International Journal on Interactive Design and Manufacturing (IJIDeM)

https://doi.org/10.1007/s12008-022-00995-x

ORIGINAL PAPER

Applying GPR-FGRA hybrid algorithm for prediction and optimization

of eco-friendly magnetorheological finishing Ti–6Al–4V alloy
Dung Hoang Tien1 · Trinh Nguyen Duy1 · Pham Thi Thieu Thoa1

Received: 7 May 2022 / Accepted: 16 July 2022

© The Author(s), under exclusive licence to Springer-Verlag France SAS, part of Springer Nature 2022

Abstract
To produce the mirror finish of the Ti–6Al–4V alloy workpieces, a newly developed polishing tool based on magnetic field
assistance. This work has developed an environmentally friendly model of magnetic finishing with a carrier fluid natural origin
combined with carbonyl and abrasive iron particles; and a new GPR-FGRA hybrid algorithm to predict, optimize and identify
critical factors affecting material removal rate and surface quality. The best surface quality with the optimization method
increased by 37.5% (Ra  0.025 μm) when choosing the optimized cutting parameter with the GPR-FGRA hybrid algorithm
compared with the Taguchi experimental analysis (Ra  0.040 μm). The analysis results by the GPR-FGRA algorithm show
that the prediction surface quality accuracy was more significant than 99.63%, and the working distances (K) had the most
remarkable influence on the polishing effect, while the result of workpiece speed was the least important. This study provides
excellent reference values for predicting, optimizing and identifying the main factors affecting surface quality and material
removal rate in a polishing model for complex, expensive materials from low-cost and environmentally friendly materials.

Keywords Magnetorheological finishing · Ti–6Al–4V alloy · GPR-FGRA · Polishing · Surface roughness · Eco-friendly

1 Introduction required. The polishing under the support of magnetic fields

is emerging as the advanced finishing method for high-gloss
Ti–6Al–4V alloy has a melting point and excellent wear resis- and precision finishes.
tance and is widely used in industry [1, 2]. With the excellent Magnetorheological finishing (MRF) forms by mixing a
properties inherent to Ti–6Al–4V alloys, such as elastic- carbonyl iron particle (CIP) and abrasive particle (AP) with
ity, high adhesion ability, and superior corrosion resistance, a carrier liquid to create a flexible polishing tool [9, 10].
Ti–6Al–4V alloy classing as one of the most challenging MRF polishing tool works like Newton liquid when without
materials to machine [3–5]. Therefore, to meet the increas- a magnetic field. When there is a magnetic field, MRF imme-
ingly stringent requirements from the industries, the finishing diately changes viscosity to a semisolid state (non-Newton
process for Ti–6Al–4V alloy machining workpiece super- liquid). When there is an impact magnetic field, this pro-
gloss poses a significant challenge for manufacturers and cess is formed the chain structure by the interaction between
technicians. Traditional surface finishing techniques such as magnetic particles that arrange themselves according to the
grinding have improved machining surface quality. However, magnetic field lines. The chain structure exhibits deformation
the poor grinding property of Ti–6Al–4V alloy leads to fast resistance when removing residual materials under differ-
wheel wear that increases costs. Besides, the grinding sur- ent machining modes. The rheological characteristics and
face quality and efficiency are lower than other materials the yield strength of the developed MR fluid affects by the
[6–8]. Therefore, to produce a super-glossy surface with very externally applied magnetic field [10, 11]. The CIP in the
high precision, a non-traditional surface finishing process is MRF creates an interlinked chain along the direction of the
magnetic field, resulting in a semisolid abrasive tool to pro-
cess hard surfaces [12]. Barman et al. [13] studied the effect
B Trinh Nguyen Duy
[email protected] of magnetorheological polishing fluid compositions on the
surface finish of Ti-alloy. Ultra-fine surface roughness was
1 Faculty of Mechanical Engineering, Hanoi University of acquired when using different types of rheological fluids
Industry, 298 Cau Dien, Bac Tu Liem District, Ha Noi,
and observed that utilising a raster path provided the best
Vietnam

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

surface finish and surface topography at a rotational tool intervals for prediction and optimization [18]. Zhang et al.
speed of 1200 rpm, a working gap of 1 mm, and a finishing [19] gave a multi-responsive GPR model and showed supe-
time of 6.30 h. Nagdeve et al. [14] developed a rotational- riority through surface reaction. Besides the advantages of
magnetorheological abrasive flow finishing process-based the GPR model, there are also limitations, such as not fully
unique tool for nano-finishing the knee joint’s femoral com- describing the influencing factors from technological param-
ponent and surface finish in the 78–89 nm was attained by eters and not identifying the most and most minor influential
considering the effect of various input process parameters. factors that affect the survey subjects. Also, optimization
The roughness parameters have obtained the values in the problem solving is preeminent but based on the objective and
range of nanometers and rendered better surface topography explicit boundary functions. It has not established sufficient
[15]. Magnetorheological (MR) fluids are used by mixing empirical relationships and has not solved the identification
with abrasive particles as finishing media for surface finish- problem thoroughly. The authors integrated fuzzy grey rela-
ing. Singh et al. developed a ball-end tool formed by MR tionship analysis (FGRA) into the GPR model to overcome
polishing medium for nano-finishing of 3D surfaces, the sur- these shortcomings. Based on the fuzzy system approach,
face roughness of the ferromagnetic workpieces was reduced the importance and uncertainty factors in the considering
from 414.1 nm to 70 nm in 100 min [16]. Parameswari weights by different criteria in the FGRA model [20, 21].
et al. [13] explored the critical quality performance of the The factors affecting the survey subjects are effectively mea-
MRF process in achieving nanolevel finish on Ti–6Al–4V sured and statistical by FGRA. This method is very suitable
flat discs. The initial surface roughness parameter signifi- for solving problems of incomplete information, little input
cantly affects the finishing. MRF is primarily used to finish data or lots of data but sporadic, uncertain and inconsistent.
optics in nanometer surface roughness without sub-surface The FGRA method is used in many fields, in which many
damage [17]. MRF has been successfully used to complete researchers have optimized the processing parameters. The
a wide variety of materials to sub-nanometer surface rough- grey relation level indicates the similarity between the com-
ness values from initial sub-micrometre surface roughness parison and reference sequences [22, 23].
values. From the above characteristics, the authors have intro-
The MRFs apply successfully to finish Ti alloy and various duced a new polishing method with MRF containing inex-
materials to sub-nanometer surface roughness values from pensive CIPs and APs. The carrier liquid is a safe and
initial sub-micrometre surface roughness values. However, eco-friendly cutting oil solution combined with a new hybrid
these studies have not addressed the issue of environmental algorithm for a super precise mirror polishing surface. In this
protection from carrier fluids derived from vegetable oils. In work, after describing the principles and operating princi-
the MRF polishing processes, the biodegradability of car- ples of the MRF model, setting up the proposed Gaussian
rier fluids in the MRF is less critical than during their use. Process Regression and Fuzzy grey relationship analysis
It depends mainly on the carrier medium used in the liq- (GPR-FGRA) hybrid algorithm. Experimental processes on
uid synthesis. Silicon oil, mineral oil, hydrocarbon oil, and the Ti–6Al–4V workpiece aim to analyze the effects of tech-
deionized water use commonly carrier fluids. In which deion- nical parameters on MRR and surface quality. The optimal
ized water can biodegrade in the environment, but with low parameters in the MRF polishing process aim to increase sur-
viscosity, APs and CIPs deposited quickly. Furthermore, the face quality and save machining time. The proposed method
deionized water components in MRF polishing are often in an provides excellent reference values to further improve the
acidic or alkaline environment to increase the magnetism in surface quality and MRR capability.
the carrier fluid. Consequently, CIP particles corrode and oxi-
dize over time, reducing liquid structures’ magnetic and yield
stress. Deionized water-based carrier fluids cannot maintain 2 Methodology and experimental design
the high temperature obtained during machining. Therefore,
necessary to implement unique methods to overcome these The created MRF polishing process by a suspension of
problems, such as adding additives to reduce sedimentation micrometre-sized (about 20–50 mm) magnetized particles in
and enhance thermal stability, creating a coating for magnetic the carrier liquid, with a magnetization particle volume per-
particles to avoid corrosion and improve the carrier liquid, centage of 30–40%. The ability to change behaviour under
etc., which will increase the cost of the synthetic fluid. From the impact of the magnetic field is a crucial feature of MRF.
the above problems and combined with biodegradation, it When no magnetic field exists, the liquid behaves like a reg-
is essential to find an environmentally friendly alternative ular liquid with properties close to the carrier liquid. MRF
carrier liquid and overcome the disadvantages of deionized acts as a semi-solid substance when there is a magnetic
water. field impact [24]. The CIPs in MRF with deionized water
The Gaussian Process Regression (GPR) algorithm is a in nature is easily oxidized and dehydrated. Depending on
powerful and flexible tool capable of providing confidence working conditions, iron particles will be oxidized quickly

