). The amplitude of sinusoidal function is 10 mm, 12.5 mm and 15 mm for
three thick plates, the period of the sinusoidal cycle is 5.5 s, respectively. The oscillate range
is 20 mm, 25 mm and 30 mm. Due to the heat source needing to be close to the surface
while leaving a certain distance for technological factors, the distance of 5 mm to the top
surface and the bottom surface is appropriate.
! "
2πt
P1 (t) = A sin +k (9)
T
The welding heat flux of the sinusoidal path is described as:

⎛ ⎞
6Q 3 x2 + (y − P1 (t))2 + (z − v1 t)2
q( x, y, z) = √ exp⎝− ⎠ (10)
πr3 π r2
According to Equation (10), the position of heat source change with time for three
thickness plates are shows in Figure 7. The y-axis is the position along the thickness of
steel plates and the z-axis is the position along the welding direction from the bottom to
the top of steel plates. The movement path of the heat source under three thicknesses is
drawn using solid lines with red, blue and green colour and the three thickness of 30 mm,
35 mm and 40 mm steel plates are indicated as dotted lines with red, blue and green colour,
respectively. The positions of the heat sources for 1 s, 5.5 s, 6 s, 7 s and 11 s are marked
in Figure 7.
3. Oscillate-stop heat source
According to the characteristics of the oscillate-stop path, the heat source not only
moves from bottom to top during the welding process but also oscillates in the direction
of the depth of the melt pool, which helps the welding bevels on both sides to obtain the
same melting depth. According to the welding practice, the oscillate-stop parameters of
the heat source for different thickness steel plates are shown in Table 6.
Table 6. Oscillate-stop parameters of welding heat source.
Thickness Internal Stop External Stop Oscillate Range (R) Oscillate Center (C) Period (T)
No.
(mm) (s) (s) (mm) (mm) (s)
1 30 1.0 1.5 20 20 5.5
2 35 1.0 1.5 25 20 5.5
3 40 1.0 1.5 30 20 5.5
222
Materials 2022, 15, 2215
Figure 7. Sinusoidal heat source movement vs. time (t) for 30 mm, 35 mm and 40 mm thick steel plates.
Assume that the oscillate range is R, the distance of the oscillate centre to the origin
of the coordinates is C, the period of the cycle is T, cycles is n. according to parameters of
the oscillate-stop path; the position of heat source relative to the steel plate with different
thickness in the welding period is described as Equation (12):
⎧
1.5 t + 10 − 5.5( n − 1) 1.5 , T (n − 1) < t ≤ T (n − 1) + 1.5
R R
⎪
⎪
⎪
⎨ C+ 2,R
(n − 1) + 1.5 < t ≤ T (n − 1) + 2.5
T
P2 (t) = (11)
⎪
⎪ − 1.5
R
t + 30 − [5.5(n − 1) + 2.5] − 1.5 R
, T (n − 1) + 2.5 < t ≤ T (n − 1) + 4
⎪
⎩
C− 2,R
T (n − 1) + 4 < t ≤ T (n − 1) + 5.5
Regarding their thickness, the oscillate centre C = 20 mm and period T = 5.5 s are
constant, so the equation can be simplified to:
⎧
1.5 ( t − 5.5n + 5.5) + 10, 5.5n − 5.5 < t ≤ 5.5n − 4
R
⎪
⎪
⎨
20 + R2 , 5.5n − 4 < t ≤ 5.5n − 3
P2 (t) = (12)
⎪
⎪ − 1.5
R
(t − 5.5n + 30) + 30, 5.5n − 3 < t ≤ 5.5n − 1.5
⎩
20 − R2 , 5.5n − 1.5 < t ≤ 5.5n
Define a function for the cycle of the equation, P(mod(t, 5.5)) means the equation cycle
one time every 5.5 s. The welding heat flux of oscillate stop path be described as:
⎛ ⎞
6Q 3 x2 + (y − P2 (mod(t, 5.5))2 + (z − v1 t)2
q( x, y, z) = √ exp⎝− ⎠ (13)
πr3 π r2
223
Materials 2022, 15, 2215
The position of the heat source changes with time for three thickness plates, as shown
in Figure 8.
Figure 8. Movement path of oscillate-stop for thickness of 30 mm, 35 mm and 40 mm.
3. Model Verification via Experiment

To verify the accuracy of the model, a cross-section is taken along the thickness (y-axis)
of the weld joint. As shown in Figure 9, the calculated melting pool and fusion line using
different heat source path modes are compared against the experimental welding joints.
The model fusion line is taken at the material melting point, 2055 K.
(1) As shown in Figure 9b, the application of the linear heat source path did not cause
sufficiently high temperatures to melt the top and bottom of the joint; only the centre
part of joint is melted, which differs to the experimental observation.
(2) The application of the sinusoidal path (see Figure 9c) leads to a fully melted weld
joint in the thickness direction. The fusion line extended to the outside of the welding
groove, but the area covered by the fusion line is smaller than that observed in the
experiments (see Figure 9a).
(3) The application of the oscillate-stop path leads to a fully melted weld joint in the
thickness direction. Additionally, the fusion line extended to the outside of the
welding groove base metal near the fusion is melted to form a solid joint, and the area
covered by the fusion line is similar to that observed in the experiments.
Quantitative analysis is carried out to measure the coordinates of the points of inter-
section between the fusion lines and top and bottom surface (marked as A, B, C and D in
Figure 9). The coordinates for points A, B, C, and D are listed in Table 7. The difference
between simulated and experimental coordinates is calculated, defined as “error” as shown
in Table 7, where Error = (simulated results-experiment result)/experiment result. The “×”
sign refers to the lack of fusion at the given location, i.e., no fusion line in the area.
224
Materials 2022, 15, 2215
Figure 9. Melt pool and fusion line of experiments and simulated welded joints of 30 mm thickness
steel plates, (a) experimental weld joint, (b) modelling using linear path heat source, (c) modelling
using sinusoidal path heat source, (d) modelling using oscillate-stop path heat source.
Table 7. Comparison of coordinates of selected points between simulated and experimental welding
melt pools.
Experiment Linear Sinusoidal Oscillate-Stop

Coordinate Error Error Error
(mm) (mm) (mm) (mm)
A (−15.9, 30) × × (−12.5, 30) 21.4% (−14.5, 30) 8.8%
B (14.8, 30) × × (12.5, 30) 15.5% (14.5, 30) 2.0%
C (6.3, 0) × × (4.5, 0) 28.6% (6.0, 0) 4.8%
D (−6.0, 0) × × (−4.5, 0) 25.0% (−6.0, 0) 0
Note: Error = (simulated result—experiment result)/experiment result.
In the case of the linear heat source path, there are no fusion lines running through
the top to the bottom of the welded joint. For the sinusoidal heat source path and the
oscillate-stop heat source path, the welding arc not only moves from bottom to top during
the welding process but also along with the weld oscillates in the direction of the depth of
the melt pool, which helps the welding bevels on both sides to obtain the same melting
depth. The “errors” of four selected points with sinusoidal heat source are: 21.4%, 15.5%,
28.6% and 25.0%, respectively, while the error of four selected points with oscillate-stop
heat source is 8.8%, 2.0%, 4.8% and 0%, respectively. The “error” of the oscillate-stop heat
source is greatly reduced. The heat source with oscillate-stop path holding on at near
surface of steel plate for more times than sinusoidal path, which helps to transfer more heat
to the surface of the joint during the welding process.
In summary, the path of the heat source can significantly affect the thermal profile of
the weld joint, hence FEA models are essential in optimizing and predicting accurately the
thermal profile, including the fusion line and shape of the melt zone in different thickness
and path conditions.
225
Materials 2022, 15, 2215
4. Evolution of Thermal Profile in Heat Affected Zone

4.1. Simulated Heat Source Paths
Heat source paths calculated using the linear heat source path, sinusoidal path and
oscillate-stop path are shown in Figure 10. In the figure, thermal profiles indicating heat
sources at 340 s, 349 s, and 357 s are shown as examples. The linear heat source moves
from the bottom to the top along the centre of the thickness of the steel plate, the sinusoidal
heat source moves along with the weld direction and the thickness of the weld pool at the
same time, and the oscillate-stop heat source path stays near the surface for a period of
time when oscillating to the internal and external position. The numbers in brackets on
the right side of every picture are the coordinates of the heat source centre of every picture.
The x axis alone the welding line and the y axis alone is the plate thickness direction. When
the welding time is 0 s, the heat source centre is in the starting point of the welding line (z)
and the centre of plate thickness (y); the coordinates of the heat source centre are the origin
coordinates (0, 0).
Figure 10. Calculated thermal profiles using three different heat source paths at different welding
times of 340 s, 349 s and 357 s. (a) linear path (b) sinusoidal path, (c) oscillate-stop path.
4.2. Thermal Cycles in Heat Affected Zone

Figure 11a is a schematic diagram indicating the positions of the selected points in
the heat affected zone (HAZ). Point 1, point 6 and point 8 are located in the groove which
belongs to the molten pool, point 2, point 3, point 4, point 5 and point 7 are located in
226
Materials 2022, 15, 2215
the base metal, the distance to the heat source of the five points is from 11 mm to 50 mm.
Figure 11b–d show the simulated thermal cycle curve of the five selected points during the
welding process with the linear path, sinusoidal heat source path and oscillate-stop heat
source path, respectively.
Figure 11. Calculated thermal cycles at selected points in heat affect zones (HAZ), (a) position of
selected points in HAZ, (b) thermal cycle curves using linear path, (c) thermal cycle curves using
sinusoidal path, (d) thermal cycle curves using oscillate-stop path.
As illustrated in Figure 11, the temperature quickly rises when the welding heat source
approaches the selected point, reaching a peak value, and then it decreases gradually as the
welding heat source moves away from the point. The peak temperature varies according
to the distance from the welding central line to the points. As listed in Table 8, the peak
temperature of point 1 with a linear heat source path is 1932 K, which does not reach the
melting point of 2055 K. The peak temperature of point 1 with sinusoidal heat source path
is 2085 K, which exceeds the melting point by 30 K, the point 1 will be melted. The peak
temperature of point 1 with the oscillate-stop heat source path is 2156 K, which exceeds
the melting point by 106 K, compared with the sinusoidal heat source path; its fusion line
advances to a position further away from the groove to form a joint with better quality. The
peak temperature values of the selected point are listed in Table 8.
227
Materials 2022, 15, 2215
Table 8. The peak temperature of selected point as shown in Figure 8.
Point
1 (K) 2 (K) 3 (K) 4 (K) 5 (K) 6 (K) 7 (K) 8 (K)
Path
Linear path 1932 1673 1513 1410 1065 1810 1640 2420
Sinusoidal path 2085 1848 1670 1551 1153 2021 1811 2295
Oscillate-stop path 2156 1919 1735 1609 1189 2410 1960 2223
Using the sinusoidal path and the oscillate-stop path, the distance between the heat
source and the selected points becomes closer when the heat source moves to the surface.
While the oscillate-stop path stays near the internal and external position for a period of
time in the cycle of movement. The welding arc can transfer more heat to the surface of
steel plates and advance the fusion line in the position. So, the temperature in HAZ with
the oscillate-stop path is higher than that with the sinusoidal path. EGW is a single pass
method for welding thick steel plates, so the stop-over of the heat source at a near-surface
position has great significance on the quality of the weld joint.
5. Recommended Heat Source Path Model for Simulating EGW Welding of Thick
Steel Plates
Three different heat source paths have been used for simulating EGW welding of thick
steel plates. In this section, the study will examine the effect of the heat source model on
the shape of weld pool and the fusion line, so that a different applicable heat source model
for simulating EGW thick plates can be defined.
The size of the cross section of the work piece gap is shown in Figure 12. The heat
source movement path for 20 mm, 25, 30 mm, 35 mm and 40 mm thick steel plate were
designed and listed in Tables 5 and 6 and Figures 7 and 8.
Figure 12. The simulated welding molten pool with three heat sauce paths. Linear heat source path
model for 20 mm thickness (a), for 25 mm thickness (b); Sinusoidal path for 30 mm thickness (c), for
35 mm thickness (d); Oscillate-stop path for 35 mm thickness (e) and for 40 mm thickness (f).
Simulations were carried out on steel plates with different thicknesses and different
heat source models, as listed below. Other modelling parameters are listed in Tables 3 and 4,
and the heat input is 157 kJ/cm for all trials.
228
Materials 2022, 15, 2215
• 20 mm, 25 mm thickness steel plates—using the linear path (Figure 12a,b);

• 30 mm, 35 mm thickness steel plates—using sinusoidal path (Figure 12c,d);
• 35 mm and 40 mm thick steel plates—using oscillate-stop path (Figure 12e,f).
For the linear path model: Figure 12a shows a fully fusion welding joint for 20 mm
thick steel plates, while Figure 12b shows the lack of fusion in the bottom of the joint for
25 mm thick steel plates. So, the maximum applicable thickness for the linear heat source
path is estimated to be 20 mm.
For the sinusoidal heat source path model: Fully fusion of the weld pool was obtained
for 30 mm thick steel plates, as shown in Figure 12c, while Figure 12d shows a lack of fusion
in the bottom of the joint for 35 mm steel plate. So, the maximum applicable thickness for
the sinusoidal function path is estimated to be under 30 mm.
For the oscillate-stop path: Figure 12e shows that a fully fusion of the weld pool is
obtained for 35 mm thick steel plates, but the lack of fusion in the bottom of the joint for
40 mm thick steel plates. So, the maximum applicable thickness for the oscillate-stop heat
source path is estimated to be around 35 mm.
6. Conclusions
1. Three different types of heat source path models (linear, sinusoidal function and
oscillate-stop) were implemented to simulate the EGW process of marine steel with a
heat input of 157 kJ/cm.
2. For EGW welding of 30 mm thickness steel plates, the model using the oscillate heat
source path predicted a more accurate thermal profile (the shape of melt pool and the
fusion line) than those using the sinusoidal and linear heat source paths.
3. The applicable heat source paths for modelling the EGW process of steel plates
with different thicknesses were investigated. The linear path model can be used
for simulating steel plate thickness up to 20 mm, a maximum thickness of 30 mm is
appropriate for the sinusoidal path, and a maximum thickness of 35 mm is appropriate
for the oscillate-stop path in EGW welding, with a heat input of 157 kJ/cm.
Author Contributions: Conceptualization, J.F. and H.D.; methodology, Q.T.; software, J.F.; validation,
J.F., B.N. and M.L.; formal analysis, Q.T.; investigation, J.F.; resources, H.D.; data curation, X.Y.;
writing—original draft preparation, J.F.; writing—review and editing, B.N., J.F.; supervision, H.D.;
project administration, B.T.; funding acquisition, B.T. All authors have read and agreed to the
published version of the manuscript.
Data Availability Statement: For further data sets, please contact the corresponding author.
Acknowledgments: All authors gratefully acknowledge the support from Nanjing Iron & Steel
United Co., Ltd. (NISCO), in particular for providing a PhD scholarship for this study.
References
1. Ichimiya, K.; Sumi, H.; Hirai, T. 460 MPa Yield Strength Class Steel Plate with JFE EWEL® Technology for Large Heat Input
Welding. JFE Tech. Rep. 2008, 5, 7–12.
2. Kaplan, M.B.; Solomon, S. A coming boom in commercial shipping? The potential for rapid growth of noise from commercial
ships by 2030. Mar. Policy 2016, 73, 119–121. [CrossRef]
3. Sasaki, K.; Suda, K.; Motomatsu, R.I.; Hashiba, Y.; Ohkita, S.; Imai, S. Development of Two-electrode Electro-gas Arc Welding
Process. Shinnittetsu Giho 2004, 90, 57–64.
4. Fritzsche, A.; Avilov, V.; Gumenyuk, A.; Hilgenberg, K.; Rethmeier, M. High power laser beam welding of thick-walled
ferromagnetic steels with electromagnetic weld pool support. Phys. Procedia 2016, 83, 362–372. [CrossRef]
5. Wang, G.; Li, J.B.; Sha, Y.Z.; Ma, J. Study on the penetration of electrogas welding. Weld. Technol. 2002, 5, 8–9.
229
Materials 2022, 15, 2215
6. Eakkachai, W.; Koei, H.; Motomichi, Y.; Kenji, S.; Kota, K.; Tadakazu, T.; Hiroshi, Y.; Tsutomu, F.; Shin, N.; Tetsuro, N.; et al.
Welding phenomena during vertical welding on thick steel plate using hot-wire laser welding method. Proc. Weld. Soc. 2015, 33,
143s–147s.
7. Haelsig, A.; Mayr, P. Energy balance study of gas-shielded arc welding processes. Weld. World 2013, 57, 727–734. [CrossRef]
8. Bai, X.; Colegrove, P.; Ding, J.; Zhou, X.; Diao, C.; Bridgeman, P.; Hönnige, J.R.; Zhang, H.; Williams, S. Numerical analysis of heat
transfer and fluid flow in multilayer deposition of PAW-based wire and arc additive manufacturing. Int. J. Heat Mass Transf. 2018,
124, 504–516. [CrossRef]
9. Aucott, L.; Huang, D.; Dong, H.B.; Wen, S.W.; Marsden, J.A.; Rack, A.; Cocks, A.C.F. Initiation and growth kinetics of solidification
cracking during welding of steel. Sci. Rep. 2017, 7, 40255. [CrossRef]
10. Aucott, L.; Huang, D.; Dong, H.B.; Wen, S.W.; Marsden, J.; Rack, A.; Cocks, A.C.F. A Three-stage mechanistic model for
solidification cracking during welding of steel. Met. Mater. Trans. A 2018, 49, 1674–1682. [CrossRef]
11. Hwang, S.-Y.; Kim, Y.; Lee, J.-H. Finite element analysis of residual stress distribution in a thick plate joined using two-pole
tandem electro-gas welding. J. Mater. Process. Technol. 2016, 229, 349–360. [CrossRef]
12. Xu, G.; Pan, H.; Wang, J. Numerical analysis model of temperature profile in swing-arc narrow gap GMAW welding. Thansactions
China Weld. Inst. 2017, 38, 55–60. (In Chinese)
13. Yuan, S.; Liu, W.; Li, L.; Jiang, X. Heat source model of temperature profile in swing-arc narrow gap MAG welding considering
sidewall fusion. Thansactions China Weld. Inst. 2018, 39, 95–99. (In Chinese)
14. Pereira, H.A.D.S.; Rodrigues, M.C.; Firmino, J.V.L.D.C. Implementation of weave patterns by path parameterization in the
simulation of welding processes by the finite element method. Int. J. Adv. Manuf. Technol. 2019, 104, 477–487. [CrossRef]
15. Wang, Z.; Oliveira, J.P.; Zeng, Z.; Bu, X.; Peng, B.; Shao, X. Laser beam oscillating welding of 5A06 aluminum alloys: Microstructure,
porosity and mechanical properties. Opt. Laser Technol. 2019, 111, 58–65. [CrossRef]
16. Guo, Z.; Saunders, N.; Schillé, J.; Miodownik, A. Material properties for process simulation. Mater. Sci. Eng. A 2009, 499, 7–13.
[CrossRef]
17. Saunders, N.; Guo, U.K.Z.; Li, X.; Miodownik, A.P.; Schillé, J.-P. Using JMatPro to model materials properties and behavior. JOM
2003, 55, 60–65. [CrossRef]
18. Rosenthal, D. The theory of moving sources of heat and its application of metal treatments. Trans. ASME 1946, 68, 849–866.
19. Lecoanet, A.; Ivey, D.; Henein, H. Simulation of the temperature profile during welding with COMSOL multiphysics® software
using Rosenthal’s approach. In Proceedings of the 2014 COMSOL, Cambridge, UK, 17 September 2014.
20. Wang, H.X.; Sun, J.S.; Wei, Y.H.; Zheng, Y.Y. Simulation of GMAW thermal process based on string heat source model. Sci. Technol.
Weld. Join. 2005, 10, 511–520. [CrossRef]
21. Eagar, T.; Tsai, N. Temperature fields produced by traveling distributed heat sources. Weld. J. 1983, 62, 346–355.
22. Rouquette, S.; Guo, J.; Le Masson, P. Estimation of the parameters of a Gaussian heat source by the Levenberg–Marquardt method:
Application to the electron beam welding. Int. J. Therm. Sci. 2007, 46, 128–138. [CrossRef]
23. Fu, G.; Gu, J.; Lourenco, M.I.; Duan, M.; Estefen, S.F. Parameter determination of double-ellipsoidal heat source model and its
application in the multi-pass welding process. Ships Offshore Struct. 2014, 10, 204–217. [CrossRef]
24. Goldak, J.; Chakravarti, A.; Bibby, M. A new finite element model for welding heat sources. Metall. Trans. B 1984, 15, 299–305.
[CrossRef]
25. Podder, D.; Mandal, N.; Das, S. Heat source modeling and analysis of submerged arc welding. Weld. J. 2014, 93, 183–191.
26. Kik, T. Computational techniques in numerical simulations of arc and laser welding processes. Materials 2020, 13, 608. [CrossRef]
[PubMed]
27. Kik, T.; Górka, J. Numerical simulations of laser and hybrid S700MC T-joint welding. Materials 2019, 12, 516. [CrossRef] [PubMed]
28. Kik, T. Heat source models in numerical simulations of laser welding. Materials 2020, 13, 2653. [CrossRef] [PubMed]
29. Hashemzadeh, M.; Chen, B. Comparison between different heat sources types in thin-plate welding simulation. In Developments
in Maritime Transportation and Exploitation of Sea Resources; Taylor & Francis: Abingdon, UK, 2013; pp. 329–335.
30. Tong, M.; Duggan, G.; Liu, J.; Xie, Y.; Dodge, M.; Aucott, L.; Dong, H.; Davidchack, R.L.; Dantzig, J.; Barrera, O.; et al. Multiscale,
multiphysics numerical modeling of fusion welding with experimental characterization and validation. JOM 2012, 65, 99–106.
[CrossRef]
31. Cai, Z.; Wu, S.; Lu, A.; Zhao, H.; Shi, Q. Line Gauss heat source model: An efficient approach for numerical welding simulation.
Sci. Technol. Weld. Join. 2001, 6, 84–88. [CrossRef]
32. Norman, A.F.; Ducharme, R.; Mackwood, A.; Kapadia, P.; Prangnell, P.B. Application of thermal modelling to laser beam welding
of aluminium alloys. Sci. Technol. Weld. Join. 1998, 3, 260–266. [CrossRef]
230
materials
Article
Exploratory Data Analysis for the Evaluation of Tribological
Properties of Wear-Resistant Surface Layers Modified with
Rare-Earth Metals
Paweł Malinowski 1 , Justyna Kasińska 2, *, Sławomir Rutkowski 3,4 and Monika Madej 2
1 Faculty of Foundry Engineering, AGH University of Science and Technology, 30-059 Krakow, Poland;
[email protected]
2 Department of Metal Science and Materials Technology, Kielce University of Technology, 25-314 Kielce,
Poland; [email protected]
3 The Faculty of Mechanics and Technology, Rzeszow University of Technology, 37-450 Stalowa Wola, Poland;
[email protected]
4 Multi-Branch Company T.S.A. Marcin Górski Sławomir Rutkowski, 37-450 Stalowa Wola, Poland
* Correspondence: [email protected]; Tel.: +48-414324470
Abstract: The role of rare Earth metals in the improvement of the properties of metals and alloys
has been analysed and described in multiple studies. Their effects on changes in microstructure
and mechanical properties are most pronounced. This paper focuses on the beneficial effect of rare
Earth metal oxides on the wear resistance of surface layers applied to castings intended for structural
elements of machinery and equipment in mining and recycling. The experiment involved modifying
prepared surfaces by adding CeO2 , Y2 O3 , and La2 O3 . Hardness measurements, a scratch test, and
tribological tests were performed under dry and fluid friction. The maximum wear track depth and
track area were measured from the surface profile. To determine correlations between the results,
exploratory data analysis was employed. Heatmaps were used to illustrate strong positive and
Citation: Malinowski, P.; Kasińska, J.;
Rutkowski, S.; Madej, M. Exploratory
negative interactions. The addition of oxides at increasing carbon content resulted in increased
Data Analysis for the Evaluation of hardness, lower coefficient of friction, and reduced track area and maximum track depth. Strong
Tribological Properties of negative interactions between the track area and maximum track depth were found. The differences
Wear-Resistant Surface Layers resulting from the test conditions (fluid and dry friction) were discussed. This study demonstrated the
Modified with Rare-Earth Metals. suitability of exploratory data analysis for analysing research results and confirmed the improvement
Materials 2022, 15, 2032. https:// of modified surface wear resistance.
doi.org/10.3390/ma15062032

Keywords: EDA—exploratory data analysis; rare Earth metal oxides; wear resistance; heatmap; correlation
Jiangshan Zhang
Received: 30 January 2022

Accepted: 7 March 2022 1. Introduction
The interest in rare Earth elements observed in recent years is driven by rising applica-
Publisher’s Note: MDPI stays neutral tion opportunities provided by new electronic technologies and nanotechnologies [1–4].
with regard to jurisdictional claims in In addition to these specialized areas of the economy, rare Earth metals (REM) are still used
published maps and institutional affil- in the metallurgical, chemical, and metal industries. Because of their physical and chemical
iations. properties, rare Earth oxides began to be used in surface engineering. Their beneficial
effect of reducing secondary dendrites and volume fraction of non-metallic inclusions
was noticed in the case of nickel-based [5,6] or iron-based [7] coating microstructures.
Another advantage was the increased corrosion resistance and improved passivation. In
their work [8], Liu et al. reported an advantageous effect of yttrium oxide Y2 O3 on the mod-
ification of the nickel-based coating, indicating, however, that the range of its application is
limited (from 0.4 to 0.6%).
conditions of the Creative Commons By influencing changes in the microstructure of steel, alloys, and metallic layers,
Attribution (CC BY) license (https:// REM directly improve tribological properties, corrosion resistance, mechanical properties,
creativecommons.org/licenses/by/ and resistance to oxidation [9–11]. Rare Earth metal additions change the parameters of
4.0/). the structure, e.g., grain size, number and size of inclusions [12–15], and the mechanical.

