Professional Documents
Culture Documents
Center for Advanced Vehicular Systems, Mississippi State University, MS 39762, USA
Simufact-Americas LLC, Plymouth, MI 48170-4347, USA
c
Department of Aerospace Engineering, Mississippi State University, MS 39762, USA
b
art ic l e i nf o
a b s t r a c t
Article history:
Received 11 November 2013
Received in revised form
29 January 2014
Accepted 11 February 2014
This paper presents an effective numerical approach for welding process parameter optimization to minimize
weld-induced distortion in structures. A numerical optimization framework based on coupled Genetic
Algorithm (GA) and Finite Element Analysis (FEA) is developed and implemented for a low and
a high delity model. Classical weakly coupled thermo-mechanical analysis with thermo-elasto-plastic
assumptions is carried out for distortion prediction of numerical models. The search for optimum process
parameters is executed by direct integration of numerical models and GA-based optimization technique. The
developed framework automatically inserts the process parameters into the simulation models, executes the
FE-based welding simulations and evaluates the required simulation output data for iterative evolutionary
optimization. The optimization results show that the proposed approach can contribute substantially to
enhance nal welded product quality while facilitating and accelerating the product design and development.
& 2014 Elsevier B.V. All rights reserved.
Keywords:
Welding simulation
Numerical design optimization
Finite element analysis
Genetic algorithm
Welding-induced distortion
1. Introduction
Arc welding is a major joining process used in automotive,
shipbuilding, and other industries. It is prominent over other
joining methods due to its competitive advantages such as
reduced cost, enhanced joint strength, and wide range of applications. However, one of the major problems of welding is weldinduced distortion in the welded assembly. Distortion affects
performance of welded structures in the form of reduced joint
strength and dimensional accuracy. Despite tremendous development in arc welding technology over the years, weld-induced
distortion is still one of the major obstacles for welding industry to
ensure adequate reliability of welded structure's performance.
Correction of distortion often requires additional after-weld
reworks, which are usually costly, time consuming, and practical
only in the most crucial applications. The best practice to minimize
or control distortion is proper welding process design through
careful selection of various welding input parameters. Several process
parameters inuence welding distortion. Better control of these
parameters will eliminate the conditions that promote distortion [1].
Control of distortion is mostly performed empirically using
experiments [2]. A set of experiments is conducted in a dened
Corresponding author.
E-mail address: mri27@msstate.edu (M. Islam).
http://dx.doi.org/10.1016/j.nel.2014.02.003
0168-874X & 2014 Elsevier B.V. All rights reserved.
55
56
C p
T
:q Q
t
57
Fig. 2. Details of lap joint experimental model (a) experimental setup and (b) experimental sample.
where the rst term is convection heat loss and h is the convective
heat transfer coefcient. The second term is emissive heat loss and
is the emissivity factor. The third term is the contact heat loss
and is the contact heat transfer coefcient. The relevant parameters are given in Table 2 and the same parameters are used in
both models. The material model used for both models is ASTM
A591 sheet metal steel. Temperature dependent thermal and
mechanical properties are considered as shown in Fig. 5.
58
Table 1
Heat source parameters.
Parameter name
1.00
2.50
3.25
4.00
Table 2
Heat transfer coefcients.
Coefcient
Value
20
100
0.6
Fig. 6. Weld pool shape comparison experimental (left) and simulation prediction
(right). (For interpretation of the references to color in this gure caption, the
reader is referred to the web version of this paper.)
59
Fig. 7. Out-of-plane distortion pattern comparison-experimental result (right) and simulation result (left).
Line 01
Line 02
Line 03
Fig. 8. Quantitative comparison of out-of-plane distortion (a) illustration of lines, (b) line 1, (c) line 2 and (d) line 03.
60
Fig. 9. Simulation predicted total distortion pattern of lap-joint model for experimental conditions.
Table 3
Description of design variables.
Design Variable
Unit
Lower bound
Upper bound
Current
Voltage
Speed
Direction
Amp
Volt
mm/s
80
8
3.5
1
250
25
10
6
The weld bead was modeled as triangular shaped with the resultant
molten weld pool shape for sub-weld 1 depicted in Fig. 10 by
dotted line.
Fig. 11 shows the typical distortion distribution over the
structure. The distortion pattern of the structure indicates that
the lower part has undergone higher distortion than the upper
part and the maximum distortion has occurred in the middle
section of the lower part with a magnitude is 0.59 mm.
4. Numerical optimization
4.1. Optimization problem formulation
4.1.1. Lap-joint model
The maximum distortion is treated as the objective function.
Through welding simulation, distortions in all N nodes are rst
calculated as the sum of square roots of nodal distortions in all
three directions. Then, the maximum distortion value is selected
and used as the objective function value for iterative optimization
via GA. Thus, the objective function is dened as
FX maxDi
q
Di dx 2i dy 2i dz 2i i 1; 2; 3 N
Welding speed (X1), arc voltage (X2), input current (X3) and
welding direction (X4) are dened as design variables. Details of
design variables are shown in Table 3.
