Beruflich Dokumente
Kultur Dokumente
Kristina Wärmefjord
With Welding Simulation for
Department of Product
and Production Development, Prediction of Deformation and
Chalmers University of Technology,
SE-412 96 Göteborg,
Sweden
Variation of a Final Assembly
e-mail: kristina.warmefjord@chalmers.se
In most variation simulations, i.e., simulations of geometric variations in assemblies, the
Lennart Karlsson influence from heating and cooling processes, generated when two parts are welded to-
Division of Computer Aided Design,
gether, is not taken into consideration. In most welding simulations, the influence from
Luleå University of Technology,
geometric tolerances on parts is not taken into consideration, i.e., the simulations are
SE-971 87 Luleå,
based on nominal parts. In this paper, these two aspects, both crucial for predicting the
Sweden
final outcome of an assembly, are combined. Monte Carlo simulation is used to generate
e-mail: lennart.karlsson@ltu.se
a number of different non-nominal parts in a software for variation simulation. The trans-
lation and rotation matrices, representing the deviations from the nominal geometry due
to positioning error, are exported to a software for welding simulation, where the effects
Rikard Söderberg from welding are applied. The final results are then analyzed with respect to both devia-
Department of Product
tion and variation. The method is applied on a simple case, a T-weld joint, with available
and Production Development,
measurements of residual stresses and deformations. The effect of the different sources of
Chalmers University of Technology,
deviation on the final outcome is analyzed and the difference between welding simula-
SE-412 96 Göteborg,
tions applied to nominal parts and to disturbed (non-nominal) parts is investigated. The
Sweden
study shows that, in order to achieve realistic results, variation simulations should be
e-mail: rikard.soderberg@chalmers.se
combined with welding simulations. It does also show that welding simulations should be
applied to a set of non-nominal parts since the difference between deviation of a nominal
John Goldak part and deviation of a non-nominal part due to influence of welding can be quite large.
Department of Mechanical [DOI: 10.1115/1.4005720]
and Aerospace Engineering,
Carleton University, Keywords: variation simulations, welding simulations, tolerances, Monte Carlo, simula-
Ottawa, tion driven design
ON K1S 5B6,
Canada
e-mail: jgoldak@mrco2.carleton.ca
1 Introduction work has although been done on spot welding and spot welding
sequences; see for example Ref. [1].
In this very first section, a short background is given and the
Welding simulation tools are used to calculate the thermal field,
need of the suggested method is explained. Further, both variation
the evolution of microstructure, and the stress–strain evolution
simulations and welding simulations are described.
during welding. Distortions and residual stresses caused by weld-
ing are often of interest when choosing a proper welding proce-
dure and designing the welded structure. The use of welding
1.1 Background. In manufacturing industry, different kinds
simulations can reduce the amount of physical testing needed in
of simulation tools are used, aiming to predict the outcome of the
the development of new welded products.
final product. Those predictions are usually done in early stages of
Sufficient accuracy and precision are two main factors control-
the product development cycle in order to compare different prod-
ling the validity of a simulation model in imitating a real world
uct concepts and process alternatives. Increased demands on sus-
problem; the more adequate those factors are, the better the simu-
tainability drives the need for an increased use of virtual methods
lation model becomes in addressing the present analysis objec-
for verification of product and production systems.
tives. Accuracy and precision are of course also of great
Among the simulation tools, variation simulation is used to pre-
importance in both variation simulations and welding simulations.
dict the geometrical variation in critical dimensions of a subas-
Therefore, this work aims to combine variation simulation and
sembly or final product. Such a simulation is based on CAD
welding simulation by taking the effect of welding into considera-
(Computer Aided Design) models or meshes in the case of non-
tion in the variation simulations and by applying the welding sim-
rigid simulations. Tolerances are added to important features in
ulations to non-nominal parts. In that way, the quality of the
the simulation model and the resulting variation in the final subas-
simulations can be further improved leading to reduced risks of
sembly is predicted using Monte Carlo simulations. In those simu-
misjudgments and thereby also a reduced scrap rate. A reduced
lations, the specific effects of the chosen joining method, besides
scrap rate will benefit sustainability, with respect to both economi-
the actual joining of the parts of course, is often neglected. Some
cal and ecological aspects. Also social sustainability is gained by
increased use of virtual tools, since this usually implies improved
Contributed by the Design Engineering Division of ASME for publication in the
JOURNAL OF COMPUTING AND INFORMATION SCIENCE IN ENGINEERING. Manuscript received
working conditions.