Materials 2022, 15, 2032
Adding La2 O3 , Gd2 O3 , Lu2 O3 particles increases the hardness of surface layers [16,17].
The authors of the work [18] showed that in the case of Si3 N4 ceramics, the addition of rare
Earth oxides reduced the friction coefficient and wear. Silicon nitride ceramics sintered
with the addition of rare Earth oxides also constitute an important class of materials for
high temperature applications. In addition to high temperature strengths, they possess
good thermal shock resistance, creep resistance, and high oxidation resistance [16,19–22].
A number of tests and devices can be applied to determine tribological properties
of materials and obtain a broad spectrum of information [23–25]. Using only selected
parameters, e.g., a friction coefficient or linear wear, and ignoring mass loss, for example,
a proper description of the tested material is impossible. In addition, the results obtained
are affected by the test conditions, e.g., temperature, friction pair, motion, medium—dry
friction and fluid friction [26].
Data analysis or elements of artificial intelligence, therefore, are increasingly used to
correctly interpret scientific results. Exploratory data analysis was introduced to support
scientific processes and use statistical methods in solving real problems. Exploratory
analysis allows for determining the correlation between the results, its type and strength,
as well as exclude or significantly limit the influence of the human factor [27,28].
Exploratory data analysis (EDA) may be defined as the art of looking at one or
more datasets in an effort to understand the underlying structure of the data contained
there [29,30].
The current paper uses the methods of exploratory data analysis to show the relation-
ships between the selected properties of the layers. The authors also wanted to demonstrate
the suitability of EDA for analysing the results of tribological tests.
2. Methodology
2.1. Surface Modification
As part of the experiment, the weld deposit was prepared using metallic powders and
rare Earth oxides (CeO2 , Y2 O3 , La2 O3 ) (Pol-Aura, Olsztyn, Poland). Nickel-based powders
(COB-ARC, Chorzow, Poland), iron-based powders (COB-ARC, Chorzow, Poland), and
powders containing chromium and tungsten carbides (COB-ARC, Chorzow, Poland) were
used to produce the deposits. The weld metals were then applied on S355 steel (Hut-Trans,
Katowice, Poland) using metal active gas (MAG) welding in combination with plasma arc
welding (PAW). Metallographic specimens were prepared from the collected samples for mi-
croscopic observation of microstructural changes. For this purpose, a Phenom XL scanning
electron microscope (Thermo Fisher Scientific /Phenom-World, Eindhoven, The Nether-
lands) was employed. Representative images of the pads made of metallic powder on a
nickel matrix (chemical composition: 0.6% C, 3% Fe, 11% Cr, 3.8% Si, the rest Ni) are shown
in Figure 1. The modification changed the morphology and arrangement of dendrites,
which was reflected in the change of wear resistance.
Figure 1. Microstructure of the Ni-based pad weld on the S355 steel: (a) non-modified and modified:
(b) CeO2 , (c) La2 O3 , (d) Y2 O3 .
232
Materials 2022, 15, 2032
2.2. Tribological Tests

Tribological tests under dry friction and fluid friction with the addition of SiO2 were
carried out for the prepared variants, in which tribological characteristics (friction coefficient,
linear wear) and wear indices (maximum wear track depth, track area) were determined.
The following tests were performed for all the welds:
• hardness test:
- micro combi tester MCT3 ANTON PAAR (Anton Paar, Corcelles-Cormondreche,
Switzerland),
- nominal loading force 2000 mN,
- loading/unloading rate = 4000 mN/min,
- Vickers indenter.
• scratch test
- micro combi tester MCT3 ANTON PAAR (Anton Paar, Corcelles-Cormondreche,
Switzerland),
- initial loading force 30 mN,
- final loading force 15,000 mN
- loading/unloading rate = 59,979.8 mN/min,
- Rockwell indenter
- indenter radius 100 μm.
• tribological tests under dry and fluid friction:
- tribometer TRB3 Anton Paar (Anton Paar, Corcelles-Cormondreche, Switzerland),
- reciprocating motion,
- amplitude: 10 mm
- frequency: 1 Hz
- number of cycles: 10,000
- friction type: dry friction/fluid friction (water solution 10% SiO2 )
- temperature: 23 ± 1◦
- humidity: 50 ± 1%
• tests for the maximum wear track depth and area measured from the surface profile:
- optical profilometer: Leica DCM8 (Leica, Wetzlar, Germany)
- the maximum track depth and the track area measured from the surface profile
were taken as measures of the sample wear.
The tribological tests allowed determining wear resistance of the welds. Table 1 shows
sample results from the experiments performed.
2.3. Exploratory Data Analysis

The data analysis was performed in three steps:
Acquisition
Wrangling
Exploration
In the first stage, exploratory data analysis (EDA) was carried out, which included
description and visualisation of the data without assuming any initial hypotheses. The de-
scription and visualisation of the data made it possible to identify trends, patterns, missing
data, outliers, etc.
Prior to exploratory data analysis, the following questions were generated:
• What are your analysis goals and outcomes?
• What tasks do you perform during analysis?
• What tools do you use?
233
Table 1. Compilation of sample test results.
Tribological Tests Hardness * Scratch Test *
Young's
Dry Friction Friction with Lubricant Maximum
Moudulus
Instrumental Penetration
Vickershardness Coefficient
Rare Hardness Depth of
Materials 2022, 15, 2032
(HV) of Friction
Sample RE Earth Coefficient Linear Wear Coefficient Linear HIT (MPa) the Indenter
(GPa)
Designation Oxygen Oxides of Friction (μm) of Friction Wear (μm) (μm)
% mas
1 - 0 0.06 39.16 - - - - - 0.181 −27.677

2 CeO2 2 0.55 47.24 - - 301.3 3192.5 207.4 0.214 −13.918
3 La2 O3 2 0.48 33.54 - - - - - 0.193 −16,982
4 Y2 O3 2 0.57 46.71 - - 301.7 3196.4 194.1 0.221 −19.903
3-1 - 0 0.78 55.61 0.35 124.26 333.9 3537.6 184.1 0.158 −7.283
3-2 CeO2 2 0.82 71.74 0.55 45.45 322.4 3415.9 171.9 0.163 −5.659
3-3 La2 O3 2 0.54 64.1 0.6 57.14 327.7 3471.9 186.5 0.186 −9.281
3-4 Y2 O3 2 0.79 71.07 0.18 96.74 428.2 4536.7 205 0.131 −6.557
4-1 - 0 0.67 55.8 0.5 60.97 335 3549.8 181.2 0.159 −7.46
4-2 CeO2 2 0.38 97.73 0.62 64.03 356.7 3779.6 189.5 0.164 −9.219
4-3 La2 O3 2 0.27 55.68 0.52 59.47 354 3750.9 184.7 0.165 −9.338
4-4 Y2 O3 2 0.69 54.15 0.5 97.26 501.4 5312.6 214.5 0.119 −5.899
234
Materials 2022, 15, 2032
The first thing to do with any data set is to read it. This is done not only to get to know
all the data collected, but also to reduce the workload during analysis. The initial data
investigation is known as exploratory data analysis or EDA and it primarily focuses on
visually inspecting the data. The main aim of EDA is to understand what data you have,
what possible trends there are, and therefore which statistical tests will be appropriate to
use [28].
In the EDA process, descriptive and visualisation analyses were performed, including
data set description (number of samples, number of not e number ( NaN )values), the re-
moval of columns with a large number of empty NaNs, the insertion of missing data using
strategy = mean, descriptive analysis (mean, standard deviation, min, max, median, 1st
quartile, 3rd quartile), the visualisation of elements in individual classes, the identification
of outliers, unsupervised learning using clustering, heat maps showing Pearson’s correla-
tions between features, and the visualisation of strong positive and negative correlations
divided into 4 classes.
The following tools and libraries were used (open source):
• Python 3.10.2,
• Jupyter Notebook 6.4.5,
• Numpy 1.22.0,
• Pandas 1.4.1,
• Matplotlib 3.5.1,
• Seaborn 0.11.2,
• Scipy 1.8.0.
Explanation of the methods used:
• Inserting missing data. When the set has empty spaces, they can be filled using several
strategies, such as inserting the mean, median, or the most common value.
• Identification of outliers. Outliers are data that do not follow the distribution of
other data. Since they are anomalies that should not be modelled, they must be
identified and removed. Outliers also have a negative effect on the Pearson’s linear
correlation coefficient.
• Cluster analysis. It is an unsupervised learning method that groups similar data using
various algorithms (K-means, Hierarchical Clustering). This method enables cluster
analysis, anomaly detection, and dimensionality reduction.
• Heatmap. A heatmap is a data visualisation technique that shows the magnitude of
a phenomenon as a colour in two dimensions. Colour variation can be due to shade
or intensity, giving visual clues as to how this phenomenon is clustered and how it
changes in space.
In preparing the dataset for analysis, new columns were introduced, the abbreviations
of which are shown in Table 2.
The set was divided into two parts:
• dry friction,
• fluid friction.
A NaN—Not a Number analysis was performed for the data set, Table 3.
Due to the small number of records (49), selected columns were removed (a large num-
ber of NaN values), i.e., Ni, Mo, Mn Co, B, W, V, WC. The remaining columns were
completed with the mean values for each column using the SimpleImputer function.
Tables 4 and 5 show the first five elements of both sets with the added mean values.
The descriptive information shown in Table 6 is then presented for the dry friction
set, which shows the calculated values of mean, standard deviation, minimum, maximum
value, and first, second, and third quartiles. Similar calculations were performed for fluid
friction, as shown in Table 7.
235
Materials 2022, 15, 2032
Table 2. Description of abbreviations.
Abbrev. Description Unit Test Type

cof1 Coefficient of friction (dry friction) Tribological tests
lw1 Linear wear (dry friction) (μm) Tribological tests
cof2 Coefficient of friction (lubricated friction) Tribological tests
lw2 Linear wear (lubricated friction) (μm) Tribological tests
vh Vickers hardness (HV) Hardness
ih Instrumental hardness (MPa) Hardness
ym Young’s modulus (GPa) Hardness
cof Coefficient of friction Scratch test
mpdi Maximum penetration depth of the indenter (μm) Scratch test
mtd1 Maximum track depth (dry friction) (μm) Leica
ta1 Track area (dry friction) (μm2 ) Leica
mwd2 Maximum track depth (lubricated friction) (μm) Leica
ta2 Track area (lubricated friction) (μm2 ) Leica
dw1 Disc wear (dry friction) (g) Mass loss
bw1 Ball wear (dry friction) (g) Mass loss
dw2 Disc wear (lubricated friction) (g) Mass loss
bw2 Ball wear (lubricated friction) (g) Mass loss
1—without additives,
2—with an addition of CeO2 ,
type
3—with an addition of La2 O3
4—with an addition of Y2 O3
Table 3. Analysis of NaN values, before and after the removal of selected columns.
NaN Values before Removal NaN Value after Removal

C 0
Si 4
Cr 0
Ni 21
Mo 24
Mn 45
Fe 8
Co 41
B 25 C 0
W 20 Si 4
V 28 Cr 0
WC 41 Fe 0
cof1 0 cof1 0
lw1 0 lw1 0
cof2 4 cof2 4
lw2 4 lw2 4
vh 2 vh 2
ih 2 ih 2
ym 2 ym 2
cof 0 cof 0
mpdi 0 mpdi 0
mwd1 0 mwd1 0
wa1 0 wa1 0
mwd2 4 mwd2 4
wa2 4 wa2 4
dw1 0 dw1 0
bw1 0 bw1 0
dw2 4 dw2 4
bw2 4 bw2 4
type 0 type 0
dtype: int64 dtype: int64
236
Materials 2022, 15, 2032
Table 4. Five elements of the set for dry friction with the added means.
No. C Si Cr Fe cof1 lw1 vh ih ym cof mpdi mtd1 ta1 dw1 bw1 Type
1 0.1 2.1 18.0 62.9 0.06 39.16 598 6330 214 0.181 −27,66763.79 52,888 0.0052 0.0003 1.0
2 0.1 2.1 18.0 62.9 0.55 47.24 301 3193 207 0.214 −13,91861.05 48,955 0.0045 0.0001 2.0
3 0.1 2.1 18.0 62.9 0.48 33.54 598 6330 214 0.193 −16,98247.23 42,889 0.0054 0.0000 3.0
4 0.1 2.1 18.0 62.9 0.57 46.71 301 3196 194 0.221 −19,90359.33 48,247 0.0053 0.0000 4.0
5 0.6 3.8 11.0 3.0 0.78 55.61 334 3538 184 0.158 −7,283 37.07 28,214 0.0033 0.0001 1.0
Table 5. Five elements of the set for fluid friction with the added means.
No. C Si Cr Fe cof2 lw1 vh ih ym cof mpdi mtd2 ta2 dw2 bw2 Type
1 0.1 2.1 18.0 62.9 0.38 58.16 598 6330 214 0.181 −27,6678.09 4939 0.0019 0.0008 1.0
2 0.1 2.1 18.0 62.9 0.38 58.16 301 3193 207 0.214 −13,9188.09 4939 0.0019 0.0008 2.0
3 0.1 2.1 18.0 62.9 0.38 58.16 598 6330 214 0.193 −16,9828.09 4939 0.0019 0.0008 3.0
4 0.1 2.1 18.0 62.9 038 58.16 301 3196 194 0.221 −19,9038.09 4939 0.0019 0.0008 4.0
5 0.6 3.8 11.0 3.0 0.35 124.26 334 3538 184 0.158 −7.283 12.37 13,033 0.0090 0.0140 1.0
Table 6. Basic information for the set—dry friction.
C vh ih ym cof mtd1 ta1 dw1 bw1

count 49 49 49 49 49 49 49 49 49
mean 0.877 597 6330 214 0.138 15.60 12,675.0 0.00289 0.00024
std 0.317 187 1982 20 0.039 23.13 22,171.4 0.00338 0.00018
min 0.100 281 2980 172 0.072 0.18 8.7 0.0010 0.00000
25% 0.800 447 4733 202 0.102 1.58 177.5 0.0007 0.00010
50% 1.000 625 6579 214 0.134 3.95 1093.0 0.0017 0.00020
75% 1.000 744 7885 225 0.165 10.66 78,109.0 0.0041 0.00030
max 1.300 942 9985 257 0.221 70.05 78,109.0 0.01860 0.00080
Table 7. Basic information for the set—fluid friction.
C vh ih ym cof mtd2 ta2 dw2 bw2

count 49 49 49 49 49 49 49 49 49
mean 0.877 597 6330 214 0.138 8.09 4938.5 0.00194 0.00082
std 0.317 187 1982 20 0.039 4.90 6094.1 0.00184 0.00195
min 0.100 281 2980 172 0.072 0.16 4.6 0.00000 0.00000
25% 0.800 447 4733 202 0.102 5.02 912.8 0.00070 0.00030
50% 1.000 625 6579 214 0.134 7.14 3008.0 0.00130 0.00050
75% 1.000 744 7885 225 0.165 10.03 5109.0 0.00220 0.01400
max 1.300 942 9985 257 0.221 28.33 30,517.0 0.00760 0.01400
In order to demonstrate the correctness of the analyses carried out with regard to the
number of tests performed with various modifiers (rare Earth oxides) and their effect on
tribological properties, a class chart was prepared. For the tests conducted, the lack of
balance between the individual classes may influence the real assessment. Figure 2 shows
the number and percentage of the elements per class.
237
Materials 2022, 15, 2032
Figure 2. Elements in each class.

Figure 3 shows the dependence of disc wear on track area for each class.
Figure 3. Dependence of dw1 on ta1 separately for each class.
Figure 4 shows the cumulative disc wear vs. track area. After removing the outliers, it
can be seen that the data follow a linear dependence.
Figure 4. Dependence of dw1 on ta1 by class.
In the next step, a cluster analysis was carried out using two methods (K-means,
Hierarchical). The analysis showed a high similarity between the two methods used. These
are unsupervised classification methods of unsupervised learning.
Figure 5 shows the analyses for dry friction (column on the left) and fluid friction
(column on the right).
238
Materials 2022, 15, 2032
Figure 5. Cluster analysis for dry and fluid friction. Determination of optimal number of clusters
using: (a,b) Elbow method; (c,d) clusters with centroids determined by K-means algorithm; (e,f) de-
termination of optimal number of clusters for Hierarchical method using dendrograms; (g,h) clusters
determined by Hierarchical algorithm.
The first row shows the optimal number of clusters determined using the so-called
Elbow method. For both sets, three clusters were determined by the breakpoint. The second
row shows the visualisation of the cluster analysis for both cases with the indication of
the centroids for all clusters. The next line shows the dendrogram plots for dry and fluid
friction on the basis of which the optimal number of clusters for the Hierarchical method
was calculated.
The two heatmaps shown above are for the dry friction and fluid friction sets
(Figures 6 and 7).
Pearson’s correlations show that in the case of dry friction, a very strong negative
correlation occurs between C-cof, C-mwd1, and C-ta1 and a very strong positive correlation
occurs between vh-ih and ta1-mwd1. Pearson’s correlation coefficient is a measure of linear
correlation between two sets of data. In contrast, for fluid friction, a very strong negative
correlation occurs between C-cof with a very strong positive correlation between ta2-mwd2.
The results are summarised in Table 8. The correlation values are shown in Table 9.
239
Materials 2022, 15, 2032
Figure 6. Heatmap for dry friction.
Figure 7. Heatmap for fluid friction.
Table 8. Significant correlations between the results.
Dry Friction Fluid Friction

Positive Negative Positive Negative
Correlation
Strong V.Strong Strong V.Strong Strong V.Strong Strong V.Strong
0.5–0.7 0.7–1 −0.5–0.7 −0.7–1 0.5–0.7 0.7–1 −0.5–0.7 −0.7–1
Cof, mwd,
C Cof
wa
Cof, mtd,
vh Ym Ih Ym Ih Cof
ta
Cof, mtd,
ih Ym Vh Ym Vh Cof
wa
ym Vh, ih mtd, ta Vh, ih
cof mtd, ta Vh, ih C Vh, ih C
mtd Cof ta Vh, ih, ym C ta
ta Cof mtd Vh, ih, ym C mtd
240
Materials 2022, 15, 2032
Table 9. Strength of association.
Correlation Negative Positive

Very strong −0.7 to −1 0.7 to 1
Strong −0.5 to −0.7 0.5 to 0.7
Several strong correlations can be observed from the heatmaps for dry and fluid
friction and for the pairwise correlations between these sets. For dry friction, the C index
shows a very strong negative correlation with cof, mtd, and ta, whereas for fluid friction,
only a very strong negative correlation with cof is observed. A strong negative correlation
was found for ih and ym against the indices from mwd and ta for dry friction. In contrast,
this correlation was not observed for fluid friction. For both sets, there is a very strong
positive correlation between mtd and ta (0.98—for dry friction, 0.93—for fluid friction). Ym
is strongly positively correlated with vh and simultaneously strongly negatively correlated
with mtd and ta.
To confirm the correlations discussed above, they are presented in the form of relation-
ships between the variables (Figures 8–10).
Figure 8. Relationships between the indices and carbon content.
241
Materials 2022, 15, 2032
Figure 9. Maximum track depth versus track area—dry friction (positive correlation (0.98)).
Figure 10. Maximum track depth versus track area—fluid friction (very strong positive correla-
tion (0.93)).
The graphs for dry friction show very strong negative correlations (−0.74, −0.80, −0.74):
Figures 8–10 show the graphs in which the layers are assigned to classes, i.e., non-
modified surface (class 1) and modified surfaces (classes 2, 3, 4). The differences between
the individual lines corresponding to the classes are visible. Especially for class 4 (Y2 O3 ),
we observe a different slope of the obtained straight lines. Analysis does not reveal a
quantitative but a qualitative relationship, indicating the effects of the additives on the
parameters determined.
Exploratory data analysis showed dependencies between individual parameters.
The proportionality of the carbon content in the welds with rare earth oxides to their
mechanical properties was demonstrated. It was confirmed that its increase resulted in an
increase in hardness. This is due to the formation of hard carbides in their microstructure
and changes in the morphology of padding welds modified with rare Earth metals. The in-
crease in hardness lowered the coefficients of friction “cof”, which directly translated into
the reduction of the surface of the track wear areas and the wear track depths. A slightly
different nature of wear was demonstrated for dry friction and fluid friction. It was shown
that despite the increase in hardness and the decrease in the friction coefficient, the remain-
ing parameters in the form of the wear track area and depth were not so strongly correlated
in the case of fluid friction. SiO2 particles were used in the latter case, which changed the
nature of the wear into tribocorrosion wear. The analyses confirmed both the complexity of
the effect that parameters such as chemical composition or test environment have on wear
processes, and the different nature of relationships between the parameters.
242
Materials 2022, 15, 2032
4. Conclusions
The authors showed that it is reasonable to use exploratory data analysis for the
evaluation of the properties of RE oxide modified pad welds. In order to optimise the
results, it is advisable to perform a number of additional experiments to expand the
database. The expanded database will be used to predict the optimal parameters (chemical
composition, percentage of rare Earth elements, padding method, etc.) while maintaining
the best wear resistance. The results obtained so far indicate dependency between geometry
indices and mechanical properties (friction coefficient, instrumental hardness).
Exploratory data analysis allowed determining directly proportional (positive cor-
relations) and inversely proportional (negative correlations) dependencies. A significant
dependency was observed for the “C” index, i.e., the effect of carbon on hardness, fric-
tion coefficient, and geometry indices (track area, track depth), which may result from
the change in the pad weld morphology after introducing RE oxides. This issue will be
addressed by the authors in their next work.
The study and analyses demonstrate that exploratory methods capture non-trivial
dependencies of the obtained results.
The analysis of the results of tribological tests is the basis for the development of new
techniques of forming surface layers and coatings with increased wear resistance. The wear
indicators analysed in this study indicate further direction of surface modification research.
The EDA results contributed to the knowledge on the influence of surface modification with
rare Earth oxides on wear resistance. The study presented in this article offers a guidance
to continue work on the technology of the surfacing process, selection of modifiers, and
welding materials for surface layers.
Author Contributions: Conceptualization, J.K. and P.M.; methodology P.M. and J.K.; software, P.M.;
validation, J.K., P.M. and M.M.; formal analysis, J.K., P.M., and M.M.; resources, S.R., J.K. and M.M.;
data curation, S.R. and P.M.; writing—original draft preparation, J.K. and P.M.; writing—review and
editing, J.K., P.M. and M.M.; visualization, P.M. and J.K.; supervision, J.K. and M.M. All authors have
read and agreed to the published version of the manuscript.
Data Availability Statement: Data sharing not applicable, all the data created for this study are
already displayed in the article.
References
1. Naumov, A.V. Review of the world market of rare earth metals. Russ. J. NonFerrous Met. 2008, 49, 2231.
2. Lundin, R.; Wilson, J.R. Rare earth metals. Adv. Mater. Processes 2000, 158, 52–55.
3. Wang, L.; Lin, Q.; Ji, J.; Lan, D. New study concerning development of application of rare earth metals. J. Alloys Compd. 2005,
408–412, 384–386. [CrossRef]
4. Committee on Assessing the Need for a Defense Stockpile National Materials Advisory Board; Division of Engineering and
Physical Science. Managing Materials for a 21st Century Military; The National Academies Press: Washington, DC, USA, 2008;
Available online: http://www.nap.edu/catalog.php?record_id=12028 (accessed on 2 January 2022).
5. Wang, K.L.; Zhang, Q.B.; Sun, M.L.; Wei, X.G.; Zhu, Y.M. Microstructure and corrosion resistance of laser clad coatings with rare
earth elements. Corros. Sci. 2001, 43, 255–267. [CrossRef]
6. Wang, X.H.; Zhang, M.; Zhou, Z.D.; Qu, S.Y. Microstructure and properties of laser clad TiC+NiCrBSi+rare earth composite
coatings. Surf. Coat. Technol. 2002, 161, 195–199. [CrossRef]
7. Zhao, G.M.; Wang, K.L. Effect of La2 O3 on corrosion resistance of laser clad ferrite-based alloy coatings. Corros. Sci. 2006, 48,
8. Liu, Q.B.; Zou, J.L.; Zheng, M.; Dong, C. Effect of Y2 O3 con-tent on microstructure of gradient bioceramic composite coating
produced by wide-band laser cladding. J. Rare Earths 2005, 23, 446.
9. Wu, C.F.; Ma, M.X.; Liu, W.J.; Zhong, M.L.; Zhang, H.J.; Zhang, W.M. Laser cladding in-situ carbide particle reinforced Fe-based
composite coatings with rare earth oxide addition. J. Rare Earths 2009, 27, 997. [CrossRef]
243
Materials 2022, 15, 2032
10. Yi, W.; Zheng, C.Q.; Fan, P.; Cheng, S.H.; Li, W.; Ying, G.F. Ef-fect of rare earth on oxidation resistance of iron base flux-ing alloy
spray-welding coating. J. Alloys Compd. 2000, 311, 65. [CrossRef]
11. Xing, X.G.; Han, Z.J.; Wang, H.F.; Lu, P.N. Electrochemical corrosion resistance of CeO2 -Cr/Ti coatings on 304 stainless steel via
pack cementation. J. Rare Earths 2015, 33, 1122. [CrossRef]
12. Kasińska, J. Wide–ranging influence of mischmetal on properties of G17CrMo5–5 cast steel. Metalurgija 2015, 54, 135–138.
13. Kasińska, J. Influence of Rare Earth Metals on Microstructure and Mechanical Properties of G20Mn5 Cast Steel. Arch. Foundry Eng.
2018, 18, 37–42.
14. Kasińska, J.; Myszka, D. Influence of rare earths metals (REM) on the structure and selected properties of grey cast iron. Metalurgija
2020, 59, 459–462.
15. Dzioba, I.; Pala, R.; Kasińka, J. Experimental—Numerical Analysis of Stress State in Front of the Crack Tip of Modified And
Unmodified G17crmo5–5 Cast Steel by Rare Earth Metals in a Brittle-Ductile Transition Region. Arch. Metall. Mater. 2016, 61,
1175–1181. [CrossRef]
16. Satet, R.L.; Hoffmann, M.J. Experimental evidence of the impact of rare-earth elements on particle growth and mechanical
behaviour of silicon nitride. J. Am. Ceram. Soc. 2000, 88, 2485–2490. [CrossRef]
17. Choi, H.J.; Lee, J.G.; Kim, Y.W. High temperature strength and oxidation behaviourof hot-pressed silicon nitride-disilicate
ceramics. J. Mater. Sci. 1997, 32, 1937–1942. [CrossRef]
18. Tatarkoa, T.; Kasiarováa, M.; Dusza, J.; Morgielb, J.; Sajgalíkc, P.; Hvizdoš, P. Wear resistance of hot-pressed Si3 N4 /SiC mi-
cro/nanocomposites sintered withrare-earth oxide additives. Wear 2010, 269, 867–874. [CrossRef]
19. Becher, P.F.; Painter, G.S.; Shibata, N.; Waters, S.B.; Lin, H.T. Effects of rare-earth(RE) intergranular adsorption on the phase
transformation, microstructureevolution and mechanical properties in silicon nitride with RE2 O3 +MgO additives: RE = La, Gd,
and Lu. J. Am. Ceram. Soc. 2008, 91, 2328–2336. [CrossRef]
20. Becher, P.F.; Shibata, N.; Painter, G.S.; Averill, F.; Van Benthem, K.; Lin, H.-T.; Waters, S.B. Observations on the influence of
secondary Me oxide additives (Me = Si, Al, Mg) on the microstructural evolution and mechanical behavior of siliconnitride
ceramics containing RE2O3(RE = La, Gd, Lu). J. Am. Ceram. Soc. 2010, 93, 570–580. [CrossRef]
21. Sharma, S.P.; Dwivedi, D.K.; Jain, P.K. Effect of La2 O3 addi-tion on the microstructure, hardness and abrasive wear behavior of
flame sprayed Ni based coatings. Wear 2009, 267, 853. [CrossRef]
22. Zhang, Z.Y.; Lu, X.C.; Luo, J.B. Tribological properties of rare earth oxide added Cr3 C2 -NiCr coatings. Appl. Surf. Sci. 2007,
253, 4377. [CrossRef]
23. Kalandyk, B. Microstructure and Abrasive Wear Resistance of 18Cr-4Ni-2.5Mo Cast Steel. Arch. Foundry Eng. 2012, 12, 81–84.
[CrossRef]
24. Kalandyk, B.; Kasińska, J. Effects of Rare Earth Metal Addition on Wear Resistance of Chromium-Molybdenum Cast Steel.
Arch. Foundry Eng. 2017, 17, 63–68.
25. Madej, M. Tribological Properties of Diamond-Like Carbon Coatings. Adv. Mater. Res. 2014, 874, 9–15. [CrossRef]
26. Styp-Rekowski, M.; Mańka, E.; Matuszewski, M.; Madej, M.; Ozimina, D. Tribological problems in shaft hoist ropes wear proces.
Ind. Lubr. Tribol. 2015, 67/1, 157–164.
27. Wuest, T.; Weimer, D.; Irgens, C.; Thoben, K.-D. Machine learning in manufacturing: Advantages, challenges, and applications.
Prod. Manuf. Res. 2016, 4, 23–45. [CrossRef]
28. Marian, M.; Tremmel, S. Current Trends and Applications of Machine Learning in Tribology—A Review. Lubricants 2021, 9, 86.
[CrossRef]
29. Pearson, R.K. Exploratory Data Analysis Using R; CRC Press: Boca Raton, FL, USA, 2018.
30. Cox, V. Translating Statistics to Make Decisions—A Guide for the Non-Statistician; Apress: New York, NY, USA, 2017.
244
materials
Article
Effect of Cerium on the Microstructure and Inclusion Evolution
of C-Mn Cryogenic Vessel Steels
Liping Wu 1,2,3 , Jianguo Zhi 2,3 , Jiangshan Zhang 1 , Bo Zhao 4 and Qing Liu 1, *
1 State Key Laboratory of Advanced Metallurgy, University of Science and Technology Beijing,
Beijing 100083, China; [email protected] (L.W.); [email protected] (J.Z.)
2 Centre of Technology Baotou Iron and Steel Co., Baotou 014010, China; [email protected]
3 Inner Mongolia Enterprise Key Laboratory of Rare Earth Steel Products Research and Development,
Baotou 014010, China
4 Institute of Metallurgical Technology, Iron and Steel Research Institute Beijing, Beijing 100083, China;
[email protected]
* Correspondence: [email protected]; Tel.: +86-010-82375255
Abstract: The effects of Cerium (Ce) were studied on the casting slab quality, microstructure,
and inclusion evolution of cryogenic vessel steel. An optical metallographic microscope, scan-
ning electron microscope, energy dispersive spectrometer, and Thermo-calc thermodynamic software
were used for characterization and analysis. The results indicated that the central segregation was
significantly improved after adding Ce and reached the lowest level when the content of Ce was
0.0009 wt.%. Meanwhile, the presence of Ce reduces the size of ferrite and improves pearlite morphol-
ogy. Ce also enables the modification of Al2 O3 and MnS + Ti4 C2 S2 inclusions into ellipsoid CeAlO3
and spherical Ce2 O2 S + Ti4 C2 S2 composite inclusions, respectively, which are easier to remove.
The formed Ce2 O2 S inclusions are fine and can work as heterogeneous nucleation points to refine the
microstructure of steel.
Citation: Wu, L.; Zhi, J.; Zhang, J.;

Keywords: cryogenic vessel steels; cerium; microstructure; inclusions
Zhao, B.; Liu, Q. Effect of Cerium on
the Microstructure and Inclusion
Evolution of C-Mn Cryogenic Vessel
Steels. Materials 2021, 14, 5262.
https://doi.org/10.3390/ma14185262
1. Introduction
Cryogenic vessel steel is an essential material for producing storage tanks and is
Academic Editor: Andrea Di Schino widely used to store and transport liquefied gases. As increasing clean energy sources
are developed and applied, cryogenic vessel storage equipment has been finding more
Received: 19 July 2021 applications [1–5]. Cryogenic vessel storage equipment requires stringent demands for
Accepted: 9 September 2021 the alloy composition, casting slab quality, and mechanical properties of the cryogenic
Published: 13 September 2021 vessel steel.
Importantly, inclusions affect the cleanliness, control, and quality of steel. During
Publisher’s Note: MDPI stays neutral the solidification-rolling process, the morphology of inclusions is changed. Rare earth
with regard to jurisdictional claims in (RE) elements present a remarkable effect on modifying inclusions and in facilitating clean
steel production, given their unique electronic layer structures. Earlier investigations
iations.
have reported that different RE elements could purify molten steel, modify inclusion
morphology, increase cryogenic toughness, and improve corrosion resistance in different
steel series [6–10]. When RE is added to extra-low sulfur (S < 0.003%) niobium-titanium
micro-alloy steel, the deoxidation and inclusion removal rates can be increased, therefore
Copyright: © 2021 by the authors. the molten steel can be more effectively purified [11]. After adding La+Ce compounds to
Licensee MDPI, Basel, Switzerland. ship plate steel, the inclusions mainly exist in the form of spheres and approximate spheres,
This article is an open access article while the sharp-angled S-Mn-Ca phase and S-O-Mn-Ca-Al phase inclusions are reduced
and the hot-rolled steel strip band structure is lightened [12]. The type of inclusions in
drill steel without Ce additions are MgAl2 O4 and (Ca, Mn)S. As the Ce content in drill
steel reaches 0.0078% (mass fraction), the types of inclusions change to Ce-O and Ce-S and
the size of the inclusions in the drill steel decrease significantly [13]. Inclusions affect the
4.0/).