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

Fig. 1 Morphology of MRFs

with different carrying fluids by
the time

Fig. 2 MRF polishing principle

or slowly with time, humidity and temperature. Magnetic also overcomes the phenomenon of ferromagnetic particles
particle oxidation directly affects MRF behaviour in practi- in the MRF oxidized and dehydrated, reducing viscosity and
cal applications because iron oxides have inferior magnetic polishing ability over time as previous work encountered.
properties compared to pure iron. MRF with CIP oxidized Figure 2 shows the principle of polishing operation by
showed a 20% decrease in magnetic effect. Figure 1 indi- MRF. The workpiece is attached to the milling machine
cates that the MRF with the proposed carrier solution has spindle via a fixture and has a rotational speed of n1 . The
a high viscosity, excellent oxidation resistance, and almost permanent magnet placed on the MFC carries material SUS
no oxidation over time. When not affected by an active mag- 304, and rotation is transmitted by motor 2. The direction of
netic field, the MRF returns to its original state. It can see that rotation of the MRF carrier is opposite to that of the milling
the proposed liquid is not only environmentally friendly but

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

MRF carrier. The micrometre size APs will remove mini-

mal machining residual due to friction between the APs and
the workpiece surface. The MRF should be set up directly
above the permanent magnet to increase the polishing force
between the APs and the workpiece. The magnetic system
of permanent magnets making of N50 type Nd-Fe-B mate-
rial. The experiment equipment in the MRF polishing system
consists of a milling machine, a permanent magnetic, a fix-
ture, a surface roughness meter, an electronic scale and an
MRF carrier.
Table 1 describes the material specifications of the com-
ponents used in the MRF polishing of Ti–6Al–4V alloy
workpieces. Table 2 shows the composition of the differ-
ent MRFs used in the polishing processes. The miscible oil
with vegetable ester-based is cutting fluid manufactured by
Vasco Jemtech Advanced Fluid Management, the CIPs with
particle size 20–30 μm from manufacturer Henan Yuhuang
Powder Metallurgy Plant. The AP (Al2 O3 ) have a particle
Fig. 3 MRF experimental equipment size of 30 μm and is manufacturing by Dengfeng Hongsheng
Abrasives.
Polishing with individual abrasive particles has many
effects that need to consider. The main influences on the
machine spindle. The working distance (K) changes by mov- MRF polishing process are the material composition and
ing the milling machine table up or down. The polishing behaviour and the ability to achieve high gloss and precision
device is located on the milling machine table to create hor- for machined surface quality. The material removal factors
izontal tooling movement when polishing. At this point, the need investigating to optimize MRF polishing elements. Pre-
polishing settings by MRF have been forming. The exper- ston was the first to adopt a mathematical model for material
imental equipment setup and determination of MRR and removal during liquid abrasive machining called Preston’s
surface roughness before and after polishing as shown in law [25].
Figs. 3 and 4.
When setting a working distance (K), the CIP move in R  p×v×k (1)
chains along the magnetic induction lines and non-magnetic
AP are affected by gravity and exerted by a pulling force. where p is the initial pressure at the contact zone, v is the
Most of the non-magnetic APs in the MRF will move relative velocity between the workpiece and the fluid in the
onto the workpiece surface as these force components act. polishing zone, and k is the Preston coefficient. The polishing
The rotation of the workpiece and the MRF carrier will method by MRF depends on the fluid properties, workpiece
generate cutting forces of APs on the workpiece surface dur- geometry, material properties and so on, as described in
ing machining. The AP receive rotation rate n2 from the Eq. (1). However, to describe the general physical properties

Fig. 4 Determination of MRR and surface roughness before and after polishing

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

Table 1 Material parameters

used in Ti–6Al–4V alloy Mumerical Description Material Size Relative Magnitude
polishing order magnetic (A/m)
permeability

1 Permanent Nd-Fe-B 100 × 80 × 1.09966 − 890,000

magnets 30 mm
2 MRF carrier SUS 304 1
3 Polishing tool MRF 5
4 Workpiece Ti–6Al–4V 32 × 32 × 1
20 mm

Table 2 MRF components

Mumerical Description Particles size Percentage of Magnetic Magnetic Specific weight

order (μm) components in the permeability sensitivity (kg/cm3 )
MRF (%) (N/A2 )

