Professional Documents
Culture Documents
Grasping Point Optimization For Sheet Metal Part Based On GSA Kriging Model in A Multi Robot Assembly System
Grasping Point Optimization For Sheet Metal Part Based On GSA Kriging Model in A Multi Robot Assembly System
Grasping Point Optimization For Sheet Metal Part Based On GSA Kriging Model in A Multi Robot Assembly System
https://doi.org/10.1007/s00170-023-10835-1
ORIGINAL ARTICLE
Received: 2 June 2022 / Accepted: 8 January 2023 / Published online: 19 January 2023
© The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2023
Abstract
Automobile flexible sheet metal parts are prone to produce non-negligible grasping deformation, and choosing an appropriate
layout of grasping points is an important means to reduce the grasping deformation for sheet metal parts. It is challenging
to optimize the layout of grasping points efficiently and accurately because of the large number of robot fingers due to a
large size sheet part. In order to improve the optimization efficiency and reduce the computational cost, this paper proposes
a two-stage optimization design method for the layout of grasping points based on GSA-Kriging surrogate model. First, the
number of robot fingers and their feasible range are determined according to the stability of robotic grasping operation and
the degree of deformation at different positions. Then, according to the position distribution of grasping points, they are
divided into two stages for optimization. Finally, GSA-Kriging surrogate model and gravitational search algorithm (GSA)
are used to find the optimal layout of grasping points. In this paper, the layout of grasping point optimization of a certain car
door sheet part is utilized as a case to validate the proposed method and we have designed an experimental system to test
the proposed grasp method on a curved sheet part. The result shows that the GSA-Kriging surrogate model is more accurate
and more stable than Kriging and other surrogate models. At the same time, the two-stage optimization method improves
the optimization efficiency while reducing the calculation cost and burden.
Keywords Sheet metal · Grasping point layout · GSA-Kriging surrogate model · Optimization method · Two-stage
13
Vol.:(0123456789)
2226 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
of the workpiece throughout the cutting process by alter- response surface model, RBF, BP neural network, and sup-
nately changing the fixture layout and clamping force. Xing port vector machine regression. In recent years, many kinds
et al. [5] proposed a new method to optimize fixture scheme of research on grasping point layout optimization design
by a non-domination sorting social radiation algorithm based on surrogate model have been carried out by scholars
(NSSRA). The case study results suggested that NSSRA at home and abroad.
has better efficiency and higher accuracy than NSGA-II. By Hamedi [13] constructed a BPNN surrogate model
combining finite element analysis with ant colony/genetic between clamping force, contact state, and the maximum
hybrid algorithms, Zhou et al. [6] optimized the positioning elastic deformation of the workpiece through finite ele-
layout of the fixture and reduced the maximum deformation ment analysis with fewer times and obtained the optimal
of sheet metal under gravity conditions. Liao [7] realized clamping force of the fixture through genetic algorithms
the optimization design of the number and position of the to reduce the maximum elastic deformation of the work-
assembly fixtures based on the optimization method coupled piece in the process of processing. Sundararaman et al.
with the finite element analysis based on genetic algorithms [14] used response surface method to model the relation-
to minimize the deformation of the workpiece under the ship between the position of positioner and fixture and
gravity effect and due to the size change of the parts. Huang the maximum deformation of the workpiece in the pro-
et al. [8] proposed a synchronous optimization method of cess of end milling and optimized the established model
clamping layout based on genetic algorithms and a para- by using genetic algorithms and particle swarm optimi-
metric simulation model to minimize the maximum defor- zation algorithm. Qin et al. [15] constructed a BPNN
mation of the workpiece after milling. Xing [9] proposed a model describing the functional relationship between
fully automated and efficient method of fixture layout opti- clamping parameters and workpiece deformation and
mization based on the combination of 3dcs simulation (for combined BPNN with genetic algorithms to develop a
dimensional analyses) and global optimization algorithms unified method to complete multi-fixture layout plan-
and applied social radiation algorithm (SRA) and GAOT to ning for sheet metal part workpiece processing to control
the proposed method, the case study results showed that the clamping deformation. Yang et al. [16] constructed the
GAOT algorithm was more suitable than SRA for generat- support vector machine regression model of the func-
ing the optimal fixture layout with excellent efficiency for tional relationship between fixture locating point and