Materials 2021, 14, 5262
toughness, corrosion resistance, and fatigue properties of steel, and reducing their size
weakens the impact of non-metallic inclusions on the properties of steel. The RE elements
La, Ce, and Nd play deoxidation and desulfurization roles in molten steel, making the
inclusions fine and diffuse in distribution [14]. Gao et al. found that after Ce treatment,
irregular Al2 O3 inclusions with a size of 10−15 μm are wrapped by rare earth and then
gradually modified into spheroidal CeAlO3 , Ce2 O3 , and Ce2 O2 S inclusions with a size of
≈5 μm, distributed into interstitial free slabs [15].
It has been proposed in the literature that RE influences the microstructure of a
casting slab. For example, with the addition of La into nonoriented electrical steel,
the equiaxed crystal ratio increased and the columnar crystals were suppressed and re-
fined [16]. The addition of La+Ce changed the solidification microstructure distribution of
thin slabs produced by compact strip production and refined the solidification structure of
430 ferritic stainless steel, thereby improving the comprehensive properties of the mate-
rial [17,18]. The addition of La had a significant effect on the distribution of arsenic by the
formation of arsenic inclusions, and various RE have been widely used in the steelmaking
process to improve steel quality [19].
However, there is still no systematic explanation regarding the mechanisms by which
RE improves the quality of casting slabs, changes the structure morphology, and modifies
the inclusions. Moreover, casting slab quality is the most important index to evaluate
the finished product. The quality of a casting slab should be analyzed from a macro
quality perspective, microstructure characteristics, and the evolution law of the inclusions.
In this study, the characteristics of inclusions and the microstructure of C-Mn cryogenic
vessel steels were comparatively studied without Ce and with the addition of different Ce
content levels.
2. Experimental
The steel samples were taken from a Inner Mongolia Baotou Iron and Steel Group
Company production site, located in Baotou, China. The production process uses a 240 t
top and bottom combined blow basic oxygen furnace (BOF), dephosphorizing with the
bottom blow, and performing automatic steelmaking with sublance 240 t ladle furnace
(LF) refining, and 240 t Ruhrstahl–Heraeus (RH) refining treatment for a minimum of
25 min. The RE ferrocerium alloy (where the mass fraction of Ce is 30% and the total
mass fraction of impurity elements including sulfur, oxygen, and phosphorus is lower
than 0.01%, the rest being Fe) were added into the ladle at the end of vacuum treat-
ment, and then re-pressed for 5 min. After re-pressing and soft blowing for greater than
8 min to homogenize the composition, it was ensured that the inclusion as fully floated
and improved the cleanliness of the molten steel. During the continuous casting pro-
cess of the slab, protective casting measures prevented an increase in nitrogen and oxy-
gen. The casting speed was 1.2 m·min−1 , and the casting slab section dimensions were
230 × 1550~2030 mm, using a 1650 mm double stand slab caster, respectively. The chemical
composition of the tundish sample was analyzed by direct reading spectrum analyzer
(ARL-4460, Thermo Fisher Scientific, Geneva, Switzerland), and the actual content of Ce in
the steel was detected by chemical analysis (ICP-MS, Plasma Mass Spectrometer, Waltham,
Massachusetts, America Perkin Elmer). Table 1 shows the chemical compositions of the
experimental steel. The hot-rolled steel strip was produced on a 2250 mm rolling mill using
controlled rolling and controlled cooling technology. The discharging temperature was
1190 ◦ C, the finish rolling temperature was 840 ◦ C, the coiling temperature was 600 ◦ C,
and the thickness of the finished product was 12 mm.
246
Materials 2021, 14, 5262
Table 1. Chemical composition (wt.%).
No. C Si Mn P S Ti Als 1 Ca Ce
I 0.18 0.16 1.45 0.010 0.003 0.017 0.040 0.0020 0
II 0.18 0.15 1.48 0.010 0.002 0.018 0.039 0.0020 0.0006
III 0.18 0.15 1.48 0.010 0.002 0.018 0.039 0.0020 0.0009
IV 0.18 0.15 1.48 0.010 0.002 0.018 0.039 0.0020 0.0013
1: Als, acid-soluble aluminum.
The casting slab selected had a size of 300 mm × 1500 mm × 230 mm. After hot acid
corrosion, casting slab quality was determined according to the standard diagrams for
macrostructure and defects in a continuous casting slab (using YB/T 4003-2016, which
is a kind of steel industry criterion in China, equal to ASTM E381-2017). In this study,
we took one piece of steel at 3/4 of the width of the casting slab, with dimensional
specifications of 100 mm × 100 mm × 230 mm, and we also chose three metallographic
samples at 1/4 distances from the inner arc, slab center, and 1/4 distance from the outer arc,
with dimensional specifications of 10 mm × 10 mm × 10 mm, following the specific
sampling plan shown in Figure 1.
Figure 1. Casting slab sampling schematic diagram.
The sampling position of the hot-rolled finished plate corresponds to the casting slab,
which is located at 3/4 of the width direction of the steel plate, and the sample size was
12 mm × 10 mm × 10 mm. The samples were mechanically ground (from #400~#1000)
and then polished. Then samples were etched using 4% nital. Optical microscope
(OM, Axio Observer D1 m, Jena, Germany) analysis was used to observe the microstruc-
tures. Inclusion distribution and morphology were determined using a scanning electron
microscope (SEM, LEO EVO, Germany) and a field emission scanning electron micro-
scope (FE-SEM, SIGMAN300, Carl Zeiss Company, Oberkochen, Germany). The com-
positions of inclusions were identified with an energy dispersive spectrometer (EDS,
GENESIS, America).
In order to more intuitively show the effect of RE elements on the denaturation of
other inclusions, Thermo-calc thermodynamic software was used to calculate the variation
law of inclusions in C-Mn cryogenic vessel steel with the addition of Ce. Combined with
the experimental results, the evolution law of various inclusions was summarized, and a
dominant zone diagram of Ce inclusions was obtained.

3.1. Effect of Ce on the Casting Slab Quality
Casting slab results, with and without Ce, are shown in Figure 2. There are apparent
differences in the observation results of the samples. It can be seen that the central segrega-
tion of the casting slabs with the Ce addition had a tendency to be ameliorated. Especially
after adding 0.0009 wt.% Ce, the shape of the central segregation line almost disappeared,
and the internal quality of the casting slab was excellent.
247
Materials 2021, 14, 5262
(a)ȱ (b)ȱ
ȱ
(c)ȱ (d)ȱ
ȱ
Figure 2. Casting slab quality results, (a) 0 wt.% Ce; (b) 0. 0006 wt.% RE; (c) 0.0009 wt.% Ce; and (d)
0.0013 wt.% Ce.
There are usually three types of central segregation, A represents the most serious,
B is the second, and C is the lightest. Each type is divided into three levels, the smaller the
number, the lighter the central segregation. The center porosity is the incompact structure
formed by the final crystal shrinkage of the casting slab. The degree of porosity was judged
by the size of the gap, and the casting slab quality was evaluated. Table 2 quantitatively
shows the judgment results for the casting slabs. It can be seen that the central segregation
of the casting slab was greatly improved after Ce treatment. The results reveal that the RE
Ce plays a vitally important role in the casting slab.
Table 2. The low magnification judgment results.
Content Central Segregation Center Porosity

0 Ce B 1.0 1.0
0.0006 wt.% Ce B 1.0 1.0
0.0009 wt.% Ce C 1.0 1.0
0.0013 wt.% Ce B 1.0 1.0
Central segregation determines the quality of a casting slab. When high temperature
stress exceeds the maximum stress that the casting slab can withstand, cracks will first form
near the central segregation zone and then extend outward gradually. Cracks indirectly
affect the uniformity of composition, the structure of hot-rolled finished product and the
judgment level of the band structure. During the period of solidification, Ce reduces
the residual sulfur and oxygen content in molten steel. High melting point RE sulfides
and oxysulfides are formed which can hinder the segregation of the alloying elements,
suppressing the accumulation of sulfides at the grain boundaries. Many scholars [20–22]
have indicated that the formation of Ce-O-S inclusions increases the heterogeneous nuclei
particles, improves the solidification structure, and reduces the center segregation of the
casting slab.
Under the same casting conditions, part of the Ce is dissolved within the molten
steel, reducing the segregation of C and Mn, while the other part of the Ce forms the
RE inclusions. The melting point of the inclusions formed by the RE elements is also
higher than that of the general metal compounds. The RE inclusions precipitate and
become dispersed within the molten steel during the cooling process. The RE inclusions
become nucleated particles. The outer layer will gradually become encased by the other
metal inclusions, thereby optimizing the solidification structure of the molten steel and
248
Materials 2021, 14, 5262
improving the quality of the casting slab. Previous researchers have indicated that adding
the appropriate amounts of RE elements into weathering steel significantly improved
central segregation [23].
The internal quality and macrostructure of the casting slab are closely related to the
microstructure of the product. Figure 3 shows the microstructure with and without and
Ce at different positions of the casting slab thickness. It can be seen that as the content of
Ce increases, the proportion of pearlite increases and the proportion of ferrite decreases,
however the microstructure tends to be refined. According to the reports, after the addition
of mixed La and Ce elements into 0.27C-1Cr steel, the original austenite grains were refined,
the grain growth rate was slowed, the nucleation rate of the phase transformation products
increased, Ac1 was unchanged, Ac3 increased, the Ms point decreased, the pre-eutectoid
ferrite or bainite precipitation period shortened, and the pearlite and bainite transformation
completion time were prolonged [24].
ȱ ȱ ȱ
ȱ ȱ ȱ
ȱ ȱ ȱ
ȱ ȱ ȱ
Figure 3. No Ce and different Ce microstructures of the casting slab in the direction of thick-
ness. (a–c) 0 wt.% Ce, respectively represents 1/4 inner arc position, casting slab center, and outer
arc 1/4; (d–f) 0.0006 wt.% Ce, respectively represents 1/4 inner arc position, casting slab center,
and outer arc 1/4; (g–i) 0.0009 wt.% Ce, respectively represents 1/4 inner arc position, casting slab
center, and outer arc 1/4; (j–l) 0.0013 wt.% Ce, respectively represents 1/4 inner arc position, casting
slab center, and outer arc 1/4.
These results mean that the addition of Ce can change the structural characteristics
of a casting slab. Due to a unique electronic layer structure, the edge-to-edge matching
model [25] was used to calculate it. The results show that containg-Ce oxide is potential
nucleant for the heterogeneous nucleation of both the δ-Fe and γ-Fe primary phases during
249
Materials 2021, 14, 5262
the solidification of steels, and it promotes grain refining. Also, Ce promotes the forma-
tion of random hexagonal close packing structures during the nuclei formation process,
and thus helps to reduce nucleation-free energy [26]. As shown in Figure 3, the grains
will be coarsened with the increase of Ce. However, when Ce addition is 0.0009 wt.%,
the pearlite and the ferrite are finer and more uniform. This is primarily due to the
addition of Ce in appropriate amounts in order to form a fine RE composite inclusion.
This promotes the nucleation of acicular ferrite in the austenite body, divides the austenite,
and refines the grain. Also, under the same casting conditions, the Ce may be responsible
for increasing the transformation temperature of austenite to pearlite. It can be concluded
that the addition of the optimum content of Ce played a vital role in the size and shape
of the microstructure of the casting slab. As discussed in previous work, the distribution
of RE on the grain boundaries indicates a refinement to pearlite layer spacing and ferrite
microstructure [27,28].
Figure 4 shows the morphology of the inclusions without and with the addition of
Ce. From Figure 4a, on the metallographic samples, islands it can be seen that MnS inclu-
sions formed near the shrinkage holes and occasionally were intermittently connected [20].
From Figure 4b, on the metallographic samples, the shape of Al2 O3 inclusions was irreg-
ular and blocky. These inclusions deteriorate matrix continuity and are origins for the
formation of cracks. For the optical microstructure of the samples, after the addition of Ce,
the casting slab inclusions were not prismatic or chain-shaped but ellipsoidal. When the
steel is hotly processed and cooled, due to the RE inclusions, the thermal expansion coef-
ficient and density were close to that of molten steel [29,30], and it could thus avoid the
stress around the inclusions, significantly reducing harm to the material. Figure 4 presents
the inclusions in the (c), (d), and (e) steel samples after adding the Ce alloy to molten steel.
The morphology of these inclusions was gradually spheroidized compared to those without
RE, which greatly reduced the damage to the steel matrix.
ȱ ȱ
(a)ȱ (b)ȱ
ȱ ȱ
(c)ȱ (d)ȱ
Figure 4. Cont.
250
Materials 2021, 14, 5262
ȱ
(e)ȱ
Figure 4. No Ce and different Ce typical inclusions morphology of the casting slab; (a) and
(b) represent 0 wt.% Ce; (c) represents 0.0006 wt.% Ce; (d) represents 0.0009 wt.% Ce; and (e)
represents 0.0013 wt.% Ce.
To determine the composition of the complex inclusion mentioned above clearly

through further investigation, the inclusions were analyzed using SEM-EDS to determine
their chemical composition, and the results are shown in Figure 5. The Al2 O3 inclusions
in the sample without Ce added were sharp-angled, while the MnS+Ti-C-S inclusions
were irregular, as shown in Figure 5a,b, respectively. The addition of RE Ce modifies
the Al2 O3 inclusion morphology, forming CeAlO3 , which are nearly spherical inclusions.
Simultaneously, the affinity of Ce to sulfur is stronger than it is to manganese [31], and thus
CeAlO3 or Ce2 O2 S spherical inclusions were formed instead of MnS + Ti-C-S inclusions.
When an appropriate amount of Ce was added to the molten steel, the Ce in the molten steel
modified the Al2 O3 and series MnS inclusions, forming Ce complex inclusions, making
their appearance approximately spherical. Some studies have indicated that the inclusions
√
with an aspect ratio less than or equal to 2 would be treated as circular, and the other
inclusions will be treated as elliptical [32]. The typical morphology and composition of
inclusions with Ce additions are shown in Figure 5c–e. The irregular Al2 O3 inclusions
or MnS + Ti-C-S inclusions were wrapped by Ce, and thus gradually transformed into
spheroidal Al-O-Ce-Mg-Ca, Al-O-Ce-C-Ca or Al-O-Ce-C-Ca inclusions. With the content
of Ce increasing in molten steel, the inclusions were gradually modified to RE inclusions,
and the size of the inclusions decreased. But when the Ce was over 0.0009 wt.%, the size of
the containing-Ce inclusions shows an increased and irregular trend. It could be concluded
that the appropriate addition of Ce modified the inclusions.
3.2. Effect of Ce on Microstructure and Inclusions in Steel Strips

Figure 6 shows the microstructures of hot-rolled steel strips with and without different
Ce levels. It can be seen that the proportion of pearlite is small and the size of the structure
is uneven in Figure 6a. As shown in Figure 6b–d, the proportion of pearlite is greater.
The proportion of pearlite and ferrite were measured quantitatively by the Photoshop
CC (Creative Cloud) image analysis tool, and the results are as shown in Figure 7. It can
be found that with the increase of the Ce element, the content of pearlite shows an in-
creasing trend, and the content of ferrite shows a decreasing trend. Meanwhile, when the
Ce was 0.0009 wt.%, the presence of Ce reduced the size of ferrite, and the microstruc-
ture was uniform and fine. It can be concluded that the microstructure in the casting
slab is directly inherited after hot rolling, such that more Ce is added, the better and
coarser the microstructure. One possible reason is that in the C-Mn cryogenic vessel steels,
Ce promotes the diffusion of C, which is conducive to the formation of pearlite, and the
formation of fine Ce inclusions improves the nucleation of pearlite and finer ferrite.
Zang et al. reported that the addition of Y element causes a large ferrite structure to
appear in 20Cr13 martensitic stainless steel [33].
251
Materials 2021, 14, 5262
Figure 5. Inclusion morphology and spectrum results of the casting slab; (a) and (b) represent 0 wt.% Ce; (c) represents
0.0006 wt.% Ce; (d) represents 0.0009 wt.% Ce; and (e) represents 0.0013 wt.% Ce.
Figure 6. Microstructures of hot rolled steel strips; (a) 0 wt.% Ce; (b) 0. 0006 wt.% Ce; (c) 0.0009 wt.%
Ce; (d) 0.0013 wt.% Ce.
252
Materials 2021, 14, 5262
Figure 7. The statistics of ferrite and pearlite in various samples, with and without Ce content.
The above results show that Ce can be a modified inclusion in the casting slab. Figure 8
shows the typical nonmetallic inclusions after the corresponding casting slab is rolled into
the finished steel strip. Figure 8a,b show the without adding Ce inclusions, the shape is
irregular and rectangular; these kind of inclusions are more harmful to the steel matrix and
also a sensitive source of crack propagation. Figure 8c–e show ellipsoidal RE non-metallic
inclusions observed in the microstructure after adding different contents of Ce, which
reduces the possibility of crack formation and propagation [34]. The size of the inclusions
decreases and tends to spheroidize, and when Ce is 0.0009 wt. %, the shape of inclusions
is spherical and their size is the smallest. These data indicate that the addition of an
appropriate quantity of Ce played an imporptant role in the size distribution and shape
deformation of the inclusions.
Combining the casting slab microstructure analysis in Figures 4 and 5, it was con-
cluded that Ce additions modified the grain size of the studied steels containing them.
Meanwhile, after adding Ce, large particles of inclusions floated up, so a typical finer spher-
ical inclusion morphology was chosen for the surface scan analysis as shown in Figure 9,
which demonstrates that the inclusions should be Ce2 O2 S. These finer RE inclusions act as
heterogeneous nucleation points [35], reducing the distribution of harmful elements in the
steel matrix.
3.3. Ce Inclusions Formed Thermodynamic Mechanisms

To illustrate the evolution mechanism of the inclusions, Thermo-Calc software was
used to calculate the formation results of S-containing inclusions in C-Mn cryogenic vessel
steel with and without the addition of RE Ce. As shown in Figures 10 and 11, it was found
that after adding Ce, the contents of Ti4 C2 S2 and MnS#2 type inclusions were significantly
reduced, forming high melting point Ce2 O2 S and Ce2 S3 inclusions, while MnS#2 is a
composite inclusion phase dominated by MnS. One possible reason is that the RE data of
the thermodynamic software is not comprehensive, which leads to certain limitations in
the calculation results.
To further explore the influences of different Ce contents on the evolution of inclu-
sions [36], the Gibbs free energy of the reactions under different Ce contents (0–100 ppm)
was calculated. The values of other elements were selected from Table 1 (No.II). The results
are shown in Figure 12.
253
Materials 2021, 14, 5262
Figure 8. Non-metallic inclusion morphorgy of hot rolled steel strips; (a) and (b) represent 0 wt.%
Ce; (c) represents 0.0006 wt.% Ce; (d) represents 0.0009 wt.% Ce; and (e) represents 0.0013 wt.% Ce.
ȱ ȱ ȱ ȱ ȱ
Figure 9. EDS results of the typical inclusions observed in the sample with 0.0009 wt.% Ce.
From Figure 12, the content of the RE elements has little effect on the precipitation
order of the inclusions. After the initial addition of a small amount of Ce, the Gibbs free
energy of each reaction was significantly reduced. However, increasing by the amount of
Ce added, the change in the Gibbs free energy gradually became smaller. According to the
relative size of the Gibbs free energy, it can be concluded that CeAlO3 is the most stable,
followed by Ce2 O3 , then Ce2 O2 S.
The formula ΔG = − RT ln K and previous research data [37–39] were used to calcu-
late the conditions for the formation of inclusions, and the results are shown in Table 3.
The phase diagram calculation results of Ce-O-S at a temperature of 1873K are shown in
Figure 13.
254
Materials 2021, 14, 5262
ȱ
Figure 10. No Ce formation of sulfide.
Figure 11. Various Ce formation of sulfide.
ȱ
Figure 12. Relation between reaction Gibbs free energy and Ce content.
255
Materials 2021, 14, 5262
Table 3. Inclusion conversion relationship and formation conditions.
Reactions Equilibrium Constant The Relation between a[O] and a[S]

2CeO2 = Ce2 O3 + [O] 0.12866 a[O] = 0.12866
Ce2 O3 + [S] = Ce2 O2 S + [O] 0.079 a[S] = 12.67 × a[O]
2CeO2 + [S] = Ce2 O2 S + 2[O] 0.01017 a[S] = 98.3284 × a[O] 2
Ce2 O2 S + [S] = 2CeS + 2[O] 0.8505 × 10−5 a[S] = 1.1758 × 105 × a[O] 2
Ce2 O2 S + 2[S] = Ce2 S3 + 2[O] 1.452 × 10−5 a[S] = 262.433a[O]
Ce2 S3 = 2CeS + [S] 0.5857 a[S] = 0.5857
ȱ
Figure 13. Ce-S-O inclusions generation diagram.
Combining Table 3 and Figure 13, the conclusions are as follows:

(1) Formation conditions of RE oxides:when a[O] > 0.232, a[S] /a[O] 2 < 98.329 the
inclusions generated are CeO2 , when a[O] < 0.232, a[S] /a[O] < 12.67, the inclusions generated
are Ce2 O3 .
(2) Formation conditions of RE oxysulfides:when a[O] < 0.232, 12.67 < a[S] /a[O] <
262.433 or a[S] < 0.5857, 98.3284 < a[S] /a[O] 2 < 117580, the inclusions generated are Ce2 O2 S.
(3) Formation conditions of RE sulfides:when a[S] < 0.5857, a[S] /a[O] 2 > 117580,
the inclusions generated are CeS, and when a[S] > 0.5857, a[S] /a[O] > 262.433, the inclusions
generated are Ce2 S3 .
3.4. Inclusion Formation Mechanisms with the Addition of Ce

RE elements have a strong mutual chemical affinity to [O] and [S] [40]. Both ele-
ments begin to aggregate and form high melting point oxides, oxysulfides and sulfides,
thus having the ability to remove the deleterious types of inclusions that are formed [13,32,41].
The combination of Ce with [O] and [S] in steel has lower Gibbs free energy. It was easy
to change the inclusion morphology from chains and strips to generate CeAlO3 , Ce2 O2 S,
and Ce2 O3 spherical inclusions [42]. The RE inclusions CeAlO3 and Ce2 O2 S in molten steel
were stable at 1873K, decreasing the mass fraction of dissolved oxygen in the molten steel,
while Ce and S formed complex RE sulfides, resulting in a decrease in CeAlO3 and increase
in Ce2 O2 S [43,44]. According to the discussion in Sections 3.2 and 3.3, Ce2 O2 S inclusion is
especially beneficial for heterogeneous nuclei. The evolution process of RE inclusions during
solidification is shown in Figure 14.
256
Materials 2021, 14, 5262
Figure 14. Schematic diagram of the evolution process of inclusions with RE-Ce alloy treatment.
Firstly, when the melting temperature is stable at 1873K, it is easy to form Al2 O3 and
MnS+Ti4 C2 S2 inclusions in molten steel. As shown in Figure 13 (Stage I), the inclusions
are primarily irregularly shaped and deteriorate the mechanical properties of steel [45].
The reaction equation is as follows:
[S] + [Mn] = MnS(S) (1)
4[Ti] + 2[C] + 2[S] = Ti4 C2 S2(S) (2)

2[Al] + 3[O] = Al2 O3(S) (3)
Secondly, when RE-Ce alloy are added to the molten steel, large-sized irregular
inclusions are also transformed into small-sized spherical or ellipsoidal CeAlO3 and Ce2 O2 S
inclusions via Ce. The reaction equation [37,38] is as follows:
[Ce] + 3[O] + [Al] = CeAlO3(S) (4)
2[Ce] + 2[O] + [S] = Ce2 O2 S(S) (5)

