Neu m0411x42f

Tuning Johnson-Cook Material Model Parameters for Impact of High
Velocity, Micron Scale Aluminum Particles
A Thesis Presented
by
Arash Alizadeh Dehkharghani
to
The Department of Mechanical and Industrial Engineering
in partial fulfillment of the requirements

for the degree of
Master of Science
in
Mechanical Engineering
Northeastern University
Boston, Massachusetts
August 2016
NORTHEASTERN UNIVERSITY
Graduate School of Engineering
Thesis Signature Page
Thesis Title: Tuning Johnson-Cook Material Model Parameters for Impact of

High Velocity, Micron Scale Aluminum Particles
Author: Arash Alizadeh Dehkharghani NUID: 001102511
Department: Mechanical and Industrial Engineering
Approved for Thesis Requirements of the Master of Science Degree
Thesis Advisor
Dr. Sinan Müftü
Signature Date
Department Chair
Dr. Hanchen Huang
Signature Date
Associate Dean of Graduate School:

Dr. Sara Wadia-Fascetti
Signature Date
To my family
ii
Contents
List of Figures v
List of Tables vii
Acknowledgments viii
Abstract of the Thesis ix
1 Introduction 1
1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 Objective and Problem Description . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 Arrangement of the Thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2 Literature Review 5
2.1 Cold Spray . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2 Particle Impact . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.3 High Strain Rate Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.3.1 Split-Hopkinson Pressure Bar Test . . . . . . . . . . . . . . . . . . . . . . 10
2.3.2 Dynamic tests in tensile and torsional loadings . . . . . . . . . . . . . . . 12
2.4 Ultra High Rate of Loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3 Modeling and Finite Element Simulation of High Velocity Impact for a Single Micron-
Scale Particle 16
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
3.2 FEM Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.1 Modeling and analysis technique . . . . . . . . . . . . . . . . . . . . . . . 17
3.2.2 Element Type and Failure Criterion . . . . . . . . . . . . . . . . . . . . . 18
3.2.3 Interaction and Contact Properties . . . . . . . . . . . . . . . . . . . . . . 20
3.2.4 Boundary Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.2.5 Material Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.6 Adiabatic Shear Instability . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.2.7 Effect of Using Temperature Dependent Material Properties . . . . . . . . 24
3.3 Mesh Convergence Study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.4 Laser Induced Single Particle Impact Experiments . . . . . . . . . . . . . . . . . . 28
iii
3.5 Summary and Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
4 Material Model for High Strain Rate Deformation of Metals 32

4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
4.2.1 Johnson-Cook Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
4.2.2 Zerilli-Armstrong Model . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
4.2.3 Voyiadjis-Abed Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.4 Preston-Tonk-Wallace Model . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.5 Khan-Huang-Liang Model . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.6 Gao-Zhang Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.2.7 Summary of All Constitutive Models . . . . . . . . . . . . . . . . . . . . 38
4.3 Bilinear Johnson-Cook Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
4.4 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
5 Optimization of Bilinear Johnson-Cook parameters 41

5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
5.2 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
5.3 Optimization Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
5.4 Optimization Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
5.5 Estimated Particle Diameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
5.6 Verification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
5.7 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
6 Conclusion 58
6.1 Summary and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
Bibliography 61
A Simulation Results Using the Classic and Bilinear Johnson-Cook Material Models 70
A.1 Simulation Results Using the Classic JC Model . . . . . . . . . . . . . . . . . . . 70
A.2 Simulation Results Using the Bilinear JC Model . . . . . . . . . . . . . . . . . . . 75
B Table of Temperature Dependent Material Properties for Aluminum 6061 79
iv
List of Figures
1.1 Comparison of gas temperature and particle velocities in thermal spray processes and
cold spray process. VPS is vacuum plasma spray and LPPS is low pressure plasma
spray [1]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Schematic diagram of a cold spray gun [1]. . . . . . . . . . . . . . . . . . . . . . 2
2.1 Schematic plot of correlation between particle velocity and deposition efficiency [2]. 6
2.2 Particle impact on a solid surface: effect of impact velocity and particle size on
features of the interaction. Regions characteristic of certain impact phenomena are
shown [3]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.3 Strain rate regimes and associated instrument and experimental conditions [4]. . . . 9
2.4 Schematic view of split-hopkinson pressure bar [4]. . . . . . . . . . . . . . . . . . 10
2.5 A compression split-Hopkinson pressure bar facility at Los Alamos National Lab
(Figure from [4]). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.6 Experimental method for high strain rate testing [4]. . . . . . . . . . . . . . . . . . 12
2.7 Schematics of dynamic tension loading [5]. . . . . . . . . . . . . . . . . . . . . . 13
2.8 (a) Schematics of Split Hopkinson Pressure Bar for dynamic punch shear tests, and
(b) sample holder in SHPB [6]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
3.1 Isometric view of the developed 3D model. . . . . . . . . . . . . . . . . . . . . . 17

3.2 Shear stress versus nominal shear strain for a typical work-hardening material during
a torsion experiment [7]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.3 rebound velocity as a function of particle impact velocity for different failure strains.
This analysis conducted by using Copper as a material properties for both particle
and substrate [8]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4 Traction separation response of interacting surfaces with a linear failure mechanism. 21
3.5 Temperature dependent material properties for Aluminum 6061 [9]. Properties
include a) Elastic Modulus; b) Poisson’s Ratio; c) Thermal Expansion; d) Thermal
Conductivity; and e) Specific Heat. . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.6 Deformed shapes of particle using bilinear JC material model for 19.75 µm particle
and impact velocity of 663 m/s. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
v
3.7 Time history analysis of particle for a) average strain; b) average strain rate; c)
average temperature; d) average yield stress. Note that the particle diameter was
19.75 µm and impact velocity was 663 m/s. Average values are the mean of all
integration point at each increment time. . . . . . . . . . . . . . . . . . . . . . . 27
3.8 Diagram of the optical setup for LIPIT [10]. . . . . . . . . . . . . . . . . . . . . . 29
3.9 particle velocity and excitation pulse energy relation [10]. . . . . . . . . . . . . . . 30
3.10 multiple exposure photograph of a particle with constant ∆t [11]. . . . . . . . . . 30
4.1 Comparison between the stress-strain rate behavior predicted by the mechanism-
based material model and experimental data for aluminum alloy 6061-T6 [12]. . . . 39
4.2 flow stress in classic JC (left) and Bilinear JC (right) . . . . . . . . . . . . . . . . 40
5.1 Stress-Strain rate behavior in a semi log plot [13]. . . . . . . . . . . . . . . . . . . 43

5.2 SEM images on which the parameter fitting is based [11] and the definition of the
dimensions D1 and D2 used in calculating the ellipticity ratio Re . . . . . . . . . . 45
5.3 Optimization process with modified steepest descent method. . . . . . . . . . . . . 46
5.4 Comparison of the deformed particle by using the initial JC parameters. . . . . . . 49
5.5 Comparison of the deformed particles by using the optimized JC parameters. . . . 50
5.6 Time history of the particle’s volume in case-4. Note that the small increase in
volume is due to thermal expansion. . . . . . . . . . . . . . . . . . . . . . . . . . 52
5.7 Comparison of the experimental particle contours with those that are computed by
using the adjusted particle diameter DSEM and the modified JC parameters reported
in Table 5.10. Note that the JC parameters for each impact speed were different as
reported in Table 5.10. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
5.8 Comparison of the experimental particle contours with those that are computed by
using the adjusted particle diameter DSEM and the average value of the modified JC
parameters reported in Table 5.10. . . . . . . . . . . . . . . . . . . . . . . . . . . 56
5.9 Comparison of the experimental coefficient of restitution with simulated results.
Lines represents the simulation data and dots are the experiments. . . . . . . . . . 57
A.1 Comparison of the deformed shape of particle between experiment (a, b) and simula-
tions (c, d, e, f) where the impact velocity. is 556 m/s . . . . . . . . . . . . . . . . 72
A.4 Strain rate contour of iso-surfaces . . . . . . . . . . . . . . . . . . . . . . . . . . 75
tions (c, d) where the impact velocity. is 556 m/s . . . . . . . . . . . . . . . . . . 76
vi
List of Tables
3.1 Temperature dependent material properties of Sapphire [14]. . . . . . . . . . . . . 24

3.2 material properties for Aluminum 6061 at room temperature. . . . . . . . . . . . . 25
3.3 Mesh convergence study for single particle impact simulation. . . . . . . . . . . . 28
4.1 Bilinear Johnson-Cook material properties for Al-6061 . . . . . . . . . . . . . . . 39
5.1 Material properties of Al-6061 [15, 9]. . . . . . . . . . . . . . . . . . . . . . . . . 42

5.2 Temperature dependent material properties of Sapphire [14]. . . . . . . . . . . . . 42
5.3 Bilinear Johnson-Cook material properties for Al-6061 obtained by SHPB test. . . 43
5.4 Bilinear Johnson-Cook parameters at each attempt of the optimization process for
the case-1, where Dp = 20.75 µm, and Vp = 175 m/s. . . . . . . . . . . . . . . . . 47
the case-2, where Dp = 24.4 µm, and Vp = 286 m/s. . . . . . . . . . . . . . . . . . 47
the case-5, where Dp = 19.75 µm, and Vp = 663 m/s. . . . . . . . . . . . . . . . . 48
5.10 Summary of the optimized values of the parameters for the bilinear JC equation . . 51
5.11 Diameter and volume estimates of the particles. Dr is the reported particle diameter
[11]. Vr is the particle volume based on Dr . VA is the particle volume computed by
Abaqus based on Dr after meshing. VSEM is the particle volume estimated based on
the post-impact SEM image of the particle. DSEM is the particle diameter based on
VSEM . DBF is the particle diameter that best fits the experimental results obtained
by using different sizes in Abaqus. E − DSEM and E − DBF are the %-errors in
particle diameters with respect to Dr . . . . . . . . . . . . . . . . . . . . . . . . . 53
A.1 Classic Johnson-Cook material properties for Al-6061 . . . . . . . . . . . . . . . . 71
vii
Acknowledgments
The current thesis is the product of contribution of many people who have been helping
me through out my life. I am unable to find the words to express my appreciation of their support.
I am deeply thankful to my supervisor, Prof. Sinan Müftü for his guidance and infinite
support in every single step of this research. His intelligence, knowledge, and wise comments have
always lit up new ways and ideas.
I would like to express my great appreciation of the inspiring work of Prof. Jae-Hwang
Lee and Ms. Wanting Xie from University of Massachusetts, Amherst, on the single particle impact
tests. This data had a remarkable influence on my entire research.
My sincere thanks to Prof. Andrew Gouldstone, and Prof. Teiichi Ando for their de-
tailed review, constructive suggestions and excellent advice during the preparation of this thesis.
The financial support of US Army (ARL W911NF-15-2-0026) which gave me the opportu-
nity and encouragement to expand my research, is gratefully acknowledged.
I would like to thank my classmates in 244H and 225B Forsyth Building for all the help,
guidance and friendship they have given me.
Last but not least, I wish I was able to thank my lovely family. I am grateful to my parents
who loved me with infinite generosity and provided me with the opportunity to be where I am.
Without them, none of this would have been possible.
viii
Abstract of the Thesis
Tuning Johnson-Cook Material Model Parameters for Impact of High

Velocity, Micron Scale Aluminum Particles
by
Arash Alizadeh Dehkharghani
Master of Science in Mechanical and Industrial Engineering
Northeastern University, August 2016
Dr. Sinan Müftü, Adviser
Cold Spray (CS) is an additive manufacturing process which uses the extreme plastic
deformation of micron scale particles to repair surface defects. This process requires accelerating
particles to very high velocities (200-1000 m/s) by a supersonic compressed gas jet at temperatures
well below particle’s melting point. Although many metals and alloys have been successfully
processed using the CS techniques, the accurate dynamic responses of individual metallic particles
related to the deformation characteristics are still largely unknown. Therefore the main objective
of this research is to investigate the mechanics of single particle impact. The outcome of this study
can be used to study multi-particle impact and ultimately study the mechanics of 3D-printed metals
using CS technology.
Numerical simulation has been used to produce the particle impact results. Simulations
show that the material experiences very high strain rates (107 -108 s−1 ) causing severe plastic
deformation. To conduct an accurate analysis in the simulations, the flow stress of the material should
predict appropriate metal behavior at that range of strain rate. In this study the Bilinear Johnson-Cook
material model has been used to predict the flow stress and Aluminum-6061 was chosen as the
particle’s material property. Simulations included the effects of high strain-rate (HSR) plasticity, heat
generation and dissipation, material damage, and surface interactions in three dimensions.
High strain rate experimental results are usually done by using Split Hopkinson Pressure
Bar (SHPB). Since these experiments have limitation on the maximum strain rate applied to the
sample, there is no accurate data for the flow stress of the material at the high rates encountered in
CS. Therefore, in this thesis a computational material model calibration has been performed for the
Bilinear JC model for the HSR applications. The optimization process uses the method of steepest
descent to find the best constants in the Bilinear JC constitutive law. The difference between ellipticity
ix
ratio of the deformed particle in simulation and experiments is used as the objective function, and the
parameters of the Bilinear JC equation are modified until the objective function is satisfied.
The optimized bilinear Johnson-Cook model was used to simulate the deformed shape
of particles. The results show a very good agreement between the simulations and single particle
impact experiments. The optimized bilinear JC-model was further verified by comparing simulation
results of the particle rebound velocity and the coefficient of restitution (COR) to experimental data.
The methodology developed in this thesis can be used to develop the model parameters for different
materials and other HSR material models.
x
Chapter 1
Introduction
1.1 Background
In recent years, research on is conducted to find materials with superior mechanical

properties to satisfy the requirements for different designs imposed by many fields and applications in
order to obtain competitive products. Some of the major modes of failure in mechanical engineering
and structural applications, such as; fatigue, fracture, wear, aging and corrosion, are very sensitive to
the properties of surfaces. These defects usually originate from outer layers and propagate inward.
This makes the surface improvement an effective approach, in increasing the durability and life of
the components. In the following, two distinct methods of coating are discussed.
In last decade, thermal spray technologies have evolved from fairly crude processes that
were relatively difficult to control into increasingly precise tools, by taking into account the properties
of the deposited material and coating requirements. The limitation of some coating types such as;
High Velocity Oxygen Fuel (HVOF) process, Plasma Spray Systems, and thermal spray coating
technology have been overcome by the cold gas spray coating technology. Figure 1.1 shows a
comparison of different coating methods by considering the temperature and velocity of the sprayed
particles [1]. In this diagram, we see that the cold spray requires much lower process temperature
with respect to other coating types. The cold spray coating technology was originally developed
in the mid-1980s at the Theoretical and Applied Mechanics division of the Russian Academy of
Sciences Institute in Novosibirisk by Dr. Anatolii Papyrin and his colleagues [16]. They successfully
deposited a wide range of pure metals, metal alloys, and composites onto a variety of substrate
materials, and demonstrated the feasibility of cold spray process for various applications.
The coating process known as Cold Spray is a material deposition process in which
1
CHAPTER 1. INTRODUCTION
Figure 1.1: Comparison of gas temperature and particle velocities in thermal spray processes and
cold spray process. VPS is vacuum plasma spray and LPPS is low pressure plasma spray [1].
relatively small particles (ranging in size from approximately 5 µm to 100 µm in diameter) in solid
state are accelerated to high velocities (typically 200-1000 m/s), and subsequently form a deposit on
a metallic substrate. Currently, Cold Spray is an advanced surface repair process, for metal surfaces
that does not induce thermal stresses in the deposited material. This makes CS uniquely compatible
with many aerospace materials. Unlike other repair processes that use heat to fuse materials together,
Cold Spray uses a compressed gas to accelerate metal powders to supersonic speeds and plastic
deformation to attach to the substrate. The resulting surface is comparable to the original substrate in
terms of the material strength.
Figure 1.2: Schematic diagram of a cold spray gun [1].
Bonding of the particles in this process is presumed to be the result of extensive plastic
deformation due to the high kinetic energy upon impact; therefore, the velocity of the particles play
the most important role in material deposition. During the process, powders are accelerated by
injection into a high velocity stream of gas which is generated through a converging-diverging nozzle.
The schematic diagram of the cold spray equipment is shown in Figure 1.2. As the process continues,
2
the particles impact the substrate and form bonds with it, resulting in a uniform, almost porosity-free
coating with high bonding strength. Note that the particles remain in the solid state and are relatively
cold, so the cohesion of the deposited material, is accomplished in solid state.
The applications of cold spray coating is mostly on the surface treatment. Wear resistance
is one of these applications. Wolfe et al. [17] used cold spray coating successfully in order to
coat 4140 alloy for wear resistance applications. The obtained results are presented in detail in [1].
Other papers [18, 19, 20] also show the potential of using cold spray coating in wear resistance
applications. Corrosion protection is also another application of Cold Spray. The advantages of
cold spray technology for deposition of corrosion-resisting layers are evident, however, a major
consideration in the selection of cold spray processing for a given application is its cost relative to
alternatives [1]. Another application of Cold Spray is to repair the damaged component. The US
Army Research Laboratory (ARL) Center for Cold Spray Technology [21] has developed a cold spray
process to reclaim magnesium components that show significant improvement over existing methods
and is in the process of qualification for use on rotor-craft. The cold spray repair has been shown to
have superior performance in the tests conducted up to date. It can be incorporated into production,
and has been modified for field repair [1]. The feasibility of using the cold spray process to repair
non-structural magnesium aircraft components, has demonstrated satisfactory results obtained from
adhesion, corrosion testing, and micro-structural analysis. In these tests, cold spray coating was not
pulled off the substrate and the coating did not fail cohesively [1]. It should be noted that cold spray
coating can be considered a young technology with respect to other types of thermal coating. The
applications of this technique are gradually increasing in the surface treatment industry. This section
has only mentioned some of the important current applications of cold spray technology, which is
used all over the world.
1.2 Objective and Problem Description
Not only many constitutive relations are not capable of predicting the behavior of material
at ultra high strain rates, but also there are no reliable experimental data to fit them onto a material
model. Therefore, in this study the main focus is to achieve a material model that can anticipate the
mechanical behavior of material in very high strain rate applications such as Cold Spray. In the Cold
Spray process the gas carries micron-scale particles with a supersonic velocity and they impact the
substrate. The process includes very large strain rates, since the duration of the process is in the order
of nano-seconds and the deformation is really large. This constitutive model, needs to be precise in a
3
wide range of impact velocities (200 m/s - 1000 m/s) and particle diameters (15 µm - 50 µm).
1.3 Arrangement of the Thesis
Material presented in this thesis is divided into six chapters. This chapter (Chapter 1)
provided a brief background on the Cold Spray technique and its applications followed by the
motivations and objectives of this research. In Chapter 2, a survey of literature of the cold spray and
the high strain rate experiments are presented. In Chapter 3, numerical simulation of single particle
impact is explained along with a comprehensive study of the techniques used to model the impact
process in Abaqus finite element software. The effect of using temperature dependent material
properties in simulations is also discussed. In Chapter 4, simulation results using classic Johnson-
Cook material model and Bilinear Johnson-Cook material model are illustrated. In Chapter 5, the
constants for Bilinear JC model is optimized to find the exact deformation of single particle impact.
In this chapter, the results of optimized bilinear JC model is also compared to the experimental data.
Finally, this thesis finishes by explaining the summary, conclusion and future works in Chapter 6.
4
Chapter 2
Literature Review
2.1 Cold Spray
Cold spray is a technique that accelerates powder particles towards a substrate using a
supersonic compressed gas jet, as mentioned in Chapter 1. Upon impact the particles experience
extreme and rapid plastic deformation. They stack up to form a solid state deposit that is well bonded
to the substrate. Particles adhere to the substrate only if their impact velocity is above a critical
value. This value varies between 500 and 900 m/s, depending on the material and the particle size.
Therefore the particle velocity prior to impact [1, 22] is the main parameter that determines what
phenomenon occurs upon the impact of spray particles, whether it would be the deposition of the
particle or the erosion of the substrate. Figure 2.1 is a schematic plot which shows the correlation
between particle velocity and deposition efficiency. Critical velocity for a given powder is defined as
the velocity that an individual particle must obtain in order to deposit after impacting the substrate [1].
In other words, for a given material, the critical velocity is the velocity at which the transition from
erosion of the substrate to deposition of the particle takes place. Experimental investigations also
reveal that successful bonding is achieved only above this velocity. However, this value is associated
with the temperature and thermo-mechanical properties of the sprayed material [22, 16], as well as
the characteristics of the substrate [23, 24]. Temperature of the particles have also an effect on both
critical velocity and deposition efficiency [1].
The coating process, depends on using high kinetic energy, provided by the supersonic
gas flow. The high velocity of the particles is obtained in a divergent-convergent nozzle. The gas
temperature along with many other process parameters, play considerable role in coating quality
from increasing deposition efficiency to the porosity control of the final deposition.
5
CHAPTER 2. LITERATURE REVIEW
efficiency with the respective spr

range of various deposition effic
different nozzle geometries (Ref
eters such as gas temperature and
investigation is based on a meth
temperature are minimized by us
and different nozzle types (Ref 7
ditions, particle velocity and tem
ing CFD. This approach assumes
and the extent of oxidation are s
ticle size for the particular feeds
2.2 Investigations of Coa