engineering applications. Milad et al. [10] proposed an opti- workpiece strain energy and compared it with the RBF
mization method based on ant colony algorithm and finite neural network model, and it was found that its accu-
element analysis to optimize the number of fixture elements racy was higher. In order to minimize the surface shape
and positioning layout with the part deviation as the optimi- error of glass laser components, Su et al. [17] proposed
zation objective. Vishnupriyan et al. [11] adopted genetic a multi-kernel support vector function regression model
algorithms to optimize fixture layout and clamping force and constructed a two-layer support vector machine
simultaneously under the condition of considering geometric regression model of device deformation response with
error, to reduce the overall machining error of the workpiece respect to ambient temperature and clamping force. Sub-
to the maximum extent. In order to improve the precision sequently, the branch and bound algorithm were used to
of precision machining of turbine blades, Wang et al. [12] optimize the clamping force. Rex and Ravindran [18]
optimized the positioning points and clamping points of the proposed a comprehensive method to optimize fixture
workpiece based on finite element analysis and neighbor- layout design, by establishing BPNN model to predict the
hood random search algorithm. deformation of the work-fixture system, so as to control
To sum up, it can be seen that the combination of finite the maximum elastic deformation of the workpiece in
element analysis and intelligent evolutionary algorithm has the whole processing process. Wang et al. [19] adopted
become the most common and practical method for grasping a method based on radial basis neural network and bat
point layout optimization design in engineering. However, optimization algorithm to optimize the positioning lay-
the direct combination of intelligent algorithm and finite out of fixture, aiming to minimize the overall deforma-
element analysis often faces huge computational pressure tion of sheet metal under gravity. Gino Dini and Franco
and requires a lot of mobilization of finite element analysis, Failli [20] proposed a computer-aided module for the
which both increases the computational cost and reduces the detection of grasps of 3D objects by using RBF. The
efficiency optimization design, while the surrogate model neural networks are properly tailored and trained in order
can largely solve this problem. Using a surrogate model to evaluate and compare the different grasping alterna-
instead of a finite element model can avoid expensive and tives according to geometrical and technological aspects
time-consuming finite element simulation analysis and of object surfaces. Qi et al. [21] based on the principle
improve the efficiency of grasping point layout optimization. of simultaneously minimizing the sum of squares and
Common surrogate models include Kriging, polynomial variance of errors at each point of the sheet metal and
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2227
13
2228 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2229
robot fingers under a given number to minimize the elas- positive correlation between the strain energy and the elastic
tic deformation of sheet part. Most of the existing studies deformation, so when the elastic deformation is the small-
evaluate the deformation degree of the sheet metal part by est, the internal strain energy is also the smallest. Therefore,
the normal overall deformation, maximum deformation, or in order to take into account the overall deformation of the
deformation at key points and ignore the deformation of sheet metal part, this paper adopts the overall strain energy
sheet metal part in other directions and overall deforma- of the sheet metal part as the evaluation index of grasping
tion. According to the functional principle of the deformed point layout optimization. The design variables are the loca-
body and the existing studies, the work done by the sheet tions of the grasping points:X = [x1 , x2 , ⋯ xi , ⋯ , xj , ⋯ , xN ].
metal part under the action of gravity and other external The (feasible domain set of grasping ) point layout is
forces is equal to the strain energy stored in it. There is a Ω = w1 , w2 , ⋯ , wi , ⋯ , wj , ⋯ , wN , t he optimization
13
2230 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
Find ∶ X = [x1 , x2 , ⋯ xi , ⋯ , xj , ⋯ , xN ]
∑K
Min ∶ U(X) = u (X)
i=1 i
(3)
s.t.
x1 ∈ w1 , ⋯ , xN ∈ wN
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2231
Table 3 Initial layout of Primary grasping Coordinate values (X × Z) Edge grasping Coordinate values (X × Z)
grasping points point point
3 GSA‑Kriging prediction model function, and the correlation function is selected as Gaussian
correlation function below:
3.1 Kriging surrogate model [ ]
( ) ∑m | i j |2
i j
R 𝜃, X , X = 𝑒𝑥𝑝 − 𝜃 |X − Xk | (5)
k=1 k | k |
The Kriging model, named after South African mining
engineer Krige, was introduced into the field of engineer- where m represents the dimension of the design vari-
ing design by Sacks et al. [28]. The Kriging model has the able, and 𝜃k represents the correlation coefficient of the k
advantages of accurate interpolation and prediction at the -th variable.