4[Ti] + 2[C] + 2[S] = Ti4 C2 S2(S) (6)
Finally, as the above reactions proceed, part of the large-size inclusions float to the
top. The other Ce inclusions generated during Stage II act as heterogeneous nucleus cores.
This result suggests that RE elements can modify inclusion morphology and purify molten
steel, further increasing casting slab quality control, as shown in Figure 13 (Stage III).
Based on the reasoning above, combined with Figure 2, the central segregation mor-
phology of the casting slab almost disappears after the addition of RE Ce. According
to in-situ statistical analysis, it was found that the addition of RE elements reduces the
segregation of easily segregated elements such as C, S, and P, significantly weakening
the central segregation defect [23,46]. This study also shows that the formed CeAlO3
and Ce2 O2 S inclusions have a low degree of mismatch with the Fe matrix, increasing the
effectiveness of the nucleation core and improving the solidification structure of C-Mn
cryogenic vessel steels. Moreover, due to the strong surface activity of the Ce element,
it is segregated in the austenite grain boundary, reducing the interface energy, weakening
the driving force for nucleation, refining the structure, and improving the quality of the
casting slab. After the addition of composite La and Ce, many high melting point (Ce,
La)2 O2 S secondary-particles can be used as an effective non-uniform nucleation core, which
increases the proportion of equiaxed crystals in the solidified structure, reduces the size of
the equiaxed crystals, and significantly refines the solidified structure of the alloy [47].
When RE atoms are dissolved in the matrix, a damping peak of solute atoms segre-
gating at the grain boundary appears in the high-temperature zone. When the RE content
exceeds the maximum solid solution of the matrix, it will react with other alloying elements,
forming a compound [48]. Therefore, it can be concluded that the addition of 0.0009wt.%
Ce reduces central segregation, accelerates the formation of needle-shaped ferrite, increases
257
Materials 2021, 14, 5262
the proportion of pearlite, and modifies Al2 O3 and MnS+Ti4 C2 S2 inclusion morphology in
the C-Mn cryogenic vessel steels.
4. Conclusions
In industrial production, when the RE amount is over 0.0015 wt.%, the nozzle of the
continuous casting mold will appear as a nodule, which is not conducive to the efficient
production of long casting times. In this study, through comparative experimental studies, it
was found that when the Ce amount is 0.0009 wt.%, the central segregation, microstructure
characteristics, and inclusion control of C-Mn cryogenic vessel steel benefited, which can
further provide theoretical guidance for the industrial production of RE steel. Based on the
above analysis, the main results can be summarized as follows:
(1) During the solidification of the RE-treated steels, the central segregation was
significantly improved after adding Ce, reaching the lowest level when the content of Ce
was 0.0009 wt.%.
(2) Comparing the microstructures of the casting slabs at different positions in the
thickness direction after the addition of Ce, the microstructure and inclusions from the
casting slab to the hot-rolled steel strip had a certain heritability, the ferrite became needle-
shaped, and the proportion of pearlite increased.
(3) Experimental observation and thermodynamic calculations may indicate that Ce
can modify the sharp-angled Al2 O3 inclusions into ellipsoid CeAlO3 composite inclusions,
and that and the formation of MnS+Ti4 C2 S2 inclusions was reduced in favor of forming
Ce2 O2 S + Ti4 C2 S2 spherical inclusions. Meanwhile, finer Ce2 O2 S inclusions can serve as
heterogeneous nucleation points and reduce the genetic harm of inclusions.
Author Contributions: Conceptualization, L.W. and J.Z. (Jianguo Zhi); methodology, L.W. and
Q.L.; software, B.Z.; validation, L.W. and B.Z.; formal analysis, L.W.; investigation, L.W. and Q.L.;
data curation, L.W. and B.Z.; writing—original draft preparation, L.W. and J.Z. (Jiangshan Zhang);
writing—review and editing, L.W. and Q.L.; visualization, L.W. and J.Z. (Jiangshan Zhang); supervi-
sion, Q.L.; project administration, J.Z. (Jianguo Zhi); funding acquisition, Q.L. and J.Z. (Jianguo Zhi).
All authors have read and agreed to the published version of the manuscript.
Funding: This research was supported by the financial support of Local Science and Technology De-
velopment Fund Projects Guided by The Central Government (2020ZY0034) and the Inner Mongolia
Natural Science Foundation (2020MS05017).
Data Availability Statement: Data sharing not applicable. No new data were created or analyzed in
Acknowledgments: The authors are grateful for the support of the financial support of Local Science
and Technology Development Fund Projects Guided by The Central Government (2020ZY0034) and
the Inner Mongolia Natural Science Foundation (2020MS05017).
References
1. Chen, Z.; Pan, J.H.; Jin, T.; Hong, Z.Y.; Wu, Y.C. Estimation of fracture toughness of 16MnDR steel using Master Curve method
and Charpy V-notch impact energy. Theor. Appl. Fract. Mech. 2018, 96, 443–451. [CrossRef]
2. Liu, D.F.; Yang, X.L.; Hou, L.F.; Cui, T.X.; Hu, Y.T.; Wei, Y.H. Research and Application of Ultralow Temperature 9Ni Steel for
LNG Storage Tank. J. Iron Steel Res. 2009, 21, 1–5.
3. Xu, L.; Li, Z.C.; Zhang, J.; Xu, Y.S. LNG land storage and transmission in China. Nat. Gas Ind. 2002, 22, 89–91.
4. Barsouma, I.; Lawala, S.A.; Simmonsb, R.J.; Rodrigues, C.C. Failure analysis of a pressure vessel subjected to an internal blast
load. Eng. Fail. Anal. 2018, 91, 354–369. [CrossRef]
5. Miao, C.J.; Zheng, J.Y.; Gao, X.Z.; Huang, Z.; Guo, A.B.; Ye, D.Y.; Ma, L. Investigation of low-cycle fatigue behavior of austenitic
stainless steel for cold-stretched pressure vessels. J. Zhejiang Univ. Sci. A 2013, 14, 31–37. [CrossRef]
6. Xin, W.B.; Zhang, J.; Luo, G.P.; Wang, R.F.; Meng, Q.Y.; Song, B. Improvement of hot ductility of C-Mn Steel containing arsenic by
rare earth Ce. Metall. Res. Technol. 2018, 115, 419–427. [CrossRef]
258
Materials 2021, 14, 5262
7. Garrison, W.M.; Maloney, J.L. Lanthanum additions and the toughness of ultra-high strength steels and the determination of
appropriate lanthanum additions. Mater. Sci. Eng. A 2005, 403, 299–310. [CrossRef]
8. Pan, F.; Zhang, J.; Chen, H.L.; Su, Y.H.; Kuo, C.L.; Su, Y.H.; Chen, S.H.; Lin, K.J.; Hsieh, P.H. Effects of Rare Earth Metals on Steel
Microstructures. Materials 2016, 9, 417. [CrossRef] [PubMed]
9. Wan, Y.; Chen, W.Q.; Wu, S.J. Effect of lanthanum content on microstructure and magnetic properties of non-oriented electrical
steels. J. Rare Earths 2013, 31, 727–733. [CrossRef]
10. Liu, H.J.; Liu, H.J.; Chen, Y.Z. Effect of RE on the structure and property of steel 45. Chin. Rare Earths 2012, 33, 24–27.
11. Lin, Q.; Ye, W.; Chen, N.; Liu, Y.H.; Guo, Y.; Chen, Z.Q.; Zhu, Q.H. Effect of rare earths on extra-low sulfur microalloy steel. J.
Chin. Rare Earth Soc. 1997, 22, 228–233.
12. Xiao, J.G.; Cheng, H.J.; Wang, F.M. Effects of rare earth in ship plate steel on its microstructure and low temperature toughness.
Chin. Rare Earths 2010, 31, 52–57.
13. Huang, Y.; Cheng, G.G.; Xie, Y. Modification Mechanism of Cerium on the Inclusions in Drill Steel. Acta Metall. Sin. 2018, 54,
1253–1261.
14. Imashuku, S.; Wagatsuma, K. Cathodoluminescence Analysis of Nonmetallic Inclusions in Steel Deoxidized and Desulfurized by
Rare-Earth Metals (La, Ce, Nd). Metall. Mater. Trans. B 2019, 51, 79–84. [CrossRef]
15. Gao, S.; Wang, M.; Guo, J.L.; Wang, H.; Zhi, J.G.; Bao, Y.P. Characterization Transformation of Inclusions Using Rare Earth Ce
Treatment on Al-Killed Titanium Alloyed Interstitial Free Steel. Steel Res. Int. 2019, 90, 1900194. [CrossRef]
16. Fan, L.F.; Zhu, R.; He, J.Z.; Lu, B. Effect of Rare Earth Element La on Texture and Inclusion of Nonoriented Electrical Steel
Produced by Thin Slab Casting and Rolling Process. ISIJ Int. 2018, 58, 2348–2353. [CrossRef]
17. Chang, H.; Jin, Z.L.; Ren, H.P.; Song, W.Z. Effect of rare earth on microstructure and non-metallicinclusions in thin slab producted
by CSP. J. Chin. Rare Earth Soc. 2010, 28, 601–606.
18. Wang, L.M.; Zhu, G.L.; Xun, J.; Qi, G.P.; Zhu, J.X.; Chen, D.X. Study of Rare Earth Effect and Mechanism in Ferrite Stainless
Steel(430). Chin. Rare Earths 2008, 28, 67–71.
19. Wang, H.P.; Jiang, S.L.; Yu, P.; Bai, B.; Sun, L.F.; Wang, Y. Distribution of Arsenic Inclusions in Rare Earth Steel Ingots. Metals 2020,
10, 146. [CrossRef]
20. Yan, Y.X.; Li, P.Z. Influence of RE addition on the central segregation in 16Mn concast slabs and the fractograph of finished
products. Iron Steel 1986, 21, 13–19.
21. Tuttle, R.B. Study of Rare Earth Additions Effect on the Solidification and Properties of 4130 Steel. JMEPG 2019, 28, 6720–6727.
[CrossRef]
22. Fruehan, R.J. The Free Energy of Formation of Ce2 O2 S and the Nonstoichiometry of Cerium Oxides. Metall. Trans. B 1979, 10,
23. Guo, H.H.; Song, B.; Mao, J.H.; Zhao, P. Effect of rare earth elements on macrosegregation in weather-resisting steel. J. Univ. Sci.
Technol. Beijing 2010, 32, 44–49.
24. Lv, W.; Zheng, D.H.; Xu, Z.Y. Influence of Rare earth on TTT Diagram for 0.27C-1Cr Steel. Acta Metall. Sin. 1993, 29, 308–311.
25. Ji, Y.P.; Kang, L.; Song, Y.Q.; Qu, W.; Ren, H.P. Crystallographic Calculation about Heterogeneous Nucleation Potency of RE2 O3 in
Liquid Steel. Rare Met. Mater. Eng. 2017, 46, 2889–2894.
26. Gao, X.Y.; Wang, H.Y.; Xing, L.; Ma, C.N.; Ren, H.P. Evolution of local atomic structure during solidification of Fe-RE (RE = La,
Ce) alloy. J. Non-Cryst. Solids 2020, 542, 120109. [CrossRef]
27. Bao, X.R.; Chen, L.; Guo, Y.P. Effect of RE on Dynamic Recrystallization and Pearlite Spacing of Heavy Rail Steel. Iron Steel 2012,
47, 76–78.
28. Jiang, J.K.; Xiao, L.F.; You, D.L.; Shu, J.S.; Ren, L.E. Effect of Rare Earth addition on Properties of Controlled Rolling Low Carbon
MnMoNb Ferrite Steel. Acta Metall. Sin. 1987, 23, 165–167.
29. Wang, L.M. Application Prospects and Behavior of RE in New Generation High Strength Steels with Superior Toughness. J. Chin.
Rare Earth Soc. 2004, 22, 48–53.
30. Li, X.; Jiang, Z.H.; Geng, X.; Chen, M.J.; Peng, L.Z. Evolution Mechanism of Inclusions in H13 Steel with Rare Earth Magnesium
Alloy Addition. ISIJ Int. 2019, 59, 1552–1561. [CrossRef]
31. Zhang, L.F. Non-Metallic Inclusions in Steel; Metallurgical Industry Press: Beijing, China, 2019; Volume 26.
32. Meng, Y.; Yan, C.L.; Yang, X.; Ju, X.H. A Statistical Analysis on the Complex Inclusions in Rare Earth Element Treated Steel. ISIJ
Int. 2020, 60, 534–538. [CrossRef]
33. Zang, Q.Y.; YJin, Y.F.; Zhang, T.; Yang, Y.T. Effect of yttrium addition on microstructure, mechanical and corrosion properties of
20Cr13 martensitic stainless steel. J. Iron Steel Res. Int. 2020, 27, 451–460. [CrossRef]
34. Pang, F.X.; Yin, S.C.; Li, J.J.; Wang, S.B. Study on deoxidization thermodynamics of trace rare earth element and formation
mechanism of inclusions in LZ50 molten steel. J. Taiyuan Univ. Technol. 2011, 42, 646–649.
35. Wang, H.P.; Yu, P.; Jiang, S.L.; Wang, Y. Effect of Heterogeneous Nucleation on Removal of Arsenic from Molten Steel by Rare
Earth Addition. Metals 2020, 10, 664. [CrossRef]
36. Wang, Y.G.; Liu, C.J. Agglomeration Characteristics of Various Inclusions in Al-killed Molten Steel Containing Rare Earth Element.
37. Waudby, P.E. Rare earth additions to steel. Int. Met. Rev. 1978, 2, 74–98.
259
Materials 2021, 14, 5262
38. Kwon, S.K.; Park, J.S.; Park, J.H. Influence of Refractory-Steel Interfacial Reaction on the Formation Behavior of Inclusions in
Ce-containing Stainless Steel. ISIJ Int. 2015, 55, 2589–2596. [CrossRef]
39. Jiang, M.Z.; Yu, Y.C.; Li, H.; Ren, X.; Wang, S.B. Effect of Rare Earth Cerium Addition on Microstructures and Mechanical
Properties of Low Carbon High Manganese Steels. High Temp. Mater. Process. 2017, 36, 145–153. [CrossRef]
40. Fan, H.Q.; Xu, W.C.; Wei, L.; Zhang, Z.H.; Liu, Y.B.; Li, Q. Relationship between La and Ce additions on microstructure and
corrosion resistance of hot-dip galvanized steel. J. Iron Steel Res. Int. 2020, 27, 1108–1116. [CrossRef]
41. Wang, L.J.; Liu, Y.Q.; Wang, Q.; Chou, K.C. Evolution mechanisms of MgO•Al2 O3 inclusions by cerium in spring steel used in
fasteners of high-speed railway. ISIJ Int. 2015, 55, 970–975. [CrossRef]
42. Wang, H.; Bao, Y.P.; Duan, C.Y.; Lu, L.; Liu, Y.; Zhang, Q. Effect of Rare Earth Ce on Deep Stamping Properties of High-Strength
Interstitial-Free Steel Containing Phosphorus. Materials 2020, 13, 1473. [CrossRef]
43. An, Y.Q.; Wang, L.Z.; Li, Y.T.; Song, G.B.; Lei, S.Z.; Tan, J. Effects of Rare Earth Cerium on the Characteristics of Inclusions.
Wide Heavy Plate 2020, 26, 11–17.
44. Li, Y.D.; Liu, C.J.; Zhang, T.S.; Jiang, M.F.; Peng, C. Liquid Inclusions in Heat-Resistant Steel Containing Rare Earth Elements.
45. Wang, C.; Han, Y.S.; Zhang, J.S.; Xiao, D.; Yang, J.; Chen, J.; Liu, Q. Behaviour of oxide inclusions and sulphur in ‘two-stage
basicity control’ refining method of Si-killed spring steel. Ironmak Steelmak 2020, 48, 466–476. [CrossRef]
46. Gao, R.Z.; Chen, H.Q. The effect of rare earth elements on the solidification characteristics and crystal structure of steel. Chin.
Rare Earths 1985, 3, 27–33.
47. Yu, Y.C.; Chen, W.Q.; Zheng, H.G. Effect of Cerium and Mischmetal on Solidification Structure of Fe-36Ni Invar Alloy. J. Chin.
Rare Earth Soc. 2012, 30, 175–180.
48. Wang, H.Y. Effect of Rare Earth on NbC Dissolution and Precipitation Behaviors in Microalloy Steels. University of Science and
Technology: Beijing, China, 2017; Volume 14.
260
materials
Article
Full-Field Temperature Measurement of Stainless Steel
Specimens Subjected to Uniaxial Tensile Loading at Various
Strain Rates
Krzysztof Żaba 1, *, Tomasz Trzepieciński 2 , Stanislav Rusz 3 , Sandra Puchlerska 1 and Maciej Balcerzak 1
1 Department of Metal Working and Physical Metallurgy of Non-Ferrous Metals, Faculty of Non-Ferrous
Metals, AGH—University of Science and Technology, al. Adama Mickiewicza 30, 30-059 Cracow, Poland;
[email protected] (S.P.); [email protected] (M.B.)
2 Department of Manufacturing and Production Engineering, Faculty of Mechanical Engineering and
Aeronautics, Rzeszow University of Technology, al. Powst. Warszawy 8, 35-959 Rzeszów, Poland;
[email protected]
3 Department of Mechanical Technology, Faculty of Mechanical Engineering, VŠB—Technical University of
Ostrava, 17 listopadu 15, CZ 708 33 Ostrava–Poruba, Czech Republic; [email protected]
Abstract: This article presents a study on the effect of strain rate, specimen orientation, and plastic
strain on the value and distribution of the temperature of dog-bone 1 mm-thick specimens during
their deformation in uniaxial tensile tests. Full-field image correlation and infrared thermography
techniques were used. A titanium-stabilised austenitic 321 stainless steel was used as test materials.
The dog-bone specimens used for uniaxial tensile tests were cut along the sheet metal rolling direction
and three strain rates were considered: 4 × 10−3 s−1 , 8 × 10−3 s−1 and 16 × 10−3 s−1 . It was
found that increasing the strain rate resulted in the intensification of heat generation. High-quality
Citation: Żaba, K.; Trzepieciński, T.;
regression models (Ra > 0.9) developed for the austenitic 321 steel revealed that sample orientation
Rusz, S.; Puchlerska, S.; Balcerzak, M.
Full-Field Temperature Measurement
does not play a significant role in the heat generation when the sample is plastically deformed. It was
of Stainless Steel Specimens Subjected found that at the moment of formation of a necking at the highest strain rate, the maximum sample
to Uniaxial Tensile Loading at Various temperature increased more than four times compared to the initial temperature. A synergistic
Strain Rates. Materials 2021, 14, 5259. effect of the strain hardening exponent and yield stress revealed that heat is generated more rapidly
https://doi.org/10.3390/ma14185259 towards small values of strain hardening exponent and yield stress.
Academic Editors: Qing Liu and Keywords: digital image correlation; mechanical properties; stainless steel; temperature; thermovi-
Jiangshan Zhang sion; uniaxial tensile test
Received: 28 June 2021

Accepted: 7 September 2021
Published: 13 September 2021 1. Introduction
Metals and their alloys have a crystalline structure, which is characterised by a regular
arrangement of atomic cores. Technical metals obtained by conventional metallurgical
methods have a polycrystalline structure [1,2]. This means that metals are made up of grains
iations.
characterised by approximately the correct crystal structure. Polycrystalline materials
consist of grains with various orientations. For many polycrystalline materials, the grain
orientations are random before any working (deformation) of the material is undertaken.
Therefore, even if the individual grains are anisotropic, the property differences tend to
average out and, overall, the material is isotropic. When a material is formed, the grains
are usually distorted and elongated in one or more directions which makes the material
anisotropic [3].
In metals subjected to plastic working, the most important defects of the crystal lattice
are dislocations and grain boundaries. Strain hardening—hardening of a material with
deformation—results from interaction and multiplication of dislocations during plastic
deformation. Grain boundaries have a much greater influence on the mechanical properties
4.0/). of metals than dislocations [4,5]. The high energy level of grain boundaries determines

Materials 2021, 14, 5259
the occurrence of many phenomena, such as continuous mobility of boundaries leading to

grain growth and lower corrosion resistance [6–8]. Grain boundaries also have a strong
influence on the ductility of metals. The degree of strain hardening of materials by grain
refining is described by the Hall–Petch relationship [9,10] according to which the yield
strength of the material increases with the degree of grain refinement. On the one hand,
grain boundaries inhibit the free movement of dislocations during grain deformation
while participating in the process of material strengthening and in the formation of the
deformation texture [1]. On the other hand, the grain boundaries are the main area of
permanent deformation during creep or superplastic flow.
The deformation of the material (elastic or plastic) occurs under the influence of a
load applied to it. Elastic deformation disappears when the load is removed. Plastic
deformation remains after the material is unloaded. There are two main mechanisms of
plastic deformation: the gliding motion of dislocations and deformation twinning [11,12].
The gliding motion does not occur simultaneously in all possible planes and directions
of slip. Deformation occurs gradually, starting from the sliding planes and directions
prioritised in relation to the direction of the applied load. During deformation, the number
of active slip planes and free dislocations that are generated during deformation increases.
Further plastic deformation requires the application of greater stress in order to initiate
new dislocations [13–15]. The process by which stress increases with deformation is called
work hardening. Plastic deformation is also accompanied by heat generation due to the
internal resistance of the material, which can be detected at the microstructure scale when
loading the material [16,17].
The investigation of heat generation and dissipation during plastic deformation was
reflected in scientific research. Bodelot et al. [18] applied infrared (IR) coupled full-field
measurements to observe the heterogeneities of kinematic and thermal data at the grain
scale of austenitic stainless steel. Specimens were subjected to a cyclic loading. It was
found that the temperature was significantly affected by diffusion and this observation
tends to confirm that temperature may not be used as a damage indicator. Boulanger
et al. [19] have focused on the determination of heat sources from a temperature field
provided by an IR camera. They separately identified the dissipative and thermo-elastic
heat sources. Thermal full-field measurements from infrared thermography were used to
study the Portevin-Le Chatelier effect [20]. The temperature field provided insights into
the dynamics of band formation and motion.
The non-contact optical methods for the measurement of temperature displacement in-
clude video extensometers [21], laser speckle correlation [22], interferometry methods [23–25]
and temperature-calibrated CCD cameras [26]. In addition to IR thermography, the other
technique that allows temperature changes to be coupled with strains is digital image corre-
lation (DIC), in which digital images of an object before and after deformation are captured
using a non-contact optic and material-independent measuring instrument, and then they
are subject to correlation analysis [23,24,27]. High-speed DIC allows deformation and dam-
age mechanisms to be analysed in a quantitative manner [28]. DIC was used to obtain strain
fields and to investigate the propagation of the Lüders band in steel specimens subjected to
the uniaxial tensile test [29]. The strain states associated with localised necking and diffuse
strains are also investigated. Results show that the cross-section of a tensile specimen must
be regarded as a structure, not as an elementary volume of material which is subjected to
the uniform load. The results of investigations of Hung and Voloshin [30] indicated that for
the uniform tension test, DIC is a very convenient and efficient tool for the measurement of
in-plane strain measurement. Feng and Xue [31] applied infrared thermography tests and
DIC tests for the display of the thermal field during mechanical tensile tests of 3D printed
bolts. Cholewa et al. [32] developed the method of calibrated infrared thermal cameras
with a stereo-vision DIC system applicable to scales of lengths both large and small. Żaba
et al. [33] used DIC and IR thermography to find the relationship between yield stress and
the Taylor–Quinney coefficient and their change with the strain rate during the stretching of
nickel-based superalloys. A coupled thermography and DIC system was calibrated using a
262
Materials 2021, 14, 5259
series of one sided heat exposure experiments performed on loaded sandwich composites.
Infrared thermography and DIC measurements are not applicable to experiments with low
conductivity materials or discontinuous materials [33,34]. Maynadier, et al. [35] developed
an infrared image correlation system equipped with a single IR camera to measure simulta-
neous thermal fields and deformation. The disadvantage of the proposed method is that
this technique requires the use of a special coating for displacement measurement using the
single camera. Chrysochoos et al. [34] proposed DIC and thermography measurements for
the determination of the mechanical energy and heat sources involved at local scale during
a heterogeneous tensile test. Thermodynamic analysis of the energy balance showed the
influence of the dissipated mechanism on the stress–strain response.
Many of studies take place at a macroscopic scale with the measurement of the average
temperature of the material [36,37]. Meanwhile, due to the grain structure of polycrystalline
metals and the occurrence of various plastic deformation mechanisms, the temperature
of the body subjected to loading is not uniform. The way to explain this phenomenon
is to take full-field measurements. In this paper, the coupled IR and DIC measurements
were conducted to assess the thermomechanical response of stainless steel strip samples
subjected to tensile load at different strain rates. An analysis of variance (ANOVA) was
used to gain information about the relationship between the strain hardening phenomenon,
yield stress, sample orientation in respect of the sheet rolling direction, strain rate, and
plastic strain of the specimen and temperature.

2.1. Material
The research material was a titanium-stabilised austenitic stainless steel, 321 (AMS
5510). Steel 321 is a steel with good formability, weldability and resistance to intergranular
corrosion that is mainly used in the chemical and aviation industries. The thickness of the
sheets was 1 mm. The test samples were cut from the sheet in three directions: the rolling
direction (RD), at an angle of 45◦ to the RD and perpendicular to the RD. The dimensions
of test samples (Figure 1) were in accordance with the ASTM E8/E8M–11 standard [38].
The requirements for the chemical composition of the test steel are shown in Table 1 and
are in accordance SAE AMS 5510 [39].
Figure 1. Dimensions (in mm) of the samples for the tensile test.
Table 1. Chemical composition of 321 steel (wt.%) [39].
C (Max.) Si Mo Mn (Max.) P + S (Max.) Cr Ni Ti (Max.) N (Max.) Fe

0.04 P
0.08 0.25–1.00 0.75 2.0 17.0–19.0 9.0–12.0 0.7 0.1 balance
0.03 S
2.2. Uniaxial Tensile Test

Dog-bone samples (Figure 1) were stretched in a uniaxial tensile test machine Zwick/Roell
Z100 (Figure 2). Three different strain rates were used in the investigations: 4 × 10− 3 s− 1 ,
8 × 10− 3 s− 1 and 16 × 10− 3 s− 1 . The tests were carried out at a temperature of 24 ◦ C.
263
Materials 2021, 14, 5259
Figure 2. Test stand.
2.3. Digital Image Correlation Technique

The digital image correlation Aramis system (GOM, Braunschweig, Germany) was
used to determine the character of the deformation of the sample in a non-contact manner
during the stretching process of samples. The measuring system consists of two essential
components. The first is a scanner, consisting of two high-resolution digital cameras
(Figure 2) positioned in relation to each other in such a way as to be able to build a spatial
image. The second element of the set is a computer with special software for the numerical
processing of images.
Proper surface preparation of the samples is necessary for DIC analysis. First, the
samples are covered with white paint, and then a random pattern of contrasting black
spots is applied. The selection of the size of the spots depends on the size of the test object
and the optical properties of the measuring apparatus. The measurement process consists
in taking a series of images of the sample by two cameras in successive stages of loading.
Correlation of images from these cameras makes it possible to determine the position of
each pixel by giving them coordinates in a three-dimensional coordinate system. Then
the images are divided into so-called deformation grids, each of which contains a unique
pattern of dots [40]. The initial setup becomes the reference step. Due to the deformation of
the sheet surface, the spots in each element of the mesh change their position in relation to
each other, which makes it possible to calculate the full field deformation in relation to the
reference position. However, in order to make such an analysis possible, the system divides
the measurement area into fields of a fixed size called facets. Local strains are obtained
from the formula: ! ! " "
l + Δld
ε eng = lim in − 1 ·100% (1)
l →0 lin
where lin is the initial dostance between two neighbouring facets and Δld is the distance
increase during uniaxial tensile test.
The dimension of standard facet is 21 × 21 pixels. The second characteristic value
is the distance between the centres of adjacent facets. The Aramis system offers many
benefits, such as:
• a stable solution for full-field analyses of test objects of just a few millimeters up to
structural components of several metres in size,
• it is a material-independent and non-contact measuring system,
• it performs high-precision measurements with a 3D measurement resolution in the
sub-micrometer range,
• it is a high-resolution for point-based and full-field measurements.
264
Materials 2021, 14, 5259
2.4. Infrared (IR) Thermal Mapping

The surface temperature of the samples during the stretching was measured using
a high-sensitivity IR Flir T640 camera (Flir Systems AB, Antennvägen 6, 187 66 Täby,
Sweden). The measurements of temperature were correlated with the measurement of
strain using the A0Aramis system. Non-invasive distance measurement was performed
with an accuracy of thermographic measurements in the range of +/− 2 ◦ C. The principal
parameters of the Flir T640 camera are listed in Table 2.
Table 2. Parameters of a Flir T640 camera.
Parameter Unit Value

Resolution pixel 2048 × 1536
Measuring range ◦C −40 ÷ +2000
Refresh rate Hz 30
Thermal sensitivity mK 30
2.5. Analysis of Variance

Quadratic multidimensional ANOVA was used to determine the relationship between
material properties (strain hardening exponent, yield stress, sample orientation), process
parameters (strain rate and percentage strain) and the maximum temperature appearing
in the sample during the stretching process. The values of the strain hardening exponent
were determined by approximating true stress-true strain curves using the well known
Hollomon’s power law.
ANOVA is a statistical method for examining observations that depend on one or
more factors acting simultaneously. Due to the different numerical ranges of the data, they
were normalised to the range [−1, +1] [41]. The min-max normalisation was applied by
means of a linear function bringing the data to a new interval (Coded Low, Coded High).
Explanatory variables should be independent of each other. The input data were fitted
with a polynomial, and the influence of individual variables on the quality of the model
was checked using backward elimination. The minimum and maximum values of the input
variables for samples made of 321 sheet are presented in Table 3. Intermediate values of
input parameters were coded proportionally in the range [−1, +1].
Table 3. Factors and levels for analysis of variance (ANOVA) of stretching of specimens of 321 steel.
Parameter Name Unit Type Minimum Maximum

Strain
A hardening - Numeric 0.35 0.392
exponent
B Yield stress MPa Numeric 341 367
C Strain rate s−1 Numeric 4 × 10− 3 16 × 10− 3
Sample
D deg. Numeric 0 90
orientation
Percentage
E % Numeric 12 117.7
strain
The regression models that were built were subjected to significance tests. The method
of backward elimination of variables was used in the analysis. It is a variable selection
procedure in which all the variables are entered into the regression equation and then
removed sequentially. The variable with the lowest partial correlation with the dependent
variable is considered for removal first. If it meets the elimination criteria, it is removed.
The basis for removing or leaving a given variable in the model is the calculation of Fisher
F statistics. The independent variable with the highest probability corresponding to the
Fisher parameter F is removed from the model if the probability p is sufficiently high
(typically p = 0.10). After the first variable is removed, the next one to be removed is the
265
Materials 2021, 14, 5259
one that has the smallest partial correlation with the dependent variable. The procedure
exits when there are no other variables in the equation that meet the removal criteria.
The significance of the regression model at a level of α = 0.05 is determined based on the
variance due to the effect of a factor and the variance due to the error term.