Bond Strength
Fig. 3 Schematic of the correlation between particle velocity and de-
To evaluate
Figure 2.1: Schematic plot of efficiency.
position correlation The between particle
transition between velocity
abrasion in theand deposition efficiency
low-velocity [2].coating quality w
range and deposition for high velocities defines the critical velocity. amount of well-bonded particle
Vcrit, critical velocity. tions were subsequently prepar
microscopy. The microstructura
2.2 Particle Impact be explained by the occurrence of local shear instabilities at par- particles and particle-particle
ticle-substrate and particle-particle interfaces due to thermal chemical etching with a solutio
softening, as first shown by Assadi et al. (Ref 18). Later
Figure 2.2 shows the classification of impact phenomenon with respect to particle diameterwork ammonia (25% in H2O), and 5 m
was aimed at refining the procedures for calculation and its ap- H2O). A selection of coating cro
and impact velocity [3].plications
Particles to different
impacting spray materials
onto (Ref 19, of
the surface 20).aBased
solidonbodythe can gated by scanning
rebound from electron micr
concept of bonding by shear instabilities and by combining the Bond strength measurement
the surface, stick to theresults
surfacefromormodeling
penetrate andinto the bulk.investigations,
experimental Often, theanalyti-impact of amode particle on ato the European
according
cal expressions were recently developed to predict the ranges of to ASTM standard C 633-01; R
surface causes a deformation or destruction of both the particle and the substrate. Insteel
optimum spray conditions with respect to the mechanical prop-
the following,
substrates were degreased
erties of spray materials,
characteristic of each region is discussed briefly. spray particle sizes, and particle tem- thermal influences by preparat
peratures (Ref 21). bond-strength samples to counte
At low velocity impacts
Another (<100
focus in m/s), small particles
cold spraying (<1.0 µm)
is the exploration of thecan
fea-stick to
bythe substrate
using adhesive DP 490 (3M, S
sibility of different feedstock materials. As shown in the latest tents of the initial feedstock and
after impact. In these cases (number
proceedings of 6theand 7 in Figure
International 2.2), adhesion
Thermal is governed by
Spray Conference byusing
the van der
a commercial analyzer o
(ITSC) conferences, quite a number of investigations
Waals and electrostatic forces, [25]. Sticking happens when the velocity of impacted particle isconcern factured by LECO (St. Joseph, M
process development for standard engineering materials and cally detached from the substrat
deal with If
between two critical velocities. different
the impactaspectsvelocity
to customize cold spraying
is below the firstforcritical
vari- velocity, particles
of the flat coupons over a sharp
ous spray powders envisaging short-term applications. To dem- using a mechanical testing mach
will attach to the surface. On the
onstrate other hand,
the versatility of theifprocess,
the impact
a couplevelocity is above the second
of such examples critical
that in 3-point bending tests.
are also given in the present article. With respect to powder de-
velocity, impacted particles will rebound. In the intermediate range there is a probability of sticking
velopments, significant attention is being paid to new material
2.3 Nozzle Developments
of particles, which dependstypes.
onItthe
has impact
been shown that cold spraying is suitable to process
velocity.
coatings on the basis of nanocrystalline alloys (Ref 22), By launching cold spraying
In region number 8 (Figure composites
nanocrystalline 2.2) adhesion can24),
(Ref 23, also happen
or bulk during
metallic a low velocity
glasses impact
trumpet-like shape were original
(Ref 25), just to mention a couple. ring to radial symmetry, the sm
of larger particles (10 < dp This
< 100 µm)
article [26].onThe
reports the rebound velocity
current “state of theof such
art” particles “standard”
in cold- decreasesnozzlewith of a slightly tr
decreasing particle size. spray process development. Taking the example of copper as the 2.7 mm, a length of the divergin
spray material, the influences of optimized spray conditions on pansion ratio of 8.8. Using the n
Larger particlescoating
(1–10 mm)quality impacting with velocities
are reported. With respect up totoabout 40 m/s on aics,
the above- plate
gastypically
and particle flow for c
mentioned understanding of the process, special emphasis is calculated (Ref 7). The numeric
rebound from the surfacegivenwithout breaking
to current but theytoleave
investigations plastic deformation
tune cold-spray parameters toon both particle
particular theand
method of charact
different applications. The given examples of coating micro- to develop new nozzle designs t
impacted surface. The size of the crater increases with increasing impact velocity. In order to predict
structures of various materials should demonstrate the versatility particle acceleration (Ref 27).
velocities and angles of ofrebound
the process.
as well as the size of the plastic prints, Hertzian theory can be
3. Results and Discu
2. Procedures 6
3.1 Evaluation of Critical
2.1 Determination of Critical Velocities
In the high-velocity gas stream
For a variety of spray materials, the critical velocity for suc- ated to significantly higher veloc
cessful impacts was determined by correlating the deposition 14). Only very small particles ar
cked region at particle-surface contact area near the axis
area between particle and surface
article
and domes by particle im-

thermore, disregarding the
lt in deviations of the drag
onsequently, failure. Here,
re similar to those in cold
act phenomena have been

very of the cold spray phe-
here is as yet no complete
on of cold spray deposition.
ore closely a circle of phe-
teractions of bodies and to
nd on particular parameters
wing different impact phe-
ext of their relationship to
rder to gain a better under- Fig. 1. Particle impact on a solid surface: Influence of impact velocity and
Figure 2.2: Particle
particleimpact
size ononfeatures
a solidof
surface: effect ofRegions
the interaction. impact velocity and particle
characteristic size on features
of certain
characteristics of impacts impact phenomena are shown. For details see text.
of the
ld sprays, are discussed ininteraction. Regions characteristic of certain impact phenomena are shown [3].
2.1. Low-velocity impacts of particles
applied [27].
The impact
At lowof large particles
impact velocities (vpat≈
(1–10 mm) higher
1–100velocities
m/s) (50–3000 m/s) illustrated in
small parti-
f the impact of particles cles (d ≈These
Figure 2.2 as ”Ballistic”. 0.1–1.0 µm) have
particles can stick to theenergy,
high kinetic surface after impact.
therefore the amount of plastic
p
Adhesion
strain and strain is governed
rate considerably by during
increases van der
the Waals
impact. and
As a electrostatic
result of that, both material
forces, e.g. [24,50]. This is typical of filters. Corresponding
hardening and heat generation happen in the material. Heating consequently results in a thermal
urface of a solid body can regions that have been studied in the literature are denoted
softening.
ck to the surface or pen- Thus, in a correct simulation it is necessary to take into account a competition between
by the numbers 6 and 7 in Fig. 1.
impact of a particleprocesses
on a of dynamic hardening
Sticking and thermal softening.
is characterized by two critical velocities. If the
destruction of both, the par- At impact impactvelocities
velocitygreater
is below
than the first critical
2000–3000 velocity,
m/s stresses particles
arising in bodies on impact,
following, different results are collected
considerably exceed the yieldbypoint
the of
surface. If the
materials. impact
Under thesevelocity
conditionsisphase
abovetransformation
faces will be classified. If the second critical velocity, impinging particles rebound. In
happens and solids behave like liquids. Locally, the flow velocities are comparable to the sound
rticles and normal impacts the intermediate range there is a probability of sticking (or
speed.
ned later scaling is difficult. This is connected with the appearance of strong shock waves. Hypervelocity impacts of
rebound) of particles, which depends on the impact velocity.
ased on two important particles
di- on semi-infinite bodies [28, 29],
The probability and on plates
of sticking has [30,
been 31]determined
have been studied
exper-extensively. The
processes occurring during
correspondingimentally,
regions haveforbeen
example,
depictedfor silica spheres
by number 2 in Figure = Hypervelocity
(dp2.2. O(1 µm)) impacts of
diameter dp of impinging impingingsize
particles of intermediate on[32]
a quartz crystal (cf. Fig.
and of micro-particles [33]2,areafter [43]).
labeled withIn this 3,4 and 5.
number
be called the size of a par- Experiments
figure the two critical velocities are of the order of 1 m/s and
done by Klinkov et al., [3] shows some penetrations when particles accelerated
certain impact phenomena about 10 m/s, respectively. The two lines represent an upper
to velocities of about 1–3 km/s by using hollow-charge explosions. They observed that some particles,
results reported in the lit- and lower limit (i.e. standard deviation) of sticking probabil-
ayed in Fig. 1 and will be ities. The true value of the probability lies between the two
ns. curves. 7
in the range of 100 µm penetrate far into the body. Penetration distances were equal to 103 –104 times
greater than diameters of the particle. This effect is called super-deep penetration which is indicated
by SDP in Figure 2.2. After impact, the material of the body collapses and therefore contains a
significant level of plastic deformation [34]. Note that in the case of 1–10 mm particles impacting a
surface at similar velocities, the depth of penetration is typically about 4–5 times the particle diameter
[35].
At moderate and low impact velocities (5–300 m/s) a repeated impact of intermediate-sized
particles (30–500 µm) on the same part of a surface, results in an erosion and destruction of the
surface. In brittle materials erosion happens when the particle is impacting on normal direction
[36, 37], however, in ductile material, the highest rate of erosion observed when the impact happens
on oblique angle [38, 39]. Klinkov et al., [3] explained that ductile materials is worn out by processes
of ductile cutting and plowing. This results in the formation of raised lips of material that are
vulnerable to removal by subsequent impacts. Nevertheless, brittle materials are worn out by a
process of fracture that is more intensive when the impact is close to normal.
Finally, impacts of micron scale particles on surfaces at velocities of approximately
200–1200 m/s, correspond to the region denoted as cold spray in Figure 2.2. Under impact conditions
corresponding with the cold spray region, particles can form strong bonding with the surface after
impact. The cold spray phenomenon is used for coating surfaces. In these applications, particles
are accelerated by a supersonic gas flow. Hence, in this process the impact velocity depends on the
temperature and pressure of the gas and also the gas type in the reservoir ahead of the nozzle.
2.3 High Strain Rate Experiments
When it comes to simulating a physical phenomenon, mechanical and thermal behavior

of a material plays a significant role, and should be determined as accurately as possible. For
example, in a creep test the variation of temperature in a process is really important or, in a crash
test the hardening and rate of applying the load are substantial information. In this section, the
dynamic behavior of a material and how this behavior can be distinguished from its static behavior
is discussed. First, the history of Split-Hopkinson Pressure Bar (SHPB) and the new developments
in this experiment are explained. In the second part, the SHPB apparatus and determination of the
dynamic material behavior are illustrated. In the next part, tensile and torsional dynamic loading
experiment based on the concept of SHPB is described. Finally, other types of dynamic experiment
and limitation of these kind of tests are briefly discussed.
8
(Fig 1). With special design, it is possible to at- measure the stress-strain response of ductile the sample. It is th
tain greater strain rates, up to about 100 s– 1, materials at a high strain rate, usually between uniaxial stress-strain
with conventional load frames. For higher approximately 50 s– 1 and 104 s– 1, depending on als at a variety of s
strain rates, other test methods are required. Ta- the sample size, over the entire stress-strain sponse of most mat
ble 1 summarizes various methods in terms of curve. Strains exceeding 100% can be achieved strain rate and the
the ranges of the strain rates that they can with the Hopkinson bar method. The maximum allows developing
achieve. strain rate that can be attained in a Hopkinson express the uniaxial
The articles in this Section describe various bar varies inversely with the length of the test strain rate and temp
CHAPTER 2. LITERATURE
methods for high strain REVIEW
rate testing. Several specimen. The maximum strain rate is also lim- one is able to prod
methods have been developed, starting with the ited by the elastic limit of the Hopkinson bars three-dimensional c
pioneering work of John Hopkinson (Ref 1) that are used to transmit the stress pulse to the merous materials.
The series of articl
pression testing begi
–8 –6 –4 –2 0 2 4 6
0 10 10 10 10 10 10 10 10
Strain rate (s–1) Table 1 Experime

Intermediate High strain Very high
Creep Quasi-static strain rates rates strain rates
strain rate testing
Applicable strain rate, s–1
Light gas gun Compression tests
Servohydraulic & Special Hopkinson or
Constant load Shock <0.1 C
screw machines servohydraulic techniques explosively
or loading 0.1– 100 S
machines driven
stress machine 0.1– 500 C
plate impact 200– 104 H
Strain versus time Constant Uniaxial Uniaxial 103– 105 T

Constant
or strain rate strain rate stress & strain & Tension tests
creep rate tests tests torsion shear tests <0.1 C
recorded tests 0.1– 100 S
Dynamic
considerations 100– 103 H
Inertia forces neglected Inertia forces important in testing 104 E
>105 F
Isothermal Adiabatic/quasi-isothermal
Shear and multiaxial test
Uniaxial & shear stress Uniaxial strain & <0.1 C
simple shear
0.1– 100 S
10– 103 T
Increasing stress levels
100– 104 H
103– 104 D
104– 107 P
Fig. 1 Strain rate regimes and associated instruments and experimental conditions
Figure 2.3: Strain rate regimes and associated instrument and experimental conditions [4].
The evolution of current Split-Hopkinson pressure bar test for conducting high strain
rate testing of materials, goes back to 1872 where John Hopkinson was investigating the effect
of impulsive forces on the iron wires [40, 41]. Three decades later, his son, Bertram Hopkinson
suggested the Hopkinson bar experiment while he was investigating the stress pulse propagation
in metal bars [42]. In these experiments he determined how to measure the pressure produced by
an explosive or impact of bullets. Of course the limitation of equipments on that era affected the
accuracy of experiments. Davis [43] in 1948 used an electronic device to record wave propagation
and improved the measurement technique. He also was the first one to find a way to measure the axial
and radial strain in the Hopkinson pressure bar experiments. One year later, Kolsky [44] modified
the mechanism of the experiment. He divided the bar into two (incident bar and transmitted bar) and
placed the specimen between these two bars. That’s why the name of the experiment changed to the
Split Hopkinson Pressure Bar (SHPB). He also obtain the stress-strain relations of materials such as
rubbers, plastics, polythene, Polymethyl methacrylate (PMMA), copper, and lead under dynamic
loading conditions.
After that, the concept of this dynamic test did not change and scientists improved the
accuracy of the experiments by minimizing the friction and radial inertia effects to produce a uniform
stress in the specimen. They also expanded the experiment to different materials and various kind
9
using either the tensile split-Hopkinson pressure discussed. tic properties are ess
bar or expanding ring test are compared and 800 °C (Ref 3). Becau
contrasted with the compression Hopkinson bar terial increases the sig
technique discussed in the section “ Stress-State Principles of the lection of a bar mate
Equilibrium during Split-Hopkinson Pressure and lower elastic mod
Bar Testing” in this article. Split-Hopkinson Pressure Bar is sometimes desirable
An alternate method of probing the mechani- tion dynamic testing o
cal behavior of materials at high strain rates, of While there is no universal standard design such as polymers or
CHAPTER 2. LITERATURE
the order of 103 s−1, is theREVIEW
Taylor rod impact for SHPB test apparatus, all facilities share selected bar material
test. This technique, named after G.I. Taylor common design elements. A compression elastic stiffnesses fro
(Ref 31), who developed the test, entails firing Hopkinson bar test apparatus consists of the GPa) to titanium (110
a solid
of loading typescylinder
such as of tension,
the material of interest
torsion, following:
compression followed by torsion, torsion followed GPa) byto magnesium (4
against a rigid target. The deformation induced finally, to polymer b
in the rod
compression, due toloading,
tri-axial the impact andin dynamic test • Two These
the Taylorindentation. long, symmetrical bars
new developments allow us to 33–
reach
35). Alternately,
shortens the rod as radial flow occurs at the im- • Bearing and alignment fixtures to allow the Hopkinson bar used to
more thanpact100% ofThe
surface. totalfractional
strain in SHPB
change in test [45]. However,
the rod the maximum
bars and striking projectile strain
to move rate that can
freely can be
be increased usin
length can then, by assuming one-dimensional while retaining precise axial alignment mitted pressure bar (R
4 −1
found inrigid-plastic
a test varies inversely
analysis, withtothe
be related thelength
dynamic • Compressed
of the specimen and can go up totube
gas launcher/gun 10ors alter-
strain rate can yield incr
nique
yield strength. By measuring the overall length nate propulsion device for accelerating a measurement sensitiv
by usingofSHPB test. Figure
the impacted cylinder2.3and
illustrates
the lengthanofappropriate
the technique
projectile, termedandtheexperiment
striker bar, tofor each range
produce of
tion of the sample str
undeformed (rear) section of the projectile, the a controlled compressive pulse in the inci- meric materials must
strain rate.
dynamic yield stress of the material can be cal- dent bar wave dispersion in th
culated (Ref 31). The Taylor test technique of- • Strain gages mounted on both bars to mea- (Ref 37, 38).
fers an apparently simplistic method to ascertain sure the stress-wave propagation in the bars The length, l, and di
2.3.1 information
Split-HopkinsonconcerningPressure
the dynamicBar Test • Associated instrumentation and data acqui- bars are chosen to mee
strength
properties of a material. However, this test rep- sition system to control, record, and analyze test validity as well as
resents an integrated test rather than a unique the stress-wave data in the bars (Ref 18) and strain level desi
A modern
experiment withSHPB apparatus
a uniform consists
stress state of two symmetrical bars (transmitted bar and incident
or strain length of the pressur
rate, as does the split-Hopkinson pressure bar. In a compression split-Hopkinson pressure one-dimensional wave
bar), bearing and fixtures
Accordingly, the Taylorfor test
alignments,
has been aused devicebar,toa accelerate the strikerbetween
sample is sandwiched bar, strain gauges to
an elas- pulse length; for exp
most prevalently as a validation experiment in tic incident and a transmitted
measure the stress wave propagation and a computer to record and analyze the data (Figure 2.4). bar (Fig. 1). The on most engineering m
concert with two-dimensional finite- element terms incident/input and transmitted/output requires approximatel
calculations. are used interchangeably throughout this ar- readily allow separatio
This article describes the techniques involved ticle to describe the two pressure bars used in flected waves for d
in measuring the high-strain-rate stress-strain the SHPB. The elastic displacements mea- should exceed a lengt
of ~20. In addition, t
desired will influence
diameter because the
require the smallest di
aspect of bar design is
tion). The third consid
Air lection of the bar leng
strain desired to be im
the absolute magnitud
to the length of the inc
bar must be at least tw
wave if the incident a
be recorded without
because the bars must
test, the displacement
interface between the s
accurately determined
ple size, for strains >3
Fig. 1 Schematic of a compression split-Hopkinson pressure bar for the split-Hopkins
Figure 2.4: Schematic view of split-hopkinson pressure bar [4].
In a compression test, samples are placed in between the incident bar and the transmitted
bar. The elastic displacements are measured in these two bars to determine the stress-strain relation
in the sample. These bars are constructed from a steel which has a high yield strength. The maximum
attainable pressure depends on the yield strength of material. The experiments are only valid if the
deformation of these two bars are elastic. It is really important to choose an appropriate length and
diameter for the pressure bars to ensure the validity of the test, and to reach the maximum amount of
desirable strain rate and also find the maximum strain in the sample. To do that, first, the stress wave
10
464 / High Strain Rate Testing
ratio of 100 or more (Ref 3). There are similar within the specimen during loading. Bar align- move along the bar axis
requirements
q
should propagate infor L/D ratios to allow
justbarone-dimension andwave ment
to make thatcannot
happenbe the
forced by overconstraining
length or should
to diameter ratio tial to apply precise d
separation
p for compression, tensile, and torsion forceful clamping of curved pressure bars in an during construction
be greater than 20.bars.
Hopkinson Second, to reach higher strain rates,
attemptthe
to diameter of thebecause
straighten them pressure
this bars should
clamp- barsbe
are often center
For proper Hopkinson bar operation, the bars ing violates the boundary conditions for one-di- length to achieve the
small enough.
must be Finally,
physicallytostraight,
retain the
free desirable amount
to move with- of totalwave
mensional strain the length
propagation in of
an the barscy-should
infinite be
straightness required.
out binding, and carefully mounted to ensure lindrical solid. Lack of free movement of the tions, as schematically
long. Foroptimal
instance to alignment.
axial reach 30%Precision
of total bar
strain, the length
align- over
bars will leaddiameter rationoise
to additional needontothe
bewave
100 orsure
morebars are mounted
ment is required for both uniform and one- forms measured on the pressure bars. Bar mo- to provide a rigid and
[4]. dimensional wave propagation within the pres- tion must not be impeded by the mounting form. Construction of
sure bars as well as for uniaxial compression bushings but rather must remain free to readily bar facility, compress
on an optical rail bea
I-beam, can be used
alignment. Figure 2 sh
sion Hopkinson bar fac
tional Laboratory, wh
to maintain accurate p
dividual mounting bra
slip bearings through
typically spaced every
in.), depending on th
ness. Mounting bracke
so that they can be i
adjust bar alignment w
Induction Heating The most common
Coil incident wave in the
Bar striker bar to impact
Mover bar. The striker bar
Gas Gun from the same materi
Bar Stop ameter as the pressure
locity of the striker b
Incident Bar the desired total strai
the test specimen. W
also be generated in
Optical Rail the adjacent detonati
Transmitted Bar
free end of the incide
(Ref 7), it is more dif
Fig. 2 A compression split-Hopkinson pressure bar facility at Los Alamos National Laboratory mensional excitation w
direct explosive loadin
LIVE GRAPH The impact of a stri
Click here to view the incident bar deve
Figure 2.5: A2.5compression split-Hopkinson pressure bar facility at Los Alamos National Lab (Figure
pressive incident wav
from [4]). εi, as denoted in Fig. 3
2.0 the bar-specimen inter
Incident wave designated εr, is refle
of the stress pulse pas
Pressure
1.5 bars are usually attached to a rigid body, for a better alignment and providing a
and, upon entering the
Transmitted wave
straight platform (for example the I-beam in figure 2.5). Figure 2.5 shows one of the compression transmitted wave, εt.
Strain gage output, V
magnitude of these thr