design sample points and has a good fitting effect for strong
non-linear problems. Standard Kriging models generally 3.2 Gravitational search algorithm (GSA)
include a regression part and a random function [29]. The
basic principles of the Kriging model are as follows. The gravitational search algorithm (GSA) is a swarm intel-
Given input vector x ∈ Rm , the mathematical expression ligence optimization algorithm based on the law of gravity
of the Kriging model can be expressed as Eq. (4). and the law of motion proposed by Esmat Rashdi et al. [30]
y(X) = F(𝛽, X) + Z(X) (4) in 2009. It seeks the optimal solution through particle posi-
tion movement of the population. The optimal solution is
where F(𝛽, X) is the global polynomial regression model, obtained when the particle moves to the optimal position.
𝛽 is the regression coefficient, and Z(X) is the local random The difference between GSA and other swarm intelli-
process model, representing a random process( with)zero gence search algorithms is that GSA mainly uses the interac-
mean and satisfying
[ ( i ) the( normal
)] distribution
( )N i 0, 𝜎 jand
2
tion force between particles to make the particles move, thus
covariance 𝑐𝑜𝑣 Z X , Z X = 𝜎 R 𝜃, X , X ,X and X are
j 2 i j
achieving the position update without the need of environ-
two design variables, 𝜎 2 is the variance of a random pro- mental information of the population. GSA has fast solving
cess, and 𝜃 is the correlation coefficient vector. In this paper, speed and strong global searching ability. The basic steps of
the polynomial regression part is selected as constant basis GSA are summarized in Table 1.
Kriging predicting model output
20 20
Expected output
18 Kriging prediction output
18
Stress deformation energy(mJ)
16
16
Kriging prediction output
14
14
12
12
10
10
8
8
6
4 6
2 4
0 5 10 15 20 25 30 35 40 4 6 8 10 12 14 16 18 20
Testing Sample set by LHS Expected output
13
2232 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
14 14
12 12
10 10
8 8
6 6
4 4
0 5 10 15 20 25 30 35 40 4 6 8 10 12 14 16 18 20
Testing Sample set by LHS Expected output
The updating rules of particle velocity and position are Under the given training sample condition, according to the
as Eq. (6) maximum likelihood estimation theory, the likelihood func-
tion is expressed as follows:
vdi (t + 1) = randi × vdi (t) + adi (t)randi ∈ [0, 1] [ ]
1 (Y − F𝛽)T R−1 (Y − F𝛽)
L= ( )n 𝑒𝑥𝑝 − (7)
1
2𝜋𝜎 2 2 |R| 2 2𝜎 2
xid (t + 1) = xid (t) + vdi (t + 1) (6)
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2233
20 20
18 18
GSA-Kriging prediction output
16 16
12 12
10 10
8 8
6 6
4 4
4 6 8 10 12 14 16 18 20 4 6 8 10 12 14 16 18 20
Expected output Expected output
18 18
16 16
BPNN prediction output
SVR prediction output
14 14
12 12
10 10
8 8
6 6
4 4
4 6 8 10 12 14 16 18 20 4 6 8 10 12 14 16 18 20
Expected output Expected output
Fig. 11 Comparison of prediction effects of several models. (a) GSA-Kriging, (b) RBF, (c) SVR, and (d) BPNN
by selecting an appropriate 𝜃 and converting it into an uncon- process 𝜎 2 could be obtained by plugging in the correspond-
strained minimization form in non-logarithmic form. Then, ing formula. Then, the GSA-Kriging model was obtained,
the optimization model of maximum likelihood estimation and the predicted value and variance of the unknown point
of related parameter 𝜃 is as Eq. (11). were obtained.