July 7, 2011; final manuscript received December 19, 2011; published online Not much work has been done in the area of combining varia-
February 10, 2012. Assoc. Editor: Satyandra Gupta. tion simulations and welding simulations. However, Lee et al. [2]
Journal of Computing and Information Science in Engineering JUNE 2012, Vol. 12 / 021002-1
Copyright VC 2012 by ASME
case, the weld joint has to be generated after the RBT is per-
formed. Once the simulation models for each case is prepared
(nominal case included), they are solved in parallel using multi-
core processors.
Fig. 3 Double ellipsoid heat source with Gaussian heat distri-
The results are post processed and exported back to the varia- bution [14]
tion simulation software for further analysis. The results, i.e., the
resulting deformations, are exported in three meshes; nominal
mesh, disturbed mesh, and deformed mesh after welding. This Table 1 Heat input parameters
data can then be gathered and analyzed. For each node, the stand-
ard deviation and mean value of the deformation for all iterations Parameter Value
are calculated. The results are illustrated using color coding.
Figure 1 shows how the data is transported between the software Heat input model a1 ¼ 12 mm
used in this paper. (Double ellipsoid a2 ¼ 6 mm
moving heat source) b ¼ 2.5 mm
c ¼ 2.5 mm
Welding net power 4800 W (24 V, 200 A)
3 Case Study Welding efficiency 85%
A simple T-joint case is simulated to demonstrate the flow of Welding speed 6.35 mm/s
data when combining variation simulations with welding simula-
tions. Deo and Michaleris [23] conducted experiments and weld-
ing simulations on “T” shaped stiffeners, commonly used in the
shipbuilding industry. The experimental setup referred to as “Case
2” in Ref. [23] will be used for this paper. The case consists of
two A36 steel plates joined by a T-joint configuration, i.e., a fillet
weld on each side of the stiffener. The plates are tack welded at
the start and stop end of the stiffener prior to welding.
In the variation simulation software, tolerances are applied
to the locators of both parts. In each locator, a tolerance of
8s ¼ 0.7 mm is applied, where s is the standard deviation of the Fig. 4 Weld joint before (left) and after RBT, two cases (centre
position of the locator, measured in its positioning direction. This and right)
tolerance can be regarded as a common tolerance, containing both
variation in the fixture due to wear or lack of repeatability and
variation in areas meeting the locator on the parts. The tolerances
are centered on nominal values. A sample of 37 details, using
those non-nominal conditions that the tolerances imply, was cre-
ated using Monte Carlo simulations. Of course, it would be desira-
ble with a larger sample size but due to computational expensive
Fig. 2 Simulation mesh and constraints for welding Fig. 5 Resulting deformation of nominal case. Deformation
simulations enhanced 103. Values in meter.
Journal of Computing and Information Science in Engineering JUNE 2012, Vol. 12 / 021002-3
Figure 5 shows the resulting deformation after welding of the To study the effect of welding for non-nominal conditions com-
nominal case. To validate the welding simulations based on nomi- pared to welding for nominal conditions more closely, the behav-
nal conditions, the results were compared to deformation and re- ior in five different nodes for one of the 37 set of non-nominal
sidual stress measurements conducted by Deo and Michaleris conditions is investigated. Only one iteration is considered and
[23]. The bowing at the center of the beam is predicted to be 4.7 therefore it is only possible to investigate the displacements, not
mm, compared to 5.7 mm obtained in the experiment. Figure 6 any variation. The positions of the nodes considered are shown in
shows a comparison between predicted and measured residual Fig. 8. Those nodes represent critical measures in this study. For
stress across the middle of the beam. The experimental values an industrial case study, the critical measures would have been
were obtained using the “blind hole drilling method.” Though, dimensions affecting functional or esthetical requirements.