1.0 the incident and trans
Hopkinson bar facilities at Los Alamos National Laboratory, where an optical rail is used to maintain
by strain gages norma
accurate pressure
0.5 bar alignment. Impact of the striker bar to incident bar happens due to propulsion point positions along t
sure bars. Figure 3 sh
of the striker. This impact, develops a horizontal compressive wave in the bar. Of course, a portion strain-gage data measu
0 for the three wave sig
of this wave reflects while the remaining reaches the specimen which they call it transmitted awave. 304L stainless steel
steel pressure bars. Th
The time of−0.5 the passage and magnitude of these elastic pulses are recorded by strain gauges normally wave signals represe
cemented at the midpoint positions along the length of the Reflected wave
two pressure bars. These wave signalspulses, from while the reflect
−1.0 Using the wave sign
the gauges on the incident and transmitted bars, can be plotted as a function of time. By integrating incident and transmitt
time, the forces and v
−1.5 faces between the pre
0 100 200 300 400 500 men can be determine
11
Time, μs deforming uniformly,
specimen is directly p
Fig. 3 Strain-gage data, after signal conditioning and amplification, from a compression split-Hopkinson pressure
bar test of a 304 stainless steel sample showing the three stress waves measured as a function of time. Note tude of the reflected w
that the wave positions in time are arbitrarily superimposed due to the time delays used during data acquisitions. within the sample is di
ica- Based on these contributions and also on an im- Hopkinson bar method and some specialized
uc- portant paper by Davies (Ref 4), Kolsky (Ref 5) compression and tension tests are discussed in
t be invented the split-Hopkinson pressure bar, the article “ High Strain Rate Tension and Com-
of which allows the deformation of a sample of a pression Tests” in this Section. An overview of
ing ductile material at a high strain rate, while shear test methods (other than the torsional
s of maintaining a uniform uniaxial state of stress Kolsky bar method) is also provided in the arti-
tes. within the sample. The basic concept of the cle “ High Strain Rate Shear Testing.”
be Kolsky apparatus involves a test sample The most important characteristic of Kolsky’s
era-
sandwiched between an input and output bar, as split-Hopkinson compression apparatus is that
onal described in detail in the article “ Classic it allows high strain rate deformation while the
for Split-Hopkinson Pressure Bar Testing” in this sample is, in fact, in dynamic equilibrium, that
ess the provides
Section. This technique mentioned plot wetocan is,
a capability determine
the stressthe forcesisand
gradient velocities
essentially zeroatalong
the two interfaces between the
at- measure the stress-strain response of ductile the sample. It is thus possible to develop the
pressure bars and the specimen. Because the specimen deformed uniformly, the strain rate within the
s – 1, materials at a high strain rate, usually between uniaxial stress-strain response of many materi-
her approximately 50 s– 1specimen
and 104 s–is
1, depending on als at to
directly proportional a variety of strainofrates.
the amplitude Because the
the reflected re- Similarly, the stress within
wave.
Ta- the sample size, over the entire stress-strain sponse of most materials depends on both the
s of the sample
curve. Strains exceeding 100% canis directly proportional
be achieved to the
strain rate andamplitude of the transmitted
the temperature, the techniquewave. The reflected wave is
can with the Hopkinson bar method. The maximum allows developing constitutive relations that
also integrated to obtain strain and is plotted against stress to give the dynamic stress-strain curve for
strain rate that can be attained in a Hopkinson express the uniaxial stress to the corresponding
ous bar varies inversely the the length of the test strain rate and temperature. From such results,
withspecimen.
eral specimen. The maximum strain rate is also lim- one is able to produce experimentally based,
the ited by the elastic limit of the In recent years,
Hopkinson bars a number of new improvement
three-dimensional constitutivehave beenfor
models designed
nu- to improve the accuracy
1) that are used to transmit the stress pulse to the merous materials.
of the experiment. Optical measurement is articles
The series of a new technique to find the
on split-Hopkinson radial strain in a sample using
com-
pression testing
linear laser line-measuring technique. begins
In this with it
method theisarticle “ Classic to attach a strain gauge to
not necessary
0 2 4 6
10 10 10
the10sample [46, 47] and this provides a significant advantage.
Strain rate (s–1) Table 1 Experimental methods for high
termediate High strain Very high
rain rates rates strain rates
strain rate testing
Applicable strain rate, s–1 Testing technique
Light gas gun Compression tests
Special Hopkinson or
Shock <0.1 Conventional load frames
ervohydraulic techniques explosively
loading 0.1– 100 Special servohydraulic frames
machines driven 0.1– 500 Cam plastometer and drop test
plate impact 200– 104 Hopkinson (Kolsky) bar in
compression
Uniaxial Uniaxial 103– 105 Taylor impact test
Constant
strain rate stress & strain & Tension tests
tests torsion shear tests <0.1 Conventional load frames
tests 0.1– 100 Special servohydraulic frames
Dynamic
considerations 100– 103 Hopkinson (Kolsky) bar in tension
Inertia forces important in testing 104 Expanding ring
>105 Flyer plate
Adiabatic/quasi-isothermal
Shear and multiaxial tests
stress Uniaxial strain & <0.1 Conventional shear tests
simple shear
0.1– 100 Special servohydraulic frames
10– 103 Torsional impact
ss levels
100– 104 Hopkinson (Kolsky) bar in torsion
103– 104 Double-notch shear and punch
104– 107 Pressure-shear plate impact
nts and experimental conditions
Figure 2.6: Experimental method for high strain rate testing [4].
2.3.2 Dynamic tests in tensile and torsional loadings
As mentioned earlier, the Hopkinson experiment is also used to determine the shear and
tensile strength of a material in a dynamic loading test. In addition to the compression version of
Kolsky SHPB, the tensile version of this experiment for tension which is Split Hopkinson Tension
12
40 K. Xia, W. Yao / Journal of Rock Mechanics and Geotechnical Engineering 7 (2015) 27e59
in Fig. 21 show the quite different damage levels, although the is to convert the external axial compression impact in external
samples in these three typical tests were approximately subjected tube into tension wave in incident bar via the external tube where
to the same incident wave. With bar/sample interfaces fully lubri- the Kolsky bar system is placed. This approach was just an
cated, the samples are completely fragmented into small pieces, extension of a compression version of Kolsky bar system and
featuring a typical splitting failure mode. This failure mode con- directly utilized the launching system of a compression bar sys-
firms 1D stress states during the dynamic tests. In contrast, with tem. However, since the whole setup was inside a solid tube, it is
friction constrain at the boundary interfaces, the splitting is con- inconvenient for instrumentation (e.g. strain gauge) and direct
strained significantly and the recovered samples feature a shear observation of the specimen deformation. Then, a new configu-
cone. It can be concluded that without proper lubrication, the ration with top-hat specimen was proposed by Lindholm and
measured strength values will be over-estimated and the failure Yeakley (1968), where a top-hat specimen is placed between the
mode will also be quite different. solid incident bar and the hollow transmission tube (Fig. 22b).
When the compression stress wave in the incident bar arrives at
5.2. Dynamic
Bar (SHTB) tension
was the specimen, a tensile load is produced on the specimen gauge
testalso developed to obtain the characteristics of materials under dynamic tensile
section and a compression stress wave is transmitted into the
For rock materials, dynamic tension tests can be approximately transmitted tube. The advantage of this design is that the spec-
loading. The initial design of dynamic tension apparatus was by Harding. In this experiment there
categorized into two approaches: direct tension (Howe et al., 1974; imen does not need to be attached to the bar ends. Nicholas (1981)
Goldsmith et al., 1976; Huang et al., 2010a) and indirect tension introduced a way to use only the compression bar system to
was a single
(Klepaczko elastic
and Brara, 2001;bar
Wu and a specimen,
et al., 2005; attached
Schuler et al., 2006; inside a hollow tube [48]. Later, the design was
achieve tensile experiments (Fig. 22c). The modifications include
Dai et al., 2008; Kubota et al., 2008; Wang et al., 2009; Zhou (i) the specimen which is threaded into the incident bar and
modified
et al., 2012). and the entire
For dynamic system
direct tension wastheestablished
method, specimen is inside the tube [49, 50]. The specimen was also
transmitted bar, and (ii) a rigid collar placed over the specimen.
subjected to dynamic tensile load which is directly generated by The function of rigid collar allows the initial compression wave to
modifed to satisfy
SHTB. The dynamic thetension
indirect tensile testincludes
method criteria. Lindholm
Brazilian disc [51] proposed the top-hat specimen and Nicholas
pass through the collar without virtually touching the specimen
(BD) method (Zhou et al., 2012), semi-circular bend (SCB) method and most of the initial compression energy is transferred into the
(Dai et al., 2008) and spalling method (Klepaczko and Brara, 2001; transmitted bar. The transmitted compression wave is reflected to
[5] introduced the threaded specimen with a rigid collar. A tensile pulse is generated in the incident
Wu et al., 2005; Kubota et al., 2008; Erzar and Forquin, 2010). be a tensile wave, and propagates back to load the specimen only
because the collar cannot support tensile load. However, in this
bar
5.2.1.by loading
Dynamic directthe endmethod
tension of the incident bar (Figure 2.7) through direct impact of a mass with a flange
design, the specimen with the collar will inevitably deform when
In order to generate directly dynamic tensile loading in spec- subjected to the initial compression wave.
atimen,
the the
endtensile
of the incident
versions bar bar
of Kolsky [48]. Another
system loading method is to release a tensile pulse which is
have been The most efficient loading method for Kolsky tension bar is
designed. The early design of dynamic tension experiments was a direct tension. The first way is to store elastic energy by stretching a
stored in the
hollow tube, incident
inside bar by
which incident andusing a clamping
transmitted bars with afixture [52]. section of incident bar in tension (Staab and Gilat, 1991; Cadoni
specimen sandwiched in between are placed (Hauser, 1966; et al., 2009). A clamp is used to divide the pre-stressed and stress
Harding and Welsh, 1983) (Fig. 22a). The principle of this device free section in the incident bar. Suddenly releasing or breaking the
Figure 2.7: Schematics of dynamic tension loading [5].
On the other hand, Goldsmith [53] was the first one who was after shear strength of
the material in a dynamic test. He used solid cylindrical Barre granite specimen in torsional split
Hopkinson bar test. Then, the torsional split Hopkinson bar with thin-walled tubes was utilized to
obtain the pure shear strength by Lipkin et al., [54, 55]. Gilat, explained more about the details of
torsional split Hopkinson bar in the ASM handbook [6]. According to Gilat, the friction and lateral
inertia effects is negligible in the torsional split-Hopkinson bar, however, it is difficult to prepare
the thin-walled specimens and mount them on the bars. Therefore, several other methods had been
Fig. 22. Schematics of four types of dynamic direct tension methods.
developed to perform the dynamic shear tests, such as, compression shear method [56], punch shear
method [57, 58] and a split Hopkinson pressure shear bar (SHPSB) [59]. The compression shear
method was used for large strain test in SHPB. The SHPSB system consists of a wedge-shaped end
incident bar and two transmitted bars, using an optical system to obtain shear strength and shear
strain.
13
Holder
Fig. 1. Schematics of punch shear devices used by: a) Mazanti and Sowers [7] and b) Stacey [5].
GENERIC TESTING PRINCIPLES
A 25 mm diameter SHPB system is utilized to apply the dynamic load for punch shear tests (Fig. 2). The SHPB consists of a
striker bar, an incident bar and a transmitted bar. The length of the striker bar is 200 mm. The incident bar is 1500 mm long
and the strain gauge station is 787 mm from the specimen. The transmitted bar is 1000 mm long and the stain gauge station is
522 mm away from the specimen. The bars are made from Maraging steel, with a yielding strength of 2.5 GPa, density 8100
kg m-3, Young’s modulus 200 GPa and one dimensional stress wave velocity 4970 m/s. A gas gun lunches the striker bar to
impact on the incident bar and generates an elastic compressive wave toward the sample. At the sample assembly, the
incident wave will be separated into two waves: an elastic tensile wave reflected back into the incident bar and a compressive
wave transmitted into the transmitted bar. The incident wave H i , reflection wave H r and transmitted wave H t are measured by
strain gauges mounted on the incident bar and the transmitter bar, respectively.
where E and A are Young’s modulus and cross-sectional area of the bars, respectively.
Strain gauge Front Rear

cover supporter
P1 P2
Striker Incident bar Transmitted bar Incident bar Transmitted bar
Sample holder
Fig. 2. Schematics of Split Hopkinson Pressure Bar for dynamic punch shear tests. Fig. 3. Schematics of the sample holder in SHPB.
(a)ends (Fig. 3)When
Using these three waves, the dynamic forces P1 and P2 on both of the the
sample
test assembly
is under can be calculated
dynamic [18]:
force equilibrium (b) P1 = P2), the inertial effect in the dynamic test can
condition (i.e.
ignored [14, 19]. In this case, the punch shear stress in the sample can be calculated using the following equation:
P1 (t ) EA[H i (t ) H r (t )] (1)
P
P2 (t ) EAH t (t ) W
(2)
SDB
Figure 2.8: (a) Schematics of Split Hopkinson Pressure Bar for dynamic punch shear tests, and (b)
where W is the punch shear stress; P = P = P is the loading force; D and B are the diameter of incident bar and
1 2
thickness of the disc specimen, respectively. It is noted here that we divide the load by the total shear area to obtain the sh
sample holder in SHPB [6]. stress, in a similar way to most other static punch shear studies. The maximum value of W is considered as the punch sh
strength W 0 of the sample tested.
A special holder is designed to support and protect the sample during dynamic punch tests. Conventional punch sh
Dynamic punch shear method is widely
systems used
for static tests to measure
usually have two kindsdynamic shear
of punch heads: strength
cylindrical of and
punch head materi-
block punch head (Fig. 1).
dynamic punch tests using SHPB, an annular holder is usually adopted [15-17]. In this paper, the stainless steel ho
als. Zhao et al. [57] measured the shear strength
consists of Bukit
of a front cover and a rearTimah
supporter, granite by using
which are jointed the
by screw to holdpunch
the sampleshear
as shown in Fig. 3. The purp
of the front cover is to reduce the bending force during tests and additional damage on samples during and after the tests.
tests in the range of 101 - 104 s−1 using
inner diameter of the rear supporter is 25.4 mm, 0.4 mm larger than the diameter of the incident bar to accommodate sh
a pneumatic
deformation. hydraulic
The incident bar serves asmachine,
the punch headwhich cansupporter
and the rear neglect the wave
is attached to the transmitted bar. The o
diameter of the entire holder is 57 mm.
propagation effects. Huang et al. [58] created a dynamic punch shear method in SHPB to measure
APPLICATIONS TO LONGYOU SANDSTONE
the dynamic shear strength of rocks (Figure 2.8a). In the dynamic design, the sample assembly is
Sample preparation
composed of a front cover, a disc sample and a rear supporter (Figure 2.8b). The purpose of the front
cover is to reduce the bending force during test and to prevent additional damage to the specimen
after test. The inner diameter of holder is 0.4 mm larger than that of bars. A Teflon adaptor is used to
connect the rear holder to the transmitted bar [58]. The hole in the rear supporter recovers the sample
after the shear test.
2.4 Ultra High Rate of Loading

Fig. 4. a) Typical virgin and tested samples; b) The rock ring and rock plug produced in a typical dynamic punch shear te
(the unit in the picture is centimeter).
In very high strain rates ( > 105 s−1 ), experiments have been done in the atomic scale
to determine the strength of materials. For example, Bringa et al. [60] utilized the large-scale
non-equilibrium molecular dynamics method to obtain the shear strength of copper at strain rates
higher than 109 s−1 . They validated the MD simulations with dynamic experimental data at the same
spatial and temporal scales, both at the lattice level and globally. Murphy et al. [61] used uniaxial
shock compression test at 100 GPa to measure the shear stress and shear strain in a single crytal
copper at 1010 s−1 strain rate. They took the advantage of X-ray diffraction method and VISAR
(velocity interferometer system for any reflector) measurement to study structure of the material
and find the information about the strength of the material directly. X-ray diffraction is a powerful
experimental technique for characterizing the state of shock-compressed material, because it gives
14
direct experimental information on crystal strain as a function of applied loading (pressure), from
which strength can be inferred [62]. In addition to experimental procedures, theoretical methods
and non-equilibrium molecular dynamic (NEMD) simulations have also been developed to calculate
the materials behavior at high strain rates. According to these studies, the strength of the material
can increase dramatically (to ∼ 1 GPa) for extreme strain rates. Comley et al. [62] also used X-ray
diffraction to measure the strain state of the compressed crystal and reached up to 35 GPa of shear
strength at a shock pressure of 181 GPa.
15
Chapter 3
Modeling and Finite Element Simulation