For the lap-joint model, X4 can take six numerical values to
represent six possible welding directions as shown in Table 4. Two
welding directions are designed with one robot and are represented by integer values 1 and 2, depending on the robots leftright or right-left movement direction, respectively. Similarly, the
remaining four welding directions are designed with two robots
and are represented by an integer from 3 to 6 depending on each
robot's left-right or right-left movement direction. For the two
robot welding cases, it is assumed that both robots will start and
stop welding at the same time.
The optimization algorithm chooses design variables automatically. As such, selected design variables do not always guarantee
X1
X2
X3
X4
Table 4
Denition of welding direction variable.
Value
1
2
3
4
5
6
No. of robots
Starting timing
1
1
2
2
2
2
Same
Same
Same
Same
Same
Same
FX;
Nc 0
FX 100N c ;
Nc 4 0
where F(X) is the original objective function value and 100Nc is the
penalty term. The penalty term increases the original objective
function value and indicates to GA that the associated model is
infeasible. An infeasible model represents poor welding quality
even though the weld-induced distortion may be small.
61
Table 5
Description of process design variables.
Table 6
Details of sub-weld order design variables.
Design variable
X1
X2
X3
X4,X5,X6
Unit
Lower bound
Upper bound
Current
Voltage
Speed
Weld order
Amp
Volt
mm/s
100
20
15
1
150
25
22
6
Welding order
Welding direction
1
2
3
4
5
6
First
Second
Third
First
Second
Third
Forward
Forward
Forward
Backward
Backward
Backward
62
Table 8
Optimization results of lap joint model.
Start
Condition
Initial Population
Optimum
(Best)
1
2
3
X1,
(Amp)
X2,
(Volt)
X3,
(mm/s)
X4
200
200
200
15
10
15
10
5
10
3
3
3
Maximum
total
distortion,
(mm)
No of
simulations
0.48
0.46
0.48
58
63
78
Yes
Stopping
Criteria?
No
Selection, Crossover & Mutation
Updated Population
Stop
Fig. 13. Flowchart of Genetic Algorithm based optimization framework.
Table 7
Details of GA parameters used in optimization
system.
Parameter name
Value
Population size
Number of generations
Number of elite candidates
Crossover fraction
Function tolerance
Mutation function
Crossover function
10
20
2
0.8
10E 06
Adaptive feasible
Scattered
5. Optimization results
Table 9
Optimization results of lower-arm model.
Optimum value of design variables
X1 (Amp)
X2 (Volt)
X3 (mm/s)
X4
X5
X6
140
20.5
22
0.35
63
Acknowledgments
Fig. 15. Illustration of the optimum welding sequence of lower arm model.
6. Conclusion
References
Due to increasing requirements for improved performance of
welded structures, it has become essential to take into account
process variability during the design phase of a welding process.
Traditional experiment based welding process variable optimization is quite expensive and is not always guaranteed to provide the
optimum parameter combination. Furthermore, such approach
cannot also effectively control several critical parameters such as
welding direction. In this context, this study introduced a robust
numerical optimization system based on integrated computational
tools, which allow automatic optimization of welding process
parameters without the requirement of expensive experiments.
The system is capable of exploring the effect of several design
variables at a time with limited modication of the simulation
model. Thereby, the developed tool can be effectively implemented for the process design purpose of a large-scale industrial
welding process. In addition, the proposed tool will also be useful
tool for performing early-stage design investigations like parametric study or sensitivity analysis. This study also widens the
automated and customized applications of CAE tools in manufacturing process design, analysis and optimization.
The illustrative example of lap joint welding specimen optimization presented in this work showed that the proposed GA-FEM
coupled method is able to search for optimum set of process
parameters, especially under the critical constraint of weld quality
requirement. In this optimization problem, a straightforward
solution approach was to run all possible 2400 (10 10 4 6)
combinations and select the best one as optimum solution.
However, it would be computationally inefcient and sometimes
infeasible. Using GA, we achieved optimum results with maximum
78 FE simulations. So, the method was certainly effective for this
case study. To examine the maturity of the developed system, a
realistic automotive structure (lower arm) with nonlinear weld
path was also investigated. In the absence of any prior work,
numerical or experimental, simulation model was calibrated based
on the authors experience only. For this case study, six design
variables were selected including welding sequence of three subwelds. Total number of combination for this study was 11520
[1] C.L. Tsai, S.C. Park, W.T. Cheng, Welding distortion of a thin-plate panel
structure, Weld. Res. (1999) S156S165.
[2] P. Michaleris, A. Debiccari, Pediction of welding distortion, Weld. Res. 76
(1997) S172S181.
[3] J.A. Goldak, M. Akhlaghi, Computational Welding Mechanics, Springer, New
York, 2005.
[4] L.-E. Lindgren, Finite element modeling and simulation of welding part 1:
increased complexity, J. Therm. Stress. 24 (2001) 141192.
[5] L.-E. Lindgren, Finite element modeling and simulation of welding part 2:
Improved material modeling, J. Therm. Stress. 24 (2001) 195231.