3.1. Experimental Investigations
Figure 3 shows the evolution of the surface temperature of the samples cut along the
sheet rolling direction of the 321 steel during its stretching with different strain rates. It
is clearly visible that even within the range of proportional deformations the zone with
the highest temperature is located approximately at the central point between the grippers
of the testing machine. As a result of heat convection caused by internal friction of the
material, an increase in the temperature of the samples was also observed in the part that
is gripped where no plastic deformation occurs. Increasing the strain rate resulted in the
intensification of heat generation. At the moment of the formation of a necking at the
highest strain rate, the maximum sample temperature increased by more than 4 times
compared to the initial temperature. Similar conclusions can be drawn for the samples cut
at an angle of 45◦ with respect to RD and perpendicular to RD (Figure 4).
Only part of the mechanical energy is converted into heat in the deformation process.
The remainder is stored in the microstructure of the material increasing the internal energy
of the material. The total energy spent on deforming an elastic-plastic material is equal to
the work undertaken on elastic (reversible) deformation and permanent (plastic) deforma-
tion. Moreover, the energy consumed on plastic deformation is divided into heat dissipated
in the forming process and energy stored in the material [42–44]. The evolution of the
microstructure during deformation depends on the type of material, the initial temperature
and the loading situation [45,46]. Changing the deformation method usually leads to a
partial or complete reconstruction of the dislocation systems formed in the previous stages
of deformation and the formation of new systems [47,48]. The interaction of the lattice
defects is related to the overlapping of their stress fields. If the overlap of stress fields
caused by defects reduces the energy of the system, then with a specific activity of the
respective slip systems, these defects will form configurations compliant with the principle
of energy minimisation [49,50].
The plasticity margin measured as the difference between ultimate tensile stress and
field stress is more pronounced in the case of 321 steel than for 17-4PH steel.
Figure 5 show the effect of percentage strain on the change in the maximum tempera-
ture of the specimens of 321 steel. These dependencies, determined by the determination
coefficient R2 , show a linear trend with a high correlation value R2 > 0.93. In the case of the
sample made from 321 steel cut at an angle of 45◦ tested at a strain rate of 16 × 10− 3 s− 1 ,
the last point corresponds to the advanced stage of sample necking. Therefore, the R2 -value
for that case is about 0.9484 (Figure 5c). In general, the higher the strain rate the steeper the
trend lines. The samples made of 321 steel show anisotropic features, which are particularly
visible at a strain rate of 8 × 10− 3 s− 1 (Figure 5b). Trend lines for samples cut at different
angles are inclined at different angles with respect to the abscissa axis. True stress-strain
curves for 321 steel determined at various strain rates are shown in Figure 6. True strain is
defined as εt : ! "
l
ε t = ln (2)
l0
where l and l0 are the current and initial gauge lengths, respectively.
266
Materials 2021, 14, 5259
Figure 3. Infrared (IR) thermography images and digital image correlation charts for 321 steel samples cut along the sheet
rolling direction, tested at the following strain rates: (a) 4 × 10− 3 s− 1 , (b) 8 × 10− 3 s− 1 and (c) 16 × 10− 3 s− 1 .
267
Materials 2021, 14, 5259
Figure 4. IR thermography images and digital image correlation charts for 321 steel samples cut
perpendicular to the rolling direction (RD), tested at the following strain rates: (a) 4 × 10− 3 s− 1 ,
(b) 8 × 10− 3 s− 1 and (c) 16 × 10− 3 s− 1 .
Figure 5. Influence of strain on temperature for 321 steel samples tested at the following strain rates: (a) 4 × 10− 3 s− 1 ,
(b) 8 × 10− 3 s− 1 and (c) 16 × 10− 3 s− 1 .
268
Materials 2021, 14, 5259
Figure 6. True stress-true strain curves for 321 steel determined at the following strain rates: (a) 4 × 10− 3 s− 1 ,
(b) 8 × 10− 3 s− 1 and (c) 16 × 10− 3 s− 1 .
3.2. Analysis of Variance (ANOVA) for 321 Steel Samples

A quadratic model containing only statistically significant elements was chosen to fit
the experimental data. The temperature response surface for the 321 steel specimens is
given by:
T = 76.71 − 18.56A − 10.02B − 4.19C + 4.71D + 48.61E − 4.97AB
+9.58AC + 8.27AD − 31.41AE − 12.99BC − 13.1BE (3)
−21.35CE − 1.22DE + 13.36E2
The significances of the influence of the individual material parameters as well as
the conditions of the stretching process and their interaction in the regression model were
determined on the basis of the ANOVA. The quality of fit of the model obtained to the
values measured for a given research plan (Equation (3)) was then determined on the
basis of the determination coefficients R2 and the F test. The significance of the model is
confirmed by an F-value of 53.30 (Table 4). This model can, therefore, be used to predict
temperature values based on material deformation. The probability that such a large
F-value can result from data noise is only 0.01%.
Table 4. Results of analysis of variance for the regression model for 321 steel samples.
Source Sum of Squares Degrees of Freedom Mean Square F-Value p-Value Meaning
Model 8991.31 14 642.24 53.30 <0.0001 significant
A—strain
hardening 26.04 1 26.04 2.16 0.1723 -
exponent
B—yield stress 39.77 1 39.77 3.30 0.0993 -
C—strain rate 0.5708 1 0.5708 0.0474 0.8321 -
D—sample 13.38 1 13.38 1.11 0.3167 -
orientation
E—plastic strain 989.76 1 989.76 82.14 <0.0001 -
AB 1.11 1 1.11 0.0920 0.7679 -
AC 9.23 1 9.23 0.7664 0.4019 -
AD 22.41 1 22.41 1.86 0.2025 -
AE 171.00 1 171.00 14.19 0.0037 -
BC 4.45 1 4.45 0.3696 0.5568 -
BE 36.85 1 36.85 3.06 0.1109 -
CE 758.38 1 758.38 62.94 <0.0001 -
DE 2.40 1 2.40 0.1992 0.6649 -
E2 66.37 1 66.37 5.51 0.0409 -
Residual 120.49 10 12.05 - - -
Cor Total 9111.80 24 - - - -
The probability values p for E, AE, CE, E2 that are less than 0.0500 prove that these
factors are statistically significant. The parameters the removal of which will not reduce the
quality of the model include, inter alia, C, AB, BC, DE. Due to the satisfactory value of the
269
Materials 2021, 14, 5259
determination coefficient for the model equal to R2 = 0.9868 (Table 5), it was decided to leave
these variables in the model. The predicted value of the coefficient of determination R2 =
0.8982 is in reasonable agreement with the adjusted coefficient of determination R2 = 0.9683
because the difference between their values is less than 0.2 [51]. The value of adequacy
precision 27.6342 is greater than 4. The adequacy of model is, therefore, confirmed.
Table 5. Fit statistics of the regression model for 321 steel samples.
Standard Deviation 3.47 R2 0.9868

Mean 54.09 Adjusted R2 0.9683
Coefficient of variation, % 6.42 Predicted R2 0.8982
- - Adequacy precision 27.6342
If the errors arising during the experimental tests are random, then the distribution
of residuals in the form of deviations between the actual values and the expected values
should be in accordance with the normal distribution. The distribution of residuals along a
straight line (Figure 7a) shows that the studentised residuals have a normal distribution [52].
Figure 7b shows a comparison of the actual values and the predicted temperature values
resulting from the introduction of independent variables to the regression model. The
distribution of data around a straight line indicates that the regression model is of good
quality [53,54].
(a) (b)
Figure 7. (a) Normal % probability vs. externally studentised residuals and (b) predicted vs. actual values of temperature
plots for 321 steel samples.
The relationship between the dependent and independent variables was analysed
using 2D response surface plots. Figures 8 and 9 show the interactive effect of two selected
parameters on the value of the maximum temperature in the sample during the stretching
test when the other input parameters are constant [55–57]. Figure 8a indicates the synergis-
tic role of the strain hardening exponent (SHE) and yield stress (YS). The heat is generated
faster towards the low values of both the SHE and YS. Figure 8b indicates that the SHE
plays a more significantly influential role in heat generation. However, the strain rate (SR)
is a significant factor, but only when considering small values of SHE. For small values of
SHE, the sample orientation in interaction with the SHE does not significantly affect the
temperature value (Figure 8c). However, if the SHE increases then the difference between
the temperature for 0◦ and 90◦ orientations increases to around 30 ◦ C. For small values of
SHE, starting from a plastic strain (PS) of about 80%, the regression model overestimates
the temperature values (Figure 8d). However, the trend observed in the experimental
270
Materials 2021, 14, 5259
studies is maintained, i.e., with an increase in PS for a given SHE value, the temperature
value increases.
(a) (b)
(c) (d)
Figure 8. Response surface plots presenting the interaction between the strain hardening exponent and (a) yield stress,
(b) strain rate, (c) sample orientation and (d) plastic strain, affecting the value of the maximum temperature of the surface of
the 321 steel sample.
The synergistic effect of both SR and YS is more complicated (Figure 9a). Similarly, in
the case of sample orientation and YS (Figure 9b), it is only for the lowest value of YS that
strain rate and orientation significantly affect the value of temperature. Figure 9c indicates
that the orientation when testing samples with different SR values plays the following role:
for orientation 0◦ , the temperature is low and increases with change in the orientation to
45◦ and 90◦ . The effect of PS and SR shown in Figure 9d is very similar to the effect of PS
and SHE on temperature (Figure 8d). A strong correlation between PS and temperature
is also confirmed in Figure 9e,f by the high isotherm concentration. Specimens cut at 90◦
heat faster during deformation than specimens cut at 0◦ (Figure 9e). The reverse effect is
observed in Figure 9f: when the YS of specimens is small, the material heats up faster.
271
Materials 2021, 14, 5259
(a) (b)
(c) (d)
(e) (f)
Figure 9. Response surface plots presenting the interaction between (a) strain rate (SR) and yield stress (YS), (b) sample
orientation and YS, (c) sample orientation and SR, (d) plastic strain (PS) and SR, (e) PS and sample orientation and (f) PS
and YS, affecting the value of maximum temperature of 321 steel sample surface.
272
Materials 2021, 14, 5259
4. Conclusions
In this paper, full-field image correlation and infrared thermography techniques were
used to analyse the effect of material properties and strain rate on heat generation in
stretched dog-bone specimens of titanium-stabilised austenitic stainless steel. Based on the
experimental results and analysis of variance, the following conclusions can be drawn:
• The heat during stretching of the samples was not generated proportionally over
the entire length of the samples. In the range of proportional deformation (before
necking formation), the zone with the highest temperature was located approximately
centrally along the distance between the grippers of the testing machine.
• At the moment of formation of a necking at the highest strain rate, the maximum sam-
ple temperature increased more than four times compared to the initial temperature.
This phenomenon was observed for all sample orientations analysed.
• The character of the effect of plastic strain on the maximum temperature generated in
the specimens was close to a linear trend in the range of strain rates analysed.
• A synergistic effect of the SHE and YS revealed that heat was generated more rapidly
towards small values of SHE and YS.
• For small values of SHE, the sample orientation in interaction with the SHE did not
significantly affect the temperature value.
• For all the orientations analysed, the temperature increased in a similar way with
increasing PS.
Author Contributions: Conceptualisation, K.Ż.; methodology, K.Ż., S.P., S.R. and T.T.; software, K.Ż.;
S.P. and M.B.; validation, K.Ż. and T.T.; investigation, K.Ż., S.P., S.R., M.B. and T.T.; data curation, K.Ż.
and T.T.; visualisation, K.Ż. and T.T.; writing—original draft preparation, T.T.; writing—review and
editing, K.Ż. and T.T. All authors have read and agreed to the published version of the manuscript.
Data Availability Statement: The data presented in this study are available on request from the
corresponding author.
References
1. Pater, Z.; Samołyk, G. Podstawy Technologii Obróbki Plastycznej Metali; Politechnika Lubelska: Lublin, Poland, 2013.
2. Di Gioacchino, F.; Edwards, T.E.J.; Wells, G.N.; Clegg, W.J. A new mechanism of strain transfer in polycrystals. Sci. Rep. 2020,
10, 10082. [CrossRef]
3. Physics on Nondestructive Evaluation. Iowa State University. Available online: https://www.nde-ed.org/Physics/Materials/
Structure/anisotropy.xhtml (accessed on 19 August 2021).
4. Yang, G.; Park, S.J. Deformation of single crystals, polycrystalline materials, and thin films: A review. Materials 2019, 12, 2003.
[CrossRef]
5. Venkataraman, A.; Sangid, M.D. Deformation Mechanisms at Grain Boundaries in Polycrystals; Final Report to U.S. Department of
Energy, Office of Basic Energy Sciences, Division of Materials Sciences and Engineering under Award#DE-SC001428; Purdue
University: West Lafayette, IN, USA, 2019.
6. Luo, H.; Sheng, H.; Zhang, H.; Wang, F.; Fan, J.; Du, J.; Liu, J.P.; Szlufarska, I. Plasticity without dislocations in a polycrystalline
intermetallic. Nat. Commun. 2019, 10, 3587. [CrossRef] [PubMed]
7. Bahrami, A.; Taheri, P. A study on the failure of AISI 304 stainless steel tubes in a gas heater unit. Metals 2019, 9, 969. [CrossRef]
8. Bansod, A.V.; Patil, A.P.; Moon, A.P.; Khobragade, N.N. Intergranular corrosion behavior of low-nickel and 304 austenitic stainless
steels. J. Mater. Eng. Perform. 2016, 25, 3615–3626. [CrossRef]
9. Hall, E.O. The deformation and ageing of mild steel: III discussion of results. Proc. Phys. Soc. 1951, 64, 747–753. [CrossRef]
10. Petch, N.J. The cleavage strength of polycrystals. J. Iron Steel Inst. 1953, 174, 25–28.
11. Roa, J.J.; Wheeler, J.M.; Trifonov, T.; Fargas, G.; Mateo, A.; Michler, J.; Jiménez-Piqué, E. Deformation of polycrystalline TRIP
stainless steel micropillars. Mater. Sci. Eng. A 2015, 647, 51–57. [CrossRef]
12. Di Gioacchino, F.; da Fonseca, J.Q. An experimental study of the polycrystalline plasticity of austenitic stainless steel. Int. J. Plast.
2015, 74, 92–109. [CrossRef]
273
Materials 2021, 14, 5259
13. Cheong, K.S.; Busso, E.P. Discrete dislocation density modelling of single phase FCC polycrystal aggregates. Acta Mater. 2004,
52, 5665–5675. [CrossRef]
14. Ichiyanagi, K.; Takagi, S.; Kawai, N.; Fukaya, R.; Nozawa, S.; Nakamura, K.G.; Liss, K.D.; Kimura, M.; Adachi, S.I. Microstructural
deformation process of shock-compressed polycrystalline aluminum. Sci. Rep. 2019, 9, 7604. [CrossRef]
15. Korsunsky, A.M. Crystal plasticity and hardening: A dislocation dynamics study. Procedia Eng. 2009, 1, 241–244.
16. Farren, W.S.; Taylor, G.I. The heat developped during plastic extension of metal. Proc. R. Soc. A 1925, 107, 422–451.
17. Taylor, G.I.; Quinney, H. The latent energy remaining in a metal after cold working. Proc. R. Soc. A 1933, 143, 307–326.
18. Bodelot, L.; Sabatier, L.; Charkaluk, E.; Dufrenoy, P. Optical and infrared coupled full-field measurements at a mesoscopic scale.
In Proceedings of the 9th International Conference on Quantitative InfraRed Thermography, Kraków, Poland, 2–5 July 2008.
19. Boulanger, T.; Chrysochoos, A.; Mabru, C.; Galtier, A. Calorimetric analysis of dissipative and thermoelastic effects associated
with the fatigue behavior of steels. Int. J. Fatigue 2004, 26, 221–229. [CrossRef]
20. Ait-Amokhtar, H.; Fressengeas, C.; Boudrahem, S. The dynamics of Portevin-Le Chatelier bands in an Al-Mg alloy from infrared
thermography. Mater. Sci. Eng. A 2008, 488, 540–546. [CrossRef]
21. Völkl, R.; Fischer, B. Mechanical testing of ultra-high temperature alloys. Exp. Mech. 2004, 44, 121–127. [CrossRef]
22. Anwander, M.; Zagar, B.G.; Weiss, B.; Weiss, H. Noncontacting strain measurements at high temperatures by the digital laser
speckle technique. Exp. Mech. 2000, 40, 98–105. [CrossRef]
23. Pan, B.; Wu, D.; Gao, J. High-temperature strain measurement using active imaging digital image correlation and infrared
radiation heating. J. Strain. Anal. Eng. Des. 2013, 49, 224–232. [CrossRef]
24. Pan, B.; Wu, D.; Wang, Z.; Xia, Y. High-temperature digital image correlation method for full-field deformation measurement at
1200 ◦ C. Meas. Sci. Technol. 2011, 22, 015701. [CrossRef]
25. Sharpe, W.N. A high-frequency high-temperature optical strain/displacement gage. Exp. Mech. 2009, 50, 227–237. [CrossRef]
26. Orteu, J.J.; Rotrou, Y.; Sentenac, T.; Robert, L. An innovative method for 3-D shape, strain and temperature full-field measurement
using a single type of camera: Principle and preliminary results. Exp. Mech. 2007, 48, 163–179. [CrossRef]
27. Tung, S.H.; Sui, C.H. Application of digital-image-correlation techniques in analysing cracked cylindrical pipes. Sadhana 2010,
35, 557–567. [CrossRef]
28. Hild, F.; Bouterf, A.; Forquin, P.; Roux, S. On the use of digital image correlation for the analysis of the dynamic behavior of
materials. In The Micro-World Observed by Ultra High-Speed Cameras; Tsuji, K., Ed.; Springer: Cham, Switzerland, 2018; pp. 185–206.
29. Wattrisse, B.; Chrysochoos, A.; Muracciole, J.-M.; Nemoz-Gaillard, M. Kinematic manifestations of localisation phenomena in
steels by digital image correlation. Eur. J. Mech. A Solids 2001, 20, 189–211. [CrossRef]
30. Hung, P.C.; Voloshin, A.S. In-plane strain measurement by Digital Image Correlation. J. Braz. Soc. Mech. Sci. Eng. 2003,
25, 215–221. [CrossRef]
31. Feng, X.; Xue, F. Characterization of 3D printed bolts based on digital image correlation and infrared thermography. Mater. Des.
2020, 191, 108641. [CrossRef]
32. Cholewa, N.; Summers, P.T.; Feih, S.; Mouritz, A.P.; Lattimer, B.Y.; Case, S.W. A technique for coupled thermomechanical response
measurement using infrared Thermography and Digital Image Correlation (TDIC). Exp. Mech. 2016, 56, 145–164. [CrossRef]
33. Żaba, K.; Trzepieciński, T.; Puchlerska, S.; Noga, P.; Balcerzak, M. Coupled thermomechanical response measurement of
deformation of nickel-based superalloys using full-field digital image correlation and infrared thermography. Materials 2021,
14, 2163. [CrossRef] [PubMed]
34. Chrysochoos, A.; Huon, V.; Jourdan, F.; Muracciole, J.M.; Peyrux, R.; Wattrisse, B. Use of full-field digital image correlation
and infrared thermography measurements for the thermomechanical analysis of material behaviour. Strain 2010, 46, 117–130.
[CrossRef]
35. Maynadier, A.; Poncelet, M.; Lavernhe-Taillard, K.; Roux, S. One-shot measurement of thermal and kinematic fields: Infrared
image correlation (IRIC). Exp. Mech. 2011, 52, 241–255. [CrossRef]
36. Curà, F.; Curti, G.; Sesana, R. A new iteration method for the thermographic determination of fatigue limit in steels. Int. J. Fatigue
2005, 27, 453–459. [CrossRef]
37. Doudard, C.; Calloch, S.; Cugy, P.; Galtier, A.; Hild, F. A probabilistic two-scale model for high-cycle fatigue life predictions.
Fatigue Fract. Eng. Mater. Struct. 2005, 28, 279–288. [CrossRef]
38. ASTM International. ASTM International. ASTM E8/E8M-11. Standard test methods for tension testing of metallic materials. In
ASTM Book of Standards; ASTM International: West Conshohocken, PA, USA, 2011.
39. SAE International. AMS 5604-Steel, Corrosion and Heat Resistant, Sheet, Strip, and Plate 18Cr-10.5Ni-0.40Ti (SAE 30321) Solution Heat
Treated; SAE International: Warrendale, PA, USA, 1991.
40. Zielecki, W.; Kubit, A.; Świ˛ech, Ł. Analiza Doświadczalna Odkształcenia Elementów Klejonych w Próbie Statycznego Oddzierania.
Pomiary Autom. Robot. 2013, 2, 71–78.
41. Trzepieciński, T.; Szpunar, M.; Kaščák, L’. Modeling of friction phenomena of TI-6AL-4V sheets based on backward elimination
regression and multi-layer artificial neural networks. Materials 2021, 14, 2570. [CrossRef] [PubMed]
42. Brennhaugen, D.D.E.; Georgarakis, K.; Yokoyama, Y.; Nakayama, K.S.; Arnberg, L.; Aune, R.E. Probing heat generation during
tensile plastic deformation of a bulk metallic glass at cryogenic temperature. Sci. Rep. 2018, 8, 16317. [CrossRef]
43. Kostina, A.; Iziumowa, A.; Plekhov, O. energy dissipation and storage in iron under plastic deformation (experimental study and
numerical simulation). Frat. Ed. Integrità Strutt. 2014, 27, 28–37. [CrossRef]
274
Materials 2021, 14, 5259
44. Rusinek, A.; Klepaczko, J.R. Experiments on heat generated during plastic deformation and stored energy for TRIP steels. Mater.
Des. 2009, 30, 35–48. [CrossRef]
45. Klitschke, S.; Trondl, A.; Huberth, F.; Liewald, M. Adiabatic heating under various loading situations and strain rates for advanced
high-strength steels. IOP Conf. Ser. Mater. Sci. Eng. 2018, 418, 012123. [CrossRef]
46. Sada, N.B. The plastic depth of heat treatment steel alloy (AISI 01) due to torsion test. J. Appl. Sci. 2005, 5, 1196–1199. [CrossRef]
47. Shen, Y.F.; Li, X.X.; Sun, X.; Wang, Y.D.; Zuo, L. Twinning and martensite in a 304 austenitic stainless steel. Mater. Sci. Eng. A 2012,
552, 514–522.
48. Song, G.S.; Ji, K.S.; Song, H.W.; Zhang, S.H. Microstructure transformation and twinning mechanism of 304 stainless steel tube
during hydraulic bulging. Mater. Res. Express 2019, 6, 1265h9. [CrossRef]
49. Rubtsov, V.E.; Kolubaev, A.V. Effect of heat generation due to plastic deformation on behavior of surface-layer material during
sliding. J. Frict. Wear 2009, 30, 324. [CrossRef]
50. Maj, M. Wpływ Kierunku Wst˛epnego Odkształcenia na Proces Magazynowania Energii w Polikryształach. Ph.D. Thesis, Instytut
Podstawowych Problemów Techniki Polskiej Akademii Nauk, Warsaw, Poland, 2007.
51. Montgomery, D.C. Design and Analysis of Experiments, 8th ed.; John Wiley & Sons Inc.: Hoboken, NJ, USA, 2013.
52. Sen, A.; Srivastava, M. Regression Analysis Theory, Methods, and Applications; Springer: New York, NY, USA, 1990.
53. Montgomery, D.C.; Peck, E.A.; Vining, G.G. Introduction to Linear Regression Analysis, 5th ed.; John Wiley & Sons Inc.: Hoboken,
NJ, USA, 2012.
54. Alexopoulos, E.C. Introduction to multivariate regression analysis. Hippokratia 2010, 14, 23–28. [PubMed]
55. Chauhan, S.R.; Dass, K. Dry sliding wear behaviour of titanium (grade 5) alloy by using response surface methodology. Adv.
Tribol. 2013, 2013, 272106. [CrossRef]
56. Khalatbari, H. Investigation of Formability of Material in Incremental Sheet Metal Forming Process. Master’s Thesis, Eastern
Mediterranean University, Farmagusta, Cyprus, 2012.
57. Venugopal, V.; Kumar, K.J.; Muralidharan, S.; Parasuraman, S.; Raj, P.V.; Kumar, K.V. Optimization and in-vivo evaluation of
isradipine nanoparticles using Box-Behnken design surface response methodology. OpenNano 2016, 1, 1–15. [CrossRef]
275
materials
Article
Study of Interaction Mechanism between Positrons and Ag
Clusters in Dilute Al–Ag Alloys at Low Temperature
Xiaoshuang Liu 1,2 , Peng Zhang 1 , Baoyi Wang 1,2 , Xingzhong Cao 1,2 , Shuoxue Jin 1 and Runsheng Yu 1, *
1 Institute of High Energy Physics, Chinese Academy of Sciences, Beijing 100049, China; [email protected] (X.L.);
[email protected] (P.Z.); [email protected] (B.W.); [email protected] (X.C.);
[email protected] (S.J.)
2 School of Nuclear Science and Technology, Chinese Academy of Sciences, Beijing 100039, China
* Correspondence: [email protected]; Tel.: +86-010-88236990
Abstract: The microstructural evolution of dilute Al–Ag alloys in its early aging stage and at low
temperatures ranging from 15 K to 300 K was studied by the combined use of Positron annihilation
lifetime spectroscopy (PALS), high resolution transmission electron microscopy (HRTEM), and
positron annihilation Coincidence Doppler broadening (CDB) techniques. It is shown that at low
temperatures below 200 K, an Ag–vacancy complex is formed in the quenched alloy, and above 200 K,
it decomposes into Ag clusters and monovacancies. Experimental and calculation results indicate
that Ag clusters in Al–Ag alloys can act as shallow trapping sites, and the positron trapping rate is
considerably enhanced by a decreasing measurement temperature.
Keywords: Al–Ag alloys; Ag clusters; positron annihilation; monovacancies; shallow trapping
Citation: Liu, X.; Zhang, P.; Wang, B.;

Cao, X.; Jin, S.; Yu, R. Study of 1. Introduction
Interaction Mechanism between Al–Ag alloys are important for the study of the fundamental processes of decom-
Positrons and Ag Clusters in Dilute position in supersaturated alloys, owing to the large difference in the atomic numbers
Al–Ag Alloys at Low Temperature. of Al and Ag (13 versus 47) [1]. Although Al–Ag alloys have limited structural appli-
Materials 2021, 14, 1451. https:// cations, Ag is a commonly used microalloying element in aluminum alloys. Adding a
doi.org/10.3390/ma14061451
small amount (from 0.1 at.% to 0.5 at.%) of Ag into aluminum-based alloys can improve
their mechanical properties, thermal stability, and stress corrosion cracking resistance [2,3].
Academic Editor: Elena Pereloma
This effect is widely seen in aluminum alloys, particularly in the high-strength alloys for
advanced aerospace and defense applications, including Al–Cu–Mg and Al–Zn–Mg-based
Received: 9 February 2021
alloys [4–6]. However, most studies on Al–Ag alloys have focused on the relatively high
Ag concentration in Al–Ag alloys. In 1942, Guinier [7] found that the Ag atoms in Al–Ag
alloys first form small clusters, then evolve into Guinier-Perston (GP) zones and further
larger precipitates in the early stage of the aging process. Guinier-Preston (GP) zones are
the early-stage solute enriched regions in aluminum alloys [8,9]. Nagai et al. [10] claimed
that the chemical composition of Ag-rich nanoclusters is AlAg3 in aged Al-2.0 at.% Ag from
iations. experimental and calculated two-dimensional angular correlation of positron annihilation
radiation (2D-ACAR) anisotropy results. In a recent study, B. Zou et al. [11] reported the
precipitation process of Ag nanoclusters in three Al–Ag alloys with different Ag contents
(Al-1 wt.%Ag, Al-5 wt.%Ag, and Al-15 wt.%Ag). However, in the above studies, little at-
tention was paid to the Al–Ag alloys with a very low Ag concentration (less than 0.5 wt.%).
Moreover, the mechanism of formation and migration of Ag clusters in Al–Ag alloys at
the early stage of aging is still unclear. The main reason for this is most probably that the
Ag clusters that form at the early stage of aging in dilute Al–Ag alloys are too small to
be detected by conventional methods. In particular, unlike most other aluminum alloys,
Attribution (CC BY) license (https:// GP zones in the Al–Ag system form immediately after quenching [12,13], which makes it
creativecommons.org/licenses/by/ difficult to detect the precipitation of Ag clusters in the early stage of aging.
4.0/).

Materials 2021, 14, 1451
Positron annihilation spectroscopy (PAS) is a highly sensitive technique for studying

subatomic scale defects, e.g., 10−5 ~10−6 density defects, in metals and semiconductors [14–16].
If a vacancy exists in the material, it will easily capture positrons, since the vacancy does not
contain positively charged nuclei. In fact, there are two typical sites in which positrons can
be trapped [17,18]. One is the deep positron trapping sites, such as vacancies and voids—
since the positron affinity of these defects is less than −1 eV [19], positrons encounter a
deep potential well from which it is thereby difficult to escape; the other type is the shallow
positron trapping sites—e.g., dislocations. The positron affinities of these defects are greater
than −100 meV [20]. Hidalgo et al. [21] studied positron annihilation in deformed copper
and found that dislocations can also trap positrons, and the positron trapping rate increased
when the temperature was below 77 K. In addition to the two above-mentioned typical
types of positron trapping sites, in recent years, some clusters have been shown to be able
to act as shallow positron trapping sites; e.g., in 2002, Huis et.al [22] found that lithium
nanoclusters are able to trap positrons in MgO, and the trapping ability is enhanced at
low temperatures. Indeed, the reason for positron trapping lies mainly in the different
chemical elements. A table of positron affinities for most elements can be found in the
work of Puska et al. [23]. The values are −4.41 and −5.36 eV for aluminum and silver,
respectively. This indicated that the Ag atom has a weak but extended attractive potential
to positrons in Al–Ag alloys. Zou et al. [11] found that the Ag clusters in Al–Ag alloys
can act as positron trapping centers and that the confinement of positrons is enhanced
with the increase in the size of Ag clusters. Moreover, Hugenschmidt et al. [24] found that
coincidence doppler broadening (CDB) is sensitive to the detection of small precipitates in
aluminum, and even a 0.1 nm thin Sn layer embedded underneath a 200 nm Al coating
leads to a significant and unambiguous fingerprint in the CDB spectrum. Nevertheless,
studies on the positron trapping mechanism in dilute Al–Ag alloys have not been reported
in literature. Moreover, it is very important to investigate the formation and evolution of
Ag clusters in Al–Ag alloys.
The main purpose of this work is to investigate the microstructural evolution of Al–Ag
alloys in the early aging stage by means of positron annihilation spectroscopy to examine
whether or not the minor Ag clusters can act as positron trapping centers and to study
the temperature dependence of positron annihilation behavior. In addition to positron
annihilation lifetime spectroscopy (PALS) and coincidence doppler broadening (CDB)
measurements, we also used high resolution transmission electron microscopy (HRTEM)
to investigate the Ag clusters at the early aging stage in Al–Ag alloys.

2.1. Materials and Preparation
The Al-0.2 wt.%Ag alloys were prepared from aluminum (99.999% purity) and silver
(99.999% purity) using a high-frequency induction furnace in a vacuum environment. After
melting, the specimens were solution treated at 823 K for 24 h in a vacuum environment,
quenched by ice water as quickly as possible (~15 ms), thinned to 1 mm, and punched
into 10 × 10 mm2 square sheets. Pure aluminum and sliver samples were also prepared
as reference specimens. All the square sheet samples were electrochemically polished to
a mirror-like surface using 25% HClO4 (Beijing Institute of Chemical Reagents, Beijing,
China) and 75% C2 H6 O (Beijing Institute of Chemical Reagents, Beijing, China) polishing
solution at room temperature (300 K).
Al-0.2 wt.%Ag alloys were then heated at 823 K for 30 min, followed by quenching
into ice water. Then, the specimens were naturally aged for 24 h at room temperature. The
specimens were measured PLAS and DBS at temperatures of 15–300 K for about 12 h. After
that, the CDB spectra were measured for about 12 h.
2.2. Characterization Methods

PALS measurements were carried out using a fast–slow coincidence system with a
time resolution of 196 ps (full width at a half maximum). A 22 Na positron source with
278
Materials 2021, 14, 1451
an intensity of 10 μCi was encapsulated by a Kapton foil (made by Nilaco Corporation,

Tokyo, Japan). The thickness of the Kapton foil was 7.5 microns. The total number
counts were 2.0 × 106 for each lifetime spectrum. The count rate of the PALS was about
100 cps. The data from the experiment were analyzed using the LT9 program [25]. After
deducting the components (12%) of the positrons annihilated in Kapton foils, one or two
lifetime components were identified in our experimental spectra. At 15–200 K, the positron
annihilation lifetime could only be deconvoluted into one lifetime component. Above
200 K, the positron annihilation lifetime was decomposed into two components.
Doppler broadening spectroscopy (DBS) was performed using a single high-purity
Ge detector (GEM35P4-76-PL, ORTEC, Oak Ridge, TN, United States) with an energy
resolution of 1.3 keV at 511.0 keV. The total number of counts for each single detector DBS
spectrum was also 2.0 × 106 . The S and W parameters were obtained from these DBS
results. The S parameter was defined as the ratio area in 510.55–511.45 keV to the total
area of the 511 keV photo peak, reflecting the ratio of the annihilation of positrons with the
low-momentum electrons. The W parameter is defined as the ratio area in 514.6–521.2 keV
and 500.8–507.4 keV to the total area, reflecting the ratio of annihilation of positrons with
the high-momentum electrons. CDB measurements were carried out using two high-purity
Ge detectors, with the efficiency of 35%. The counting rate was about 100 cps, and each
spectrum took 12 h to ensure no less than 106 counts in each CDB spectrum. To obtain
supplementary information about the Al–Ag sample microstructure, a TecnaiF30 high
resolution transmission electron microscopy (HRTEM, FEI, Hillsboro, OR, United States)
operating at 300 kV was utilized.