of High Velocity Impact for a Single
Micron-Scale Particle
3.1 Introduction
In this chapter, the characteristic of finite element model, used to analyze the single
particle impact is expressed in detail. First, the creation of geometry by using Python scripts and the
advantages of explicit time integration in simulating the dynamic impact are explained. Then, the
element type used in simulations and features associated with this type of elements are discussed.
In this section, the criterion and circumstances of element failure are also expressed. Next, the
properties of contact between the particle and substrate are defined. Then, the boundary conditions
and the use of symmetry in the problem are presented. In the Section 3.2.7 temperature dependent
material properties and its effects in simulation are covered. Study on the mesh independence in
single particle, dynamic impact simulations is also conducted. Finally, the chapter is finished by
explaining the single particle impact experiments, preformed at UMass Amherst University under
supervision of Dr. Jae-Hwang Lee [10].
16
CHAPTER 3. MODELING AND SIMULATION
3.2 FEM Model
Supersonic impact of a spherical particles on a flat substrate has been simulated by using
commercial finite element analysis software, ABAQUS/Explicit 6.13-2 with Lagrangian formula-
tion. Fully coupled thermal-stress analysis was performed with the brick elements, that have both
displacement and temperature degrees of freedom. The interactions between the particle and the
substrate is defined by using the general contact algorithm. The properties of this interaction includes
the definition of friction in tangential behavior, hard contact in normal direction, heat generation due
to mechanical interaction and cohesive behavior. In the following sections, each feature is explained
in detail.
3.2.1 Modeling and analysis technique
The whole process of modeling an impact simulation is controlled by a Python script. The
script creates the non-dimensional 3D model, assigns material property, defines contact, meshes the
parts, launches the analysis, and gets all the necessary results from output files. In this study only a
quarter of the whole shape is simulated due to the symmetry. The particle is assumed to be a perfect
sphere and the substrate a perfect cylinder. In the Python script, all dimensions are with respect to the
particle diameter. For example the height and the radius of the cylinder is 12.5 times larger than the
particle diameter, and the mesh size is 25 times smaller than particle diameter. Schematic isometric
view of the assembled model can be find in Figure 3.1.
rp
25 rp
Figure 3.1: Isometric view of the developed 3D model.
17
The explicit dynamics analysis procedure uses the central difference integration rule to
integrate the equations of motion for the body, so that’s why it is conditionally stable without
adding any kind of damping to the system. An explicit dynamic analysis is computationally efficient
for the analysis of large models with relatively short dynamic response times. It is also stable in
complicated contact problems, large deformation and inelastic dissipation (to generate heat from
plastic deformation) with respect to implicit dynamic analysis which uses the direct-integration
algorithm. However, the accuracy of the results needed to be checked due to some simplification
in formulation of dynamic explicit algorithms. In the case of impact simulation, it is reasonable to
use the explicit package because this phenomena happens in a relatively short time, it involves large
deformation due to high velocity impact and it is a complicated contact problem.
3.2.2 Element Type and Failure Criterion
The element type used in this work is C3D8RT. This is a first order, 3D continuum element
with 8 nodes and one integration point. The displacement and temperature variation is linear along
each direction. The elements with reduced integration are referred to uniform strain or centroid strain
elements. Reduced integration elements decrease running time of simulations, especially in three
dimensional problems. However, the elements can be distorted in such a way that the strains at the
integration point become zero, which in turn, leads to uncontrolled distortion of the mesh. This
problem is known as Hourglassing [63]. In Abaqus/Explicit, first-order, reduced-integration elements
have the capability to control this issue. Hourglass control, attempts to minimize the excessive
distortion without introducing extra constraints on the element’s physical response.
It was mentioned that particle impact at supersonic velocities, involves large deformation,
therefore, it is necessary to add a damage control mechanism for the model. The progressive damage
and failure for ductile metals involves damage initiation and damage evolution. Damage initiation can
be detected by using shear strain, forming limit diagram (FLD) [64], or Marciniak-Kuczynski (M-K)
[65] criteria. After damage initiation, the material stiffness is degraded progressively according
to the specified damage evolution response. The progressive damage models allow for a smooth
degradation of the material stiffness, which makes them suitable for dynamic situations. In this
study shear criterion is utilized to determine the onset of damage due to shear band localization. The
P ∆ε̄p
criterion for damage initiation is met when w = ε̄p f is equal to one. ∆ε̄p is an increment of
the equivalent plastic strain (PEEQ), ε̄p f is the failure shear strain and w is the damage parameter
which increases monotonically at each increment during the analysis.
18
described as a function of both the driving imperfections and the homogeneous response of a work-
hardening, strain-rate hardening and thermally softening material. We will describe a simple generalization
of this scheme to formulate a macroscopic failure model for damage due to shear localization and the
incorporation of this more general model within the context of finite element analysis techniques. We will
exercise the model by way of example problems designed to highlight strengths and weaknesses in the
current formulation and to understand its appropriateness for describing failure for a certain class of
materials
CHAPTER and 3.
driving conditions.
MODELING AND Finally we will summarize our model and describe some of its current
SIMULATION
assumptions as well as general directions for further development.
Observations of material failure during explosive and impact problems, often shows
2. Summary of a criterion for adiabatic shear
localized deformation that may either dominate or contribute to failure. This mechanism, which
isThe ideas,
called upon which
adiabatic shearthe numericalhas
instability, model
beenfor initiation
known of an adiabatic
and studied for many shear band
years [7].inFigure
this paper
3.2 is
based, were developed in two papers some eight to ten years ago, (Wright, 1992, 1994) and have also been
illustrates
discussed in the typicalcomprehensive
a recent response of a summary
work-hardened material.
(see Wright, In slow
2002). loading,
However, we (no
onlyadiabatic
recently heating)
recognized
that the earlier papers actually contained the beginnings of an idea that could be developed into a com-
the material may show continuous hardening out to large shear strains, as indicated by the upper curve
putational failure model. These ideas will be outlined in this section; for further details, the reader may refer
tointhe original
Figure 3.2.papers.
However, plastic work involves heat generation and as a result of that the temperature
Consider the dynamic response of a thin-walled tube in torsion, as experienced by a specimen in a
of the material
Kolsky increases.
bar experiment. Fig.Since metals tend
1 illustrates thetotypical
be softened by increasing
response temperature,material.
of a work-hardening eventuallyIntheslow
loading,
flow stress will reach a maximum at γmax , followed by strain softening, as indicated schematically to
or if adiabatic heating is somehow suppressed, the material may show continuous hardening out
large shear strains, as indicated by the upper curve in the figure. However, plastic working will heat the
by the middle
material, curve. metals
and because In a perfect
tend tomaterial
soften with
with perfectly
increasinguniform distributions
temperature, of stress,
eventually the flowstrain, andwill
stress
reach a maximum
temperature, softening at c max may , followed by strain softening, as indicated schematically by the middle
stress continue indefinitely. In a work hardened material, the adiabatic stress, curve.
In a perfect material with perfectly uniform distributions of stress, strain, and temperature, softening may
reaches indefinitely.
continue a maximum at some temperature and plastic strain.
Fig. 1. Shear stress versus nominal shear strain for a typical work-hardening material during a torsion experiment.
Figure 3.2: Shear stress versus nominal shear strain for a typical work-hardening material during a
torsion experiment [7].
The considered material failure in these simulations depend on the local equivalent plastic
strain. A constant failure strain, εf , is used as the criterion for material failure. Different values
of εf were tested by Yildirim et al., [8] to find an appropriate one which prevents excessive mesh
distortion but does not alter impact behavior significantly. They observed that (Figure 3.3) when the
failure shear strain is in the interval of [2, ∞) impact behavior is not significantly altered and mesh
distortion is greatly reduced.
Element removal is also activated in the numerical simulation when the shear failure
criterion is met at an integration point. As a result of failure, all the stress components will be set to
19
Figure 6.2: Deformed mesh structure of the particle and the substrate (a), and rebound velocity as
Figurea function
3.3: rebound velocity
of particle impactasvelocity
a function
(b) foroffailure
particle εf = 1,velocity
impact
strains εf = 2, and
fornodifferent failure strains.
failure condition
(εf = ∞). Particle and substrate are both copper.
This analysis conducted by using Copper as a material properties for both particle and substrate [8].
Investigated parameters and material combinations are provided in Table 6.2. Oxygen free,
zero and
highthat integration
conductivity point copper,
(OFHC) fails. Inaluminum
FEA, when all of the
1100-H12, integration
316L points at any
steel, commercially one
pure section of
(CP)
an element fail, the element is removed from the mesh. In the case of first-order reduced-integration
titanium and molybdenum were considered in this work as particle and substrate materials. These
solid element, removal of the element takes place as soon as its only integration point fails.
materials were selected as they differ in key properties, as follows: Copper is a low strength and
There is also another method to control the excessive distortion in simulation which is
high density material; aluminum is a low strength and low density material; steel and
mostly utilized in metal forming simulations and it is called arbitrary lagrangian eulerian (ALE).
molybdenum
Yildirim et al., [8]are
andhigh strength
Assadi and[22]
et al., highreported
density material; and titanium
that, adaptive is a highcauses
remeshing, strength and low shape
unphysical
of the out-flowing jet of material at the interface which is far from reality. Therefore, in this study
adaptive remeshing is ignored.
3.2.3 Interaction and Contact Properties
The general contact algorithm is used for modeling the contact and interaction between the
particle and substrate. For the normal behavior of interaction properties, hard contact is specified as
a pressure overclosure relationship, which minimizes the penetration of the particle surface into the
substrate surface at the constraint locations and does not allow the transfer of tensile stress across
the interface. When the impact happens and surfaces are in contact, the contact pressure can be
transmitted between surfaces. The surfaces can also separate if the contact pressure reduces to zero.
Separated surfaces could come into contact again if the clearance between them reduces to zero.
In the tangential direction, the basic form of the Coulomb friction model is utilized for the
friction. In this model, two contacting surfaces can carry shear stresses up to a certain magnitude
20
across their interface before they start sliding relative to one another. It is also assumed that the
friction coefficient is the same in all directions (isotropic friction).
In fully coupled temperature-displacement simulations, heat can be generated due to the
dissipation of energy created by the mechanical interaction of contacting surfaces. The source of heat
in these kind of analyses is frictional sliding and it is distributed between the interacting surfaces of
particle and substrate equally.
The definition of cohesive behavior is utilized to study the bonding/rebound behavior of
particles impacting a substrate. This approach introduces an interfacial cohesive strength parameter
(σc ) into the finite element model in order to model the bonding upon impact. Traction-separation
relationship is assumed for the behavior of surfaces in contact. This theory assumes a linear elastic
behavior followed by the initiation and evolution of damage in the interface. The initiation occurs
at (δc , Tc ) and the evolution can be predicted by different kinds of relations, however, in this study
it is assumed that this relationship is linear until separation reaches its ultimate value δf . Figure
3.4 shows a typical traction-separation response with a linear failure mechanism. In the case of
impact simulations, deboning happens when the kinetic energy of the particle, overcomes the fracture
energy or energy dissipated due to failure Gc . Although, in most sections of this study the physical
bonding has not been taken into the account and the particles are rebounding after impact, this option
is available in the Python script and can be switched on or off depending on the application.
Traction
Tc
Gc
δc δf Separation
Figure 3.4: Traction separation response of interacting surfaces with a linear failure mechanism.
3.2.4 Boundary Conditions
Based on the symmetry in the problem, only a quarter of the complete shape is modeled.
Therefore symmetry constraints are applied on each side of the sectioned body. The bottom part of
21
the substrate is also set to be fixed. Both particle and substrate are initially at room temperature. The
particle has an initial velocity which is the same as impact speed in vertical direction. In this study
the gravitational force acting on the particle is ignored.
One of the demanding aspects of high velocity impact simulations is the possibility of
wave reflection from the boundaries and its effect on the solution. It has been also mentioned by
Assadi et al., [22] that the waves induced by the particle impact could have a notable effect on the
results. In our simulations the substrate needs to be large enough that the reflected waves do not
return the interface of substrate and particle during impact. In addition, the duration of the numerical
analysis needs to be set long enough to make sure that the particle will detach completely and the
total energy in the system will become stable. Thus, by taking both of these points into account,
the radius and the height of the cylindrical substrate are set to be 25 times larger than the radius of
particle, with this size, the simulation time can be up to 200 ns [8]. With such large substrate the
reflected waves reach the impact zone only after the rebound of particle.
3.2.5 Material Properties
The properties of material used in the simulations are provided in this section. All material
properties used in the finite element simulations were taken from material databases and the literature
[9, 15, 14]. Manes et al., [13, 66] calculated Bilinear Johnson-Cook parameters for Aluminum 6061
using the experimental and theoretical data. This is done by fitting Bilinear Johnson-Cook model
constants (A, B, n, C1 , C2 and m) to the experimental data and using nonlinear multiple regression.
All material properties are temperature dependent. Figure 3.5 shows the variation of each property
with respect to temperature for Aluminum 6061. The respective table for these properties can be find
in Appendix B. Table 3.1 depicts the material properties of Sapphire as a function of temperature.
Note that the density is the only material property that is not defined as a function of temperature due
to some limitation with the explicit package of ABAQUS software.
3.2.6 Adiabatic Shear Instability
Studies show that thermal softening cause shear instability at the interface of contact. This
could be the reason for bonding between the particle and substrate [22]. According to Bever et al.[67]
a majority of the plastic work at the tip of a propagating crack is converted to heat while a small
percentage is actually stored in the material due to dislocation interaction [67]. Numerous studies
have been conducted to measure the amount of heat generated due to plasticity. For instance, Rosakis
22
Appendix A
The following five figures represent the temperature dependent material properties for Al-6061,
extracted from MPDB database [9] and implemented in Abaqus simulations.
a) b)
c) d)
e)
Figure 3.5: Temperature dependent material properties for Aluminum 6061 [9]. Properties include
a) Elastic Modulus; b) Poisson’s Ratio; c) Thermal Expansion; d) Thermal Conductivity; and e)
17 | P a gHeat.
Specific e
23
Table 3.1: Temperature dependent material properties of Sapphire [14].
20◦ C 500◦ C 1000◦ C

Density, kg/m3 3980
Elastic Modulus, GPa 416 390 364
Poisson0 s ratio 0.231 0.237 0.244
Thermal conductivity, W/mK 33 11.4 7.22
Specific heat, J/kgK 755 1165 1255
Thermal Expansion coefficient, 10−6 K−1 4.6 7.1 8.1
et al. [68] used the Split Hopkins Pressure Bar in conjunction with an infrared detector to investigate
the fraction of plastic work converted to heat. In their experiments, the SHPB loads the specimen
dynamically with a controllable strain rate. Loading time is very short, so it can be assumed that the
impact is totally adiabatic. Since there is no time for conductive, radiative or convective heat loss, it
is expected that the dynamic loading apparatus will provide highly accurate measurements of the
converted plastic work fraction.
Assuming adiabatic conditions the converted plastic work fraction, β, can be defined as
follows [68],
ρcp Ṫ = β Ẇp (3.1)
The left hand side is the amount of heat flux added to the system, ρ is the density, cp is the specific
heat capacity, T is the temperature, ( ˙ ) represents the time derivative and Wp is the plastic dissipated
energy which can be defined as,
Z t
Wp = σ ε̇p dt (3.2)
0
Based on the literature [68, 69, 70], this inelastic heat fraction is in the range of 85% to
95% depending on the material and it is mostly constant in high strain rate applications.
3.2.7 Effect of Using Temperature Dependent Material Properties
In finite element simulations the obtained results rely on the accuracy of the material model
and material properties. One of the key factors that alters the behavior of material response is heating.
Therefore, the temperature dependence of material properties should be considered in the simulations.
24
The material model that is used for simulating the particle impact is the Bilinear Johnson-Cook
model. As presented in Chapter 4, temperature is one of the main factors that helps softening of
the stress strain relationship in this material model. In the following, the role of using temperature
dependent material properties is discussed.
Figure 3.5 shows the variation of each material property (elastic modulus, Poisson’s ratio,
thermal expansion, thermal conductivity and specific heat) with respect to temperature for Aluminum
6061. On the other hand, Table 3.2 depicts the material properties of Al-6061 at room temperature.
The following is an analysis of the temperature dependent material response for Al-6061. Other
simulation properties such as substrate material properties, particle diameter, impact velocity, and
inelastic heat fraction are kept constant in order to focus on the effects of temperature dependent
material properties in simulations.
Table 3.2: material properties for Aluminum 6061 at room temperature.
Density, kg/m3 2700

Elastic Modulus, GPa 68.9
Poisson0 s ratio 0.33
Thermal conductivity, W/mK 220
Specific heat, J/kgK 904
Thermal Expansion coefficient, 10−6 K−1 23.6
The simulated deformed shape of the particle after impact is illustrated in Figure 3.6. In
these simulations, the diameter of the particle was 19.75 µm and impact velocity was 663 m/s. The
particle was made of Aluminum and substrate from Sapphire. In this picture the red curve represents
the deformed shape of particle based on constant material properties and the blue curve represents
the same shape but temperature dependent material properties. Both shapes are nearly identical,
however, the red curve deformed a bit more. It was observed taht as the impact velocity increases,
the difference between two curves becomes more and more evident.Here in Figure 3.6 it is seen that
the red curve (temperature independent material properties) spreads more in the vertical direction
and compressed more in the longitudinal direction. This kind of deformation was not expected from
a material using temperature dependent material properties (blue curve), since their strength should
be degraded due to temperature softening which causes more deformation.
In order to answer this apparent dilemma an in-depth analysis has been carried out. Figure
3.7 is the time history analysis of an impact simulation. Blue and red plots represent data for the
25
bilinear temp dep

bilinear temp indep
15
10
Y
5
−5
−10 −5 0 5 10
X
Figure 3.6: Deformed shapes of particle using bilinear JC material model for 19.75 µm particle and
impact velocity of 663 m/s.
temperature dependent and temperature independent material properties, respectively. Based on the
time history plots the average values for strain and strain rate of temperature independent material
properties simulation are slightly higher (Figure 3.7 a, b). On the other hand, the temperature plot
shows that average temperature increases slower with temperature independent material properties
(Figure 3.7 c).
Considering the Bilinear Johnson-Cook relationship explained in Section 4.3, higher strain
and strain rate can help the hardening of material resulting in a higher yield stress. However,
increasing the temperature of material results in softening and decreasing the yield stress of material.
Depending on the intensity of temperature effect or the strain and strain rate effect, the final result
of material strength could be harder, neutral or softer as compared to the simulation with constant
material properties. In the case of Aluminum 6061, the combination of strain hardening, strain rate
hardening and temperature softening, causes the temperature dependent material properties to behave
slightly harder. This idea can also be shown by considering the average yield stress of the material
(Figure 3.7 d). The average yield stress is slightly higher for the case of temperature dependent
material properties. That’s why particle using temperature dependent material properties deformed
slightly less in Figure 3.6.
Previously, it was mentioned that the particle’s temperature increases slowly in the case of
using temperature dependent material properties. This is due to the effect of specific heat. Based
on Figure 3.5 e the specific heat increases with temperature. This implies that as the temperature
26
increases, the material needs more heat energy to increase 1 unit of temperature. Therefore, for a
given amount of plastic work hardening, as converted to heat, the increase of temperature is relatively
lower with respect to the case of temperature independent material.
a) b)
c) d)
Figure 3.7: Time history analysis of particle for a) average strain; b) average strain rate; c) average
temperature; d) average yield stress. Note that the particle diameter was 19.75 µm and impact
velocity was 663 m/s. Average values are the mean of all integration point at each increment time.
27
3.3 Mesh Convergence Study
It is well-known that element size strongly affects the results of numerical simulations.
Therefore, in order to find an appropriate mesh size, a mesh convergence study was carried out for
single particle impact. A very dense mesh has been generated on the impact area and the element
sizes have been increased getting far from the impact zone. ”C3D8RT” linear hexahedron, 8 node,
reduced integration elements have been used to mesh the particle and the substrate. The size of the
elements in the impact zone area and particle have been chosen as functions of the particle diameter
which are dp /10, dp /15, dp /20, dp /25, and dp /30. The results are summarized in Table 3.3.
Table 3.3: Mesh convergence study for single particle impact simulation.
Mesh Total no. Simulation Space Vr

size of Nodes time, mins required, GB m/s
dp /10 36,074 2 0.47 20.6

dp /15 120,191 5 1.56 23.62
dp /20 283,083 14 3.68 19.7
dp /25 535,311 35 6.95 19.3
dp /30 928,170 81 12.05 19.0
It was observed that the energy dissipated by damage and removing the failed elements are
relatively high due to large element sizes in first and second cases. As the element size decreases this
energy becomes more stable. In addition, Table 3.3 shows that the rebound velocity results converged
in the last three cases. Also, by considering the computational time and disk space required for a
single analysis, the average element size near the impact zone in the substrate and for the particle
was selected as dp /25. Moreover, the experimental results also reveal that the rebound velocity for
this case is 19.2 m/s which has a good agreement with the selected mesh density case.
3.4 Laser Induced Single Particle Impact Experiments
Laser-induced projectile impact test (LIPIT) [10] is illustrated in Figure 3.8, used to
accelerate a micron scale particle , to study the high strain rate phenomenon in particle impacts. In
this experimental setup an excitation laser pulse is generated by a pulsed laser system that has an
optical parametric oscillator. The pulse, results in ablation and expansion of the gold film in the form
28
of gas. The polymer layer, which is between the gold layer and a particle expands and ejects the
particle. Those particles can accelerate up to 4 km/s. Figure 3.9 illustrates the relation between the
pulse energy and particle velocity [10].
Figure 3.8: Diagram of the optical setup for LIPIT [10].
Figure 3.10 shows images of a particle moving toward a substrate. The velocity of particles
are measured by capturing their positions with two pulses that have a specific phase delay.
The excitation laser pulse generates a supersonic gas plume surrounded by a shock front
trailed by the particle. Deceleration of the particles by air is negligible until the particles overtake
the shock front. Therefore, by placing a target sample at a position where the shock front arrives
before the particle, the flight speed can be used as the impact speed. As demonstrated in Figure 3.9
the kinetic energy of the particles is linearly proportional to the input energy after subtracting the
energy used for ablation. More information on the experiment can be found in Lee et al. [10, 71].
29
OMMUNICATIONS | DOI: 10.1038/ncomms2166
c
v = 2.3 (E –0.08)0.5 S
Speed of beads, v
Secondary 2
delay P
(km s–1)
group
Speed of sound
/2
0
Plate
0.0 0.5 1.0
Excitation pulse energy, E (mj)
Figure 3.9: particle velocity and excitation pulse energy relation [10].
d
Beam Polarizer Primary delay
splitter (optical fiber)
Capability of the UMass micro-ballistic system
A B
Target Excitation pulse
= 800 nm
12-exposure
micrograph Δx=45.9
Probe μm
pulse 3.4 km s–1
Δt=100 ns
= 400 nm
To microscope v=459 m/s
Mach cone
Laser-induced projectile impact test. (a) Diagram of the optical setup for LIPIT. (b) The speed of the m-projectiles is prop
excitation pulse energy for a given thickness of the absorbing polymer layer. The error bar represents the maximum and
n value and the green line is a fitting curve. (c) Double exposure photograph with a 22-ns interval at 250 ns after the exci
200 mm. (d) Expanded section of (c). Here, each m-projectile appears twice; for example, the m-projectile marked as ‘A’ is
‘B’ and the corresponding speed is 3.4 kmTrajectory: 0.51 mm
s 1. Only the m-projectiles outside of the shock front present a Mach cone,
eed exceeds the speed of sound. The sizes of the m-projectiles appear differently depending on their distance to the micr
bar is 100 mm. Δx=47.3 μm
Δt=100 ns
v=473 m/s
Figure 3.10: multiple exposure photograph of a particle with constant ∆t [11].
30
3.5 Summary and Conclusions
A 3D finite element analysis has been developed in order to predict the deformed shape
of a particle impacting a substrate. Abaqus/Explicit 6.13-2 has been used to perform the numerical
analysis. Python scripts are used to create the input file, post-process the Abaqus outputs and preform
a series of numerical tests to accurately detect the outer shape of particle. In this chapter, it is found
that temperature dependent material properties for Aluminum 6061 does not change the deformed
shape of particle significantly. Furthermore, the mesh convergence study revealed that reasonable
convergence can be obtained by using an element size of dp /25 for the particle and in the impact
zone of the substrate.
31
Chapter 4
Material Model for High Strain Rate