( 2 1∕n )
� |R|
𝜎
min (11)
𝜃>0 2 4 Proposed approach
In this paper, GSA was used to search and optimize
In this paper, a two-stage grasping point layout optimiza-
the above models. When the optimal value of the relevant
tion design method based on GSA-Kriging model is pro-
parameter 𝜃 was obtained, the undetermined coefficient vec-
posed. The research is carried out for sheet metal part, and
tor 𝛽 of the regression term and the variance of the random
the optimal grasping point layout is found to minimize the
13
2234 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2235
3 2.5
2.5 2
1 2 3 4 5 6 7 8 9 10 2 3 4 5 6 7
Testing Sample set by LHS Expected output
6
GSA-Kriging prediction output
6
Stress deformation energy(mJ)
5.5
5.5
5
5
4.5
4.5
4
4
3.5
3.5 3
3 2.5
2.5 2
1 2 3 4 5 6 7 8 9 10 2 3 4 5 6 7
Testing Sample set by LHS Expected output
Fig. 13 Comparison of prediction effect between Kriging and GSA-Kriging. (a) Kriging model and (b) GSA-Kriging model
13
2236 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
6.5 6.5
GSA-Kriging prediction output
6 6
5 5
4.5 4.5
4 4
3.5 3.5
3 3
3 3.5 4 4.5 5 5.5 6 6.5 3 3.5 4 4.5 5 5.5 6 6.5
Expected output Expected output
6.5 6.5
6 6
SVR prediction output
5.5 5.5
5 5
4.5 4.5
4 4
3.5 3.5
3 3
(c)SVR (d)BPNN
Fig. 14 Comparison of prediction effects of GSA-Kriging, RBF, SVR, and BPNN. (a) GSA-Kriging, (b) RBF, (c) SVR, and (d) BPNN
GSA population size is set to 100, the maximum iteration model is 0.5973, and the correlation coefficient (R2) is 0.967.
step is set to 800, and the initial gravitational constant G0 Its prediction output is shown in Fig. 10.
is set to 20. The lower limit of hyperparameters 𝜃 is [0.001 At the same time, in order to further evaluate the accuracy
0.001 0.001 0.001 0.001 0.001 0.001 0.001], and the upper of the GSA-Kriging model in predicting the strain energy of
limit is [10 10 10 10 10 10 10 10]. the car door sheet with different grasping point layout, the
Finally, the optimized hyperparameter 𝜃 is [0.34 0.021 same training samples and test samples are used to estab-
0.001 0.008 0.004 0.618 0.018 0.164], and the optimized lish the radial basis function model, support vector machine
hyperparameter is constructed to obtain the GSA-Kriging (SVM) regression model, and BP neural network model. The
model. The root mean square error (RMSE) of GSA-Kriging comparison of prediction accuracy of each model is shown
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2237
2.45
P1 (286.4, 154.4), P2 (480.6, 503.6), P3 (140.6, 768.8), and
P4 (93.4, 482.6), and the coordinates of the edge grasping
2.4 points are P5 (795, Z5), P6 (805, Z6), P7 (X7, 1370), and P8
(− 193, Z8), a total of 4 design variables.
2.35 Similarly, 110 groups of training sample points and 10
groups of test sample points are generated through LHS,
2.3 and the strain energy of the car door sheet part is calculated
through finite element analysis, so as to construct the Krig-
2.25
0 50 100 150 200 250 300 350 400 ing model. The hyperparameter 𝜃=[0.013 0.392 1.076 0.185]
Evolutionary generation in the model. Then, GSA is used to optimize the hyperpa-
rameters of the Kriging model. The population size of GSA
Fig. 15 The second stage grasping point layout optimization curve is set to 100, the maximum number of iterative steps is set
to 400, and the initial gravitational constant G0 is set to 20.
The lower limit of hyperparameters 𝜃 is [0.001 0.001 0.001
in Table 4 and Fig. 11. The more consistent the green shaded 0.001] and the upper limit is [10 10 10 10]. Finally, the
figure is with the triangle under the dotted line, the better the optimized hyperparameter 𝜃 is [0.094 7.489 6.698 0.006],
model fitting effect is. and the optimized hyperparameter is constructed to obtain
By comparing with other models, it can be found that the
GSA-Kriging model has higher accuracy when using the
same training sample set and test sample set. Therefore, the
GSA-Kriging model is selected as a surrogate model for the
first stage of the grasping point layout optimization.
Then, GSA is used to optimize the first stage of the
grasping point layout optimization model. The GSA pop-
ulation size is set to 100, the maximum iteration step is
set to 800, and the initial gravitational constant G0 is set
to 20. The optimization curve of the first stage is shown
in Fig. 12.
Finally, the optimal positions of the four grasping points
on the primary datum plane are P1 (286.4, 154.4), P2 (480.6,
503.6), P3 (140.6, 768.8), and P4 (93.4, 482.6). At this time,
the strain energy of the car door sheet metal part predicted
by the model is 2.681 mJ. The optimal grasping point layout
is substituted into the finite element software for analysis,
and the strain energy of the car door sheet metal part under Fig. 16 Optimal grasping point layout
this layout is 2.802 mJ, with an error of 4.32%, which meets
the requirements of general engineering accuracy. the GSA-Kriging model. The predicted output of the Krig-
ing model and the GSA-Kriging model is shown in Fig. 13.