1 CIP 20–30 40 μ  5.03 × 10–4 χ m  103 ρ C  7.8 × 10–3

2 AP 30 12 χ’m  103 ρ A  3.965 ×
10–3
3 Miscible oil 45 ρ A  0.9187 ×
cutting fluid 10–3
4 α-cellulose 3

of MRF processes, it is necessary to use an extended equation. the fluid given by φ and the magnetic strength field provided
The Equation of. The equation of Zhang et al. [26] includes by H.
process parameters that depend on the physical properties of The distance between the workpiece surface and the car-
the MR fluid through a complex relationship, as shown by rier for the liquid at each position is given by the actual
Eq. (2) for the pressure distribution. distance K. The actual distance h depends on the depth of
cut ae and the MRF height h0 , as shown in Eq. (3).
⎡ ⎤
H K  h 0 − ae (3)
−2μ × u × x × η0 μp − μ f
p⎣ + 3φμ0 μ f × × Hd H⎦
h2 μ p + 2μ f
0 Experimental to analyze the consistency between material
(2)
removal and the listed MRF parameters. Factors investigated
during MRF polishing, including MRF carrier speed, work-
where η0 is the kinematic viscosity of the fluid and u is the piece speed, the working distance (K) and feed rate, are
velocity of the fluid, η0  v×ρ shows that the initial viscosity shown in Tables 3 and 5.
of the liquid η0 can be calculated from the liquid density ρ Nowadays, machining process studies strive to exploit
and the kinematic viscosity ν. The parameters μ is related environmental safety features while preserving machining
to the magnetic field, μ0 is the magnetic permeability of the properties such as good abrasion resistance, low volatility,
vacuum, μf is the magnetic permeability of the liquid, and maintaining viscosity and ability to work at high tempera-
μp is the magnetic permeability of the magnetic particles in tures, non-toxic, easy to biodegradable, etc. Studies of [27]

Table 3 Parameter level of MRF

polishing technology according Level Workpiece speed (nw ) Working distances (K) MRF carrier speed Feed rate (F)
to the Taguchi method (rpm) (mm) (nMRF ) (mm/min)
(rpm)

1 530 1 25 250
2 720 1,25 30 500
3 980 1,5 35 750
4 1330 1,75 40 900

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

Table 4 Components by weight

of miscible water cutting Mumerical Material Percentage Mumerical Material Percentage
fluid-based vegetable oils order (%) order (%)

1 Vegetable oils 5–20 6 Anti-rust 0–3

agents
2 Synthetic esters 15–35 7 Extreme 0–3
of vegetable pressure
oils agents
3 Waters 5–15 8 Aseptic 0–2
4 Emulsifying 2–12 9 Defoamer 0–2
agents
5 Dispersing 0–2 10 Alkaline 15–30
agents reserve

Table 5 Experiment parameters for polishing Ti–6Al–4V alloy by MRF

Experiments Factor and level MRR Average surface

order (mg/min) roughness value
Ram (μm)
Workpiece speed MRF carrier speed Feed rate Working distance
(nw ) (nMRF ) (F) (K)

1 1 1 1 1 13.33 0.041
2 1 2 2 2 11.66 0.059
3 1 3 3 3 10 0.089
4 1 4 4 4 7.67 0.098
5 2 1 2 3 14.15 0.044
6 2 2 1 4 12.83 0.050
7 2 3 4 1 10.50 0.057
8 2 4 3 2 9.21 0.079
9 3 1 3 4 15.30 0.067
10 3 2 4 3 13.26 0.080
11 3 3 1 2 10.83 0.081
12 3 4 2 1 9.33 0.091
13 4 1 4 2 15.92 0.040
14 4 2 3 1 14.02 0.059
15 4 3 2 4 11.17 0.062
16 4 4 1 3 9.60 0.071

for a vegetable oil-based water-soluble cutting liquid have Table 6 Constraints on cutting parameters
solved the above requirements. On that basis, the per cent
Mumerical order Constraint factor Constraint range
components by weight of the cutaneous fluid, as described
in Table 4. 1 Working distance (mm) 0.3 < K < 5
Taguchi experiment method is applied in polishing by 2 Feed rate (mm/min) 2 < f < 3500
MRF to save many experiments and time from getting the
3 Spindle speed (rpm) 100 < n < 2100
optimal target group [28]. According to this approach, the
4 Carrier speed (rpm) 0 < n < 60
impact levels of the survey factors are established based on
the orthogonal array. In this work, the MRF polishing exper-
iments for Ti–6Al–4V titanium alloy were set up according
to the Taguchi experimental design, as shown in Table 5.
Together with the constraints for cutting parameters are
described in Table 6. Before polishing, the workpieces were ground on a single crystal diamond grinding wheel, giving a
surface roughness of about Ra  0.49 μm.

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

Before polishing, the workpiece cleaning with ethanol and and optimizes the ability of MRR and surface quality when
acetone. The MRR (mg/min) was determined by a high- polishing by MRF, as depicted in Fig. 5.
resolution electronic balance (0.1 mg) as expressed by the The GPR-FGRA hybrid algorithm includes the steps
equation: described as follows:

m × 103
MRR  (4)
T
• Step 1. Set up an experimental model according to Taguchi
where m: the mass difference before and after polishing analysis. The experiment processes are analyzed based on
(g), T: polishing time (min). survey factors and levels.
The surface quality obtained after polishing is determined • Step 2. Conduct experiment steps and data collection.
by the average value of three measurements at different loca- Ti–6Al–4V alloy polishing steps constructing in the L16
tions using the Mitutoyo SJ-400 roughness meter. Surface orthogonal array.
morphology of the workpiece after polishing observed by • Step 3A. Give empirical relationships and random values
electron microscopy. by algorithm GPR and FGRA.
• Step 3B. Experimental analysis by the Taguchi method
affecting the polishing process of Ti–6Al–4V alloy. The
3 Modeling prediction and optimization experiment analysis results decide to repeat the experiment
of MRF polishing process Al–6Ti–4V alloy or proceed with the previously established investigations.
• Step 4A. Set up the prediction model and optimize surface
3.1 Set up GPR-FGRA hybrid algorithm quality and MRR when polishing Ti–6Al–4V alloy. The
for predicting and optimizing the polishing analysis results give the optimal parameters from specific
process Ti–6Al–4V alloy polishing conditions.
• Step 4B. Given the smallest and largest impact factors
Prediction and optimization for the polishing process in polishing surface quality and MRR. The experimental
Al–6Ti–4V alloy belong to the experiment problem. With the results give the problem of surface quality optimization
approach from Taguchi’s empirical analysis, the GPR-FGRA and MRR. This process makes it cost-effective to polish
hybrid algorithm applying provides predictive parameters the Ti–6Al–4V alloy.