Deformation of Metals
4.1 Introduction
This chapter gives a brief overview of different constitutive laws which can predict the
behavior of material at very high strain rates.
4.2 Background
One of the most important requirements to mimic a real world application in a numerical
analysis is to have an accurate constitutive model. The goal of this study is to find a proper
material model which can model material behavior in cold spray application. Particle spraying is a
dynamic phenomena which involves large and nonlinear deformation, high strain rate plasticity, and
temperature spike at the impact interface which could cause local melting and bonding between a
particle and substrate. An appropriate constitutive law should consider all of these effects at the same
time. There are a few material models that can predict the behavior of material in such conditions,
including Johnson-Cook (JC) [72, 73], Zerilli and Armstrong (ZA) [74, 75], Voyiadjis and Abed
(VA) [76], Preston-Tonk-Wallace (PTW) [77], Khan-Huang-Liang (KHL) [78, 79, 80], Gao and
Zhang (GZ) [81]. In the following, a brief introduction to each material model is given.
32
CHAPTER 4. MATERIAL MODEL FOR HIGH STRAIN RATE DEFORMATION OF METALS
4.2.1 Johnson-Cook Model
Johnson and Cook [72, 73] proposed this material model in 1983. This is a phenomeno-
logical model which is widely used in most computer codes for static and dynamic analysis, such
as; metal cutting [82, 83], cold spray [8], crash tests [84] and etc. It is relatively easy to calibrate in
the experiments and can predict the behavior of the material flow stress at different strain rates and
temperatures.
Johnson-Cook model was the first model that defined the flow stress as a function of plastic
strain, strain rate, and temperature. Equation 4.1 is the definition of flow stress based on Johnson and
Cook. Here A, B, n, C, and m are material constants, εp is the equivalent plastic strain, ε˙p is the
plastic strain rate, ε˙0 is the strain rate at which the material constants are obtained and it is known as
reference strain rate. T ∗ is the homologous temperature and defined in equation 4.2, where, Tm is
the melting temperature, Tr is the reference temperature, and T is the absolute temperature.
!
ε˙p
σ = A + Bεnp 1 + C ln 1 − T ∗m (4.1)
ε˙0

 0

 T ≤ Tr
∗ T −Tr
T = Tm −Tr Tr < T < Tm (4.2)


1 T ≥ Tm

This model has been used for simulations to predict the dynamic behavior of materials in the literature
regularly. The advantage of the JC model is that the strain, strain rate and temperature are considered
at the same time and they are directly proportional to the flow stress predicted by this model. However,
it has been indicated that the model may not be accurate when the strain rate of the system is over
104 s−1 [81, 85, 78]. This inaccuracy is somewhat related to experimental limitations. For example
in Hopkinson pressure bar test, the strain guages cannot capture the rate if it is greater than 104 s−1 .
Therefore these JC material constants calibrated to predict the behavior of materials in a specific
range. In addition, the dependency of work hardening on the logarithm of strain rate is linear in this
model which is a simplified relation. For these reasons, this model is not usually able to accurately
predict the material behavior at very high strain rates.
33
4.2.2 Zerilli-Armstrong Model
The initial Zerilli-Armstrong model is a physically based model proposed for different
metals, that are sensitive to strain rate and temperature [74, 75]. The ZA model has different forms
of constitutive relations for body-centered-cubic (BCC) materials and face-centered-cubic (FCC)
materials. The constitutive equation for BCC materials such as tantalum is expressed as follows,

σ = C0 + C1 exp − C3 T + C4 T ln ε˙p + C5 εnp (4.3)
where, C0 , C1 , C3 , C4 , C5 , and n are material constants, εp is the equivalent plastic strain, ε˙p is the
plastic strain rate, and T is the temperature. The first term C0 is related to Hall–Petch relation and
equals to σ0 + kd−1/2 , where d is the grain size of the material and k is the micro-structural stress
intensity. In this model, it is presumed that the strain hardening is independent of temperature and
strain rate. The exponential term in equation 4.3 is used to describe the thermal stress component
based on experimental observation. This definition is inappropriate as the thermal stress component
goes to zero only when the temperature tends to infinity. It is noted that, the thermal stress component
must disappear at the melting point of material. In order to resolve this issue, the model was modified
by Abed et al. [86].

σ = C2 εnp 1 − X 1/2 − X + X 3/2 + C6 (4.4)

ε˙0
X = C4 T ln (4.5)
ε˙p
where, C2 , C4 , C6 are material constants, ε˙p is the plastic strain rate, and ε˙0 is the reference strain
rate.
The modified ZA model generally improves the results at temperatures above 300 K.
However, the work hardening is expressed independent of temperature and strain rate just like the
initial ZA model. This is why these two models are not appropriate to model the materials whose
responses are strongly depended on temperature and strain rate.
34
4.2.3 Voyiadjis-Abed Model
To improve the prediction capability of flow stress at high strain rates and temperatures,
Voyiadjis and Abed [76] modified the ZA model as follows,
1/p
1/q
σ = Ŷ 1 − (β1 T − β2 T ln ε̇p ) + Bεnp + Ya (4.6)
where, Ŷ , β1 , β2 , Ya , B, p, q and n are material constants, ε˙p is the plastic strain rate, ε˙0 is
the reference strain rate and T is the temperature. The last two terms are athermal components
of flow stress and they are the same forms as the ZA model in equation 4.3. The first term is
thermal flow stress and it is related to the strain rate and temperature. It was modified from the
ZA model and derived by using the concept of thermal activation energy as well as the dislocation
interaction mechanism where, the mobile dislocation density evolution was also taken into account.
By modifying the thermal component of flow stress, the prediction capability at high strain rates and
temperatures is improved as they reported in [76].
4.2.4 Preston-Tonk-Wallace Model
This model was developed by Preston et al. [77] to describe the behavior of materials at
very high strain rates. It is a complex constitutive model proposed based on the dislocation motion
during plastic deformation. According to the theory, thermal activation mechanism of dislocations
has a significant influence on the deformation by weak shocks that cause strain rates up to 105 s−1 .
However, the strain rate in explosively driven deformations or in high-velocity impacts is sometimes
much higher than 105 s−1 , and thus the plastic constitutive model based on only the thermal activation
mechanism can result in a significant error. In order to model the material behavior accurately at a
strain rate up to 1012 s−1 , Preston et al. proposed a plastic constitutive model considering nonlinear
dislocation drag effects that are predominant in a strong shock regime. The model is given by
equation 4.7.
  

1 τbs − τby  pθψ 
τb = τbs + s0 − τby ln 1 − 1 − exp − p exp − h i 
p s0 − τby  (s − τb ) exp p τbs −bτy 
0 y s0 −b τy
(4.7)
where, τb is a normalized flow stress and equals to τ /G ( τ is the shear stress and G is the shear
35
modulus ). τbs and τby are the normalized work hardening saturation stress and normalized yield stress,
respectively. The variables, p, q, and s0 are dimensionless material constants. Following equations
define τbs and τby .
( )
β
τbs = max s0 − s0 − s∞ erf κ Tb ln γ ζ̇/ε̇ , s0 ε̇/γ ζ̇ (4.8)
( " )
y2 β
τby = max y0 − y0 − y∞ erf κ Tb ln γ ζ̇/ε̇ , min y1 ε̇/γ ζ̇ , s0 ε̇/γ ζ̇ (4.9)
where the material constants s0 and s∞ are the values that τbs takes at zero temperature and very
high temperature, respectively. y0 and y∞ have analogous interpretations. κ and γ are dimensionless
material constants. Scaled temperature Tb is also defined by T , where the absolute temperature is T
Tm
and the melting temperature is Tm . The parameter ζ̇ is defined by
1/4 1/2
1 4πρ G
ζ̇ = (4.10)
2 3M ρ
where ρ is the density and M is the atomic mass. ζ̇ −1 has the meaning of the time required for
a transverse wave to cross an atom. Shear modulus G is taken to be a function of density and
temperature; G(ρ, T ) = G0 (ρ)(1 − αTb), where α is the G dependency on scaled temperature and
G0 is the shear modulus at absolute temperature. Constants y0 , y1 , y2 , y∞ , s0 , s1 , s∞ , κ, γ, p, and β
are fitted from experiments. In 2009 Kim and Shin [85] modified the existing PTW to be applicable
in wide ranges of strain, strain rate, and temperature.
4.2.5 Khan-Huang-Liang Model
Khan et al. [79, 80] introduced this model based on the JC model. They considered the
work hardening as a coupled function of strain and strain rate. Later, Huh et al. [87, 88] presented
the modified version of KHL model. This model correlated better to the experimental results in
compression with the KHL model. Equation 4.11 represents the flow stress in the modified KHL
model.
!
ln ε̇∗ n1 n0

∗ p
σ= A+B 1− ε 1 + C(ln ε̇ ) (1 − T ∗m ) (4.11)
ln D0p
36
where ε̇∗ = ε̇/ε̇ref , n1 = a (1 − T ∗b ), m = m1 + m2 .lnε̇p and D0p is chosen to be 106 per second.
The material constants of the proposed model have some physical meaning. A is the yield stress at
the quasi static state. B and n0 represent the work hardening at the quasi static strain rate, C and
p are the coefficients of the strain rate hardening. a and b represent the change of work hardening
rate with respect to the temperature and finally, m1 and m2 are the strain rate dependent thermal
softening parameters, respectively.
4.2.6 Gao-Zhang Model
This physically based model was proposed by Gao and Zhang in 2011 [81] for deformation
of material response at strain rates higher than 104 s−1 . As mentioned above, some of the previous
models do not have the ability to predict the correct material behavior (e.g. copper) at strain rates
above 104 s−1 . Unlike the other models, the density of dislocations is not assumed to be constant in
this model. In fact, it is defined as a function of equivalent plastic strain, strain rate and temperature.
The flow stress in this model is given as follows,
σ = σth + σath (4.12)
v
u c1 T ( " −c1 T #)
u ε̇ ε̇ ε̇
σth = C
b t 1 + tanh c0 log 1 − exp −k0 ε
ε̇s0 ε̇s0 ε̇s0
(4.13)
( 1/q )1/p
ε̇
1 − −c2 T ln
ε̇0
h i1/2
σath = σG + B 1 − exp (−ka0 ε) (4.14)
where σth and σath are thermal and non-thermal stress components, respectively. C
b is reference
thermal stress. k0 , c0 , c1 , c2 , p and q are all material constants for thermal stress component. ε̇0 and
ε̇s0 are reference strain rate and saturated strain rate, respectively. σG is the stress due to initial
defects, B and ka0 are material constants for non-thermal stress component.
37
4.2.7 Summary of All Constitutive Models
There are numerous of material models which might predict the behavior of material in
high strain rate applications. Among all, only six of them are described in this section. Some of
them are simple models, while others require many material constants. Determining these constants
are quite complicated and requires a large number of experiments. Rahmati et al. [89] compared
all six different material models mentioned above in his paper. They utilized cold spray experiment
data, and built six different user subroutines for each material model for Abaqus software. Pure
Copper material properties was used for both particle and substrate in simulations. According to this
paper, jetting phenomenon can only be estimated by the JC and the PTW models. VA, modified KHL
and modified ZA models did not predict the flow stress over a wide range of strains and strain rates.
Thus, they cannot predict the cold spray process accurately. GZ model overestimates the flow stress
for copper and cannot anticipate the deformed shapes of particles. This model was also unable to
predict the critical velocity of particles attaching to the substrate. JC model is the simplest model
among all six material models, and it is very popular in high strain rate applications. Therefore,
for almost all metals, material constants have been reported in the literature. However, this model
underestimates the flow stress at very high strain rates. Therefore, for impact simulations using this
criteria overpredict the deformation. In addition, in this model the strain rate and temperature effects
are uncoupled which implies that the strain rate sensitivity is independent of temperature.
4.3 Bilinear Johnson-Cook Model
As mentioned in the previous section, the classic JC material model only works fine in the
low or intermediate strain rates regimes. the bilinear JC model compensates this issue by adding an
extra condition for high strain rates. is really high. Equations 4.15 and 4.16 explain more about this
material model.
!
ε˙p
σ = A + Bεnp 1 + C ln 1 − T ∗m (4.15)
ε˙0

 C ε̇ < ε̇
1 p c
C= (4.16)
 C2 εp ≥ ε̇c
The proportionality constants C1 , C2 and the reference strain rate ε̇0 are determined
experimentally. The coefficient C and the reference strain rate ε̇0 control the strain-rate dependent
38
is related to the hydrostatic component of stress, This hydrostatic stress accelerates void
and that this porosity increase directly affects leads to the eventual coalescence o
material yielding. ductile failure of the bar. Failure was
In these observations it was assumed that the correspond to the loss of load-carrying
material surrounding a void was incompressible. this displacement-controlled simulation.
Gurson8 proposed a pressure-sensitive macroscopic The bar was assumed to have the follo
yield surface that relates void growth to the evolu- ial properties: E = 20.7 GPa, υ = 0.3, yi
CHAPTER 4. MATERIAL tion of microscopic
MODEL FOR HIGH(pointwise physicalRATE
STRAIN quantities of 690 MPa,OF
DEFORMATION withMETALS
a linear hardening modu
the matrix material) and macroscopic quantities to MPa. The initial void fraction was ass
account for the behavior of void-containing solids. equal to 0.050.
Here, macroscopic refers to the average values of The initial and deformed shapes of
hardening behavior. Experiments show that as the plastic strain rate increases,specimen
physical quantities, which represent the material
the relation between
are shown in Fig. 11, which
stress and log of strain rateaggregate behavior.
is not linear anymoreAs defined
(Figure by Gurson,
4.1). the yield the regions of predicted high void growt
surface for a ductile material is: of rate-dependence is shown in Fig. 1
Manes et al. presented these 2 coefficients as shown in Table 4.1. Theyincreasedused the experimental
loading rate resulted in an
q   3p   2
Φ =   + 2q1 f cosh  – 1 + q2f  = 0, normalized axial load (actual axial
results [90, 66] and optimized the  σ 0parameters
  2aσ 0ballistic
for  impact applicationyield
[13].strength), with softening similar
(12) independent Gurson model results.
where σo is the tensile flow stress of the micro- Also shown in Fig. 12 is the conven
scopic matrix material, q and p are the equivalent ticity solution, which does not e
500 1500
●
Prediction Prediction Drag
1400
● Data ● Data contro
450 Drag-control ●
Stress (MPa)
1300
Stress (MPa)
●
400 Discrete- ● 1200
obstacle- ●
controlled ●● ● 1100
● ● Discrete-
350 ● ● ● ●●● ● obstacle-
● ● ● ● 1000 controlled
300 900
10–4 10–2 100 102 104 10–4 10–3 10–2 10–1 100 101 10
Strain rate (s–1) Strain rate (s–1)
Figure 9. Comparison between the stress-strain rate behav- Figure 10.Comparison between the stress-strain
ior predicted by the mechanism-based material model and predicted by the mechanism-based material mo
experimental data for aluminum alloy 6061-T6. Regions of experimental data for titanium alloy Ti-6Al-4V.
the stress-strain rate curve that are dominated by discrete- the stress-strain rate curve that are dominated b
Figure 4.1: Comparison between the stress-strain
obstacle plasticity rate behavior
and drag-controlled predicted
plasticity are shown. by the mechanism-based
obstacle plasticity and drag-controlled plasticit
material model and experimental data for aluminum alloy 6061-T6 [12].
6-12 Engineering Research Development and Technology

Lesuer et al. [12] explained that the deformation in terms of strain rate is controlled by two
sequential processes; the cutting (or bypassing) of obstacles by dislocations, and the drag on moving
dislocations by phonons or electrons. They believe that different rates represent the deformation
kinetics associated with discrete obstacles or drag. The Figure 4.1 shows the regions of the stress
versus strain rate curve that are dominated by discrete-obstacle plasticity or by drag-controlled
plasticity.
Table 4.1: Bilinear Johnson-Cook material properties for Al-6061
A, M P a B, M P a n C1 C2 m Tm ε̇0 ε̇c
JC-4 [66] 270 154.3 0.2215 0.002 0.1301 1.34 925 1 597.2
JC-5 [13] 270 138.2 0.1792 0.002 0.1301 1.34 925 1 597.2
Figure 4.2 shows the flow stress as two functions of strain and strain rate. The definitions
for JC-1 to JC-5 can be found in Tables A.1 and 4.1. Left picture is the classic JC plotted for a
39
JC-5 [3] 270 138.2 0.1792 0.002 0.1301 1.34 925 597.2
Table 1: constants for bilinear JC flow stress

Figure 2 represents
constant temperature theand
flow
the stress as twoisfunctions
right picture of JC.
the bilinear strain
Theand strain rate.
transition fromLeft picturerate
low strain is to
the
classic JC plotted in three different temperatures and the right picture is the bilinear JC. The
high strain rate is clear in the right picture. The jump in this figure can be interpreted as an extra
transition from low strain rate to high strain rate is clear in the right picture. The jump in this
hardening
figure can beforinterpreted
the material.
as an extra hardening for the material.
Figure 2: flow stress in classic JC (left) and Bilinear JC (right)

Figure 4.2: flow stress in classic JC (left) and Bilinear JC (right)
If we consider the simulation one more time, we will understand that in the first 25 ns of the
impact (before rebounding happens) the strain rate of the particle is in its maximum amount.
4.4 Conclusion
2|P a g e
In this chapter various material models used to represent the material behavior of high
strain rates. The introduced Bilinear JC model is going to be used in the next chapter to simulate
impact of Al-6061 particles on Sapphire substrate.
40
Chapter 5
Optimization of Bilinear Johnson-Cook

parameters
5.1 Introduction
The goal of this chapter is to calibrate the material constants in the bilinear Johnson-Cook
(JC) model for Al-6061 particle impact simulations. To this end, a multi-variable parameter fit
procedure was performed by using the results of the impact experiments performed by the UMass-
Amherst group [11] and finite element simulations.
In particular, a modified version of the method of steepest descent was used to find the
values of the six JC-model parameters by fitting the aspect ratio of the experimentally obtained
particle shapes to the simulated ones (Section 5.4). Results were further improved by considering
the possibility of size variability of the original particles (Section5.5). The average JC-parameters
obtained by the parameter fit were shown to be very effective in predicting the deformed particle
shapes for a wide range of 175 – 699 m/s impact velocities (Section 5.6).
The computational process for this section was done in Python and Abaqus. The analysis
is set to be dynamic explicit, and 3D elements are utilized to mesh the particle and substrate. Each
element has 8 nodes and 1 integration point, and all material properties are defined temperature
dependent except the density.
41
CHAPTER 5. OPTIMIZATION OF BILINEAR JOHNSON-COOK PARAMETERS
5.2 Background
In the optimization procedure, temperature dependent material properties are obtained

from the MPDB [9] database software, presented in Chapter 3 were used as mentioned earlier. Tables
5.1 and 5.2 depict other properties used for Aluminum-6061 and Sapphire.
Table 5.1: Material properties of Al-6061 [15, 9].
Density, kg/m3 2700

Reference Temperature, K 293
Melting Temperature, K 925
Plastic heat conversion ratio 0.9
Table 5.2: Temperature dependent material properties of Sapphire [14].
20◦ C 500◦ C 1000◦ C