Table 6 Optimal layout of Primary grasping Coordinate values (X × Z) Edge grasping Coordinate values (X × Z)
grasping points point point
13
2238 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
Number of variables 8 4 12
Number of model samples 460 120 ≈800
Calculate time of a sample (s) 80 s 65 s 110 s
The total time (h) 12.4 h ≈24.4 h
The radial basis function model, support vector machine prediction accuracy of each model is compared in Table 5
regression model, and BP neural network model are respec- and Fig. 14.
tively constructed with the same training samples and test The GSA-Kriging model is optimized by GSA. The popula-
samples, and their accuracy is compared and analyzed. The tion size of GSA is set to 100, the maximum iteration step is
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2239
13
2240 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
16 16
15 15
14 14
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Testing Sample set by LHS Testing Sample set by LHS
Fig. 22 Deformation of sheet part before and after optimization of grasping point position. (a) Initial layout and (b) optimal layout
13
The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242 2241
13
2242 The International Journal of Advanced Manufacturing Technology (2023) 125:2225–2242
Code availability The code during the current study are available from 14. Sundararaman K, Padmanaban K, Sabareeswaran M (2016) Optimi-
the corresponding author on reasonable request. zation of machining fixture layout using integrated response surface
methodology and evolutionary techniques. P I Mech Eng C-J Mec
Declarations 230(13):2245–2259. https://doi.org/10.1177/0954406215592920
15. Qin G, Wang Z, Rong Y (2017) A unified approach to multi-fixtur-
Ethics approval Not applicable. ing layout planning for thin-walled workpiece. P I Mech Eng B-J
Eng 231(3):454–469. https://d oi.o rg/1 0.1 177/0 95440 54155 85240
Consent to participate Not applicable. 16. Yang Y, Wang ZQ (2017) Prediction model for aeronautical thin-
walled part fixture layout optimization based on SVR. Comput Integr
Consent for publication All the authors have reached agreement for Manuf Syst 23(06):1302–1309. https://doi.org/10.13196/j.cims.2017.
publication. 06.016
17. Su J, Cao E, Hong Q (2014) Optimization of fixture layouts of
Competing interests The authors declare no competing interests. glass laser optics using multiple kernel regression. Appl Opt
53(14):2988–2997. https://doi.org/10.1364/ao.53.002988
18. Rex FMT, Ravindran D (2017) An integrated approach for optimal
fixture layout design. P I Mech Eng B-J Eng 231(7):1217–1228.
https://doi.org/10.1177/0954405415590991
References 19. Wang ZQ, Yang Y, Yang B, Kang YG (2016) Optimal sheet metal
fixture locating layout by combining radial basis function neural
1. Krishnakumar K, Melkote SN (2000) Machining fixture layout network and bat algorithm. Adv Mech Eng 8(12):217–233. https://
optimization using the genetic algorithm. Int J Mach Tools Manuf doi.org/10.1177/1687814016681905
40(4):579–598. https://doi.org/10.1016/s0890-6955(99)00072-3 20. Dini G, Failli F (2000) Planning grasps for industrial robotized
2. Kashyap S, Devries WR (1999) Finite element analysis and opti- applications using neural networks. Robot Comput Integr Manuf
mization in fixture design. Struct Multidiscip O 18(2–3):193–201. 16(6):451–463. https://doi.org/10.1016/s0736-5845(00)00021-1
https://doi.org/10.1007/s001580050120 21. Qi ZC, Zhang KF, Li Y, Cheng H (2015) Analysis and optimi-
3. Li GL, Du SC, Huang DL, Zhao C, Deng YF (2019) Elastic mechanics- zation for locating errors of large wing panel during automatic
based fixturing scheme optimization of variable stiffness structure drilling and riveting. Acta Aeronautica et Astronautica Sinica
workpieces for surface quality improvement. Precis Eng 56:343–363. 36(10):3439–3449. https://doi.org/10.7527/S1000-6893.2015.