Fig. 5 The structure of the

GPR-FGRA hybrid algorithm

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

• Step 5. Problem solve of identifying and establishing the weaknesses of conventional polishing methods, such as giv-
relationship between network parameters by the FGRA ing experimental relationships when using the GPR-FGRA
algorithm. hybrid algorithm to increase the ability to predict surface
• Step 6. Predict the surface quality, MRR and polishing quality, MRR and that reliability of prediction and optimiza-
interval confidence using MRF according to GPR. tion. Applying GPR-FGRA hybrid algorithm to predict and
• Step 7. Optimization and prediction parameters compared optimize the polishing model with target functions set from
with surface texture parameters and critical threshold the GPR algorithm increases the proposed model’s flexibil-
MRR. Predicted parameters more significant or smaller ity, predictability, and optimization. Adding a new stage to
than the prediction threshold will go to step 8 and back to the polishing model: The experimental polishing model has
step 2, respectively. determined the influence of technology modes on surface
• Step 8. Stop the polishing device when the predicted quality, thereby validating the reliability of the experiment
parameter equals or exceeds the critical value. When the implementation method and data collection.
survey parameters approach, the limit value will give warn-
ings. 3.2 Convergence and optimization solution
of GPR-FGRA
GPR-FGRA hybrid algorithm approach combined
with Taguchi experimental parameters when polishing Hybrid algorithm GPR-FGRA proposed when applied to
Ti–6Al–4V alloy, the disadvantages of the traditional polish- multimodal functions (F10–F12) and non-modal (F1–F9)
ing method are reduced and add new stages. To overcome the [29], as described in Table 7.

Table 7 Nonmodal Benchmark

functions and polynomial Func Description Range Global
equation with MRR and Ra minimum

Nonmodal benchmark functions

F-1 n  x ∈ [−100, 100] 0
xi2
i1

F-2 
B  2 x ∈ [−10, 10] 0
100 × xi2 − xi+1 + (1 − xi )2
i1

F-3 y × [sin(4x)] + 1.1x × [sin(2y)] x, y ∈ [−10, 10] −19.8625

F-4 
n x ∈ [−5.12, 5.12] 0
xi2 − 10 × cos(2π xi ) + 10 × n
i1
       
F-5 1
−0.2 30 B 1
−0.2 30 B x ∈ [−30, 30] 8.8818e−16
i1 cos(2π x i )
2
i1 x i
−20 × e −e
F-6 
n x ∈ [−10, 10] −100.22365
cos(xi ) × xi2 + xi
i1

F-7 n 
 √ x ∈ [−10, 10] −10
|xi | − 10. cos |xi |
i1
 
F-8 sin2

x 12 +x 22 −0.5
x1 , x2 ∈ [−10, 10] 0
 1
0.001× x 12 +x 22 +1
+ 2
 2
F-9 
n
xi 
n xi x ∈ [−10, 10] 0.29289
1− cos √ + 4000
i
i1 i1

Polynomial equation with MRR and surface quality when polishing Ti–6Al–4V alloy
F-10 MRR  −7.60409 + 0.0133498 × nw – 9.86072 × K + 1.76568 × nMRF – 0.0158805 ×
F – 0.0000149638 × nw 2 + 0.00319454 × nw × K + 0.000372297 × nw ×nMRF –
0.00000274302 × nw × F + 1.96309 × nMRF 2 – 0.456901 × K ×nMRF + 0.0175629 ×
K × F – 0.0186909 × nMRF 2 –0 .000326395 × nMRF × F + 0.00000460377 × F 2
F-11 Ra  −0.840419 + 0.000481304 × nw + 0.391957 × K + 0.036819 × nMRF –
0.000553489 × F – 1.73741E-7 × nw 2 – 0.0000710729 × nw × K – 0.00000273123 ×
nw ×nMRF – 2.50774E−8 × nw × F – 0.020918 × nMRF 2 – 0.456901 × K ×nMRF –
0.0109927 × K × F –0 .000336342 × nMRF 2 – 000000371069 × nMRF × F +
1.54043E−7 × F 2

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

The following formula determines the stop criteria of the F-11 increases, thereby reducing the quality of solutions. The
algorithm: average optimization value (AO) corresponding to the func-
tion F-10 is 19.1. However, the AO value with the function
f End (G P R − F G R A[M]) − Glo ≤ S.C A (5) F-11 achieved by GPR-FGRA is 0.025. The results show
that the quality obtained by GPR-FGRA is superior to the
where f End (G P R − F G R A[M]) is the value corre- algorithms mentioned in most cases. Therefore, the proposed
sponding to the solution, which is best obtained at the last hybrid algorithm GPR-FGRA is highly applicable for iden-
iteration of the algorithm, and S.CA is the stop criterion of tification and optimization when polishing Ti–6Al–4V alloy.
the algorithm. This criterion will converge with tolerances
10−6 and 10−3 corresponding to non-modal and multimodal
functions, respectively.
The standard deviation with the obtained results is shown 4 Results and discussion
as follows:
Predicting and optimizing MRR and surface quality is deter-
 1
1 mined through the experimental process. Aim to confirm
n 2
S.AD  [xi − x]2 (6) the accuracy of the prediction model. The experiments were
n
i1 repeated three times with different machining conditions.
Used the initial two datasets for training procedures and the
where x i is the solution result vector run by the algorithm third set of tests to predict the MRR and surface quality.
and x is the average value of solutions determined by the The prediction diagram of MRR and surface quality
following formula: according to different cutting modes is shown in Figs. 7 and
8. The results of MRR prediction, as described in Fig. 7,
1
n
x xi (7) show that the prediction error is less than 4.24% (the pre-
n diction accuracy is more than 95.76%). The accuracy of the
i1
GPR-FGRA hybrid algorithm for surface roughness predic-
The results include the best, mean and worst values and tion is higher than 99.63%, as illustrated in Fig. 8. Thus, the
standard deviation (S.AD) after 50 times of independent predicted parameters from the GRP-FGRA algorithm close
implementation with the proposed GPR-FGRA and then with measurements and provide the confidence interval of
compared with GSA, ABC, BBO, PSO and TLBO algorithms the proposed prediction is 98%, as shown in Figs. 7 and 8.
as shown in Table 8. Figure 9 shows the SEM surface morphology before and
Table 8 shows the proposed GPR-FGRA for the best per- after finishing with MRF proposed for some experiments
formance. The GPR-FGRA clearly shows its superiority and polishing Ti–6Al–4V. Figure 9 shows the results of surface
effectiveness in finding a search solution and has a lower stan- roughness. Experimental results show that corresponding
dard deviation than other algorithms. Furthermore, in the case to different polishing technology modes will have different
of the worst solution, GPR-FGRA has shown its excellence MRR and surface quality. It can see that the polishing exper-
amongst the different algorithms by benchmark functions. iments all yielded a significantly improved surface quality
Hence, with non-modal objective functions, in terms of solu- compared to the original grinding surface. Figure 10 depicted
tions and quality, GPR-FGRA shows superiority to other the surface roughness (Ra) before and after finishing the
well-known algorithms. However, this superiority does not experiments. The workpiece surfaces Ti–6Al–4V alloy ini-
manifest in all benchmark functions following the conver- tial (Ra ≈ 0.35 μm) significantly improve with Ra < 0.1 μm
gence capability. The convergence feature of GPR-FGRA under different test conditions.
was investigated and compared with the optimal algorithms Experimental results show that experiment number 13
of GSA, BBO, PSO, ABC and PLBO. Figure 6 shows the gives the best surface quality corresponding to polishing dis-
excellent convergence rate of GPR-FGRA compared to pre- tance K  1 mm, workpiece speed 1330 rpm, MRF carrier
viously mentioned algorithms. speed 40 rpm and feed rate F  500 mm/min. A mirror surface
When applying algorithms for polynomial functions with Ra  0.040 μm is produced by these polishing parameters.
MRR and surface roughness (F-10, F-11) based on time, the With experiment 4, polishing distance K  1.75 mm, spindle
average value (mean value of all algorithm runs) and the opti- speed 530 rpm, MRF carrier speed 40 rpm and feed rate F
mal average value (average number of peaks found above 50  900 mm/min, for maximum roughness but still shown the
runs with S.CA  1000), as shown in Table 9. The results mirror surface Ra  0.098 μm after this polishing process.
show that the possibility of polynomial functions with MRR Experimental procedures showed different surface qualities
and surface roughness is not as high as non-modal functions. produced by other technological parameters when polish-
The complexity of the benchmark functions from F-10 and ing Ti–6Al–4V alloy. However, the polishing surfaces are all