Density, kg/m3 3980
Elastic Modulus, GPa 416 390 364
Poisson0 s ratio 0.231 0.237 0.244
Thermal conductivity, W/mK 33 11.4 7.22
Specific heat, J/kgK 755 1165 1255
Thermal Expansion coefficient, 10−6 K−1 4.6 7.1 8.1
The Johnson-Cook material model is used to characterize the dependence of the yield
stress σ on the plastic strain εp , rate of plastic strain ε̇p and temperature T , as follows,
! m !
ε˙p T − Tr
σ = A+ Bεnp 1 + C ln 1− (5.1)
ε˙0 Tm − Tr
where A is the static yield stress of material, parameters B, and n are the strain-hardening parameters,
C controls the strain rate hardening, ε˙0 is the reference strain rate, Tr is a reference temperature
and Tm is the melting temperature of the material, and m is the temperature exponent. The classic
JC-model has 5 experimentally determined constants A, B, n, C, m. Recent split-Hopkinson bar
experiments for Al-6061 show that strain-rate hardening increases dramatically at high strain rates
42
[12]. Manes et al. [66, 13] summarized this observation in the context of Johnson-Cook model [72]
by modifying the
A. Manes et al.strain rate
/ Procedia coefficient
Engineering C as follows,
10 (2011) 3477–3482 3481
1.5 1.5
Experimental data Experimental data
Lesure et al. Lesure et al.
1.4 J-C [5] 1.4 Optimized model
1.3 1.3
σ/σst (-)
1.2
σ/σst (-) 1.2
1.1 1.1
1 1
0.9 -4 -2 0 2 4
0.9 -4 -2 0 2 4
10 10 10 10 10 10 10 10 10 10
-1 -1
Strain rate (s ) Strain rate (s )
Fig. 3. Stress-strain rate behavior in the semi-log plane: C=0.002 and ε0 =1 s (left); stress-strain rate behavior in the semi-log
--11
Figure 5.1: Stress-Strain rate behavior in a semi log plot [13].

plane: C=0.1301 and ε0 =597.2 s--11 (right). In both cases the solid line is saturated at the value of 1 because in LS-DYNA ε0
represents the threshold under which the strain rate effects are negligible (the second term of the equation 2 is equal to the unity).
1000

 C ε̇p < ε̇c
Engineering stress (MPa)
800 1
C= (5.2)
 C2 εp ≥ ε̇c
600
C1 and C2 are two experimentally determined coefficients that show the additional increase in the
400
yield stress when the plastic strain rate ε̇p is greater than an experimentally determined critical plastic
200
strain rate ε̇c . In the small strain rate region (ε̇p < ε̇c ), the yield stress is a weak function of strain
Experimental
0
rate with less than 2% effect on the yield stress. On the other hand, the yield stress can increase
Numerical
0 0.1 0.2 0.3 0.4 0.5
more than 200%
Engineering strain (-)in the high0.0
strain rate region. Thus the strain rate effects in the bilinear JC-model
278.2
Fig. 4. Comparison between experimental end numerical stress-strain curves (up to 50% of deformation): ~4600 s-1,
are characterized by C , C and ε̇ . The bilinear Johnson-Cook
~3000 s , material parameters for Al-6061, -1
1 2 c
Z ~1900 s-1, © static (left); temperature increment distribution (K) in the specimen at ~4100 s-1 and 5 mm of stroke.
obtained
For what concerns byratesplit-Hopkinson
the strain bar usually,
sensitivity identification, test areonly
given in Table
the parameter 5.3.
C is considered
as an optimization variable, while ε0 is set equal to 1 s-1. In Figure 3 there is the comparison between the
experimental stress-strain rate 5.3:
Table resultsBilinear
and the Johnson-Cook
J-C model obtainedmaterial
with the properties
strain rate sensitivity
for Al-6061 obtained by SHPB test.
coefficients taken from [5]. The experimental data are shifted according to the requirement of ε0 =1 s-1
and normalized respect to the stress obtained in the test at the lowest strain rate.
A, MPa B, MPa n C1 C2 m Tm ε̇0 ε̇c
270 154.3 0.2215 0.002 0.1301 1.34 925 1 597.2
43
5.3 Optimization Process
Express the bilinear-JC model parameters in the following vector,
n o
x = A, B, C2 , m, n, ε̇c (5.3)
Define the ellipticity ratio R for the deformed particle as follows,
D1
R= (5.4)
D2
where the dimensions D1 and D2 are demonstrated on a deformed particle in Figure 5.2. The
ellipticity ratio can be defined for the experimental measurements (Re ) and for the simulations
(Rs ). Note that ellipticity ratio of the simulations Rs depends on the chosen JC parameters, that is
Rs = Rs (x). The following objective function (normalized least square error) is defined to measure
the error between the two ellipticity ratios,
(Re − Rs )2
E(x) = × 100 (5.5)
Re2
For a given deformed particle (e.g. Figure 5.2) the objective (error) function is minimized by using a
modified version of the method of steepest decent [91, 92]. The x vector which minimizes the error
(Equation 5.5) is found iteratively. The values given in Table 5.3 are used as initial conditions. The
gradient of the objective function is determined numerically with respect to each parameter xi of the
x vector. The parameter xi which should be optimized is identified by comparing all the gradients.
The one with the maximum absolute value is modified as follows,
(k+1) (k) (k)

xi = xi −n
b i xi (5.6)
where the correction coefficient is determined from,

1 ∂E
n
b=r (5.7)
PN ∂E 2 ∂xi
i=1 ∂xi
44
and N is the length of the x vector. The procedure outlined by equations 5.6 and 5.7 is repeated
until the error is minimized for each element of the x vector. Figure 5.3 illustrates the optimization
process.
D1
D2
Figure 1 : SEM images on which the parameter fitting is based [4], and the definition of the dimensions D 1 and D 2 used in
calculating the ellipticity ratio R e .
Figure 5.2: SEM images on which the parameter fitting is based [11] and the definition of the
dimensions D1 and D2 used in calculating the ellipticity ratio Re .
5|Page
45
‐ Slightly Modified Method of Steepest descent is used for optimization process
Initial Values for JC parameters
Calculate , ∈ , , , , , ,
Find critical parameter: Max{ }

N
Is a local
Y Remove this parameter All parameters
minimum? from x vector removed from x?
N Y
Update the critical parameter
END
,
∑
Figure 5.3: Optimization process with modified steepest descent method.
5.4 Optimization Results
In this section, the optimized parameters for the bilinear-JC model are presented. UMass-
Amherst experiments that we considered (Figure 5.2) consist of six different particles impacting a
sapphire substrate at six different velocities. The reported particle diameters Dr were determined
from SEM images before impact to be 20.75, 24.4, 24.3, 22.7, 19.75 and 23.4 µm. The impact
velocities of these particles were measured to be 175, 286 416, 530, 663 and 699 m/s, respectively.
Particles were collected post impact and their morphology was imaged by SEM . We digitized these
images by using Matlab. The ellipticity ratio of Re was determined from these images. Details of the
optimization procedure outlined in the previous section are given below for each case.
Figure 5.4 shows the computed and measured particle shapes by using the bilinear-JC
model parameters reported by Manes et al. [66] (Table 5.3). The detailed progression of the
optimization procedure is outlined in Table 5.4 to Table 5.9. The columns marked in green, indicate
the parameter on which optimization is performed. For example in Table 5.9, attempt-1 identified C2
as the critical parameter, followed by n in attempt-2. All parameters converged to a very tight error
tolerance value after attempt-3.
46
Table 5.4: Bilinear Johnson-Cook parameters at each attempt of the optimization process for the
case-1, where Dp = 20.75 µm, and Vp = 175 m/s.
A, MPa B, MPa n C2 m ε̇c E(%)
initial 270 154.3 0.2215 0.1301 1.34 597.2 0.1482

Attempt1 270 154.3 0.2215 0.027 1.34 597.2 1.47E-04
Attempt3 270 154.3 0.2215 0.027 1.74 597.2 1.90E-06
initial 270 154.3 0.2215 0.1301 1.34 597.2 1.536

Attempt1 270 154.3 0.2215 0.011 1.34 597.2 1.21E-04
Attempt3 270 154.3 0.2215 0.011 1.44 597.2 2.02E-06
initial 270 154.3 0.2215 0.1301 1.34 597.2 3.122

Attempt1 270 154.3 0.2215 0.027 1.34 597.2 1.61E-03
Attempt3 270 154.3 0.24 0.027 1.34 597.2 5.50E-04
initial 270 154.3 0.2215 0.1301 1.34 597.2 2.472

Attempt3 270 154.3 0.2215 0.055 1.34 597.2 2.43E-04
47
initial 270 154.3 0.2215 0.1301 1.34 597.2 15.103

Attempt1 270 154.3 0.2215 0.03 1.34 597.2 0.0759
Attempt2 270 154.3 0.2215 0.027 1.34 597.2 0.0034
Attempt3 270 154.3 0.26 0.027 1.34 597.2 3.47E-05
initial 270 154.3 0.2215 0.1301 1.34 597.2 17.736

Attempt1 270 154.3 0.2215 0.03 1.34 597.2 0.0728
Attempt2 270 154.3 0.2215 0.027 1.34 597.2 0.0197
Attempt3 270 154.3 0.27 0.027 1.34 597.2 7.16E-09
The particle shapes computed by using the optimized bilinear JC-parameters are reported
in Figure 5.5. These shapes show a remarkable resemblance to the overall shape of the actual,
deformed particles. Nevertheless, there is a difference in the “volumes” of the two shapes that cannot
be reconciled by parameter fitting. This point will be addressed in more detail below.
Table 5.10 gives a summary of the optimized bilinear JC parameters for all six cases
considered in this work. The C2 parameter is the most sensitive parameter that was adjusted by the
optimization algorithm. The algorithm did not find A, B and ε̇c to have any effect on the minimization
of the objective function. The C1 value reported in Table 5.10 was not part of the optimization
process and was not included in the x vector which is mentioned earlier. This parameter corresponds
to the low strain rate regime in which we can barely find a material point in our simulations. The
average values of all of the parameters are reported in Table 5.10.
48
a) Dp = 20.75 Pm, and Vp = 175 m/s b) Dp = 24.4 Pm, and Vp = 286 m/s
c) Dp = 24.3 Pm, and Vp = 416 m/s d) Dp = 22.7 Pm, and Vp = 530 m/s
e) Dp = 19.75 Pm, and Vp = 663 m/s f) Dp = 23.4 Pm, and Vp = 699 m/
Figure 1 : Comparison of the deformed particle by using the initial JC parameters.
Figure 5.4: Comparison of the deformed particle by using the initial JC parameters.
49
a) Dp = 20.75 Pm, and Vp = 175 m/s b) Dp = 24.4 Pm, and Vp = 286 m/s
c) Dp = 24.3 Pm, and Vp = 416 m/s d) Dp = 22.7 Pm, and Vp = 530 m/s
e) Dp = 19.75 Pm, and Vp = 663 m/s f) Dp = 23.4 Pm, and Vp = 699 m/s
Figure 2 : Comparison of the deformed particles by using the optimized JC parameters.
Figure 5.5: Comparison of the deformed particles by using the optimized JC parameters.
50
By using the modified steepest descent method, we were able to reduce the objective
function to a negligible amount. However, the match between the deformed shapes of particles in the
experiments and the optimized simulations is not yet exact. Error introduced while measuring the
size of the particle from an oblique SEM image could be one of the reasons for this mismatch. In
addition, there could be some error in the initial volume and/or diameter of the particles. In the next
section we investigate the effect of volume and particle size.
Table 5.10: Summary of the optimized values of the parameters for the bilinear JC equation
A, MPa B, MPa n C1 C2 m ε̇c
Manes et al. [66] 270 154.3 0.2215 0.002 0.1301 1.34 597.2
Case1 : Dp = 20.75µm, Vp = 175m/s 270 154.3 0.2215 —– 0.027 1.74 597.2

Case2 : Dp = 24.40µm, Vp = 286m/s 270 154.3 0.2215 —– 0.011 1.44 597.2
Case3 : Dp = 24.34µm, Vp = 416m/s 270 154.3 0.24 —– 0.027 1.34 597.2
Case4 : Dp = 22.74µm, Vp = 530m/s 270 154.3 0.2215 —– 0.055 1.34 597.2
Case5 : Dp = 19.75µm, Vp = 663m/s 270 154.3 0.26 —– 0.027 1.34 597.2
Case6 : Dp = 23.40µm, Vp = 699m/s 270 154.3 0.27 —– 0.027 1.34 597.2
average 270 154.3 0.239 0.002 0.029 1.42 597.2

Standard Deviation 0 0 0.0215 0 0.0143 0.1602 0
5.5 Estimated Particle Diameters
In the previous section it was mentioned that the volume of the particles needed to be
adjusted. First, we need to prove that the volume of the particle does not change after the impact.
Then, by calculating the volume based on the deformed shape, the initial diameter of the particle can
be computed.[93]
It is known that, the volume of a metal undergoing plastic deformation remains unchanged
during deformation, and the volume that is experiencing elastic deformation will change as follows,
∆V
= trace(ε) (5.8)
V0
However, this elastic volume dilatation is recoverable. Moreover, according to the molecu-
lar dynamic (MD) simulations insignificant amount of the FCC structures of the Aluminum convert
51
Case 5: D p = 24.40 Pm, V p = 286 m/s 270 154.3 ---- 0.011 1.44 0.2215 597.2
Case 6: D p = 20.75 Pm, V p = 175 m/s 270 154.3 ---- 0.027 1.74 0.2215 597.2
average 270 154.3 0.002 0.029 1.42 0.239 597.2
standard deviation 0 0 0 0.0143 0.1602 0.0215 0
CHAPTER
Table 115.: Diameter
OPTIMIZATION OFofBILINEAR
and volume estimates the particles. D rJOHNSON-COOK PARAMETERS
is the reported particle diameter [4]. V r is the particle
volume based on D r . V A is the particle volume computed by Abaqus based on D r after meshing. V SEM is the particle
volume estimated based on the post-impact SEM image of the particle. D SEM is the particle diameter based on V SEM . D BF
is the particle diameter that best fits the experimental results obtained by using different sizes in Abaqus. E-D SEM and E-
to BCCD BFduring deformation.
are the %-errors in particleTherefore,
diameters with we expect
respect to D r . the volume of the particle to remain unchanged.
Case
Figure 5.6 D r history
shows time V rof the particle
VA V SEM computed
volume D SEM E-Dcase-4.
for SEM D
The E-D BF
BF change in the com-
6 20.75 4677.92 4658.17 3646.23 19.05 8.2 19.15 7.7
puted particle
5 volume
24.40in the first 100
7606.21 ns is insignificant.
7491.36 6085.58 Note that7.4
22.60 the increase
22.20in volume
9.0 is due to
4
thermal expansion24.30
which is7513.07
expected7491.36 5098.80
to diminish 21.30
at steady 12.3
state. Therefore, 21.10 13.2and expect
we conclude
3 22.70 6124.58 6104.18 3841.69 19.39 14.6 19.10 15.9
that the initial
2 and19.75
post-impact particle
4033.67 volumes
4016.63 to remain
3261.21 the same.
18.36 7.0 19.00 3.8
1 23.40 6708.82 6686.48 4463.24 20.38 12.9 21.20 9.4
Case 3, Vi=530 m/s

6.60E+03
6.50E+03
Volume, μm3
6.40E+03
6.30E+03
6.20E+03
6.10E+03
6.00E+03
0 20 40 60 80 100 120
Time, ns
Figure 5 : Time history of the particle’s volume in case-3. Note that the small increase in volume is due to thermal
expansion.
Figure 5.6: Time history of the particle’s volume in case-4. Note that the small increase in volume is
11 | P a g e
due to thermal expansion.
By taking into consideration that the volume of a particle should not change, we can
estimate an initial diameter for the particle, based on its final volume. To this end, first, an image
processing technique is used to measure the post-impact volume of the particles. The border of a
particle is extracted as a 2D plot. Then, the approximate volume (VSEM ) is calculated by rotating
this 2D picture about its middle axis. Note that this is an approximation of the volume, because, it
is assumed that the particle is perfectly symmetrical. Nevertheless, a particle diameter (DSEM ) is
computed based on this volume. The error (E − DSEM ) between this diameter and the reported
diameter (Dr ) is on the order of 10% as shown in Table 5.11. It is reasonable to state that this error
has contribution from both the initial measurements and post-impact volume assessment procedure.
Second, the original particle diameter is estimated by making it as a fitting parameter.

The simulations were run with different initial particle diameters until the error with respect to the
post-impact SEM images is minimized. This is done by visual inspection. The particle diameter that
52
gives the smallest error is deemed the best fit diameter DBF . Table 5.11 shows that in general the
error (E − DBF ) is also on the order of 10%. More interesting is the fact that the estimated particle
diameters DBF and DSEM , which are obtained by completely different approaches, are very close
to one another.
Figure 5.7 shows the simulation results, based on estimated DSEM particle diameters. In
each case the red curves represent computed particle shape of the optimized bilinear JC parameters
reported in Table 5.10. The blue curves represent the particle shapes obtained from the SEM images.
The volumetric error mentioned in the previous section is reduced remarkably well.
Table 5.11: Diameter and volume estimates of the particles. Dr is the reported particle diameter [11].
Vr is the particle volume based on Dr . VA is the particle volume computed by Abaqus based on Dr
after meshing. VSEM is the particle volume estimated based on the post-impact SEM image of the
particle. DSEM is the particle diameter based on VSEM . DBF is the particle diameter that best fits
the experimental results obtained by using different sizes in Abaqus. E − DSEM and E − DBF are
the %-errors in particle diameters with respect to Dr .
Case Dr Vr VA VSEM DSEM E − DSEM DBF E − DBF
1 20.75 4677.92 4658.17 3646.23 19.05 8.2 19.15 7.7

2 24.40 7606.21 7491.36 6085.58 22.60 7.4 22.20 9.0
3 24.30 7513.07 7491.36 5098.80 21.30 12.3 21.10 13.2
4 22.70 6124.58 6104.18 3841.69 19.39 14.6 19.10 15.9
5 19.75 4033.67 4016.63 3261.21 18.36 7.0 19.00 3.8
6 23.40 6708.82 6686.48 4463.24 20.38 12.9 21.20 9.4
53
a) Vp = 175 m/s b) Vp = 286 m/s
c) Vp = 416 m/s d) Vp = 530 m/s
e) Vp = 663 m/s f) Vp = 699 m/s
Figure 3 : Comparison of the experimental particle contours with those that are computed by using the adjusted particle
diameter DSEM and the modified JC parameters reported in Table 10. Note that the JC parameters for each impact speed
were different as reported in Table 10.
Figure 5.7: Comparison of the experimental particle contours with those that are computed by using
the adjusted particle diameter DSEM and the modified JC parameters reported in Table 5.10. Note
that the JC parameters for each impact speed were different as reported in Table 5.10.
54
5.6 Verification
Next, the average values of the optimized bilinear JC parameters reported in Table 5.10 are
used to simulate the impact process. DSEM values reported above were used as the diameters of the
undeformed particles. Figure 5.8 shows the comparison of these simulations with the experimentally
obtained contours. The match is generally very reasonable.
Further verification is obtained by comparing the coefficient or restitution (COR) as a
function of impact velocity in experiment and simulation (Figure 5.9). In these simulation the average
values of the modified bilinear JC parameters reported in Table 5.10 utilized in conjunction with the
adjusted particle diameter DSEM available in Table 5.11. The figure is plotted in the logarithmic
scale. Experimental data is illustrated by colored dots, and each color represents a specific particle
diameter. Simulations are depicted by two lines. The lines are the upper limit and lower limit of the
particle sizes available in this figure. Both experiment and simlation show that the results are not a
function of particle diameter. Results show a good resemblance between simulation and experiment.
5.7 Conclusion
In this chapter a combined experimental and simulation approach was presented in order to
find the parameters of a bilinear-JC flow stress relationship for Al-6061. It was shown that by using
the modified method of steepest decent and by allowing the particle volume to be a fit parameter,
remarkably good match between experimental and theoretical results can be obtained.
55
a) Vp = 175 m/s b) Vp = 286 m/s
c) Vp = 416 m/s d) Vp = 530 m/s
e) Vp = 663 m/s f) Vp = 699 m/s
Figure 4 : Comparison of the experimental particle contours with those that are computed by using the
adjusted particle diameter DSEM and the average value of the modified JC parameters reported in Table 10.
Figure 5.8: Comparison of the experimental particle contours with those that are computed by using
the adjusted particle diameter DSEM and the average value of the modified JC parameters reported
in Table 5.10.
56
Coefficientofrestitution
Impactspeeds50Ͳ950m/s
Measurements:Particlediameters15– 25Pm(UMassͲAmherst,May2016)
Simulations:15and25Pmparticlediameter(Calibrated,bilinearJCmodel)
Dp = 15 μm
Dp = 25 μm
Figure 5.9: Comparison of the experimental coefficient of restitution with simulated results. Lines
represents the simulation data and dots are the experiments.
57
Chapter 6
Conclusion
6.1 Summary and Conclusion
In this thesis, an investigation of micron-scale particles, impacted on substrate at supersonic

velocities is presented. The major goal of this work was to study the results of single particle impact
in both experiments and computer simulations to find a material model that can predict the behavior
of material under very high strain rates, based on those observations. To this end, the finite element
method has been used to simulate this physical phenomenon. In these 3D analyses, high strain
rate plasticity, temperature dependent material properties, heat generation and contact algorithm
have been used. Since, impact velocities vary from 50 to 1000 m/s and diameter of particles are in
the range of 20 to 50 micro-meters, simulating all combinations of particle diameters and impact
velocities are very time consuming. Therefore, Python scripts were used to automate all the process
of creating a geometry, running the simulation to creating the outputs and updating input file for the
later simulations.
The main focus of this thesis was to find a material flow model that is suitable to simulate
mechanics of Al-6061 particles on a sapphire substrate. The classic Johnson-Cook and the bilinear
Johnson-Cook material models were considered to this end. The equation and constants of classic
JC constitutive law, were obtained using Split Hopkinson Pressure Bar. In these experiments, the
equation was optimized to give an appropriate results in the range of 10−4 to 104 s−1 . The bilinear
JC model, was proposed to predict the material behavior under ultra high strain rates. In the high
velocity impact simulation, it was observed that the strain rate go beyond 108 in certain regions
near the impact interface. Therefore, the classic Johnson-Cook model was not able to predict the
deformed shape of particles after the impact. As the classic-JC model does not account for high
58
CHAPTER 6. CONCLUSION
strain rate strengthening, it is not suitable for particle impact simulations.