https://doi.org/10.1016/j.precisioneng.2019.01.004 0221
4. Kulankara K, Satyanarayana S, Melkote SN (2002) Iterative fix- 22. Aderiani AR, Wrmefjord K, Sderberg R (2020) Optimal design
ture layout and clamping force optimization using the genetic of fixture layouts for compliant sheet metal assemblies. Int J
algorithm. J Manuf Sci Eng-Trans Asme 124(1):119–125. https:// Adv Manuf Technol 110(7):2181–2201. https://doi.org/10.1007/
doi.org/10.1115/1.1414127 s00170-020-05954-y
5. Xing YF, Hu M, Zeng H (2015) Fixture layout optimization based 23. Yu KG (2019) Robust fixture design of compliant assembly
on a non-domination sorting social radiation algorithm for auto- process based on a support vector regression model. Int J Adv
body parts. Int J Prod Res 53(11):3475–3490. https://doi.org/10. Manuf Technol 103(1–4):111–126. https://d oi.o rg/1 0.1 007/
1080/00207543.2014.1003662 s00170-019-03488-6
6. Zhou T, Xiong ZQ, Yao W, Qin Y (2016) Flexible tooling layout 24. Lu C, Wang Y (2017) Positioning variation analysis for the
optimization for thin-walled workpieces based on improved ant sheet metal workpiece with N-2-1 locating scheme. Int J Adv
colony algorithm. J Propuls Technol 37(6):1165–1174. https://d oi. Manuf Technol 89(9–12):3021–3035. https://doi.org/10.1007/
org/10.13675/j.cnki.tjjs.2016.06.022 s00170-016-9284-y
7. Liao YG (2003) A genetic algorithm-based fixture locating positions 25. Xiong CH (2007) Fundamentals of robotic grasping and fixturing.
and clamping schemes optimization. Proc IME Part B-J Eng Manuf World Scientific
217(8):1075–1083. https://doi.org/10.1177/095440540321700805 26. Xiong CH, Li YF, Ding D, Xiong YL (1999) On the dynamic
8. Huang Q, Srijana Y, Gao S, Xu ZW, Wang XY (2018) Analysis of stability of grasping. Int J Robot Res 18(9):951–958
adaptive clamping force of fixture based on finite element method. 27. Wan XJ, Yang JQ, Zhang HJ (2018) Optimization of fixture lay-
IOP Conference Series: Mater Sci Eng 423(1):1–13. https://doi. out based on error amplification factors. J Comput Inf Sci Eng
org/10.1088/1757-899X/423/1/012116 18(4):041007. https://doi.org/10.1115/1.4040607
9. Xing YF (2017) Fixture layout design of sheet metal parts based 28. Sacks J, Welch WJ, Mitchell TJ (1989) Design and analysis of
on global optimization algorithms. J Manuf Sci E-T Asme computer experiments. Stat Sci 4(4):409–423
139(10):1–10. https://doi.org/10.1115/1.4037106 29. Nielsen HB, Lophaven SN, Sondergaard J (2002) DACE, a MAT-
10. Milad K, Maryam GS, Abdolreza O (2020) Multi-objective opti- LAB kriging toolbox
mization of auto-body fixture layout based on an ant colony algo- 30 Rashedi E, Nezamabadi-Pour H, Saryazdi S (2009) GSA: a gravi-
rithm. P I Mech Eng C-J Mec 234(6):1137–1145. https://doi.org/ tational search algorithm
10.1177/0954406219891756
11. Vishnupriyan S, Majumder MC, Ramachandran KP (2011) Opti- Publisher's note Springer Nature remains neutral with regard to
mal fixture parameters considering locator errors. Int J Prod Res jurisdictional claims in published maps and institutional affiliations.
49(21):6343–6361. https://doi.org/10.1080/00207543.2010.532167
12. Wang H, Huang LJ (2015) Integrated analysis method of thin- Springer Nature or its licensor (e.g. a society or other partner) holds
walled turbine blade precise machining. Int J Precis Eng Man exclusive rights to this article under a publishing agreement with the
16(5):1011–1019. https://doi.org/10.1007/s12541-015-0131-0 author(s) or other rightsholder(s); author self-archiving of the accepted
13. Hamedi M (2005) Intelligent fixture design through a hybrid sys- manuscript version of this article is solely governed by the terms of
tem of artificial neural network and genetic algorithm. Artif Intell such publishing agreement and applicable law.
Rev 23(3):295–311. https://doi.org/10.1007/s10462-004-7187-z
13