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

Table 8 Optimal solution obtained on non-modal functions

Func Characteristics Optimization algorithms

GSA BBO PSO ABC TLBO GPR-FGRA

F-1 Best 5.8265 4.4548 4.15 e−14 0.0369 3.25 e−6 0

Medium 15.2532 17.3493 8.98 e−12 0.4283 9.35 e−5 0
Worst 24.0363 52.3655 1.84 e−10 1.4497 0.0011 0
S.AD 4.0362 10.7951 2.69 e−11 0.3498 1.81 e−4 0
F-2 Best 0.0085 0.0305 3.14 e−16 5.62 e−7 6.82 e−5 0
Medium 0.348 1.1483 375 e−9 0.0016 2.43 e−2 0
Worst 1.1295 5.5835 4.96 e−8 0.0175 0.1656 0
S.AD 0.2915 1.3058 1.04 e−8 0.0036 0.0325 0
F-3 Best − 19.7734 − 19.7338 − 19.8625 − 19.8624 − 19.8625 − 19.8625
Medium − 18.5456 − 18.3465 − 17.0658 − 19.8624 − 19.8623 − 19.8625
Worst − 17.4263 − 16.8586 − 11.6432 − 19.8622 − 19.8584 − 19.8626
S.AD
0.5189 0.6635 1.9550 4.46 e−10 5.72 e−04 1.68 e−06
F-4 Best 1.0313 0.0112 0 1.98 e−6 0 0
Medium 4.2553 0.6125 1.3720 0.3132 2.53 e−13 0
Worst 7.7232 1.7141 4.9718 1.6948 5.02 e−12 0
S.AD 1.5538 0.4819 1.3154 0.4332 8.79 e−13 0
F-5 Best 4.8954 1.4845 5.33 e−8 0.1133 0.0028 8.88 e−16
Medium 6.1778 2.8952 0.1716 0.4016 0.0452 8.88 e−16
Worst 7.3043 4.7716 1.6463 1.3312 0.9551 8.88 e−16
S.AD 0.5293 0.5990 0.4348 0.2695 0.1333 0
F-6 Best − 200.4396 − 200.4472 − 200.4476 − 200.4476 − 200.4473 − 200.4474
Medium − 199.0867 − 199.9673 − 150.7991 − 200.0686 − 200.4472 − 200.4475
Worst − 193.2685 − 198.2015 − 89.8413 − 181.5917 − 200.4475 − 200.4470
S.AD
1.3388 0.5862 42.3978 2.6665 1.56 e−12 1.07 e−04
F-7 Best − 45.7485 − 87.5372 − 76.7760 − 76.2008 − 80.2418 − 100
Medium − 39.1212 − 77.646 − 65.9420 − 65.3394 − 73.1428 − 100
Worst − 32.0332 − 71.1917 − 61.2958 − 59.7663 − 67.8235 − 100
S.AD
3.5163 3.8413 4.2107 3.4835 2.8902 0
F-8 Best 3.61 e−5 2.80 e−5 0 0 1.23 e−4 0
Medium 0.0074 0.0403 0.0162 9.46 e−10 0.0165 0
Worst 0.0266 0.0446 0.0448 4.71 e−8 0.0403 0
S.AD 0.0065 0.0124 0.0218 6.66 e−9 0.0105 0
F-9 Best 0.3206 0.0167 0 4.38 e−4 6.35 e−5 0
Medium 0.5895 0.0578 0.0398 0.1269 0.0336 0
Worst 0.7472 0.1128 0.1228 0.4573 0.0458 0
S.AD 0.1002 0.0223 0.0314 0.1068 0.0165 0

mirror-like, thereby demonstrating the effectiveness of the the surface quality obtained from our experiments is sig-
proposed polishing method. nificantly improved compared to the method proposed by
The polishing experiments with MRF performed by Parameswari et al. [30]. Our study resulted in the best sur-
Parameswari et al. [30] apply to the polishing process of face quality corresponding to Ra  40 nm, and the surface
Ti–6Al–4V alloy from grinding surfaces with an average quality improved by 57.44% compared with the proposed
surface roughness Ra  0.415 mm. After polishing by method by Parameswari et al. [30].
Parameswari et al. [30], the workpiece obtained the best sur- Figures 11 and 12 show a 3D image visually describing
face quality with Ra  94 nm (0.094 μm). As described in the relationship between input parameters and survey param-
Figs. 9 and 10, our experimental results show that the surface eters Ra and MRR given by the GPR-FGRA. Figure 11a, d
quality from the proposed method is significantly improved illustrate the relationship between working distance (K) and
compared to before polishing. The experiment processes cor- workpiece speed, MRF carrier speed. The slight MRR values
responding to different polishing technology modes will have correspond to a high working distance (K), workpiece speed,
different MRR and surface quality. However, it can see that and low MRF carrier speed. The MRR value increases with
decreasing working distance (K) and increasing workpiece