Figure 1 shows a typical HRTEM image of Al–Ag alloy sample after aging at room
temperature for 24 h. The inset in Figure 1 shows the Fast Fourier Transform (FFT) of the
HRTEM image. The HRTEM presented almost prefect crystallographic image. There were
no defects or clusters that could be observed in the HRTEM image. The reason may be
that the atomic size difference between Al (1.43 Å) and Ag (1.44 Å) is negligible. Thus,
it was difficult to distinguish clusters using HRTEM. Moreover, the concentration of Ag
clusters in Al-0.2 wt.%Ag alloys is too low (less than 2 ×10−4 ). The HRTEM was not
sufficiently sensitive to detect such a low content of clusters. It is thus understandable
that most previous studies focused on samples with a high Ag content after aging at high
temperature [26], where the sizes of GP zones in the studied samples were more than 10 nm.
However, as proved by the present study, it is difficult to detect Ag cluster in Al–Ag alloys
with very low Ag content after aging at room temperature, even with HRTEM. The FFT of
the HRTEM image however has streaks which may indicates the presence of some very fine
linear defects or clusters. To make further clarification, positron annihilation spectroscopy
is thereafter applied for characterization of the nanostructure in dilute Al–Ag alloys.
The positron annihilation lifetime of quenched Al-0.2% Ag alloys at different tempera-
tures ranging from 15 K to room temperature was obtained by LT9 analysis. At 15–200 K,
the positron annihilation lifetime could only be deconvoluted into one lifetime component
τ m about 150 ps after deducting the source components. We measured the positron an-
nihilation value of pure Al at room temperature. The observed value was 150 ps, which
was very close to that in the Ag–Vacancy complex (146 ps) [27]. In addition, the positron
annihilation lifetime in the vacancy is 231 ps [28], which is much larger than the positron
annihilation lifetime at 15–200 K. It is worth noting that the positron annihilation lifetime
at 15–200 K is very close to that in aluminum and Ag–V complexes. This provides evidence
that the V–Ag complexes were formed after being quenched, which is consistent with
the conclusion of Nagai et al. [27] who showed that quenched-in vacancies are bound
to Ag atoms in Al–Cu–Mg–Ag alloys with very low Ag contents. Furthermore, Shasha
Zhang et al. [29] reported that the Ag atoms have a strong attraction interaction with the
vacancy in Al–Ag alloys. That may be the origin of the formation of Ag–V complexes in
Al–Ag alloys. In 1988, F. Maury et al. [30] conducted electrical resistivity measurements
279
Materials 2021, 14, 1451
on dilute Al–Ag alloys after electron irradiation. They reported that the Ag atoms in the
dilute Al–Ag alloys are easily trapped by vacancies to form Ag–V complexes when the
temperature is below 180 K, and Ag–V complexes can be recovered above 180 K. This
coincides rather well with our current positron annihilation measurement results.
Figure 1. High resolution transmission electron microscopy (HRTEM) image of the nanostructure of
Al-0.2 wt.%Ag sample aged at room temperature after quenching. The inset view is the Fast Fourier
Transform (FFT) of the HRTEM image.
As shown in Figure 2, the positron annihilation lifetimes in Al–Ag alloys changed

from one component of ~150 ps to two components of 118 ps and 195 ps over 200 K as the
temperature increased. Most probably, in our studied dilute Al–Ag alloys, Ag–V complex
formed as they were quenched, and after 200 K, the Ag–V complex decomposed into Ag
clusters and vacancies. With Ag atoms gradually separated from the binding of the vacan-
cies, the size of the open space near Ag–V complexes increased, resulting in the appearance
of a long positron annihilation lifetime of 195 ps (τ 2 ). Indeed, above 200 K, the short lifetime
(τ 1 ) of around 118 ps was close to the positron annihilation lifetime of Ag clusters, and the
long lifetime of 195 ps was close to the positron annihilation lifetime of monovacancies [31].
This provides evidence that the Ag–V complexes are decomposed into monovacancies and
Ag clusters at 200 K. Thereafter, part of the injected positrons annihilated in the vacancies,
while others annihilated on Ag clusters. Furthermore, as can be seen from Figure 2, the
relative intensity (I1 and I2 ) for τ 1 and τ 2 gradually decreased from 52.7 to 44 and increased
from 47.3% to 56%, respectively, as the temperature increased from 200 K to 300 K. That
indicates that fewer positrons annihilated with the electrons around Ag clusters and more
positrons annihilated with the monovacancies as the temperature increased. Thus, the
ability of Ag clusters to capture positrons is enhanced at a lower temperature.
Single detector DBS results are shown in Figure 3. It is evident that the S increased
from 0.494 to 0.509 as the temperature increased from 15 K to 200 K. The W showed an
opposite trend, decreasing from 0.112 to 0.097. This suggests the same conclusion as that of
the PALS results; i.e., Ag atoms gradually separated from the Ag–V complex and resulted
in an increase in the size of the open space near Ag–V complexes at 200–300 K. It is worth
noting that the S value started to increase slowly after 200 K. In particular, from 250 K,
the S value basically remained stable. This indicates that almost all Ag–V complexes
decomposed into Ag clusters and monovacancies until 250 K.
280
Materials 2021, 14, 1451
Figure 2. Positron annihilation lifetimes (a) and their relative intensities (b) of Al-0.2 wt.%Ag as a
function of temperature. τ 1 , τ 2 and τ m represent a short lifetime, long lifetime and mean lifetime in
the positron annihilation spectrum. I1 and I2 represent the relative intensity of the short lifetime and
long lifetime.
Figure 3. S and W values for the Al-0.2% Ag alloys at different measurement temperatures.
Figure 4 shows the DBS results with the S-W plot of Al–Ag alloys at different tempera-
tures. Clearly, the S-W characteristic points are in a straight line, and the S-W characteristic
points move from Ag towards Al with increasing temperature. This indicates that the
surrounding elements in which positrons were trapped and annihilated gradually changed
from Ag clusters (surrounding Ag–V complexes) to aluminum (surrounding monova-
cancies) as the temperature increased. This is in consistent with the above-mentioned
conclusion from the positron annihilation lifetime results.
281
Materials 2021, 14, 1451
Figure 4. S-W plot for the Al-0.2 wt.%Ag alloys at different temperatures. The arrow shows the
increase in the measurement temperature, and the movement of the S-W characteristic points from
Ag towards Al. The red line is presented as a visual aid.
Figure 5 shows the ratio curves for CDB spectra of Al-0.2% Ag at different temper-
atures compared to pure Al; the ratio curve for pure Ag measured at room temperature
is also added for comparison. A CDB ratio curve can reflect the momentum distribution
of core electrons of certain elements annihilated with positrons, due to the significantly
enhanced signal-to-noise ratio of at least 105 :1 for a typical CDB spectrum. As shown in
Figure 5, there is a narrow peak at around 11 × 10−3 m0 c in the CDB ratio curve of pure
Ag, which is consistent with the experimental results of Y. Nagai [32] and B. Zou [11]. The
curves in Figure 5 show a bump at 11 × 10−3 m0 c at low temperatures, which disappears at
high temperatures. In particular, at 200 K, the characteristic peak at 14 × 10−3 m0 c moves
to 15 × 10−3 m0 c. Combined with the positron annihilation lifetime results, this gives
evidence that Ag atoms gradually become free from binding with a vacancy and the Ag–V
complexes decomposed to Ag clusters and monovacancies at 200 K. When the temperature
reached 300 K, however, the characteristic peak for Ag clusters in Al-0.2% Ag was hardly
observable; there was only a very small ratio fluctuation around 1.0.
Figure 5. Coincidence doppler broadening (CDB) ratio curves for Al-0.2 wt.%Ag at different temper-
atures. The statistical errors are omitted for clarity of representation.
282
Materials 2021, 14, 1451
When the temperature increased above 200 K, the value of characteristic peak grad-
ually decreased with the increase in temperature. That is, above 200 K, more positrons
annihilated with low-momentum electrons around monovacancies and fewer positrons
were trapped and annihilated with high-momentum inner shell 4d electrons of Ag clusters.
In other words, the Ag clusters can trap positrons, but the positron trapping ability of Ag
clusters decreases with the increase in temperature. At room temperature, the Ag clusters
could barely trap positrons any more. Based on such temperature-dependent behavior, we
may deduce that the Ag clusters can act as shallow positron trapping sites to trap positrons,
and the trapping ability increases with the decrease in temperature. The conclusion is
consistent with the results of B. Zou et al. [11].
Puska et al. [23] have reported that a minimum radius of embedded clusters is neces-
sary for positron trapping to occur. There is one bound state if the radius of the precipitate
is larger than the critical radius, rc = 5.80a
√ 0 , where ΔA is the positron affinity difference
ΔA
between the two kinds of metals, and a0 is the Bohr radius. For the Al–Ag alloy, the critical
radius is about 5.9 a0 with A+ (Al) = −4.41 eV and A+ (Ag) = −5.36 eV. This means that
precipitates should contain at least ~6 Ag atoms.
We attempted to quantitatively describe the positron trapping mechanism in Al–Ag
alloys at varied temperatures utilizing the three state trapping model [20,33]. At 15–200 K,
there were two positron trapping states in the Al–Ag alloy: free annihilation (bulk state)
and deep trapping (Ag–V complex). At 200–300 K, there were three positron states in
the Al–Ag alloy: free annihilation (bulk state), shallow trapping (Ag clusters), and deep
trapping (vacancies). For spherical defects, the positron trapping rate and detrapping rate
capture rate can be written as [19,34]
κ( T ) = r0 Ce−γT (1)
m − 1 −1 3 − 1 − Eb
δ( T ) = π 2 C κ ( T )(k B T ) 2 Eb 2 e k B T (2)
4 2
where γ is the trapping rate of positrons by the trapping sites at 0 K and r0 is a constant
representing the specific trapping rate of such trapping sites. C means the concentration of
trapping sites, Eb means the binding energy of the trapping sites and positrons. The above
equations show that the trapping rate is proportional to the concentration of trapping sites
(either shallow or deep trapping). In contrast, the detrapping rate is independent of the
trapping site concentration but related to the binding energy.
Using the above equations, we calculated the temperature dependence of the positron
detrapping rate as a function of temperature for a range of binding energies from 0.01 eV
to 1 eV. The calculated results are shown in Figure 6. It can be seen that the higher the
binding energy, the smaller the detrapping rate. When the binding energy reaches 1 eV, the
detrapping rate remains at around 0. This suggests that the higher the binding energies, the
harder it is for the positrons to escape from the trapping sites. As regards the temperature
dependence, it is clear that an increased temperature facilities the detrapping of positrons
from trapping sites for all the binding energies except Eb = 1 eV. It was reported that
the positron binding energy for Ag clusters is lower than 40 meV [11]; therefore, Ag
clusters could only act as shallow trapping defects for positrons, and the low temperature
enhancement of the effect of Ag on positron trapping is thereby understandable.
283
Materials 2021, 14, 1451
Figure 6. The temperature dependence of the positron detrapping rate at different binding energies.
4. Conclusions
In summary, we conducted a positron annihilation spectroscopy study on the early
stage of precipitation in dilute Al–Ag alloys. The interaction mechanism between positrons
and Ag clusters as well as monovacancies in Al–Ag alloys at varying temperatures were
extensively studied. Several conclusions can be drawn. Firstly, a monovacancy in quenched
Al–Ag alloys can easily combine with Ag atoms to form Ag–V complexes at 15–200 K,
and within this temperature range, the Ag–V complex effect is predominant for positron
annihilation in the quenched Al-0.2% Ag alloy. Secondly, above 200 K, Ag atoms gradually
escape from binding with vacancies and migrate to form Ag clusters. Meanwhile, Ag
clusters can act as shallow trapping defects, and the positron detrapping from Ag clusters
is enhanced at higher temperatures. Last but not least, the addition of a small amount
(0.2 wt.%) of Ag has a great influence on the positron trapping mechanism; in other
words, low-temperature positron annihilation spectroscopy is an important method to
investigate the microstructural evolution of defects and precipitates in the early aging stage
of dilute alloys.
Author Contributions: Conceptualization, X.L.; methodology, P.Z. and B.W.; software, X.C. and
S.J.; formal analysis, X.L.; investigation, X.L.; resources, S.J.; data curation, X.L.; writing—original
draft preparation, X.L.; writing—review and editing, B.W., X.C., S.J. and R.Y.; project administration,
X.C. and R.Y.; funding acquisition, X.C. and R.Y. All authors have read and agreed to the published
Funding: This research was funded by the National Natural Science Foundation of China, Nos.
11675188, 11875055 and Beijing Natural Science Foundation, Nos. 11705212.
Data Availability Statement: Data sharing not applicable, all the data created for this study are
already displayed in the article.
Conflicts of Interest: We declare that we have no financial and personal relationships with other
people or organizations that can inappropriately influence our work, and there is no professional or
other personal interest of any nature or kind in any product, service and/or company that could be
construed as influencing the position presented in, or the review of, the manuscript.
284
Materials 2021, 14, 1451
References
1. Zhang, Z.Z.; Bourgeois, L.; Rosalie, J.M.; Medhekar, N.V. The bi-layered precipitate phase zeta in the Al-Ag alloy system. Acta
Mater. 2017, 132, 525–537. [CrossRef]
2. Auld, J.H.; Vietz, J.T.; Polmear, I.J. T-phase Precipitation induced by the Addition of Silver to an Aluminium–Copper–Magnesium
Alloy. Nature 1966, 209, 703–704. [CrossRef]
3. Polmear, I.J. A Trace Element Effect in Alloys based on the Aluminium-Zinc-Magnesium System. Nature 1960, 186, 303–304.
[CrossRef]
4. Lumley, R.N.; Polmear, I.J.; Morton, A.J. Development of mechanical properties during secondary aging in aluminium alloys.
Mater. Sci. Technol. 2005, 21, 1025–1032. [CrossRef]
5. Marceau, R.K.W.; Sha, G.; Ferragut, R.; Dupasquier, A.; Ringer, S.P. Solute clustering in Al-Cu-Mg alloys during the early stages
of elevated temperature ageing. Acta Mater. 2010, 58, 4923–4939. [CrossRef]
6. Nakamura, J.; Matsuda, K.; Kawabata, T.; Sato, T.; Nakamura, Y.; Ikeno, S. Effect of Silver Addition on the beta ’-Phase in
Al-Mg-Si-Ag Alloy. Mater. Trans. 2010, 51, 310–316. [CrossRef]
7. Guinier, A. The mechanism of precipitation in aluminium-silver alloy crystals during structural hydrogenation. C. R. Hebd.
Seances Acad. Sci. 1942, 214, 34–37.
8. Dubey, P.A.; Schonfeld, B.; Kostorz, G. Shape and internal structure of guinier-preston zones in Al-Ag. Acta Metall. Mater. 1991, 39.
[CrossRef]
9. Baur, R.; Gerold, V. Existence of a metastable miscibility gap in aluminium-silver alloys. Acta Metall. 1962, 10, 637–645. [CrossRef]
10. Nagai, Y.; Toyama, T.; Tang, Z.; Inoue, K.; Chiba, T.; Hasegawa, M.; Hirosawa, S.; Sato, T. Interactions between Fermi surfaces and
Brillouin zone boundaries and phase stability of embedded metallic nanoparticles. Phys. Rev. B 2009, 79, 21405. [CrossRef]
11. Zou, B.; Qi, N.; Liu, Z.W.; Chen, Z.Q.; Liu, H.Q.; Yi, D.Q.; Tang, Z. Enhanced positron trapping by Ag nanoclusters at low
temperatures: A challenge of positron sensitivity to quantum dots. J. Appl. Phys. 2017, 121, 125103. [CrossRef]
12. Alkassab, T.; Haasen, P. Early stages of precipitation in dilute Al-Ag alloys. Z. Met. 1993, 84, 248–250.
13. Nicholson, R.B.; Nutting, J. The metallography of precipitation in an A1-16% Ag alloy. Acta Metall. 1961, 9, 332–343. [CrossRef]
14. Wang, C.; Ma, B.; Cao, X.Z.; He, S.; Han, J.B.; Wei, M.; Evans, D.G.; Duan, X. Bridge-type interface optimization on a dual-
semiconductor heterostructure toward high performance overall water splitting. J. Mater. Chem. A 2018, 6, 7871–7876. [CrossRef]
15. Ning, X.; Cao, X.Z.; Li, C.; Li, D.M.; Zhang, P.; Gong, Y.H.; Xia, R.; Wang, B.Y.; Wei, L. Modification of source contribution in
PALS by simulation using Geant4 code. Nucl. Instrum. Methods Phys. Res. Sect. B Beam Interact. Mater. Atoms. 2017, 397, 75–81.
[CrossRef]
16. Cao, X.Z.; Zhu, T.; Jin, S.X.; Kuang, P.; Zhang, P.; Lu, E.Y.; Gong, Y.H.; Guo, L.P.; Wang, B.Y. Detection of helium in irradiated
Fe9Cr alloys by coincidence Doppler broadening of slow positron annihilation. Appl. Phys. A Mater. Sci. Process. 2017, 123, 177.
[CrossRef]
17. Nagai, Y.; Takadate, K.; Tang, Z.; Ohkubo, H.; Sunaga, H.; Takizawa, H.; Hasegawa, M. Positron annihilation study of vacancy-
solute complex evolution in Fe-based alloys. Phys. Rev. B 2003, 67, 224202. [CrossRef]
18. Iakubov, I.T.; Pogosov, V.V. Positron binding energy in monovacancy of metals and work functions in stabilized jellium model.
In Positron Annihilation Icpa-10, Pts 1 and 2; He, Y.J., Cao, B.S., Jean, Y.C., Eds.; Transtec Publications Ltd.: Zurich-Uetikon,
Switzerland, 1995; Volume 175, pp. 169–172. [CrossRef]
19. Smedskjaer, L.C.; Manninen, M.; Fluss, M.J. An alternative interpretation of positron-annihilation in dislocations. J. Phys. F Metal
Phys. 1980, 10, 2237–2249. [CrossRef]
20. Hidalgo, C.; Linderoth, S.; Dediego, N. Positron-trapping mechanism at dislocations in Zn. Phys. Rev. B 1987, 36, 6740–6745.
[CrossRef] [PubMed]
21. Riceevans, P.; Hlaing, T.; Chaglar, I. Evidence for Temperature Dependence of Positron Trapping Rate in Plastically Deformed
Copper. Phys. Rev. Lett. 1976, 37, 1415–1418. [CrossRef]
22. Van Huis, M.A.; van Veen, A.; Schut, H.; Falub, C.V.; Eijt, S.W.H.; Mijnarends, P.E.; Kuriplach, J. Positron confinement in
embedded lithium nanoclusters. Phys. Rev. B 2002, 65, 085416. [CrossRef]
23. Puska, M.J.; Lanki, P.; Nieminen, R.M. Positron affinities for elemental metals. J. Phys. Condes. Matter 1989, 1, 6081–6093.
[CrossRef]
24. Hugenschmidt, C.; Pikart, P.; Stadlbauer, M.; Schreckenbach, K. High elemental selectivity to Sn submonolayers embedded in Al
using positron annihilation spectroscopy. Phys. Rev. B 2008, 77. [CrossRef]
25. Kansy, J. Microcomputer program for analysis of positron annihilation lifetime spectra. Nucl. Instrum. Methods Phys. Res. Sect. A
Accel. Spectrometers Detect. Assoc. Equip. 1996, 374, 235–244. [CrossRef]
26. Alexander, K.B.; Legoues, F.K.; Aaronson, H.I.; Laughlin, D.E. Faceting of GP zones in an Al-Ag alloy. Acta Metall. 1984, 32,
2241–2249. [CrossRef]
27. Nagai, Y.; Honma, T.; Tang, Z.; Hono, K.; Hasegawa, M. Coincidence Doppler broadening study of the pre-precipitation stage of
an Al-Cu-Mg-Ag alloy. Philos. Mag. A Phys. Condens. Matter Struct. Defects Mech. Prop. 2002, 82, 1559–1572. [CrossRef]
28. Raju, P.; Gupta, R.; Siegel, W. Annihilation of a positron in a vacancy in aluminum. Phys. Rev. B 1980, 22, 4572–4589. [CrossRef]
29. Zhang, S.S.; Yao, Z.J.; Kong, X.S.; Chen, L.; Qin, J.Y. First-principles calculations of solute-vacancy interactions in aluminum. Chin.
Physics. B 2020, 29. [CrossRef]
285
Materials 2021, 14, 1451
30. Maury, F.; Lucasson, A.; Lucasson, P.; Dimitrov, C. Mixed-interstitial recovery in AlTi and AlAg dilute alloys. J. Phys. F Metal
Phys. 1988, 18, 657–675. [CrossRef]
31. Folegati, P.; Makkonen, I.; Ferragut, R.; Puska, M.J. Analysis of electron-positron momentum spectra of metallic alloys as
supported by first-principles calculations. Phys. Rev. B 2007, 75, 054201. [CrossRef]
32. Nagai, Y.; Tang, Z.; Ohkubo, H.; Takadate, K.; Hasegawa, M. Elemental analysis of positron affinitive site in materials by
coincidence Doppler broadening spectroscopy. Radiat. Phys. Chem. 2003, 68, 381–386. [CrossRef]
33. Seeger, A. Study of defects in crystals by positron-annihilation. Appl. Phys. 1974, 4, 183–199. [CrossRef]
34. Brandt, W.; Paulin, R. Positron diffusion in solids. Phys. Rev. B Solid State 1972, 5, 2430–2435. [CrossRef]
286
materials
Article
Integrating a Top-Gas Recycling and CO2 Electrolysis Process
for H2-Rich Gas Injection and Reduce CO2 Emissions from an
Ironmaking Blast Furnace
Yichao Hu 1 , Yinxuan Qiu 1 , Jian Chen 2 , Liangyuan Hao 3 , Thomas Edward Rufford 1 , Victor Rudolph 1
and Geoff Wang 1, *
1 School of Chemical Engineering, The University of Queensland, St. Lucia 4072, Australia;
[email protected] (Y.H.); [email protected] (Y.Q.); [email protected] (T.E.R.);
[email protected] (V.R.)
2 College of Mechanical and Electrical Engineering, Central South University, Changsha 410083, China;
[email protected]
3 The Strategy Research Institute, HBIS Group Co., Ltd., Shijiazhuang 050023, China; [email protected]
Abstract: Introducing CO2 electrochemical conversion technology to the iron-making blast furnace
not only reduces CO2 emissions, but also produces H2 as a byproduct that can be used as an auxiliary
reductant to further decrease carbon consumption and emissions. With adequate H2 supply to
the blast furnace, the injection of H2 is limited because of the disadvantageous thermodynamic
characteristics of the H2 reduction reaction in the blast furnace. This paper presents thermodynamic
analysis of H2 behaviour at different stages with the thermal requirement consideration of an iron-
making blast furnace. The effect of injecting CO2 lean top gas and CO2 conversion products H2 –CO
Citation: Hu, Y.; Qiu, Y.; Chen, J.; gas through the raceway and/or shaft tuyeres are investigated under different operating conditions.
Hao, L.; Rufford, T.E.; Rudolph, V.; H2 utilisation efficiency and corresponding injection volume are studied by considering different
Wang, G. Integrating a Top-Gas reduction stages. The relationship between H2 injection and coke rate is established. Injecting
Recycling and CO2 Electrolysis 7.9–10.9 m3 /tHM of H2 saved 1 kg/tHM coke rate, depending on injection position. Compared with
Process for H2 -Rich Gas Injection and
the traditional blast furnace, injecting 80 m3 /tHM of H2 with a medium oxygen enrichment rate
Reduce CO2 Emissions from an
(9%) and integrating CO2 capture and conversion reduces CO2 emissions from 534 to 278 m3 /tHM.
Ironmaking Blast Furnace. Materials
However, increasing the hydrogen injection amount causes this iron-making process to consume
2022, 15, 2008. https://doi.org/
more energy than a traditional blast furnace does.
10.3390/ma15062008
Academic Editor: Miguel Ángel Keywords: blast furnace; hydrogen injection; gas utilisation efficiency; energy consumption; CO2 emission
Sanjuán
Received: 31 January 2022

Published: 8 March 2022 1. Introduction

Traditional blast furnace (BF) iron making relies on carbon and contributes to over 70%
of CO2 emissions in the iron and steel industry [1]. In a blast furnace, coke is converted into
published maps and institutional affil- a high-temperature CO gas and performs an exothermic reaction with iron ores, resulting
iations. in a large amount of CO and CO2 leaving the furnace with top gas. Typically, every tonne
of hot metal (tHM) produced from a traditional BF requires about 500 kg/tHM carbon and
generates around 1.2 tonnes of CO2 emissions [2,3]. Hence, there were various attempts for
a clean iron-making process to reduce CO2 emissions [4–7]. One of the approaches is using
Copyright: © 2022 by the authors. alternative reductants produced from renewable energies to replace carbon. BF operation
Licensee MDPI, Basel, Switzerland. with hydrogen as an auxiliary reducing agent was extensively investigated because of
This article is an open access article its specific advantages over CO [8–12]. Compared to CO reduction that generates CO2 ,
distributed under the terms and reducing iron ores by hydrogen only forms water vapor. Kinetically, hydrogen enables
conditions of the Creative Commons a higher gas flow rate, and a faster reduction in iron ores and productivity than only
CO does [13,14]. The higher thermal conductivity of hydrogen helps in heat transfer
efficiency between solid and gas phases [15]. In addition, the hydrogen reduction in iron
4.0/).

Materials 2022, 15, 2008
ores suppresses the strong endothermic direct reduction in iron ores. However, hydrogen
reduction is thermally more disadvantageous than CO reduction. Due to this endothermic
reduction in iron ores by hydrogen, hydrogen addition changes the energy supply of the
BF, and it is only useful to a certain extent [16–18].
Few previous experiments and mathematical models have investigated the maximal
or optimal hydrogen injection to the BF, and results are controversial and need further
investigation. A thermogravimetric experiment showed the efficiency of hydrogen on
reduction rate is neglected when its content is lower than 5%, and H2 content should be 5%
to 15% at reduction temperatures between 700 and 1000 ◦ C [13,19]. Wang et al. performed
a pulverisation experiment at 900 ◦ C with 70% N2 and found that the reduction degree of
burdens was more than 90% when H2 content was higher than 20% [20]. On the basis of re-
duction experiments, Lyu et al. reported that the appropriate H2 content lies between 5% to
10% in terms of the reduction rate, gas utilisation, and reasonable distribution of the energy
in the BF in the CO and H2 mixture [17]. Nogami et al. used a multiphase fluid dynamic
model to simulate the effect of hydrogen injection with 2.5% oxygen enrichment [8]. They
demonstrated that the coke rate decreases linearly, and the maximal hydrogen injection can
reach 43.7%.
The application of hydrogen in large-scale BF iron-making processes is limited due to
its supply in terms of cost, availability, storage, and transportation [21,22]. The traditional
BF contains a low level of hydrogen content because it only generates from the blast air
moisture and the volatiles of pulverised coal. One opportunity for BF hydrogen enrichment
is injecting hydrogen-rich (so-called H2 -rich) gas or hydrogen-bearing materials from
external sources, including natural gas, fuel oil, coke oven gas (COG), reformed gas, and
waste plastics (Cn Hm ) [23–26]. Previous studies adopted photocatalysis to produce CO and
H2 from CO2 . However, the productivity of photocatalysis is less than 1000 μmol CO/gCO2
and 19 μmol H2 /gCO2 [27,28]. It is challenging for photocatalysis to meet the large-scale
CO2 conversion requirement in the iron-making process. Electrochemical CO2 conversion is
one option that can recycle carbon into blast furnace gas (BFG) as CO [29]. The high energy
demand is considered to be a major difficulty for electrochemical reduction in CO2 [30].
However, it provides an added opportunity for a BF because hydrogen is coproduced in the
electrolyser, which can be used for iron-ore reduction. In low-temperature electrolysis cells,
CO2 reduction is carried out in aqueous solutions, and different levels of current density
can be used to produce hydrogen at various concentrations [31–33]. As an additional
benefit, a pure oxygen stream is generated as another byproduct during electrolysis, which
can be used directly for oxygen enrichment to the BF [34]. In this study, we use CO2
capture and utilisation (CCU) technology to provide a reliable on-site H2 -rich gas for the
BF iron-making process while avoiding CO2 emissions. Renewable energy can be coupled
to CO2 electrolysis to achieve further emission reduction.
This study aims to use a modelling method to predict hydrogen involvement in the BF
and determine the CO2 emission reduction potential. An alternative method to determine
the hydrogen utilisation efficiency is developed. First, thermodynamic and thermal balance
models are introduced; then, hydrogen injection is quantitatively studied to determine the
optimal injection position and volume by considering key BF performance indices. Lastly,
the effect of hydrogen injection on coke consumption, gas utilisation efficiency, and the
energy consumption of the iron-making system are investigated.