Since, there is no large scale experiment equipment to measure the yield stress of material
as a function of strain rate, this relationship should be obtained by using an optimization technique.
The bilinear JC model has six different constants. By using the method of steepest descent, and
considering the least square error of the ellipticity ratio of deformed particles as an objective function,
we were able to optimize the bilinear JC model for the high strain rate applications.
6.2 Future Work
In this work, impact of a single particle onto a substrate has been simulated under various
scenarios by using the finite element method. The material parameters of the bilinear JC model are
optimized, leading to very good agreement between the experiments and simulations. This work can
be extended in a number of ways. The following is some recommendations for future work, some of
which are already under progress:
I. In this study, impact of single Aluminum particle to Sapphire substrate was studied. The results
of the bilinear JC model can be used to model the single Aluminum particle to Aluminum
substrate. The cohesive zone model (CZM) can also be applied to these simulation for
studying the bonding/rebound situation. The cohesive stress in these cases can be a function of
temperature, pressure or both. The resultant simulations can be compared to experimental data
for validations and further investigations such as interface strength.
II. Although mechanics of single particle impact, reveals valuable information about the particle
deformation, multi-particle impact simulations could be a reasonable goal to achieve. By
simulating the multi-particle impact, a lot of new areas of studies will be opened. For instance,
a metal plate with a surface defect can be simulated. Then the cold spray coating can be
applied on top. Finally, the strength of the new plate can be measured based on the tensional
force, torsional force and fatigue loading. These results can be compared to the results of a
plate with defects to reveal new ideas and improve the conditions of coating and spraying in
real world applications.
III. In section 4.2, six different constitutive laws were explained. All of the models are capable
of predicting the behavior of material at very high strain rates. One of the interesting future
studies, could be the modeling of all constitutive laws and trying to compare the result of all
59
CHAPTER 6. CONCLUSION
simulation results. Most probably some of them need to be optimized to give proper results. In
that case, the same method of optimization, explained in chapter 5 can be applied.
60
Bibliography
[1] V. K. Champagne, The cold spray materials deposition process: fundamentals and applications.
Elsevier, 2007.
[2] F. Gärtner, T. Stoltenhoff, T. Schmidt, and H. Kreye, “The cold spray process and its potential
for industrial applications,” Journal of Thermal Spray Technology, vol. 15, no. 2, pp. 223–232,
2006.
[3] S. V. Klinkov, V. F. Kosarev, and M. Rein, “Cold spray deposition: significance of particle
impact phenomena,” Aerospace Science and Technology, vol. 9, no. 7, pp. 582–591, 2005.
[4] G. T. Gray, “Classic split-hopkinson pressure bar testing.” Materials Park, OH: ASM Interna-
tional, 2000., pp. 462–476, 2000.
[5] T. Nicholas, “Tensile testing of materials at high rates of strain,” Experimental Mechanics,
vol. 21, no. 5, pp. 177–185, 1981.
[6] A. Gilat, “Torsional kolsky bar testing.” Materials Park, OH: ASM International, 2000., pp.
505–515, 2000.
[7] S. Schoenfeld and T. Wright, “A failure criterion based on material instability,” International
Journal of Solids and Structures, vol. 40, no. 12, pp. 3021–3037, 2003.
[8] B. Yildirim, S. Muftu, and A. Gouldstone, “Modeling of high velocity impact of spherical
particles,” Wear, vol. 270, no. 9, pp. 703–713, 2011.
[9] I. JAHM Software Inc. (1998) Mpdb material database software. [Online]. Available:
https://www.jahm.com/pages/about mpdb.html
61
BIBLIOGRAPHY
[10] J.-H. Lee, D. Veysset, J. P. Singer, M. Retsch, G. Saini, T. Pezeril, K. A. Nelson, and E. L.
Thomas, “High strain rate deformation of layered nanocomposites,” Nature communications,
vol. 3, p. 1164, 2012.
[11] J.-H. Lee and W. Xie, “Single aluminum particle impact on rigid substrate,” Unpublished
Internal Document, 2015.
[12] D. R. Lesuer, G. Kay, and M. LeBlanc, “Modeling large strain, high rate deformation in metals,”
Engineering research, development and technology, 1999.
[13] A. Manes, L. Peroni, M. Scapin, and M. Giglio, “Analysis of strain rate behavior of an al 6061
t6 alloy,” Procedia Engineering, vol. 10, pp. 3477–3482, 2011.
[14] M. MUNRO, “Evaluated material properties for a sintered alpha-alumina,” Journal of the
American Ceramic Society, vol. 80, no. 8, pp. 1919–1928, 1997.
[15] material database website. [Online]. Available: https://www.matweb.com
[16] A. Papyrin, V. Kosarev, S. Klinkov, A. Alkhimov, and V. M. Fomin, Cold spray technology.
Elsevier, 2006.
[17] D. E. Wolfe, T. J. Eden, J. K. Potter, and A. P. Jaroh, “Investigation and characterization of

cr3c2-based wear-resistant coatings applied by the cold spray process,” Journal of Thermal
Spray Technology, vol. 15, no. 3, pp. 400–412, 2006.
[18] K. Spencer, D. Fabijanic, and M.-X. Zhang, “The use of al–al 2 o 3 cold spray coatings to
improve the surface properties of magnesium alloys,” Surface and Coatings Technology, vol.
204, no. 3, pp. 336–344, 2009.
[19] C. Hager, J. Sanders, S. Sharma, A. Voevodin, and A. Segall, “The effect of temperature
on gross slip fretting wear of cold-sprayed nickel coatings on ti6al4v interfaces,” Tribology
International, vol. 42, no. 3, pp. 491–502, 2009.
[20] S. R. Bakshi, D. Wang, T. Price, D. Zhang, A. K. Keshri, Y. Chen, D. G. McCartney, P. H.

Shipway, and A. Agarwal, “Microstructure and wear properties of aluminum/aluminum–silicon
composite coatings prepared by cold spraying,” Surface and Coatings Technology, vol. 204,
no. 4, pp. 503–510, 2009.
[21] ARL. Army research laboratory. [Online]. Available: http://www.arl.army.mil
62
BIBLIOGRAPHY
[22] H. Assadi, F. Gärtner, T. Stoltenhoff, and H. Kreye, “Bonding mechanism in cold gas spraying,”
Acta Materialia, vol. 51, no. 15, pp. 4379–4394, 2003.
[23] P.-H. Gao, C.-J. Li, G.-J. Yang, Y.-G. Li, and C.-X. Li, “Influence of substrate hardness on
deposition behavior of single porous wc-12co particle in cold spraying,” Surface and Coatings
Technology, vol. 203, no. 3, pp. 384–390, 2008.
[24] J.-G. Legoux, E. Irissou, and C. Moreau, “Effect of substrate temperature on the formation
mechanism of cold-sprayed aluminum, zinc and tin coatings,” Journal of Thermal Spray
Technology, vol. 16, no. 5-6, pp. 619–626, 2007.
[25] Y. Rollot, S. Régnier, and J.-C. Guinot, “Simulation of micro-manipulations: Adhesion forces
and specific dynamic models,” International journal of adhesion and adhesives, vol. 19, no. 1,
pp. 35–48, 1999.
[26] X. Li, P. Dunn, and R. Brach, “Experimental and numerical studies on the normal impact of
microspheres with surfaces,” Journal of aerosol science, vol. 30, no. 4, pp. 439–449, 1999.
[27] K. L. Johnson and K. L. Johnson, Contact mechanics. Cambridge university press, 1987.
[28] I. H. Livingstone, K. Verolme, and C. J. Hayhurst, “Predicting the fragmentation onset velocity
for different metallic projectiles using numerical simulations,” International journal of impact
engineering, vol. 26, no. 1, pp. 453–464, 2001.
[29] G. Stradling, G. Idzorek, B. Shafer, H. Curling, M. Collopy, A. H. Blossom, and S. Fuerstenau,

“Ultra-high velocity impacts: cratering studies of microscopic impacts from 3 km/s to 30 km/s,”
International Journal of Impact Engineering, vol. 14, no. 1, pp. 719–727, 1993.
[30] A. J. Piekutowski, “Debris clouds produced by the hypervelocity impact of nonspherical

projectiles,” International journal of impact engineering, vol. 26, no. 1, pp. 613–624, 2001.
[31] R. Tennyson and C. Lamontagne, “Hypervelocity impact damage to composites,” Composites

Part A: Applied Science and Manufacturing, vol. 31, no. 8, pp. 785–794, 2000.
[32] F. K. Schäfer, T. Geyer, E. E. Schneider, M. Rott, and E. Igenbergs, “Degradation and destruction
of optical surfaces by hypervelocity impact,” International journal of impact engineering,
vol. 26, no. 1, pp. 683–698, 2001.
63
BIBLIOGRAPHY
[33] C. J. Hayhurst, H. J. Ranson, D. J. Gardner, and N. K. Birnbaum, “Modelling of microparticle

hypervelocity oblique impacts on thick targets,” International journal of impact engineering,
vol. 17, no. 1, pp. 375–386, 1995.
[34] S. Kiselev and V. Kiselev, “Superdeep penetration of particles into a metal target,” International
journal of impact engineering, vol. 27, no. 2, pp. 135–152, 2002.
[35] A. Charters, “High-speed impact,” Scientific American, vol. 203, pp. 128–140, 1960.
[36] D. Wheeler and R. Wood, “Solid particle erosion of diamond coatings under non-normal impact
angles,” Wear, vol. 250, no. 1, pp. 795–801, 2001.
[37] V. Heuer, G. Walter, and I. Hutchings, “A study of the erosive wear of fibrous ceramic compo-
nents by solid particle impact,” Wear, vol. 225, pp. 493–501, 1999.
[38] A. Stevenson and I. Hutchings, “Wear of hardfacing white cast irons by solid particle erosion,”
Wear, vol. 186, pp. 150–158, 1995.
[39] J. Hansen, “Relative erosion resistance of several materials,” in Erosion: Prevention and Useful
Applications. ASTM International, 1979.
[40] J. Hopkinson, On the rupture of iron wire by a blow,(Article 38), Original Papers of John
Hopkinson. Cambridge University Press, 2014, vol. 1.
[41] ——, Further experiments on the rupture of iron wire, (Article 39), Original Papers of John
Hopkinson. Cambridge University Press, 2014, vol. 1.
[42] B. Hopkinson, “A method of measuring the pressure produced in the detonation of high
explosives or by the impact of bullets,” Philosophical Transactions of the Royal Society of
London. Series A, Containing Papers of a Mathematical or Physical Character, vol. 213, pp.
437–456, 1914.
[43] R. Davies, “A critical study of the hopkinson pressure bar,” Philosophical Transactions of the
Royal Society of London A: Mathematical, Physical and Engineering Sciences, vol. 240, no.
821, pp. 375–457, 1948.
[44] H. Kolsky, “An investigation of the mechanical properties of materials at very high rates of
loading,” Proceedings of the Physical Society. Section B, vol. 62, no. 11, p. 676, 1949.
64
BIBLIOGRAPHY
[45] S. Nemat-Nasser, “Introduction to high strain rate testing.” Materials Park, OH: ASM Interna-
tional, 2000., pp. 427–428, 2000.
[46] S. W. James, R. P. Tatam, S. R. Fuller, and C. Crompton, “Monitoring transient strains on a gun
barrel using fibre bragg-grating sensors,” Measurement Science and Technology, vol. 10, no. 2,
p. 63, 1999.
[47] F. Gascón, A. Bayón, and A. Varadé, “Measurement of longitudinal vibrations in a slender rod
by optical heterodyne interferometry,” European Journal of Mechanics-A/Solids, vol. 17, no. 1,
pp. 167–177, 1998.
[48] J. Harding, E. Wood, and J. Campbell, “Tensile testing of materials at impact rates of strain,”
Journal of Mechanical Engineering Science, vol. 2, no. 2, pp. 88–96, 1960.
[49] F. E. Hauser, “Techniques for measuring stress-strain relations at high strain rates,” Experimental
Mechanics, vol. 6, no. 8, pp. 395–402, 1966.
[50] J. Harding and L. M. Welsh, “A tensile testing technique for fibre-reinforced composites at
impact rates of strain,” Journal of Materials Science, vol. 18, no. 6, pp. 1810–1826, 1983.
[51] U. Lindholm and L. Yeakley, “High strain-rate testing: tension and compression,” Experimental
Mechanics, vol. 8, no. 1, pp. 1–9, 1968.
[52] G. Staab and A. Gilat, “A direct-tension split hopkinson bar for high strain-rate testing,”
Experimental Mechanics, vol. 31, no. 3, pp. 232–235, 1991.
[53] W. Goldsmith, J. Sackman, and C. Ewerts, “Static and dynamic fracture strength of barre
granite,” in International Journal of Rock Mechanics and Mining Sciences & Geomechanics
Abstracts, vol. 13, no. 11. Elsevier, 1976, pp. 303–309.
[54] J. Lipkin, D. Grady, J. Campbell et al., “Dynamic flow and fracture of rock in pure shear,” in The
18th US Symposium on Rock Mechanics (USRMS). American Rock Mechanics Association,
1977.
[55] J. Lipkin, K. Schuler, and T. Parry, “Dynamic torsional failure of limestone tubes,” Sandia
Labs., Albuquerque, NM (USA); Oxford Univ.(UK). Dept. of Engineering Science, Tech. Rep.,
1979.
65
BIBLIOGRAPHY
[56] D. Rittel, S. Lee, and G. Ravichandran, “A shear-compression specimen for large strain testing,”
Experimental Mechanics, vol. 42, no. 1, pp. 58–64, 2002.
[57] J. Zhao, H. Li, and Y. Zhao, “Dynamic strength tests of the bukit timah granite,” Singapore:
Nanyang Technological University, 1998.
[58] S. Huang, X. Feng, and K. Xia, “A dynamic punch method to quantify the dynamic shear
strength of brittle solids,” Review of Scientific Instruments, vol. 82, no. 5, pp. 157–160, 2011.
[59] J. Zhao, “An overview of some recent progress in rock dynamics research,” Advances in rock
dynamics and applications. CRC Press, USA, pp. 5–33, 2011.
[60] E. Bringa, K. Rosolankova, R. Rudd, B. Remington, J. Wark, M. Duchaineau, D. Kalantar,

J. Hawreliak, and J. Belak, “Shock deformation of face-centred-cubic metals on subnanosecond
timescales,” Nature materials, vol. 5, no. 10, pp. 805–809, 2006.
[61] W. Murphy, A. Higginbotham, G. Kimminau, B. Barbrel, E. Bringa, J. Hawreliak, R. Kodama,

M. Koenig, W. McBarron, M. Meyers et al., “The strength of single crystal copper under
uniaxial shock compression at 100 gpa,” Journal of Physics: Condensed Matter, vol. 22, no. 6,
p. 065404, 2010.
[62] A. Comley, B. Maddox, R. Rudd, S. Prisbrey, J. Hawreliak, D. Orlikowski, S. Peterson,

J. Satcher, A. Elsholz, H.-S. Park et al., “Strength of shock-loaded single-crystal tantalum [100]
determined using in situ broadband x-ray laue diffraction,” Physical review letters, vol. 110,
no. 11, p. 115501, 2013.
[63] T. Belytschko, J. S.-J. Ong, W. K. Liu, and J. M. Kennedy, “Hourglass control in linear and
nonlinear problems,” Computer Methods in Applied Mechanics and Engineering, vol. 43, no. 3,
pp. 251–276, 1984.
[64] S. P. Keeler and W. A. Backofen, “Plastic instability and fracture in sheets stretched over rigid
punches,” Asm Trans Q, vol. 56, no. 1, pp. 25–48, 1963.
[65] Z. Marciniak and K. Kuczyński, “Limit strains in the processes of stretch-forming sheet metal,”
International journal of mechanical sciences, vol. 9, no. 9, pp. 609IN1613–612IN2620, 1967.
[66] A. Manes, D. Lumassi, L. Giudici, and M. Giglio, “An experimental–numerical investigation on

aluminium tubes subjected to ballistic impact with soft core 7.62 ball projectiles,” Thin-Walled
Structures, vol. 73, pp. 68–80, 2013.
66
BIBLIOGRAPHY
[67] M. B. Bever, D. L. Holt, and A. L. Titchener, “The stored energy of cold work,” Progress in
materials science, vol. 17, pp. 5–177, 1973.
[68] A. Rosakis, J. Mason, and G. Ravichandran, “The conversion of plastic work to heat around a
dynamically propagating crack in metals,” Journal of the Mechanical Behavior of Materials,
vol. 4, no. 4, pp. 375–386, 1993.
[69] M. A. Meyers, Dynamic behavior of materials. John wiley & sons, 1994.
[70] R. Kapoor and S. Nemat-Nasser, “Determination of temperature rise during high strain rate
deformation,” Mechanics of Materials, vol. 27, no. 1, pp. 1–12, 1998.
[71] D. Veysset, A. J. Hsieh, S. Kooi, A. A. Maznev, K. A. Masser, and K. A. Nelson, “Dynamics of

supersonic microparticle impact on elastomers revealed by real–time multi–frame imaging,”
Scientific reports, vol. 6, 2016.
[72] G. R. Johnson and W. H. Cook, “A constitutive model and data for metals subjected to large
strains, high strain rates and high temperatures,” in Proceedings of the 7th International
Symposium on Ballistics, vol. 21. The Hague, The Netherlands, 1983, pp. 541–547.
[73] ——, “Fracture characteristics of three metals subjected to various strains, strain rates, temper-
atures and pressures,” Engineering fracture mechanics, vol. 21, no. 1, pp. 31–48, 1985.
[74] F. J. Zerilli and R. W. Armstrong, “Dislocation-mechanics-based constitutive relations for