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

Fig. 6 Ability to converge with

different algorithms

Table 9 Optimal solution with

polynomial functions with MRR Func Characteristics Optimization algorithms
and surface roughness
GSA BBO PSO ABC TLBO GPR-FGRA

F-10 Time reached 25.5 31 37.5 40 41.5 24.01

OA value 16 17.5 18 18.2 18.5 19.1
F-11 Time reached 26 31.5 38.2 41 42.3 24
OA value 0.038 0.035 0.033 0.031 0.028 0.025

speed and MRF carrier speed. It can find that the MRR values speed increase. The cutting speed between the APs and the
decrease as the working distance (K) increases. As our previ- workpiece increases, thereby increasing the ability to remove
ous study [28] mentioned, the higher working distance leads residual material.
to the APs force acting on the surface machining more minor The results obtained in Fig. 11b, c show that the relation-
in the MRF process; therefore, MRR will reduce. K  1 mm ship between the feed rate, the MRF carrier speed and the
at the working distance for the highest material removal rate workpiece speed to MRR value is quite complex. Figure 14b
capacity. When the workpiece speed and the MRF carrier illustrates that the workpiece speed in the range of 1000–1200

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

Fig. 7 MRR prediction under

different polishing experiments
conditions

Fig. 8 Predictive results of

surface quality under different
polishing conditions

Fig. 9 Surface roughness and morphology after MRF polishing of Ti–6Al–4V alloy under different experimental conditions

(rpm) and the MRF carrier speed of 25–30 rpm gives a MRR when the workpiece speed is in the range of 800–1000
small MRR value. With the workpiece speed in the range (rpm) with a feed of 300–500 (mm/min) for small MRR val-
of 1000–1200 (rpm) and the MRF carrier speed increased ues, whereas increasing the feed rate between 300 and 500
to 35–40 rpm, the MRR value increases rapidly. Figure 11d (mm/min) for high MRR. The relationship between work-
shows the influence of workpiece speed and feed rate on the piece speed, feed rate and MRF carrier speed to MRR is

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

The analysis results in Fig. 12 show that the relation-

ship between technological parameters and surface quality
is quite complicated. However, it can see that the slight sur-
face roughness corresponding to the range of the working
distance, workpiece speed, feed rate and MRF carrier speed
is 1–1.2 mm, 500–600 rpm, 25–30 (mm/min) and 25–35 rpm,
respectively. The analysis results show that when the work-
ing distance is low (K  1–1.2 mm), the impact force of the
APs on the workpiece increases, thereby giving an excellent
ability to remove the workpiece material leading to surface
roughness decreases. In comparison, the high rotation of the
MRF carrier speed will cause uneven material removal as
the abrasive particles to move away from the centre due to
centrifugal force in the MRF process. Therefore, the surface
Fig. 10 Surface roughness of workpiece before and after 90 min of quality degrades with higher MRF carrier speed.
polishing under different test conditions To determine the influence of technological factors such
as workpiece speed, polishing distance, MRF carrier speed
and feed rate to Ra and MRR. The factors affecting MRR
complex. Still, it seems that a high MRR value corresponds
and Ra are determined through the GPR-FGRA coefficient
to a significant workpiece speed, feed rate and MRF car-
with comparison and reference matrix as described below:
rier speed. With these factors increased, the relative cutting
speed between the APs and the workpiece increases, thereby
improving the ability to remove excess workpiece material.

Fig. 11 Effect of workpiece speed, working distance, MRF carrier and F to MRR

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

Fig. 12 Effect of workpiece speed, working distance, MRF carrier and F to Ra

⎡ ⎡⎤ ⎤
Y1 (x) 530 530 530 530 720 720 720 720 980 980 980 980 1330 1330 1330 1330
⎢Y ⎥ ⎢
⎢ 2 (x) ⎥ ⎢ 1
⎢ ⎥ 1, 25 1, 50 1, 75 1 1, 25 1, 50 1, 75 1 1, 25 1, 50 1, 75 1 1, 25 1, 50 1, 75 ⎥
⎥
⎢ Y3 (x) ⎥  ⎢ ⎥
⎢ ⎥ ⎢ 25 30 35 40 30 25 40 35 35 40 25 30 40 35 30 25 ⎥ (8)
⎢ ⎥ ⎢ ⎣ 750 ⎦
⎥
⎣ Y4 (x) ⎦ 250 500 750 900 750 900 250 500 750 900 500 250 500 250 900
Y (x) 13, 33 11, 66 10 7, 67 14, 15 12, 83 10, 50 9, 21 15, 30 13, 26 10, 83 9, 33 15, 92 14, 02 11, 17 9, 60

⎡ ⎡⎤ ⎤
Z 1 (x)
530 530 530 530 720 720 720 720 980 980 980 980 1330 1330 1330 1330
⎢ Z (x)⎥ ⎢
⎢ 2 ⎥ ⎢ 1
⎢ ⎥ ⎢ 1, 25 1, 50 1, 75 1 1, 25 1, 50 1, 75 1 1, 25 1, 50 1, 75 1 1, 25 1, 50 1, 75 ⎥⎥
⎢ Z 3 (x)⎥  ⎢ 25 ⎥
⎢ ⎥ ⎢ 30 35 40 30 25 40 35 35 40 25 30 40 35 30 25 ⎥ (9)
⎢ ⎥ ⎥
⎣ Z 4 (x)⎦ ⎣ 250 500 750 900 750 900 250 500 750 900 500 250 500 250 900 750 ⎦
Z (x) 0, 041 0, 059 0, 089 0, 098 0, 044 0, 050 0, 057 0, 079 0, 067 0, 080 0, 081 0, 091 0, 040 0, 059 0, 062 0, 071

Figures 9 and 10 illustrate four factors of fuzzy mem-

bership classes that affect polishability through the output and Ra. At the lowest level is the workpiece speed change
parameters Ra and MRR. The fuzzy member layer results (0.5921), followed by MRF carrier speed (0.7383) and the
analysis showed that with MRR, the highest index corre- feed rate (0.7487), with the shift of polishing distance K valid
sponds to the workpiece speed (0.7654), then MRF carrier the largest (0.7805). Thus surface quality will be affected by
speed (0.7148) and feed rate (0.6982). Meanwhile, the work- polishing distance (K) being the largest and the workpiece
ing distance has the lowest value (0.6185). There is a speed being the smallest. Whereas MRR tends opposite to Ra,
significant difference in the analytical results between MRR the workpiece speed has the most significant effect, while the
working distance K has the slightest influence on MRR.