Here, we studied the effect of H2 -rich gas injection through BF tuyeres at the raceway
position and/or shaft tuyeres, as shown in Figure 1. The BF is studied as the main subsystem
when changing the hydrogen injection condition. The CCU and gas heating subsystem is
used as a black box to provide necessary input information to BF. The BFG composition
and flow rate from BF are the input for the CCU unit, and the CCU unit provides H2 -rich
reducing gas as the input to the BF. The operating conditions for this BF to produce 1 tonne
hot metal (1 tHM) are listed in Table 1 and were kept constant throughout simulations. The
288
Materials 2022, 15, 2008
overall CO content in the gas injectant was maintained at 200 m3 /tHM. The productivity
of this BF is 238 tHM per hour.
Figure 1. Blast furnace process with CO2 capture, conversion, and H2 -rich gas injection.
Table 1. Operating conditions of the simulation.
Operating Parameters
PCI rate (kg/tHM) 137
Blast temperature, ◦ C 1052
Humidity of hot blast, g/m3 12.93
Top gas temperature, ◦ C 161
Hearth injection temperature, ◦ C 1250
Shaft injection temperature, ◦ C 900
As shown in Figure 1, hot oxygen-enriched blast and pulverised coal is injected

through the tuyeres. The upper limit of oxygen enrichment rate for the blast was set at 14%
to maintain the stable operation of the large-scale BFs. After drying and dust removal, some
BFG is combusted in the hot blast stoves to heat cold blast, and in the gas heating device
to provide high-temperature gas injectants. The rest of BFG enters the amine absorption
CO2 capture unit to provide a CO2 -rich stream that is processed in an electrochemical CO2
conversion unit to produce a CO and H2 stream containing, for example, 30% vol. H2 and
70% vol. CO. The CO2 -lean stream from the top of the CO2 capture unit contained a mix
of CO, H2 , and N2 that is exported to other processes in the integrated steel mill. We did
not in this study consider additional separation of CO and H2 from the N2 in this stream
for recycling back to the BF. Oxygen enrichment is required with H2 -rich gas injection
to provide heat to the BF and enrich BFG for CO2 capture [35]. The BF can take another
advantage from the CCU unit, as the electrolyser produces pure oxygen in another effluent
stream. Besides BFG, a small amount of the basic oxygen furnace gas (i.e., BOFG) from
289
Materials 2022, 15, 2008
steel making is usually combusted as fuel to a hot blast stove. Following assumptions of
the BF, CCU unit, gas heating device and hot stove are made in this study:
• Degree of indirect reduction Ri depends on the reducing gas concentration of BF bosh
gas, which is estimated by an empirical equation, as shown in Equation (1) [36]:
Ri = 0.2777 + 0.0051 × %(Reducing gas) (1)

where %(Reducing gas) is the proportion of H2 and CO in the total amount of gas entering
the bosh and shaft.
• For the amine absorption CO2 capture unit: 30% monoethanolamine (MEA) concen-
tration was used for CO2 capture in this study. The capture unit recovered 90% CO2
in BFG, and the CO2 purity was >99%; the general thermal energy requirement for
capture was assumed to be around 1000 kWh/tCO2 (3.6 GJ/ tCO2 ) [37–39]. Any
additional CO2 captured and not converted was assumed to be released as per current
operation or could be sent to CO2 storage routes, shown as CO2 letdown in Figure 1.
• The electrochemical CO2 conversion unit was treated in the model as a simplified
input–output model. Assumptions for the material and energy balance in the CO2
conversion unit were based on laboratory demonstration data with additional inputs
from literature sources. Briefly, the model was based on multiple two-cell vapour fed
electrolyser stacks with the capacity to treat 50 tCO2 per day; further details can be
found in our other report [33]. The current density of the electrolyser was altered from
2.68 V at 0 A/m2 to 3.59 V at 1862 A/m2 to produce the H2 -rich gas with different
H2 /CO compositions.
• The electricity consumption for CO2 conversion is proportional to the H2 generation,
which can be estimated as in Equation (2) [33]:
Econv = 10.75VH2 gen + 1282 (2)

where Econv is the power required for the CO2 conversion unit, kWh/tHM; VH2 gen is the
amount of H2 generated by the conversion unit, m3 /tHM;
• efficiency of the gas heating device was 85%;
• The hot blast stove system uses two stoves on-gas and one stove on-blast, and the
efficiency of the hot blast stoves was 75%.
As indicated in Figure 1, CO2 capture and conversion units use renewable energy to
avoid their own CO2 emissions. The type of renewable power used by the industry depends
on availability and cost, such as solar power [40,41]. Besides solar power, industries can
use thermal–electrical materials to recover a large amount of waste heat in an integrated
steel mill to provide electricity for CCU units [42]. In addition, using the lower heating
value energy to generate electricity in the steel mill and the on-site power plant can help to
minimise the renewable power periodic availability problem.
To achieve the objectives of this study, two mathematical models were developed.
As a reducing agent, H2 -rich gas needs to fulfil the thermodynamic requirement of the
reduction reaction to capture oxygen in iron ores. H2 -rich gas also needs to provide enough
heat in the shaft for keeping an effective reduction process. First, hydrogen behaviour in
different parts of the BF was analysed. A thermodynamic model for hydrogen reduction
was built to determine hydrogen utilisation efficiency. This model provides a guideline of
the proper hydrogen injection concentration. Then, a thermal balance model was used to
limit the hydrogen injection temperature and volume.
The optimal hydrogen injection amount and position were determined by increasing
the reduction potential in the coke consumption and CO2 emissions, increasing gas utilisa-
tion efficiency and lowering the energy consumption. A static mass balance model of the
BF was used to calculate the above parameters.
290
Materials 2022, 15, 2008
2.1. Thermodynamic Calculations of H2 -Rich Gas Injection BF

The reduction behaviours of injected gas are discussed in different parts of the BF to
determine the reducing gas utilisation.
2.1.1. Raceway
In the BF raceway, the main reactions considered in this study were carbon combustion,
coke solution loss reaction between coke and CO2 , water–gas reaction between coke and
moisture in the hot blast, which can be described as shown in Equations (3)–(5):
Solution loss reaction C (s) + CO2 (g) → 2CO (g) ΔH0 = 172,430 kJ/kmol (3)
Water–gas reaction C (s) + H2 O(g) → CO (g) + H2 (g) ΔH0 = 124,190 kJ/kmol (4)
ΔG0 = 134,542 − 142.28T (5)

With excess coke existing in the BF bottom, it could be assumed that CO and H2
combustion was negligible. This could be justified by the assumption that, if a small
amount of H2 reacts with O2 to form water in the raceway, the generated water vapour
reacts with coke and turn back to H2 . Therefore, this process can be simplified as heating
the hearth gas injectant, as shown in Equation (6):
T
hearth
ΔQhearth = Vhearth ·Cp_hearth dt (6)
25
where Thearth is hearth gas injection temperature, ◦ C; Cp_hearth denotes specific heat capacity
of the gas injected to BF, kJ/m3 ·◦ C; Vhearth is hearth injection volume, m3 /tHM; and
ΔQhearth is the sensible heat carried by the gas injected to BF, kJ/tHM.
2.1.2. Dripping, Cohesive, and High-Temperature Zones over 1000 ◦ C

The primary reaction in the dripping zone is a direct reaction between coke and FeO.
Hot gas containing H2 and CO that passes through the cohesive zone reacts with molten
FeO or semi-molten FeO to form H2 O and CO2 . As the temperature was over 1350 ◦ C in
the dripping zone, the amount of CO2 was negligible due to the solution loss reaction. At
the high-temperature zone over 1000 ◦ C, some reduced FeO and Fe3 O4 were still in the
solid state, and H2 could pass through their surface. Almost all the H2 O produced by H2
reduction rapidly participates in water–gas reaction at the presence of coke to form CO and
H2 over 1273 K (1000 ◦ C). Therefore, the reduction reaction in this section was essentially
the direct reduction in iron by coke. H2 injected through the tuyeres at the raceway mainly
catalyses direct reduction and heats the molten or semi-molten burden.
2.1.3. Shaft Zone Temperature between 800 and 1000 ◦ C

According to the thermodynamics of iron oxide reduction and dynamics studies, H2
has better reducing capability than that of CO above 800 ◦ C [43,44]. At the same time, the
extent of coke solution loss reaction and the water–gas reaction was less than that in the
higher temperature zone. In this temperature zone, H2 reacts with various iron oxides to
generate H2 O, and the formed H2 O is not gasified into H2 by carbon completely. Therefore,
it is the primary zone to improve H2 utilisation efficiency.
The H2 -rich reducing gas utilisation rate and volume requirement vary in different
ferric oxides reduction stages. The iron oxides reduction reactions are at a nonequilibrium
state in the BF. When the temperature is above 570 ◦ C, the reduction in ferric oxides by CO
and H2 in the BF occurs in the following sequences: 1/2 Fe2 O3 → 1/3 Fe3 O4 (Stage I) →
FeO (Stage II) → Fe (Stage III) [45]. The gas produced by the reduction in the latter stage is
the reducing gas for the previous stage. Heat is gradually transferred to solid materials
during the gas ascending. At the same time, part of reducing gas reacts with iron oxides and
converts into CO2 and H2 O, and finally forms top gas at around 150 to 250 ◦ C when leaving
the BF. There are 25% of the total oxygen elements removed during the reduction of Fe3 O4
291
Materials 2022, 15, 2008
into FeO, and the remaining 75% of oxygen elements were removed in reducing FeO to
Fe. Therefore, the reduction process from FeO to Fe is the key step. The required reducing
gas amount is n kmol for CO, and m kmol for H2 to produce 1 kmol iron. The value of n
and m is the excess coefficient. The reduction reactions and thermodynamic parameters
in Stage III for CO and H2 are expressed as in Equations (7)–(12), and Equations (13)–(17),
respectively [46]:
Stage III : FeO (s) + nCO (g) → Fe (s) + CO2 (g) + (n − 1)CO (g) (7)
ΔH03CO = −13,190 kJ/kmol

ΔG03CO = −22,800 − 24.26T, kJ/kmol (8)
ΔG03CO −22,800 − 24.26T
lnKIIICO = − =− (9)
RT RT
pCO2 ϕCO2 × P0 1
KIIICO = = = (10)
pCO ϕCO × P0 n−1
where K is the reaction equilibrium constant; ϕ denotes the fraction of gas component.
The minimal CO required for Stage III is described in Equation (11):
1
n = 1+ (11)
KIIICO
The utilisation efficiency of CO in Stage III, η3CO , is described in Equation (12):
ϕ3CO2 1 1 KIIICO 1
η3CO = = ϕ3CO = = = (12)
ϕ3CO + ϕ3CO2 ϕ3CO2 +1 1
K3CO +1 1 + KIIICO n
FeO (s) + mH2 (g) → Fe (s) + H2 O(g) + (m − 1)H2 (g) (13)

ΔH03H2 = 28,010 kJ/kmol
ΔG03H2 = 23,430 − 16.16T, kJ/kmol (14)
ΔG03H2 23,430 − 16.16T

lnKIIIH2 = − =− (15)
RT RT
The minimal H2 required for Stage III is calculated as in Equation (16):
1
m = 1+ (16)
KIIIH2
The utilisation efficiency of H2 in Stage III, η3H2 is described in Equation (17):
ϕ3H2 O KIIIH2 1
η3H2 = = = (17)
ϕ3H2 + ϕ3H2 O 1 + KIIIH2 m
According to the theoretical thermochemical calculations, 50% of H2 and CO partici-

pates in the water–gas shift reaction Equation (18) between 600 and 1400 ◦ C [47]. Therefore,
the heat consumed by the water–gas shift reaction at temperatures above 820 ◦ C is balanced
by the heat generated at the temperature below 820 ◦ C, as calculated by Equation (19).
However, H2 promotes iron ore reduction by CO via the water gas shift reaction when the
temperature is over 820 ◦ C [48]. The CO2 generated reacts with H2 to reform CO, which
participates in FeO reduction reaction again and improve the utilisation efficiency of CO.
Water–gas shift reaction H2 O(g) + CO (g) = H2 (g) + CO2 (g) (18)
ΔH0 = −41,325 kJ/kmol
292
Materials 2022, 15, 2008
ΔG0 = −33,447 + 30.56T, kJ/kmol (19)

The heat effect of FeO reduction by H2 and CO gas is calculated by Equation (20):
XCO η3CO ΔH03CO + XH2 η3H2 ΔH03H2

ΔH = (20)
XCO η3CO + XH2 η3H2
where Xi is the proportion of CO or H2 in the reducing gas entering the BF shaft.

The gas produced by the reduction in Stage III is the reducing gas for Stage II. The
reduction reactions and thermodynamic parameters in Stage II for CO and H2 are expressed
in Equations (21)–(28):

Stage II : 1
3 Fe3 O4 (s) + (n − 1)CO (g) + CO2 (g) → FeO (s) + 43 CO2 (g) + n − 43 CO (g)
(21)
ΔH02CO = 22,400 kJ/kmol
ΔG02CO = 35,380 − 40.16T, kJ/kmol (22)

1
3 Fe3 O4 (s) + (m − 1)H2 (g) + H2 O (g) → FeO (s) + 4
3 H2 O (g) + m − 4
3 H2 (g)
(23)
ΔH02H2 = 63,600 kJ/kmol
ΔG02H2 = 71,940 − 73.62T, kJ/kmol (24)
ΔG02CO 35,380 − 40.16T

lnKIICO = − =− (25)
RT RT
ΔG02H2 71,940 − 73.62T
lnKIIH2 = − =− (26)
RT RT
ϕCO2 × P0 4
KIICO = = 3
(27)
ϕCO × P 0
n− 4
3
4
KIIH2 = 3
(28)
m− 4
3
The minimal CO and H2 volume required for Stage II is calculated as shown in
Equations (29) and (30), respectively:
! "
4 1
n= 1+ (29)
3 KIICO
! "
4 1
m= 1+ (30)
3 KIIH2
The utilisation efficiency of CO and H2 in Stage II is calculated as shown in
ϕ2CO2 KIICO 4
η2CO = = = (31)
ϕ2CO + ϕ2CO2 1 + K2CO 3n
ϕ2H2 O KIIH2 4
η2H2 = = = (32)
ϕ2H2 + ϕ2H2 O 1 + K2H2 3m
In the first stage of iron ores reduction, the transformation of Fe2 O3 to Fe3 O4 is very
rapid due to the very high equilibrium constant of Fe2 O3 reduction above 600 K, as shown
in Equations (33) and (35):

Stage I : 1
2 Fe2 O3 (s) + n − 43 CO (g) + 43 CO2 (g) → 13 Fe3 O4 (s) + 32 CO2 (g) + n − 3
2 CO(g)
(33)
ΔH01CO = −43,280 kJ/kmol
293
Materials 2022, 15, 2008
ΔG01CO = −52,131 − 41.0T, kJ/kmol (34)

1
2 Fe2 O3 (s) + m − 4
3 H2 (g) + 43 H2 O (g) → 1
3 Fe3 O4 (s) + 2 H2 O(g) +
3
m− 3
2 H2 (g)
(35)
ΔH01H2 = −21,810 kJ/kmol
ΔG01H2 = −15,547 − 74.4T, kJ/kmol (36)

The gas produced by the reduction in Stage II provides the reducing gas for Stage I.
These reactions only require a low concentration of reducing gas to proceed. The minimal
CO and H2 volume required for Stage I is shown in Equations (37) and (38), respectively.
With utilisation efficiency close to 100%, Fe2 O3 reduction is an irreversible reaction.
! "
3 1
n= 1+ (37)
2 KICO
! "
3 1
m= 1+ (38)
2 KIH2
The utilisation efficiency of CO and H2 in Stage I is described by Equations (39) and (40),
respectively:
ϕ2CO2 3
η1CO = = (39)
ϕ2CO + ϕ2CO2 2n
ϕ2H2 O 3
η1H2 = = (40)
ϕ2H2 + ϕ2H2 O 2m
The overall gas utilisation efficiency for H2 -rich reducing gas in the BF is calculated as
in Equation (41) below:
ϕCO2 + ϕH2 O
η= = XCO ηCO + XH2 ηH2 (41)
ϕCO + ϕCO2 + ϕH2 + ϕH2 O
Assuming the water generated in the Fe2 O3 reduction is reacted with CO, in which H2
performs only as a catalyst of CO reduction of Fe2 O3 . The water in top gas is determined
by H2 utilisation efficiency in FeO and Fe3 O4 , which was calculated as in Equation (42):

VH2 O = ∑ VH2 η3H2 + ∑ VH2 1 − η3H2 η2H2 + ∑ VH2 1 − η3H2 1 − η2H2 η1H2 (42)
Since FeO reduction is the key step, the theoretical overall H2 utilisation efficiency was
calculated as shown in Equation (43). The highest theoretical H2 utilisation efficiency can
be obtained with the minimal H2 requirement value on the basis of the thermodynamic
requirement in Stage III, and this highest value is determined by temperature. Due to
thermal restrictions and excess H2 injected, the actual gas utilisation efficiency can only
approach this theoretical value. The actual thermodynamic utilisation efficiency of H2 is a
function of the amount of H2 introduced to the BF, as shown in Equation (43):

ηH2 = η3H2 + 1 − η3H2 η2H2
+ 1 − η3H 2 )(1 − η2H2 η1H2 (43)
=m 1
+ 1− m 1 4
3m + 1 − 1
m 1 − 4
3m 2m
3
As the FeO reduction is the key step in the indirect reduction process, the thermo-
dynamic requirement of gas entering the BF shaft to produce 1 tHM is calculated as in
Equations (44) and (45):
1000(Fe)HM (1 − Rd ) 22.4
Vbosh_shaft = × (44)
η3 × %(Reducing gas) 56
Rd = 1 − R i (45)
294
Materials 2022, 15, 2008
where Vbosh_shaft is the amount of gas raised from BF bosh after direct reduction and the
gas injected through the shaft tuyeres, m3 /tHM; [Fe]HM is the proportion of iron content in
hot metal; Rd is the degree of direct reduction.
2.2. Thermal Calculations of H2 -Rich Gas Injection BF

As the heat carrier, the H2 -rich gas injected through raceway tuyeres needs to compen-
sate for the required energy in the lower furnace and maintain the theoretical combustion
temperature at a reasonable range. The gas injected through the shaft also needs to satisfy
the heat requirement in the upper furnace. The energy of H2 -rich gas includes the oxidation
heat release from the iron ore reduction and sensible heat. The oxidation heat release
depends on gas utilisation efficiency and gas composition. The injection temperature
determines the sensible heat. Thermal calculations for determining the amount of H2 -rich
gas were developed by a static mass and energy model of the iron-making process.
The thermal balance for this iron-making process is developed in the lower and upper
furnaces, divided by the shaft gas injection position. In this work, the lower furnace
included BF raceway, dripping zone, and cohesive zone. The thermal balance of the lower
furnace is shown in Equation (46) below:
Qcc + Qblast + Qhearth + Qcoke + Qore

(46)
= HCO2 + HH2 Ode + HPCde + Qbosh + HdA + HdFe + HS + QHM + Qslag + Qloss_l
where the heat income in the lower furnace includes: Qcc = combustion heat of coke
and pulverised coal in front of tuyeres, kJ/tHM; Qblast = sensible heat of the hot blast,
kJ/tHM; Qhearth = sensible heat of H2 -rich gas injection to the hearth, kJ/tHM; Qcoke = heat
of the coke brings to the lower part of BF, kJ/tHM; Qore = sensible heat of the iron ores
into the lower part of the BF, kJ/tHM; and the heat expenditure includes: HCO2 = heat
consumption of solution loss reaction due to the possible CO2 in the hearth injection gas,
kJ/tHM; HH2 Od = heat consumption of water decomposition in front of tuyeres, kJ/tHM;
HPCde = heat consumption of pulverised coal decomposition in front of tuyeres, kJ/tHM;
Qbosh = heat brought to the shaft by bosh gas, kJ/tHM; HdA = heat consumption by direct
reduction of alloy element, kJ/tHM; HdFe = heat consumption by direct reduction of FeO,
kJ/tHM; HS = heat consumption by desulphurisation, kJ/tHM; QHM = sensible heat of hot
metal, kJ/tHM; Qslag = sensible heat of slag, kJ/tHM; and Qloss_l = heat loss in the lower
furnace, kJ/tHM.
With H2 -rich gas injection to BF hearth, the raceway adiabatic flame temperature
(RAFT) is calculated as in Equation (47):
Qcoke + Qblast + Qhearth + Qcc − QH2Ode − QPCde − HCO2

RAFT = raceway raceway raceway (47)
VH2 × CH2 + VCO × CCO + VN2 × CN2
raceway
where Qcoke is the heat brought to the raceway by coke, kJ/tHM; Vi is the gas volumes
of H2 , CO and N2 in the raceway, m3 /tHM, respectively.
The energy input and output of the BF shaft can be expressed as in Equations (48) and (49),
respectively:
Qshaft + Qbosh + Qind + Qore_shaft = Qtop + Qcoke + Qore + Qloss_s (48)
Qind = QFe2 O3 + QFe3 O4 + QFeO (49)

where Qshaft is the heat carried by H2 -rich gas injected into the shaft, kJ/tHM; Qind is the
heat generation by iron oxides reduction by H2 and CO, kJ/tHM; Qore_shaft is the sensible
heat carried by iron ores entering the BF top, kJ/tHM; Qtop , kJ/tHM is heat loss in the
shaft, kJ/tHM.
295
Materials 2022, 15, 2008
The heats provided by the reduction reactions from Fe2 O3 to Fe3 O4 , Fe3 O4 to FeO,
and FeO to Fe by CO and H2 are calculated as from Equations (50)–(52), respectively:
more ×
w(Fe2 O3 )ore
QFe2 O3 = 56 × 160
rico × ΔH01CO + riH2 × ΔH01H2 (50)
3
2 × more ×
w(Fe2 O3 )ore
QFe3 O4 = 56 × 160
rico × ΔH02CO + riH2 × ΔH02H2 (51)
3

2×more ×w(Fe2 O3 )ore m ×w(FeO)ore
QFeO = 56 × (1 − Rd ) + ore 72
160 (52)
rico × ΔH03CO + riH2 × ΔH03H2
where rico and riH2 are the degree of indirect reduction by CO and H2 , as shown in
VH2O
riH2 = × 56 (53)
22.4 × 1000(Fe)HM
riCO = Ri − riH2 (54)

3.1. Results of the Thermodynamic Model
At a medium oxygen enrichment rate (9%), the nitrogen content in the BF shaft is
calculated at around 35%. Figure 2a shows the heat effect results based on Equation (20).
The FeO reduction reaction transforms from an exothermic into an endothermic process
when H2 content increases to 25% around 900 ◦ C. A similar phenomenon shows up when
the H2 reaches 20% around 1000 ◦ C. With more H2 participating in the reduction at high
temperatures, it causes a severe negative effect on the thermal energy supply to the BF.
Therefore, the shaft gas injection temperature should not be too high to reduce its endother-
mic heat effect and require less preheating in the gas preheating device. The higher oxygen
enrichment (less N2 content) enables higher H2 content in the BF, as shown in Figure 2b. It
is suggested that with 20% N2 entering the BF shaft, the H2 content should be lower than
25% to avoid too much heat consumption by its reduction reaction.
According to Equations (11), (16), (29) and (30), the theoretical minimal H2 and CO
requirement with temperature are shown in Figure 3. From this figure, Stage III for iron
ores reduction requires more H2 when the temperature is over 625 ◦ C. Stage III requires
more CO over 650 ◦ C. In this study, since the gas injection temperature is kept above
820 ◦ C to ensure the high reducing capability of H2 , the thermodynamic key step for
iron ores reduction is Stage III, and the other stages are proceeded with excess reducing
gas. Compared with Fe3 O4 , FeO reduction requires more reducing gas to proceed, which
determines the minimal amount of gaseous mixture. Fe3 O4 and Fe2 O3 reductions are
carried out with excess reducing gas. In addition, the amount of H2 required for FeO
reduction at high temperatures is less than the amount of CO required in the reaction.
Thermodynamically, injecting H2 content to replace some amount of CO would reduce
the total amount of gas mixture and reduce the fuel requirement for the gas preheating.
However, CO content in the BF should be enough to meet its thermal condition.
According to Figure 4, gas utilisation efficiency for H2 and CO was similar in Stage II
when the temperature was above 820 ◦ C, but H2 utilisation efficiency was much higher than
that for CO in Stage III. Although the utilisation efficiency of CO decreases as temperature
increases, the FeO reduction reaction would still be promoted with increasing H2 content
due to the effect of the water–gas shift reaction.
On the basis of the calculation from Equation (41), gas utilisation efficiency at different
H2 contents in the reducing gas for FeO and Fe3 O4 reduction is shown in Figure 5.
296
Materials 2022, 15, 2008
Figure 2. Heat effect of FeO reduction in H2 –CO–N2 gas mixture at (a) 35% N2 and (b) 20% N2 . H2
content entering blast furnace shaft is indicated by different colours.
Figure 3. Variation in minimal gas consumption for iron ores reduction in pure H2 or CO with
temperature.
Figure 4. Variation in gas utilisation efficiency for iron ores reduction in pure H2 or CO with temperature.
297
Materials 2022, 15, 2008
Figure 5. Gas utilisation efficiency at different H2 /CO ratio with temperature in (a) FeO and
(b) Fe3 O4 reductions.
Figure 5a presents the gas utilisation efficiency of FeO reduction. Due to the endother-
mic reaction of the H2 reduction, utilisation efficiency of H2 -rich reducing gas at 1000 ◦ C
for the FeO reduction increased from 23% with no H2 to around 30% with 50% H2 by
the equal interval. When H2 content was less than 30% in reducing gas, gas utilisation
efficiency in FeO reduction decreased with the increase in temperature. In contrast, gas
utilisation efficiency rose with temperature when H2 content is more than 40% in reducing
gas. Hence, H2 content in the reducing gas should not be too low to hinder improvement
in the gas utilisation efficiency. As shown in Figure 5b, since Fe3 O4 reductions by CO
and H2 are endothermic, gas utilisation efficiency increases with temperature. The effect
of increases in H2 content for Fe3 O4 reduction is less than FeO reduction in terms of gas
utilisation efficiency. At 1000 ◦ C, gas utilisation efficiency increases by less than 4% when
H2 content increases from 0% to 50%. The shaft gas injection temperature should be higher
than 820 ◦ C to promote gas utilisation in FeO and Fe3 O4 reduction, especially focusing on
FeO reduction.
3.2. BF Simulation Conditions and Validation

Without gas injection, the model developed in this work can be used for a traditional
BF. The measured data collected from a 2500 m3 BF were used to validate this proposed
model. The comparison of the industrial data and the model predictions is summarised in
Table 2. The coke rate and top gas components were compared because they are essential
measurable parameters that indicate the overall performance of an iron-making BF. The
chemical composition of raw material data is shown in Tables 3–5. In general, results in this
simulation show a similar trend as in practice, and the model was capable of estimating
the overall iron-making process. The slight difference in top gas composition is because
industrial data contain a slight amount of O2 and CH4 in top gas, which does not exist in
this model.
Table 2. Quantitative validation of the model.
Parameter Prediction Industrial Top Gas Prediction Industrial

Coke rate, kg/tHM 386 386 CO, % 25.1 24.9
Blast, Nm3 /tHM 1060 1089 CO2 , % 21.1 20.0
Slag rate, kg/tHM 364 373 H2 , % 1.2 0.8
Burden input, kg/tHM 1676 1676 N2 , % 50.0 53
RAFT, ◦ C 2205 2195 Rd 0.46 -
298
Materials 2022, 15, 2008
Table 3. Chemical composition of raw material and dust, w/w%.
Composition Tfe FeO SiO2 CaO MgO TiO2 S Al2 O3

Sinter 1 55.85 9.59 5.25 10.35 2.19 0.17 0.03 2.5
Sinter 2 55.85 9.2 5.26 10.27 2.2 0.27 - 2.5
Pellet 61.92 1.46 5.31 1.45 - 1.58 - 0.98
Ti ore 40.27 - 9.07 2.05 - 10.78 - 1.69
Dust 39.80 2.58 4.05 2.04 0.83 0 - 1.31
Table 4. Chemical composition of coke and pulverised coal, w/w%.
Composition Fixed C H2 O FeO CaO SiO2 Al2 O3 MgO N O H S

Coke 86.06 4.50 1.6 0.39 4.49 3.69 0.35 0.30 0.21 0.33 1.80
PCI 71.15 0 0.03 1.98 8.4 7.93 0.18 0.34 3.16 2.10 0.30
Table 5. Chemical composition of hot metal, w/w%.
Composition Fe C Si Mn P S Ti
Hot metal 95.14 4.12 0.34 0.32 0.136 0.023 0.129
3.3. Effect of H2 Injection on Coke Consumption Rate

The effect of injecting H2 -rich gas to hearth and/or shaft on coke rate is shown in
Figure 6. The highest H2 injection rate is limited to 160 m3 /tHM to ensure a balanced energy
distribution in the BF shaft. Injecting H2 to the BF hearth shows less coke consumption
than injecting it to the shaft. This is because it provides more sensible heat to the lower
part of the furnace to compensate the heat supplied by coke combustion. The relationship
between H2 injection and coke consumption rate at 9% oxygen enrichment rate is given as
in Equation (55). Injecting 7.9~10.9 m3 /tHM H2 can reduce the coke consumption rate by
1 kg/tHM, depending on the injection position.
K = 351Xshaft + 341Xhearth − (0.0913Xshaft + 0.1275Xhearth ) × H2inj (55)
where Xshaft and Xhearth are the proportion of H2 -rich gas injected to the shaft and hearth,
respectively; and H2 is the total H2 injection volume to the BF, m3 /tHM.
Figure 6. (a) Coke consumption rate at different H2 injection volumes at different injection positions;
(b) effect of H2 injection to hearth on RAFT at 9% oxygen enrichment rate.
The effect of H2 injection on RAFT at a constant oxygen enrichment rate is shown in

Figure 6b. Injecting 20 m3 /tHM H2 can reduce RAFT by around 16 ◦ C. Increasing oxygen
299
Materials 2022, 15, 2008
enrichment can achieve thermal compensation to maintain a stable RAFT and reduce coke
consumption, as shown in Figure 7. Compared to the cases without thermal compensation,
further reduction in coke consumption is obtained at 309 kg/tHM with 12.4% oxygen
enrichment at 160 m3 /tHM H2 injection.
Figure 7. Coke consumption rate at different H2 injection volumes at different oxygen enrichment
rates for thermal compensation. RAFT was maintained at 2096 ◦ C in this test.
3.4. Effect on H2 Utilisation Efficiency