[6] L.-E. Lindgren, Finite element modeling and simulation of welding. part 3:
efciency and intetration, J. Therm. Stress. 24 (2001) 305334.
[7] D. Camilleri, T.G.F. Gray, Computationally efcient welding distortion simulation techniques, Model. Simul. Mater. Sci. Eng. 13 (2005) 13651382.
[8] S.B. Brown, H. Song, Finite element simulation of welding of large structures,
ASME J. Eng. Ind. 114 (1992) 441451.
[9] P. Michaleris, A. Debiccari, Predictive technique for buckling analysis of thin
section panels due to welding, J. Ship Prod. 12 (1996) 269275.
[10] G.H. Jung, C.L. Tsai, Plasticity-based distortion analysis for llet welded thinplate T-joints, Weld. J. 83 (2004) 177187.
[11] D. Deng, W. Liang, H. Murakawa, Determination of welding deformation in
llet-welded joint by means of numerical simulation and comparison with
experimental measurements, J. Mater. Process. Technol. 183 (2007) 219225.
[12] D. Deng, H. Murakawa, Prediction of welding distortion and residual stress in a
thin plate butt-welded joint, Comput. Mater. Sci. 43 (2008) 353365.
[13] D. Deng, H. Murakawa, M. Shibahara, Investigations on welding distortion in
an asymmetrical curved block by means of numerical simulation technology
and experimental method, Comput. Mater. Sci. 48 (2010) 187194.
[14] D. Deng, H. Murakawa, FEM prediction of buckling distortion induced by
welding in thin plate panel structures, Comput. Mater. Sci. 43 (2008) 591607.
[15] D. Deng, H. Murakawa, W. Liang, Numerical simulation of welding distortion
in large structures, Comput. Methods Appl. Mech. Eng. 196 (2007) 46134627.
[16] H. Murakawa, D. Deng, N. Ma, J. Wang, Applications of inherent strain and
interface element to simulation of welding deformation in thin plate structures, Comput. Mater. Sci. 51 (2011) 4352.
[17] J. ERD, S. JHD, A review on optimization of welding process, Proced. Eng. 38
(2012) 544554.
[18] K.Y. Benyounis, A.G. Olabi, Optimization of different welding processes using
statistical and numerical approaches a reference guide, Adv. Eng. Softw. 39
(2008) 483496.
[19] I. Voutchkov, A.J. Keane, A. Bhaskar, T.M. Olsen, Weld sequence optimization:
The use of surrogate models for solving sequential combinatorial problems,
Comput. Methods Appl. Mech. Eng. 194 (2005) 35353551.
[20] J.E.V. Petera, R.P. Dwight, Numerical sensitivity analysis for aerodynamic
optimization: a survey of approaches, Comput. Fluids 39 (2010) 373391.
[21] M.H. Kadivar, K. Jafarpur, G.H. Baradaran, Optimizing welding sequence with
genetic algorithm, Comput. Mech. 26 (2000) 514519.
64
[22] J. Song, J. Peters, A. Noor, P. Michaleris, Sensitivity analysis of the thermomechanical response of welded joints, Int. J. Solids Struct. 40 (2003)
41674181.
[23] J. Song, J.Y. Shanghvi, P. Michalers, Sensitivity analysis and optimization of
thermo-elasto-plastic processes with applications to welding side heater
design, Comput. Methods Appl. Mech. Eng. 193 (2004) 45414566.
[24] A.H. Eras, F.O. Sonmez, Design optimization of spot-welded plates for maximum fatigue life, Finite Elem. Anal. Des. 47 (2011) 419423.
[25] Y. Zhang DT, Optimization of spot-welded structures, Finite Elem. Anal. Des.
37 (2001) 10131022.
[26] T.-L. Teng, C.-P. Fung, P.-H. Chang, W.-C. Yang, Analysis of residual stresses
and distortions in T-joint llet welds, Int. J. Press. Vessel. Pip. 78 (2001)
523538.
[27] L. Gannon, Y. Liu, N. Pegg, M. Smith, Effect of welding sequence on residual stress
and distortion in at-bar stiffened plates, Mar. Struct. 23 (2010) 385404.
[28] T. Schenk, M. Doig, G. Esser, I.M. Richardson, Inuence of clamping support
distance on distortion of welded T joints, Sci. Technol. Weld. Join. 15 (2010)
575582.
[29] T. Schenk, I.M. Richardson, M. Kraska, S. Ohnimus, A study on the inuence of
clamping on welding distortion, Comput. Mater. Sci. 45 (2009) 9991005.
[30] L.-E. Lindgren, Numerical modeling of welding, Comput. Methods Appl. Mech.
Eng. 195 (2006) 67106736.
[31] J. Goldak, A. Chakravarti, M. bibby, A new nite element model for welding
heat source, Metall. Trans. B 15B (1984) 299305.
[32] L.-E. Lindgren, Computational Welding Mechanics, Woodhead Publishing,
Cambrige, 2007.