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

Fig. 13 The results analysis of

the fuzzy grey relationship with
MRR

Fig. 14 The results analysis of

the fuzzy grey relationship with
Ra

As shown in Figs. 13 and 14, the grey Euclidean relation According to the GPR-FGRA algorithm, the proposed
layer of the four elements in the polishing process, show optimization model gives the optimal parameters for the best
negligible differences. However, it still shows that with MRR, surface quality. The results of different optimization mod-
the grey Euclidean relation layer of the workpiece speed is els as described in Table 10. The surface quality obtained
highest. The grey Euclidean relation layer of the working with the GPR-FGRA proposed algorithm is significantly
distance (K) is the highest with Ra. Therefore, when it is improved (Ra  0.025 mm) compared with the Taguchi
necessary to improve the surface quality, we should focus experimental method (Ra  0.040 mm). Thus, the surface
on changing the working distance (K), then the MRF carrier quality increased by 37.5% when choosing the optimized
speed and feed rate. Meanwhile, changing workpiece speed cutting parameter with the GPR-FGRA hybrid algorithm.
has the most influence on MRR compared to changing the
remaining parameters.
The fuzzy grey relationship layer of the four factors up to
the surface polishability, as shown in Figs. 13 and 14. The 5 Conclusions
analysis provides an overview to evaluate survey factors that
affect the MRR and Ra. The fuzzy grey relationship layer This study introduced a new polishing method with MRF
for the Ra of the four elements (workpiece speed, working containing inexpensive CIPs and APs. The carrier liquid is
distance, MRF carrier speed and feed rate) corresponds to a safe and eco-friendly cutting oil solution combined with a
0.6983, 0.8057, 0.7818, and 0.7817. Whereas with MRR, new GPR-FGRA hybrid algorithm for an ultra-precise mirror
the fuzzy grey relationship layer to technological parameters polishing surface Ti–6Al–4V alloy. According to work done,
is shown by 0.7995, 0.7122, 0.7678 and 0.7555. Therefore, the main conclusions as description follows:
when MRF polishes Ti–6Al–4V alloy, the working distance
must first be chosen when super smooth surface quality is
required. Workpiece speed is the priority when increasing
productivity and reducing polishing time. • Introduced ultra-precise machining models for expensive,
difficult-to-machine materials from highly commercial

123
 International Journal on Interactive Design and Manufacturing (IJIDeM)

Table 10 The best experimental

and optimized cutting parameter Optimize surface quality nw K nMRF F Ra
results (rpm) (mm) (rpm) (mm/min) (μm)

Best cutting parameters by 1330 1 40 500 0.040

Taguchi experiment
Optimal cutting parameters by 1350.28 1.15 28.18 404.85 0.025
GPR-FGRA

and environmentally friendly materials. The polishing car- References

rier liquid is established based on miscible oil cutting fluid
of natural origin, low-cost CIPs and APs. 1. Yin, S., Nguyen, D., Chen, F., Tang, Q., Duc, L.A.: Application of
compressed air in the online monitoring of surface roughness and
• A new GPR-FGRA hybrid algorithm proposes for predict- grinding wheel wear when grinding Ti–6Al–4V titanium alloy. Int.
ing and optimizing MRF polishing of Ti–6Al–4V alloy. J. Adv. Manuf. Technol. 101, 1315–1331 (2018)
From analytical results with the proposed algorithm, the 2. Nguyen, D., Yin, S., Tang, Q., Son, P.X., Duc, L.A.: Online moni-
quality obtained by GPR-FGRA is superior to the algo- toring of surface roughness and grinding wheel wear when grinding
Ti–6Al–4V titanium alloy using ANFIS-GPR hybrid algorithm and
rithms mentioned in most cases. Thereby shows the high
Taguchi analysis. Precis. Eng. 55, 275–292 (2019)
applicability of GPR-FGRA for the polishing problem and 3. Ju, J., Li, J., Yang, C., Wang, K., Kang, M., Wang, J.: Evolution
application to other prediction and optimization objects. of the microstructure and optimization of the tensile properties of
• The polishing Ti–6Al–4V material yielded the fuzzy grey the Ti–6Al–4V alloy by selective laser melting and heat treatment.
Mater. Sci. Eng. A 802, 140673 (2021)
relationship layer for the Ra of the four elements (work-
4. Tümer, D., Güngörürler, M., Havıtçıoğlu, H., Arman, Y.: Investiga-
piece speed, working distance, MRF carrier speed and tion of effective coating of the Ti–6Al–4V alloy and 316L stainless
feed rate) corresponding to 0.6983, 0.8057, 0.7818, and steel with graphene or carbon nanotubes with finite element meth-
0.7817. Whereas with MRR, the fuzzy grey relationship ods. J. Mater. Res. Technol. 9, 15880–15893 (2020)
5. Siva Surya, M., Gugulothu, S., Prasanthi, G.: Optimization of cut-
layer to technological parameters is shown by 0.7995, ting parameters while turning Ti–6Al–4V using response surface
0.7122, 0.7678 and 0.7555. Therefore, when MRF pol- methodology andmachine learning technique. Int. J. Interact. Des.
ishes Ti–6Al–4V alloy, the working distance must first Manuf. (IJIDeM) 15, 453–462 (2021)
be chosen when super smooth surface quality is required; 6. Qian, N., Fu, Y., Chen, J., Khan, A.M., Xu, J.: Axial rotating heat-
pipe grinding wheel for eco–benign machining: a novel method
workpiece speed is the priority when increasing produc- for dry profile-grinding of Ti–6Al–4V alloy. J. Manuf. Process. 56,
tivity and reducing polishing time. 216–227 (2020)
• The analysis results with the proposed MRF polishing 7. Zhao, B., Jiang, G., Ding, W., Xiao, G., Huan, H., Wang, Y.,
model get out superfine Ti–6Al–4V alloy surface can be et al.: Characterisation of the wear properties of a single-aggregated
cubic boron nitride grain during Ti–6Al–4V alloy grinding. Wear
produced with the surface roughness Ra  0.025 μm 452–453, 203296 (2020)
according to the parameters K  1.15 mm, nw  8. Siva Surya, M.: Optimization of turning parameters while turn-
1350.28 rpm, nMRF  28.18 rpm and F  404.85 mm/min. ing Ti–6Al–4V titanium alloy for surface roughness and material
The optimization surface quality increased by 37.5% and removal rate using response surface methodology. Mater. Today
Proc. 62, 3479–3484 (2022)
gave the prediction accuracy greater than 99.63% when 9. Luo, H., Yin, S., Zhang, G., Liu, C., Tang, Q., Guo, M.: Opti-
choosing the optimized and prediction with the GPR- mized pre-thinning procedures of ion-beam thinning for TEM
FGRA hybrid algorithm, respectively. Hence the polishing sample preparation by magnetorheological polishing. Ultrami-
process using MRF and the proposed algorithm with croscopy 181, 165–172 (2017)
10. Duy Trinh, N., Nhat Tan, N., Quang, N.M., Thi Thieu Thoa, P., Duc,
inexpensive CIP and AP particles and vegetable oil car- L.A.: Application of magnetic liquid slurries and fuzzy grey anal-
rier liquid has excellent potential to create an ultra-fine ysis in polishing nickel-phosphorus coated SKD11 steel. Particul.
Ti–6Al–4V surface without affecting the environment and Sci. Technol. 40, 401–414 (2021)
high efficiency. 11. Kumar, S., Jain, V.K., Sidpara, A.: Nanofinishing of freeform
surfaces (knee joint implant) by rotational-magnetorheological
abrasive flow finishing (R-MRAFF) process. Precis. Eng. 42,
165–178 (2015)
12. Das, M., Barman, A.: Force analysis during spot finishing of tita-
nium alloy using novel tool in magnetic field assisted finishing
process. Int. J. Precis. Technol. 8, 190 (2019)
13. Barman, A., Das, M.: Toolpath generation and finishing of bio-
titanium alloy using novel polishing tool in MFAF process. Int. J.
Declarations Adv. Manuf. Technol. 100, 1123–1135 (2019)