In the case of injecting H2 to hearth, bosh gas composition is shown in Figure 8a. As
the H2 injection rate increased from 0 to 160 m3 /tHM at a constant CO injection rate, CO
content in bosh gas drops due to less coke consumption. N2 content also decreased since
less hot blast is required for carbon combustion. There was no H2 O in the bosh gas because
H2 O that formed from the iron oxide reduction by H2 reacted with coke to generate H2
again at high temperature. The amount of top gas and its composition is described in
Figure 8b. With more H2 injection, the moisture content in the top gas increases slightly
from 2.72% to 3.12% when the H2 injection rate reaches from 0 to 100 m3 /tHM. This is
because H2 utilisation efficiency significantly dropped from 72.6% to 22.9%, as shown in
Figure 9a. CO content in top gas decreased because of less CO in the bosh gas and increased
CO utilisation efficiency. CO2 concentration was reduced by less than 2% because there
was a significant drop of CO in the bosh gas and increase in CO utilisation efficiency is very
limited. The H2 injection promotes CO utilisation with the presence of the water–gas shift
reaction. However, the overall effect of the water–gas shift reaction was very limited across
the whole BF. The comprehensive gas utilisation efficiency gradually decreased from 43.6%
to 39.0% when H2 injection reached 160 m3 /tHM.
H2 and CO utilisation efficiency after thermal compensation is shown in Figure 9b.
Results in this simulation reflect a similar trend as in the literature results [8,49]. Oxygen
enrichment increased, and the nitrogen composition in the blast decreased with H2 injection.
Hence, there was more increase in reducing gas concentration than that in Figure 9a. With
less N2 dilution and stronger indirect reduction, CO utilisation with thermal compensation
was enhanced by 5% with H2 injection from 0 to 160 m3 /tHM.
By H2 increasing in the BF, CO utilisation increased, and H2 utilisation efficiency
decreased. One reason is that water–gas shift reaction Equation (18) would tend to proceed
to the right-hand side, and both H2 and CO2 are generated in the upper furnace below
1000 ◦ C. With the regeneration of H2 , FeO reduction in Equation (7) can be considered to
be proceeding in two successive stages, as shown in Equations (13) and (18):
From Equation (13) : FeO (s) + H2 (g) → Fe (s) + H2 O(g)
From Equation (18) : H2 O(g) + CO (g) = H2 (g) + CO2 (g)
300
Materials 2022, 15, 2008
From Equation (7) : Overall : FeO (s) + CO (g) → Fe (s) + CO2 (g)
Figure 8. Variations of (a) bosh gas and (b) top gas composition with hydrogen injection. (H2 injection
at 0 m3 /tHM indicates the operating condition of a traditional BF with oxygen enrichment rate at 9%).
Figure 9. Variations in gas utilisation efficiency with hydrogen injection (a) before and (b) after
thermal compensation.
Due to the smaller size and high diffusivity of H2 , the reaction in Equation (13) has an
advantage over the reaction in Equation (7). Thus, FeO reduction by CO was promoted by
increased H2 content.
As depicted in Figure 10, with H2 injection and thermal compensation, the degree of
indirect reduction increased because H2 reduction replaced part of the direct reduction and
oxygen enrichment enhanced reducing gas atmosphere. Since the direct reduction was
a huge endothermic reaction process in the lower furnace, less direct reduction reduces
coke consumption. Compared to the higher H2 injection volume, injecting 0 to 80 m3 /tHM
H2 generates more effect on the degree of indirect reduction, from 0.107 to 0.125. Further,
the injection of more than 50 m3 /tHM H2 significantly increased the indirect reduction of
CO. Considering gas utilisation efficiency, H2 injection should be around 50–80 m3 /tHM to
maintain smooth BF operation and avoid too much excess H2 in top gas.
301
Materials 2022, 15, 2008
Figure 10. Variations in degree of indirect reduction by CO and H2 with H2 injection and thermal
compensation.
3.5. Effects on CO2 Emissions and Energy Consumption

In this work, the CO injected into the BF comes from the CO2 captured in top gas,
which reduces CO2 emission compared to a traditional BF process. The emission from CCU
can be negligible because it applies renewable electricity. The total emission of this iron-
making process includes uncaptured CO2 , flue gases from BFG combusted in the preheater
and the hot blast stoves, which is shown in Figure 11a. The main CO2 emission reduction
comes from the hot stoves. Compared with the traditional BF iron-making system that
lacks CCU and gas injection, CO2 emission dropped from 534 to around 272 m3 /tHM with
160 m3 /tHM H2 injection in this system. When increasing H2 content from 0 to 80 m3 /tHM,
CO2 emission only decreased by 20 m3 /tHM because more BFG was consumed to preheat
the injection gas. Additionally, injecting H2 -rich gas into the shaft showed more CO2
emission reduction capability than injecting H2 into a hearth or both tuyeres, as shown
in Figure 11b.
Figure 11. CO2 emissions of iron-making process with H2 injection (a) with different CO2 emission
sources for shaft injection case; (b) at different injection positions (injection temperature for shaft and
hearth was 900 and 1250 ◦ C, respectively).
302
Materials 2022, 15, 2008
The energy consumption of the process described in Figure 1 was calculated on the
basis of static mass and energy balance, using Equation (56). The standard coal coefficient
for each substance in a kilogram of coal equivalent per ton of hot metal (kgce/tHM) is used
in this analysis [50].
Enet = Ecoke + EPC + EBOFG + Eblast + Ecap + Econv + Ewater − Eexport − EO2 (56)
where Enet is the net energy consumption of the process; Ecoke is the energy input by coke;
EPC is the energy input by pulverised coal; EBOFG is the energy input by BOFG to the hot
blast stoves; Eblast is the energy carried by the blast to hot blast stoves; Ecap is the electricity
required for CO2 capture unit; Econv is the electricity required for CO2 electrolyser, as
calculated by Equation (2); Ewater is the energy carried by the make-up water required at
the humidifier for H2 generation in the electrolsyer; Eexport is the energy carried by the BFG
that is exported to other processes in the integrated steel mill; and EO2 is the energy carried
by the oxygen that is generated in the electrolyser and exported to other processes in the
integrated steel mill.
Energy consumption results are shown in Figure 12. In a traditional BF, carbon re-
sources from coke and pulverised coal injection provide the primary energy input, which
accounts for 80% in total. When hydrogen is injected into the BF from 0 to 160 m3 /tHM,
carbon only accounts for 65% to 50% of the total energy consumption. With H2 -rich gas injec-
tion from 0 to 160 m3 /tHM, net energy consumption increased from 541 to 698 kgce/tHM,
which is higher than that of the traditional BF (504 kgce/tHM) because electricity required
for CO2 conversion to generate H2 kept increasing from 158 to 369 kgce/tHM.
Figure 12. Energy consumption for H2 -rich gas injection BF and traditional BF process (energy carried
by the make-up water needed for the CO2 conversion and by BOFG was less than 1% of total energy
consumption and thus not indicated in this figure).
In summary, increasing hydrogen injection volume can reduce coke consumption and
CO2 emissions. For coke consumption and CO2 emission reduction, the hydrogen injection
amount should be as much as possible, as long as it satisfies the energy balance in the BF.
In this case, the hydrogen injection amount should be 160 m3 /tHM. However, injecting
too much H2 significantly reduces its utilisation efficiency and increases the net energy
consumption of this process. Further study is recommended to develop a multiobjective
optimisation model to balance these effects of hydrogen injection. In general, injecting H2
at around 80 m3 /tHM may consume less energy and suppress CO2 emissions under the
simulation conditions.
Energy consumption to produce 1 tonne of hot metal in the case of injecting 80 m3 /tHM
H2 with 200 m3 /tHM CO is indicated in Figure 13. Compared to the traditional BF as
303
Materials 2022, 15, 2008
indicated by Table 2, coke consumption decreased by 43 kg/tHM. CO2 emission dropped

from 534 m3 /tHM for a traditional BF to 278 m3 /tHM in this case (including gas heating
flue gas, capture unit CO2 letdown, and stack flue gas). However, electricity consumption
in the CO2 capture unit and electrolyser is one of the largest energy inputs. The economic
impact of this CCU technology as an auxiliary system to the BF is highly recommended for
future investigation.
Figure 13. Energy and material consumption of the process with 80 m3 /tHM hydrogen injection.
Relative flow size estimated by energy and demonstrated as bar width.
4. Conclusions
In this study, a thermodynamic model was used to determine the utilisation efficiency
of H2 -rich gas by considering H2 behaviour at different reduction stages. A static mass and
energy balance model of the BF was adopted with this thermodynamic model. The effect of
injecting H2 -rich gas on BF performance was determined in terms of its coke consumption,
CO2 emissions, and energy consumption. Under these simulation conditions, the major
findings of this study were:
1. The desired shaft gas injection temperature should not exceed 1000 ◦ C to suppress
the endothermic FeO reduction reaction by H2 -rich gas.
2. Injecting H2 to BF hearth has a better effect on coke rate reduction than that of injection
to the shaft. The lowest H2 consumption to save 1 kg of coke was estimated to be
7.9 m3 /tHM.
3. H2 utilisation efficiency dropped significantly with increasing H2 content, and the
increase in CO utilisation efficiency was limited. Further research should focus on
improving H2 utilisation efficiency with a high H2 injection rate.
4. Considering H2 utilisation efficiency and the degree of indirect reduction by H2 and
CO, the proper H2 injection rate should be from around 50 to 80 m3 /tHM.
5. Introducing H2 -rich gas injection can reduce CO2 emissions of the iron-making process
by up to 262 m3 /tHM compared with a traditional BF. However, injecting too much
H2 would hinder CO2 emission reduction due to its requirement of preheating outside
the BF.
6. The energy consumption of this proposed process was higher than that of the tradi-
tional BF. Although coke consumption was reduced by 43 kg/tHM more than that
of the traditional BF, net energy consumption increased with the amount of injected
304
Materials 2022, 15, 2008
hydrogen due to the high electricity consumption in the CO2 capture and electrolyser.
Developing a CO2 conversion unit with higher efficiency but less energy consumption
is strongly recommended.
Author Contributions: Conceptualization, Y.H., T.E.R., V.R. and G.W.; methodology, Y.H. and G.W.;
software, Y.H.; validation, Y.H. and L.H.; formal analysis, Y.H.; investigation, Y.H., Y.Q., J.C. and L.H.
resources, L.H.; data curation, Y.H.; writing—original draft preparation, Y.H.; writing—review and
editing, Y.H., Y.Q., J.C., L.H., T.E.R. and G.W.; visualization, Y.H.; supervision, T.E.R., V.R. and G.W.;
project administration, G.W. and L.H.; funding acquisition, L.H. All authors have read and agreed to
the published version of the manuscript.
Funding: This research was funded by HBIS (China) grant number ICSS2017-04.
Data Availability Statement: Data sharing not applicable. No new data were created or analysed in
Acknowledgments: The authors gratefully acknowledge HBIS Group Company Limited for funding
support and the scholarship support from the University of Queensland.
Nomenclature
Abbreviations
BF Blast furnace
PCI Pulverised coal injection
CCU CO2 capture and conversion unit
RAFT Raceway adiabatic flame temperature
tHM Tonne of hot metal
Roman and Greek symbols
E Energy, kWh/tHM or kgce/tHM
V Volume of gas, m3 /tHM
T Temperature, ◦ C
Ci Specific heat capacity, kJ/m3 ·◦ C
Ki Reaction equilibrium constant of reduction stage i; i = I, II, III
G Gibbs free energy, kJ/mol
ϕ Volume fraction of gas component
ηiCO Utilisation efficiency of CO in stage i, i = 1, 2, 3
ηiH2 Utilisation efficiency of H2 in stage i, i = 1, 2, 3
n Minimal CO required for iron ores reduction, mol
m Minimal H2 required for iron ores reduction, mol
Xi Volume reaction of CO or H2 in reducing gas entering the BF shaft
η Overall gas utilisation efficiency for H2 -rich reducing gas in the BF
%(Reducing gas) Proportion of H2 and CO in the total amount of gas entering the bosh and shaft.
Q Sensible heat of material, kJ/tHM
H Enthalpy of material, kJ/tHM
Rd Degree of direct reduction
ri Degree of indirect reduction
(Fe)HM Iron content in hot metal
ΔH0 Enthalpy of reaction, kJ/tHM
w Weight fraction of solid material
H2inj Total H2 injection volume to the BF, m3 /tHM
305
Materials 2022, 15, 2008
References
1. Xu, K. Low carbon economy and iron and steel industry. Iron Steel 2010, 45, 1–12.
2. Pardo, N.; Moya, J.A. Prospective scenarios on energy efficiency and CO2 emissions in the European Iron & Steel industry. Energy
2013, 54, 113–128. [CrossRef]
3. Babich, A.; Senk, D. New Trends in Coal Conversion: Combustion, Gasification, Emissions, and Coking; Woodhead Publishing: Sawson,
UK, 2019; pp. 367–404.
4. Danloy, G.; Berthelemot, A.; Grant, M.; Borlée, J.; Sert, D.; Van Der Stel, J.; Jak, H.; Dimastromatteo, V.; Hallin, M.; Eklund, N.;
et al. ULCOS-Pilot testing of the Low-CO2 Blast Furnace process at the experimental BF in Luleå. Rev. Métall.-Int. J. Metall. 2009,
106, 1–8. [CrossRef]
5. Knop, K.; Hallin, M.; Burstrom, E. ULCORED SP 12 Concept for minimized CO2 emission. Metall. Res. Technol. 2009, 106, 419–421.
[CrossRef]
6. Material Economics. Industrial Transformation 2050: Pathways to Net-Zero Emisisons from EU Heavy Industry. 2019. Available
online: https://materialeconomics.com/publications/industrial-transformation-2050 (accessed on 8 January 2022).
7. Chisalita, D.-A.; Petrescu, L.; Cobden, P.; van Dijk, H.E.; Cormos, A.-M.; Cormos, C.-C. Assessing the environmental impact of
an integrated steel mill with post-combustion CO2 capture and storage using the LCA methodology. J. Clean. Prod. 2019, 211,
1015–1025. [CrossRef]
8. Nogami, H.; Kashiwaya, Y.; Yamada, D. Simulation of Blast Furnace Operation with Intensive Hydrogen Injection. ISIJ Int. 2012,
52, 1523–1527. [CrossRef]
9. Usui, T.; Kawabata, H.; Ono-Nakazato, H.; Kurosaka, A. Fundamental Experiments on the H2 Gas Injection into the Lower Part
of a Blast Furnace Shaft. ISIJ Int. 2002, 42, S14–S18. [CrossRef]
10. Pinegar, H.K.; Moats, M.S.; Sohn, H.Y. Process Simulation and Economic Feasibility Analysis for a Hydrogen-Based Novel
Suspension Ironmaking Technology. Steel Res. Int. 2011, 82, 951–963. [CrossRef]
11. Wagner, D.; Devisme, O.; Patisson, F.; Ablitzer, D. A laboratory study of the reduction of iron oxides by hydrogen. arXiv 2008,
arXiv:08032831.
12. Spreitzer, D.; Schenk, J. Reduction of Iron Oxides with Hydrogen—A Review. Steel Res. Int. 2019, 90, 1900108. [CrossRef]
13. Qie, Y.; Lyu, Q.; Li, J.; Lan, C.; Liu, X. Effect of Hydrogen Addition on Reduction Kinetics of Iron Oxides in Gas-injection BF. ISIJ
Int. 2017, 57, 404–412. [CrossRef]
14. Zuo, H.-b.; Wang, C.; Dong, J.-j.; Jiao, K.-x.; Xu, R.-s. Reduction kinetics of iron oxide pellets with H2 and CO mixtures. Int. J.
Miner. Metall. Mater. 2015, 22, 688–696. [CrossRef]
15. Chu, M.; Nogami, H.; Yagi, J.-I. Numerical Analysis on Injection of Hydrogen Bearing Materials into Blast Furnace. ISIJ Int. 2004,
44, 801–808. [CrossRef]
16. Yilmaz, C.; Wendelstorf, J.; Turek, T. Modeling and simulation of hydrogen injection into a blast furnace to reduce carbon dioxide
emissions. J. Clean. Prod. 2017, 154, 488–501. [CrossRef]
17. Lyu, Q.; Qie, Y.; Liu, X.; Lan, C.; Li, J.; Liu, S. Effect of hydrogen addition on reduction behavior of iron oxides in gas-injection
blast furnace. Thermochim. Acta 2017, 648, 79–90. [CrossRef]
18. Watakabe, S.; Miyagawa, K.; Matsuzaki, S.; Inada, T.; Tomita, Y.; Saito, K.; Osame, M.; Sikström, P.; Ökvist, L.S.; Wikstrom,
J.-O. Operation Trial of Hydrogenous Gas Injection of COURSE50 Project at an Experimental Blast Furnace. ISIJ Int. 2013, 53,
2065–2071. [CrossRef]
19. Qie, Y.; Lyu, Q.; Liu, X.; Li, J.; Lan, C.; Zhang, S.; Yan, C. Effect of Hydrogen Addition on Softening and Melting Reduction
Behaviors of Ferrous Burden in Gas-Injection Blast Furnace. Met. Mater. Trans. A 2018, 49, 2622–2632. [CrossRef]
20. Wang, Y.; He, Z.-J.; Zhang, W.-l.; Zhu, H.-b.; Zhang, J.-h.; Pang, Q.-h. Reduction behavior of iron-bearing burdens in hydrogen-rich
stream. Iron Steel 2020, 55, 34–40.
21. Andersson, J.; Grönkvist, S. A comparison of two hydrogen storages in a fossil-free direct reduced iron process. Int. J. Hydrog.
Energy 2021, 46, 28657–28674. [CrossRef]
22. Hydrogen Council. Path to Hydrogen Competitiveness: A Cost Perspective. 2020. Available online: https://hydrogencouncil.
com/en/path-to-hydrogen-competitiveness-a-cost-perspective/ (accessed on 8 January 2022).
23. Uribe-Soto, W.; Portha, J.-F.; Commenge, J.-M.; Falk, L. A review of thermochemical processes and technologies to use steelworks
off-gases. Renew. Sustain. Energy Rev. 2017, 74, 809–823. [CrossRef]
24. Da Rocha, E.P.; Guilherme, V.S.; de Castro, J.A.; Sazaki, Y.; Yagi, J.-I. Analysis of synthetic natural gas injection into charcoal blast
furnace. J. Mater. Res. Technol. 2013, 2, 255–262. [CrossRef]
25. Halim, K.S.A. Theoretical Approach to Change Blast Furnace Regime with Natural Gas Injection. J. Iron Steel Res. Int. 2013, 20,
40–46. [CrossRef]
26. Trinkel, V.; Kieberger, N.; Bürgler, T.; Rechberger, H.; Fellner, J. Influence of waste plastic utilisation in blast furnace on heavy
metal emissions. J. Clean. Prod. 2015, 94, 312–320. [CrossRef]
27. Domingo-Tafalla, B.; Martínez-Ferrero, E.; Franco, F.; Palomares-Gil, E. Applications of Carbon Dots for the Photocatalytic and
Electrocatalytic Reduction of CO2 . Molecules 2022, 27, 1081. [CrossRef]
28. Zhong, H.; Sa, R.; Lv, H.; Yang, S.; Yuan, D.; Wang, X.; Wang, R. Covalent organic framework hosting metalloporphyrin-based
carbon dots for visible-light-driven selective CO2 reduction. Adv. Funct. Mater. 2020, 30, 2002654. [CrossRef]
306
Materials 2022, 15, 2008
29. Küngas, R. Review—Electrochemical CO2 Reduction for CO Production: Comparison of Low- and High-Temperature Electrolysis
Technologies. J. Electrochem. Soc. 2020, 167, 044508. [CrossRef]
30. Parvanian, A.M.; Sadeghi, N.; Rafiee, A.; Shearer, C.J.; Jafarian, M. Application of porous materials for CO2 reutilization: A
Review. Energies 2022, 15, 63. [CrossRef]
31. Delacourt, C.; Ridgway, P.; Kerr, J.; Newman, J. Design of an Electrochemical Cell Making Syngas (CO + H2 ) from CO2 and H2 O
reduction at Room Temperature. J. Electrochem. Soc. 2007, 155, B42. [CrossRef]
32. Lu, Q.; Jiao, F. Electrochemical CO2 reduction: Electrocatalyst, reaction mechanism, and process engineering. Nano Energy 2016,
29, 439–456. [CrossRef]
33. Garg, S.; Li, M.; Idros, M.N.; Wu, Y.; Wang, G.; Rufford, T.E. Is Maximising Current Density Always the Optimum Strategy in
Electrolyser Design for Electrochemical CO2 Conversion to Chemicals? 2021. Available online: https://chemrxiv.org/engage/
chemrxiv/article-details/61a7235f7848050092a110b4 (accessed on 8 January 2022).
34. De Ras, K.; Van de Vijver, R.; Galvita, V.V.; Marin, G.B.; Van Geem, K.M. Carbon capture and utilization in the steel industry:
Challenges and opportunities for chemical engineering. Curr. Opin. Chem. Eng. 2019, 26, 81–87. [CrossRef]
35. De Castro, J.A.; Takano, C.; Yagi, J.-i. A theoretical study using the multiphase numerical simulation technique for effective use of
H2 as blast furnaces fuel. J. Mater. Res. Technol. 2017, 6, 258–270. [CrossRef]
36. Zhang, W. Fundamental Study and Process Optimization of the Oxygen Blast Furnace Ironmaking. Doctoral Dissertation,
Northeastern University, Shenyang, China, 2015.
37. Kwak, N.-S.; Lee, J.H.; Lee, I.Y.; Jang, K.R.; Shim, J.-G. A study of the CO2 capture pilot plant by amine absorption. Energy 2012,
47, 41–46. [CrossRef]
38. Bui, M.; Adjiman, C.S.; Bardow, A.; Anthony, E.J.; Boston, A.; Brown, S.; Fennell, P.S.; Fuss, S.; Galindo, A.; Hackett, L.A.; et al.
Carbon capture and storage (CCS): The way forward. Energy Environ. Sci. 2018, 11, 1062–1176. [CrossRef]
39. Choi, J.; Cho, H.; Yun, S.; Jang, M.-G.; Oh, S.-Y.; Binns, M.; Kim, J.-K. Process design and optimization of MEA-based CO2 capture
processes for non-power industries. Energy 2019, 185, 971–980. [CrossRef]
40. Hari, S.R.; Balaji, C.P.; Karunamurthy, K. Carbon Capture and Storage Using Renewable Energy Sources: A Review. IOP Conf. Ser.
Earth Environ. Sci. 2020, 573, 12004. [CrossRef]
41. Aresta, M.; Dibenedetto, A.; Angelini, A. The use of solar energy can enhance the conversion of carbon dioxide into energy-rich
products: Stepping towards artificial photosynthesis. Philos. Trans. R. Soc. London. Ser. A Math. Phys. Eng. Sci. 2013, 371, 20120111.
[CrossRef]
42. McBrien, M.; Serrenho, A.C.; Allwood, J. Potential for energy savings by heat recovery in an integrated steel supply chain. Appl.
Therm. Eng. 2016, 103, 592–606. [CrossRef]
43. Mazanek, E.; Jasienska, S.; Brachucy, A.; Bryk, C. The influence of hydrogen in the gas mixture of hydrogen and CO on the
dynamics of the reduction process. Metal. Odlew. 1982, 8, 53–70.
44. Piotrowski, K.; Mondal, K.; Lorethova, H.; Stonawski, L.; Szymanski, T.; Wiltowski, T. Effect of gas composition on the kinetics of
iron oxide reduction in a hydrogen production process. Int. J. Hydrog. Energy 2005, 30, 1543–1554. [CrossRef]
45. Jozwiak, W.; Kaczmarek, E.; Maniecki, T.; Ignaczak, W.; Maniukiewicz, W. Reduction behavior of iron oxides in hydrogen and
carbon monoxide atmospheres. Appl. Catal. A Gen. 2007, 326, 17–27. [CrossRef]
46. Wu, S.; Wang, X.; Zhang, J. Ferrous Metallurgy; Metallurgical Industry Press: Beijing, China, 2019; pp. 46–48.
47. Bernasowski, M. Theoretical Study of the Hydrogen Influence on Iron Oxides Reduction at the Blast Furnace Process. Steel Res.
Int. 2014, 85, 670–678. [CrossRef]
48. Li, J.; Wang, P.; Zhou, L.; Cheng, M. The Reduction of Wustite with High Oxygen Enrichment and High Injection of Hydrogenous
Fuel. ISIJ Int. 2007, 47, 1097–1101. [CrossRef]
49. Tang, J.; Chu, M.; Li, F.; Zhang, Z.; Tang, Y.; Liu, Z.; Yagi, J. Mathematical simulation and life cycle assessment of blast furnace
operation with hydrogen injection under constant pulverized coal injection. J. Clean. Prod. 2021, 278, 123191. [CrossRef]
50. China GB/T 2589-2020; General Principles for Calculation of the Comprehensive Energy Consumption. Standardization Adminis-
tration of China: Bijing, China, 2020.
307
MDPI
St. Alban-Anlage 66
4052 Basel
Switzerland
Tel. +41 61 683 77 34
Fax +41 61 302 89 18
www.mdpi.com
Materials Editorial Office

E-mail: [email protected]
www.mdpi.com/journal/materials
MDPI
St. Alban-Anlage 66
4052 Basel
Switzerland
Tel: +41 61 683 77 34
www.mdpi.com ISBN 978-3-0365-6431-9

Metallurgical Process Simulation and Optimization: Qing Liu and Jiangshan Zhang

Uploaded by

Copyright:

Available Formats

Metallurgical Process Simulation and Optimization: Qing Liu and Jiangshan Zhang

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Metallurgical Process Simulation and Optimization: Qing Liu and Jiangshan Zhang

Uploaded by

Copyright:

Available Formats

Metallurgical

ISBN 978-3-0365-6430-2 (Hbk)

About the Editors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii

Preface to ”Metallurgical Process Simulation and Optimization” . . . . . . . . . . . . . . . . . ix

Jiangshan Zhang, Yuhong Liu and Qing Liu

ZhanpengTie, HaiyanTang, KaiminWang, HongshengMiao, SenCai, FenqiangXian

Peng Zhao, Rongxun Piao and Zongshu Zou

Michał Moskal, Piotr Migas and Mirosław Karbowniczek

Shaopei Duan, Baokuan Li and Wenjie Rong

Xiaodong Deng, Jianli Li and Xiao Xie

Lei Liu, Xiuli Han, Mingduo Li and Di Zhang

Paweł Malinowski, Justyna Kasińska, Sławomir Rutkowski and Monika Madej

Qing Liu and Jiangshan Zhang

Materials 2022, 15, 8421. https://doi.org/10.3390/ma15238421 1 https://www.mdpi.com/journal/materials

principles, accompanied by SEM, TEM, and XRD characterization methods. An optimized

Academic Editors: Qing Liu and

Materials 2022, 15, 7458. https://doi.org/10.3390/ma15217458 5 https://www.mdpi.com/journal/materials

1.2. Inclusion Motion Characteristics

1.3. Background of the Research

Table 1. Chemical composition of selected steel (SPFH590) (wt%).

Table 2. Thermo-physical properties of SPFH590 steel.

Table 3. Process Parameters.

Process Parameters Values

The Governing Equations

£ = 0 i f Temperature ( To ) < Solidus Temperature( Ts)

£ = 1 i f Temperature ( To ) > Solidus Temperature( Ts)

Figure 2. (a) Schematic of the domain and (b) domain meshing.

Figure 3. Grid independence test.

4. Results and Discussion

In Figure 9, the inclusion concentration proﬁle is shown in the whole domain of

Figure 10. Alumina inclusion motion in the mold.

Abstract: Cracking con-rod is an advanced high-precision connecting structure based on brittle

Materials 2022, 15, 3698. https://doi.org/10.3390/ma15103698 29 https://www.mdpi.com/journal/materials

2. Production Process and Basic Parameters of C70S6 Steel

Table 1. Chemical composition of steel C70S6, wt%.

Table 2. Process parameters for the continuous casting of C70S6 steel.

Frm = Frp (1)

where m denotes the tundish model, and p is the prototype.

Lm /Lp = λ, Vm /Vp = λ3 , um /up = λ0.5 , Qm /Qp = λ2.5 (3)

where V is tundish volume and Q is the volumetric ﬂowrate of molten steel.

Table 3. Calculation formula of ﬂow characteristic parameters of molten steel in tundish.

Parameters Symbol and Unit Formula

Volume fraction of dead zone θd , % 1 − tav /tf

3.2. Mathematical Model

Table 4. Governing equations of the mathematical model.

Governing Equations Symbols

1. Continuity equation ρ is density of molten steel, ui

3.3. Assumptions and Boundary Conditions

3.3.2. Boundary Conditions and Solution Methods

Table 5. Boundary conditions and material parameters.

Boundary Condition and

4. Experimental Results and Discussion

Table 6. Flow characteristic parameters of the two experimental schemes.

Case Outlet No. tmin /s tpeak /s tav /s θd /% θp /% θm /% Stmin /s Stav /s

Figure 3. Comparison of RTD curves between cases A and B.

Figure 4. Ink trajectory comparison of cases A (a) and B (b).

4.2. Validation of CFD Modelling and Mesh Independence Test

Table 7. Mesh independence study and model validation.

4.3. Results of Numerical Simulation