material dynamics calculations,” Journal of Applied Physics, vol. 61, no. 5, pp. 1816–1825,
1987.
[75] R. Armstrong and F. Zerilli, “Dislocation mechanics aspects of plastic instability and shear
banding,” Mechanics of materials, vol. 17, no. 2, pp. 319–327, 1994.
[76] G. Z. Voyiadjis and F. H. Abed, “Microstructural based models for bcc and fcc metals with
temperature and strain rate dependency,” Mechanics of Materials, vol. 37, no. 2, pp. 355–378,
2005.
[77] D. L. Preston, D. L. Tonks, and D. C. Wallace, “Model of plastic deformation for extreme
loading conditions,” Journal of Applied Physics, vol. 93, no. 1, pp. 211–220, 2003.
67
BIBLIOGRAPHY
[78] R. Liang and A. S. Khan, “A critical review of experimental results and constitutive models for
bcc and fcc metals over a wide range of strain rates and temperatures,” International Journal of
Plasticity, vol. 15, no. 9, pp. 963–980, 1999.
[79] A. S. Khan and R. Liang, “Behaviors of three bcc metal over a wide range of strain rates and
temperatures: experiments and modeling,” International Journal of Plasticity, vol. 15, no. 10,
pp. 1089–1109, 1999.
[80] ——, “Behaviors of three bcc metals during non-proportional multi-axial loadings: experiments
and modeling,” International journal of plasticity, vol. 16, no. 12, pp. 1443–1458, 2000.
[81] C. Gao and L. Zhang, “Constitutive modelling of plasticity of fcc metals under extremely high
strain rates,” International Journal of Plasticity, vol. 32, pp. 121–133, 2012.
[82] S. Jaspers and J. Dautzenberg, “Material behaviour in conditions similar to metal cutting: flow
stress in the primary shear zone,” Journal of Materials Processing Technology, vol. 122, no. 2,
pp. 322–330, 2002.
[83] D. Umbrello, R. M’saoubi, and J. Outeiro, “The influence of johnson–cook material constants
on finite element simulation of machining of aisi 316l steel,” International Journal of Machine
Tools and Manufacture, vol. 47, no. 3, pp. 462–470, 2007.
[84] M. Iqbal, S. Rai, M. Sadique, and P. Bhargava, “Numerical simulation of aircraft crash on
nuclear containment structure,” Nuclear Engineering and Design, vol. 243, pp. 321–335, 2012.
[85] J.-B. Kim and H. Shin, “Comparison of plasticity models for tantalum and a modification of
the ptw model for wide ranges of strain, strain rate, and temperature,” International Journal of
Impact Engineering, vol. 36, no. 5, pp. 746–753, 2009.
[86] F. H. Abed and G. Voyiadjis, “A consistent modified zerilli-armstrong flow stress model for bcc
and fcc metals for elevated temperatures,” Acta Mechanica, vol. 175, no. 1-4, pp. 1–18, 2005.
[87] H. Huh, H. Lee, and J. Song, “Dynamic hardening equation of the auto-body steel sheet with
the variation of temperature,” International Journal of Automotive Technology, vol. 13, no. 1,
pp. 43–60, 2012.
[88] H. J. Lee, J. H. Song, and H. Huh, “Dynamic tensile tests of auto-body steel sheets with the
variation of temperature,” in Solid State Phenomena, vol. 116. Trans Tech Publ, 2006, pp.
259–262.
68
BIBLIOGRAPHY
[89] S. Rahmati and A. Ghaei, “The use of particle/substrate material models in simulation of
cold-gas dynamic-spray process,” Journal of thermal spray technology, vol. 23, no. 3, pp.
530–540, 2014.
[90] A. Gilioli, A. Manes, and M. Giglio, “Calibration of a constitutive material model for al-6061-t6
aluminum alloy,” Proceeding of the ACE X, pp. 8–9, 2010.
[91] L. Grace and M. Altan, “Characterization of anisotropic moisture absorption in polymeric

composites using hindered diffusion model,” Composites Part A: Applied Science and Manu-
facturing, vol. 43, no. 8, pp. 1187–1196, 2012.
[92] L. Pun, “Introduction to optimization practice,” Wiley New York, Tech. Rep., 1969.
[93] A. Shahabi, “An atomistic study on configuration, mechanics and growth of nanoscale filaments,”
PhD dissertation, Northeastern University, 2016.
[94] T. Eden and J. Schreiber, “Finite element analysis of cold spray particle impact,” Cold Spray
Action Team Meeting (6/22-6/23), 2015.
69
Appendix A
Simulation Results Using the Classic and

Bilinear Johnson-Cook Material Models
A.1 Simulation Results Using the Classic JC Model
In this section, the deformed shapes of the particles in the experiments are compared
to simulations. The impact experiments were performed in UMass at Amherst [10]. In these
experiments [11], Aluminum 6061 particles and a sapphire substrate were used. The idea behind
these experiments was to study only the deformation of the particles. The diameters of the particles in
these experiments were in the range of 19 to 24 micrometers and the impact velocities were between
200 and 1100 m/s.
Simulations were conducted by using Abaqus/explicit 6.13-2. All experimental conditions
were simulated to duplicate these experiments as closely as possible. Dynamic explicit was used
in simulate that includes coupled temperature and displacement degrees of freedom. In these
simulations, Three different configurations for the particle and the substrate were compared:
I. Deformable particle and substrate (Young’s modulus of substrate is set to 1000 times greater
than the particle’s).
II. Deformable particle and rigid substrate.
III. Deformable elastic-plastic particle and perfectly elastic substrate.
These modelling approaches are referred to as the type-I, type-II and type-III models, here after.
Elastic plastic behavior with exponential isotropic hardening is utilized for the mechanical properties
70
APPENDIX A. SIMULATION RESULTS USING THE CLASSIC AND BILINEAR JC MODEL
of the particle in these simulations. There are three sets of Johnson-Cook parameters that can be
found in the literature for material behavior of Al-6061. The information about these three can be
seen in Table A.1. Other material properties are reported in Chapter 3.
Table A.1: Classic Johnson-Cook material properties for Al-6061
A, M P a B, M P a n C m Tm ε̇0
JC-1 [12] 324 114 0.42 0.002 1.34 925 1
JC-2 [94] 200 203.4 0.35 0.011 1.34 925 1
JC-3 [13] 270 154.3 0.2215 0.002 1.34 925 1
Deformed shapes of particles are presented in Figures A.1, A.2 and A.3, using three
different modeling approaches and the three JC parameters (From Table A.1). All simulations results
compared to the single particle impact experiments [10]. In each figure, (a) and (b) show the scanning
electron microscope (SEM) image of the particle after impact. Figure (c) represents the type-II
analysis using JC-1 parameters (Table A.1) as a material model. In each figure, (d), (e) and (f)
display the results for type-III modeling approach using JC-1, JC-2, and JC-3 as a material models,
respectively.
we found that the shock waves produced by the impact is very large for the type-I analysis,
where the particle and substrate are deformable but the Young’s modulus of the substrate is much
greater than the particle. This requires very small time increments putting a large computational
cost on simulations in terms of time and capacity. For instance, a typical analysis by using type-III
approach, JC-3 material model and taking the advantage of multiprocessing completes in 15 to 20
minutes, however, type-I analyses takes 5 to 6 hours. Therefore, type-I was not pursued any further.
Although simulations reveal reasonable match between the experiment and analysis at
slow impacts, as the impact velocity increases, the finite element predictions become less accurate.
In general, simulations predicts more deformation than what is observed in the experiments. It can
be concluded that the flow stress predicting by the classic Johnson-Cook equation needs to be more
sensitive to the strain rate. Therefore, to compensate the excessive deformation, the flow stress should
provide higher yield stress. In other words, the amount of hardening predicted by JC due to plastic
deformation is not sufficient to reproduce the exact deformations as experiments.
71
d p ȝP9 p = 556 m/s, M p = 9.8x10-12 kg, KE p = 2.8x10-6 J
10ȝm 10ȝm
a) Front view of deformed particle b) Top view of deformed particle
9.1 9
30.4 31.4
c) Type II with JC-1 d) Type III with JC-1
8.8 8.8
31 30.6
e) Type III with JC-2 f) Type III with JC-3
Figure A.1: Comparison of the deformed shape of particle between experiment (a, b) and simulations
(c, d, e, f) where the impact velocity. is 556 m/s
72
d p ȝP9 p = 691m/s, M p = 7.6x10-12 kg, KE p = 3.6x10-6 J
10ȝm
10ȝm
a) Front view of deformed particle
b) Top view of deformed particle
6 5.6
35.1 35.9
5.6 5.5
34.8 35.9
73
d p ȝP9 p = 859m/s, M p = 1.1x10-11 kg, KE p = 8.3x10-6 J
10ȝm
10ȝm

3.7 3.6
60 57.1
3.8 3.4
55.3 58.4
74
The simulation results show the limitations of the Johnson-Cook model at high velocities.
According to Johnson and Cook papers [72, 73], this material model is calibrated for application
in the range of 10−4 − 104 s−1 . However, simulation results reveals that due to very high shear
deformation, the strain
Figure 3 rate usuallycontour
is a transparent goesofup
strain 108atsa−1
to rate in the
specific contact
increment time. interface regions.
This plot shows that Figure A.4 is
almost all of the particle elements are in the second regime of stress-strain rate plot. It means that
a transparent contour ofbilinear
the new strainJohnson-Cook
rate at a specific
flow stress increment time.
can provide more Thisforplot
hardening showsmodel.
the material that almost all of the
This extra hardening will prevent the particle from severe deformation and it helps to match the
one of theε̇solutions
particle elements deformed
are in the second
shapes regime
of particles of stress-strain
in simulations and experiments.rate plot,bewhere
This could
p > ε̇c .
which could help to anticipate the behavior of the dynamic impact phenomenon precisely.
Figure 3: Strain rate contour of iso-surfaces

Figure A.4: Strain rate contour of iso-surfaces
2. Results
Therefore,Phase-I
a modified material model is needed to predict the precise deformation in high
experiments of UMass Amherst University [6] is used as a reference to compare the
simulation results with experiments. In these experiments Aluminum 6061 and sapphire is used
strain rates. Manes
for et
theal. [13,and66]
particle suggested
substrate a BilinearA Johnson-Cook
material respectively. summary of material material model
characteristics are for Aluminum
available in table 2.
6061 and they calibrated the constants for their specific application with the method of optimization.
In the next section, the simulations results using the bilinear Johnson-Cook are compared to the
experiments.
A.2 Simulation Results Using the Bilinear JC Model
The deformed shapes of particles are presented in Figures A.5, A.6 and A.7, using two
Table 2: Experiments material characteristics
bilinear JC parameters (Table 4.1). All simulations results compared to the single particle impact
3|P a g e

experiments conducted at UMass-Amherst by Professor J. H. Lee [11]. In each Figure, (a) and (b)
refer to the SEM images of the particle after the impact, (c) represents the analysis using a set of
JC parameters that Manes et al. [66] found for Aluminum 6061 material and using Split Hopkinson
pressure bar test. In the same picture, (d) displays the results for another set of JC parameters which
is optimized by Gilioli et al. [90] to find the exact deformation shape of a bullet impact on the
helicopter tail rotor transmission.
Schematically, it is clear that the particles simulated by using the bilinear JC material
model, deformed less than simulations performed by using the classic JC material model. Therefore,
75
we conclude that the extra term which is added to the classic JC equation ”C2 ” can play a significant
role in high strain rate applications and it can compensate for the underestimation of classic JC
material model.
To compare classic the JC results with bilinear JC model, a new parameter is defined. The
Ellipticity ratio is defined as the maximum vertical distance to the maximum horizontal distance
in the front view of deformed particle shape. This study shows that in three different cases that
comparison have been made, the bilinear JC model improved the ellipticity ratio by 89%, 138% and
246% for the cases where the impact velocity was 556 m/s, 691 m/s and 859 m/s, respectively.
dp = 18.5μm, Vp = 556 m/s, Mp = 9.8x10-12 kg, KEp = 2.8x10-6 J
10μm 10μm
12.4 12.4
22.8 22.9
c) Type III with JC-4 d) Type III with JC-5
(c, d) where the impact velocity. is 556 m/s
76
dp = 17.5μm, Vp = 691m/s, Mp = 7.6x10-12 kg, KEp = 3.6x10-6 J
10μm
10μm
9.7 9.7
26.6 26.8
77
dp = 20.0μm, Vp = 859m/s, Mp = 1.1x10-11 kg, KEp = 8.3x10-6 J
10μm
10μm

7.6 7.5
38.1
37.8
78
Appendix B
Table of Temperature Dependent

Material Properties for Aluminum 6061
Data is taken from MPDB material database software [9].
Temperature Elastic Modulus Poisson’s Thermal Thermal Specific Heat

Ratio Expansion Conductivity
K GPa 10−6 /K W/m K J/kg K
10 76.5951 0.3239 0.2525 14.3824 1.5609
20 76.5621 0.3240 0.5045 28.4645 8.8829
30 76.4967 0.3242 0.9477 41.0270 33.4402
40 76.4011 0.3243 1.7028 52.2049 81.9529
50 76.2777 0.3245 2.8316 62.1332 149.1508
60 76.1285 0.3248 4.3375 70.9469 223.9242
70 75.9555 0.3250 6.1652 78.7811 297.9033
80 75.7608 0.3253 8.2007 85.7707 368.3660
90 75.5463 0.3255 10.2179 92.0509 433.5356
100 75.3137 0.3258 11.8261 97.7565 492.5810
110 75.0647 0.3262 13.1962 103.0226 545.6162
120 74.8011 0.3265 14.3611 107.9842 593.7009
130 74.5243 0.3268 15.3513 112.7788 637.9319
140 74.2358 0.3271 16.1946 116.8674 677.6165
79
APPENDIX B. TEMPERATURE DEPENDENT MATERIAL PROPERTIES FOR AL-6061
150 73.9370 0.3274 16.9163 120.6998 712.0432

160 73.6291 0.3278 17.5390 124.2883 742.6290
170 73.3136 0.3281 18.0826 127.6453 771.5068
180 72.9913 0.3284 18.5646 130.7830 798.7493
190 72.6634 0.3287 18.9998 133.7137 824.4276
200 72.3309 0.3290 19.4002 136.4498 848.6113
210 71.9946 0.3293 19.7756 139.0036 871.3683
220 71.6554 0.3296 20.1328 141.3873 892.7655
230 71.3138 0.3298 20.4761 143.6133 912.8678
240 70.9707 0.3301 20.8074 145.6938 931.7389
250 70.6264 0.3303 21.1257 147.6413 949.4410
260 70.2816 0.3306 21.4275 149.4679 966.0347
270 69.9364 0.3308 21.7067 151.1861 981.5790
280 69.5914 0.3310 21.9547 152.8080 996.1318
290 69.2465 0.3312 22.1677 154.3460 1009.7491
300 68.9021 0.3314 22.3724 155.8133 1022.4857
310 68.5580 0.3315 22.5752 157.5203 1034.3947
320 68.2143 0.3317 22.7762 159.1733 1045.5277
330 67.8709 0.3319 22.9753 160.7722 1055.9351
340 67.5275 0.3320 23.1726 162.3171 1065.6656
350 67.1838 0.3321 23.3680 163.8080 1074.7662
360 66.8395 0.3323 23.5616 165.2448 1083.2828
370 66.4941 0.3324 23.7533 166.6276 1091.2597
380 66.1470 0.3325 23.9431 167.9564 1098.7395
390 65.7977 0.3327 24.1311 169.2311 1105.7634
400 65.4454 0.3328 24.3173 170.4518 1112.3714
410 65.0892 0.3329 24.5016 171.6185 1118.6016
420 64.7284 0.3331 24.6840 172.7311 1124.4907
430 64.3620 0.3333 24.8646 173.7898 1130.0742
440 63.9889 0.3335 25.0433 174.7943 1135.3858
450 63.6079 0.3337 25.2202 175.7449 1140.4577
460 63.2179 0.3339 25.3952 176.6414 1145.3208
80
470 62.8176 0.3342 25.5684 177.4839 1150.0045

480 62.4055 0.3345 25.7397 178.2723 1154.5365
490 61.9803 0.3349 25.9091 179.0067 1158.9432
500 61.5403 0.3353 26.0767 179.6871 1163.2495
510 61.0839 0.3357 26.2425 180.3134 1167.4786
520 60.6094 0.3362 26.4064 180.8858 1171.6525
530 60.1150 0.3368 26.5684 181.4040 1175.7914
540 59.5987 0.3374 26.7286 181.8683 1179.9144
550 59.0588 0.3382 26.8869 182.2785 1184.0388
560 58.4929 0.3390 27.0434 182.6347 1188.1804
570 57.8992 0.3399 27.1980 182.9368 1192.3538
580 57.2752 0.3409 27.3570 183.1850 1196.5717
590 56.6187 0.3420 27.5120 183.3790 1200.8456
600 55.9274 0.3433 27.6650 183.5191 1205.1855
610 55.1986 0.3446 27.8170 183.6051 1209.5998
620 54.4300 0.3461 27.9680 183.6371 1214.0954
630 53.6187 0.3478 28.1180 183.6151 1218.6778
640 52.7622 0.3496 28.2660 183.5390 1223.3509
650 51.8576 0.3515 28.4130 183.4089 1228.1173
660 50.9019 0.3537 28.5590 183.2247 1232.9778
670 49.8923 0.3560 28.7040 182.9865 1237.9321
680 48.8256 0.3585 28.8480 182.6943 1242.9780
690 47.6988 0.3613 28.9900 182.3481 1248.1122
700 46.5085 0.3642 29.1310 181.9478 1253.3295
710 45.2514 0.3674 29.2710 181.4935 1258.6236
720 43.9243 0.3708 29.4100 180.9852 1263.9864
730 42.5235 0.3745 29.5480 180.4228 1269.4085
740 41.0455 0.3785 29.6840 179.8064 1274.8790
750 39.4867 0.3828 29.8200 179.1359 1280.3854
760 37.8433 0.3873 29.9550 178.4115 1285.9137
770 36.1116 0.3922 30.0880 177.6330 1291.4486
780 34.2720 0.3974 30.2210 176.8004 1296.9731
81
790 32.3480 0.4030 30.3520 175.9139 1302.4689

800 30.3240 0.4089 30.4830 174.9732 1307.9159
810 28.1950 0.4152 30.6120 173.9786 1313.2929
820 25.9560 0.4218 30.7410 172.7800 1315.3000
830 23.6040 0.4289 30.8690 171.3400 1319.2000
840 21.1330 0.4365 30.9950 169.6600 1322.9000
850 18.5400 0.4444 31.1210 167.6900 1326.1000
82

Neu m0411x42f

Uploaded by

Copyright:

Available Formats

Neu m0411x42f

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

Neu m0411x42f

Uploaded by

Copyright:

Available Formats

Tuning Johnson-Cook Material Model Parameters for Impact of High

Velocity, Micron Scale Aluminum Particles

Arash Alizadeh Dehkharghani

The Department of Mechanical and Industrial Engineering

in partial fulfillment of the requirements

Thesis Title: Tuning Johnson-Cook Material Model Parameters for Impact of

Author: Arash Alizadeh Dehkharghani NUID: 001102511

Department: Mechanical and Industrial Engineering

Approved for Thesis Requirements of the Master of Science Degree

Associate Dean of Graduate School:

List of Tables vii

Abstract of the Thesis ix

4 Material Model for High Strain Rate Deformation of Metals 32

5 Optimization of Bilinear Johnson-Cook parameters 41

B Table of Temperature Dependent Material Properties for Aluminum 6061 79

3.1 Isometric view of the developed 3D model. . . . . . . . . . . . . . . . . . . . . . 17

5.1 Stress-Strain rate behavior in a semi log plot [13]. . . . . . . . . . . . . . . . . . . 43

3.1 Temperature dependent material properties of Sapphire [14]. . . . . . . . . . . . . 24

4.1 Bilinear Johnson-Cook material properties for Al-6061 . . . . . . . . . . . . . . . 39

5.1 Material properties of Al-6061 [15, 9]. . . . . . . . . . . . . . . . . . . . . . . . . 42

A.1 Classic Johnson-Cook material properties for Al-6061 . . . . . . . . . . . . . . . . 71

Tuning Johnson-Cook Material Model Parameters for Impact of High

In recent years, research on is conducted to find materials with superior mechanical

Figure 1.2: Schematic diagram of a cold spray gun [1].

1.2 Objective and Problem Description

1.3 Arrangement of the Thesis

2.1 Cold Spray

efficiency with the respective spr

2.2 Investigations of Coa

and domes by particle im-

act phenomena have been

2.3 High Strain Rate Experiments

When it comes to simulating a physical phenomenon, mechanical and thermal behavior

Strain rate (s–1) Table 1 Experime

Strain versus time Constant Uniaxial Uniaxial 103– 105 T

Figure 2.4: Schematic view of split-hopkinson pressure bar [4].

magnitude of these thr

2.3.2 Dynamic tests in tensile and torsional loadings

Figure 2.7: Schematics of dynamic tension loading [5].

GENERIC TESTING PRINCIPLES

Strain gauge Front Rear

Striker Incident bar Transmitted bar Incident bar Transmitted bar

2.4 Ultra High Rate of Loading

Modeling and Finite Element Simulation

3.2 FEM Model

3.2.1 Modeling and analysis technique

Figure 3.1: Isometric view of the developed 3D model.

3.2.2 Element Type and Failure Criterion

3.2.3 Interaction and Contact Properties

3.2.4 Boundary Conditions

3.2.5 Material Properties

3.2.6 Adiabatic Shear Instability

Table 3.1: Temperature dependent material properties of Sapphire [14].

20◦ C 500◦ C 1000◦ C

ρcp Ṫ = β Ẇp (3.1)

3.2.7 Effect of Using Temperature Dependent Material Properties

Table 3.2: material properties for Aluminum 6061 at room temperature.

Density, kg/m3 2700

bilinear temp dep