Conflict of interest The authors have nothing to disclose and no conflict

of interests regarding the publication of this paper.

123
International Journal on Interactive Design and Manufacturing (IJIDeM)

14. Nagdeve, L., Jain, V.K., Ramkumar, J.: Preliminary investigations 24. Hema Latha, K., Usha Sri, P., Seetharamaiah, N.: Design and
into nano-finishing of freeform surface (femoral) using inverse manufacturing aspects of magneto-rheological fluid (MRF) clutch.
replica fixture. Int. J. Adv. Manuf. Technol. 100, 1081–1092 (2019) Mater. Today Proc. 4, 1525–1534 (2017)
15. Barman, A., Das, M.: Magnetic field assisted finishing process for 25. Shahinian, H., Cherukuri, H., Mullany, B.: Evaluation of fiber-
super-finished Ti alloy implant and its 3D surface characterization. based tools for glass polishing using experimental and computa-
J. Micromanuf. 1, 251659841878550 (2018) tional approaches. Appl. Opt. 55, 4307–4316 (2016)
16. Singh, H., Prakash, C., Singh, S.: Plasma spray deposition of 26. Fengdong, Z., Xuejun, Z., Jingchi, Y.: Mathematics model of mag-
HA-TiO2 on β-phase Ti–35Nb–7Ta–5Zr alloy for hip stem: char- netorheological finishing. In: Proceedings of the SPIE (2000)
acterization of bio-mechanical properties, wettability, and wear 27. Pal, A., Chatha, S.S., Sidhu, H.S.: Experimental investigation on
resistance. J. Bionic Eng. 17, 1029–1044 (2020) the performance of MQL drilling of AISI 321 stainless steel using
17. Rao, P.S., Jain, P.K., Dwivedi, D.K.: Optimization of key process nano-graphene enhanced vegetable-oil-based cutting fluid. Tribol.
parameters on electro chemical honing (ECH) of external cylin- Int. 151, 106508 (2020)
drical surfaces of titanium alloy Ti 6Al 4V. Mater. Today Proc. 4, 28. Nguyen, D., Wu, J., Quang, N.M., Duc, L.A., Son, P.X.: Applying
2279–2289 (2017) fuzzy grey relationship analysis and Taguchi method in polishing
18. Kong, D., Chen, Y., Li, N.: Gaussian process regression for tool surfaces of magnetic materials by using magnetorheological fluid.
wear prediction. Mech. Syst. Signal Process. 104, 556–574 (2018) Int. J. Adv. Manuf. Technol. 112, 1675–1689 (2021)
19. Zhang, J., Wang, T., Xu, L., Wang, P., Zhang, M., Xu, M.: Develop- 29. Duc, L.A., Li, K., Nguyen, T.T., Yen, V.M., Truong, T.K.: A new
ment and validation of a prognostic model based on the albumin- effective operator for the hybrid algorithm for solving global opti-
to-fibrinogen ratio (AFR) and gamma-glutamyl transpeptidase-to- misation problems. Int. J. Syst. Sci. 49, 1088–1102 (2018)
platelet ratio (GPR) in hepatocellular carcinoma patients. Clin. 30. Parameswari, G., Jain, V.K., Ramkumar, J., Nagdeve, L.: Experi-
Chim. Acta 511, 107–116 (2020) mental investigations into nanofinishing of Ti6Al4V flat disc using
20. Olabanji, O.M., Mpofu, K.: Appraisal of conceptual designs: coa- magnetorheological finishing process. Int. J. Adv. Manuf. Technol.
lescing fuzzy analytic hierarchy process (F-AHP) and fuzzy grey 100, 1055–1065 (2019)
relational analysis (F-GRA). Results Eng. 9, 100194 (2021)
21. Zeng, J.E.Y., Jin, Y., Zhang, B., Huang, Z., Wei, K., et al.: Heat
dissipation investigation of the power lithium-ion battery module
Publisher’s Note Springer Nature remains neutral with regard to juris-
based on orthogonal experiment design and fuzzy grey relation
dictional claims in published maps and institutional affiliations.
analysis. Energy 211, 118596 (2020)
22. Roozbahani, A., Ghased, H., Hashemy Shahedany, M.: Inter-basin
Springer Nature or its licensor holds exclusive rights to this article
water transfer planning with grey COPRAS and fuzzy COPRAS
under a publishing agreement with the author(s) or other rightsholder(s);
techniques: a case study in Iranian Central Plateau. Sci. Total Env-
author self-archiving of the accepted manuscript version of this article
iron. 726, 138499 (2020)
is solely governed by the terms of such publishing agreement and appli-
23. Siva Surya, M., Vepa, K.S., Karanam, M.: Optimization of
cable law.
machining parameters using ANOVA and grey relational analy-
sis while turning aluminium 7075,". Int. J. Recent Technol. Eng.
8, 5682–5686 (2019)

123