more residual stress measurements would be needed to make a In Fig. 9, the deviation in x-, y- and z-direction in each node af-
sufficient comparison. ter welding using non-nominal conditions are shown (white bar)
As mentioned earlier, a sample containing 37 different non- as well as the deviation after welding using nominal conditions
nominal conditions for the parts was created using Monte Carlo (gray bar). Further, the sizes of the added deviations (the black
simulation. For each set of non-nominal conditions, welding simu- bar) in each node are illustrated.
lation is applied. The goal is to investigate if there is a difference, The figure shows that the difference between welding using
larger than the applied tolerance, between welding simulations non-nominal conditions and nominal conditions can be consider-
applied to nominal conditions and welding simulations applied to able. For example, in node V in y-direction, the welding for nomi-
non-nominal conditions. It is also of interest to investigate how nal conditions results in almost no deviation from nominal at all,
the influence of welding affects the variation simulations. while the deviation for non-nominal condition amounts to almost
In Fig. 7, the results concerning displacement, i.e., deviation 4 mm. For some other nodes and directions, there are only minor
from nominal values, are illustrated. In Fig. 7(a), the displacement differences. It seems to be difficult to find some pattern in the
after welding, using nominal conditions, can be seen. In Fig. 7(b), behaviors, and of course, the behavior in one node depends on
the mean value of the displacements for non-nominal conditions, what happens in other nodes of the assembly.
before welding, is shown. Since the tolerances are centered on To summarize, there are large differences between the effect of
zero, the mean deviation is very small; observe the different scale welding applied to nominal conditions and the effect of welding
in this picture compared to Figs. 7(a) and 7(c). In Fig. 7(c), the applied to non-nominal conditions and it is important to be aware
mean value of the displacements for non-nominal conditions after of these differences when conducting welding simulations.
welding is illustrated. Since the mean effect of the tolerances is In Fig. 10, the values of 6s, where s is the standard deviation,
almost negligible, Figs. 7(a) and 7(c) are very similar. for the nodal displacement of all nodes of the final assembly are
Journal of Computing and Information Science in Engineering JUNE 2012, Vol. 12 / 021002-5
Acknowledgment
This work was financed by VINNOVA (the Swedish Govern-
mental Agency for Innovation Systems) and carried out in parallel
with the work in the (also VINNOVA funded) Faste Laboratory
and Wingquist Laboratory. The VrWeld software used for the
simulations in this research was provided by Goldak Technolo-
gies, Inc. The RD&T software used for the simulations in this
research was provided by RD&T Technology.
References
[1] Wärmefjord, K., Söderberg, R., Lindkvist, L., 2010, “Strategies for Optimiza-
tion of Spot Welding Sequence With Respect to Geometrical Variation in Sheet
Metal Assemblies,” Proceedings of ASME International Mechanical Engineer-
ing Congress and Exposition.
[2] Lee, D., Kwon, K. E., Lee, J., Jee, H., Yim, H., Cho, S. W., Shin, J.-G., and
Lee, G., 2009, “Tolerance Analysis Considering Weld Distortion by Use of Pre-
generated Database,” ASME J. Manuf. Sci. Eng., 131, pp. 041012.1–041012.6.
[3] Xiong, C., Rong, Y., Koganti, R., Zaluzec, M., and Wang, N., 2002,
“Geometric Variation Prediction in Automotive Assembling,” Assem. Autom.,
22(3), pp. 260–269.
[4] Ramesh, R., Mannan, M. A., and Poo, A. N., 2000, “Error Compensation in
Machine Tools—A Review, Part II: Thermal Errors,” Int. J. Mach. Tools
Manuf., 40, pp. 1257–1284.
[5] Yang, H., and Ni, J., 2003, “Dynamic Modeling for Machine Tool Thermal
Error Compensation,” ASME J. Manuf. Sci. Eng., 125, pp. 245–254.
[6] Zhu, J., Ni, J., and Shih, A. J., 2008, “Robust Machine Tool Thermal Error
Modeling Through Thermal Mode Concept,” ASME J. Manuf. Sci. Eng., 130,
Fig. 10 (a) 6s for non-nominal conditions before welding. pp. 061006.1–061006.9.
[7] Kang, Y., Chang, C.-W., Huang, Y., Hsu, C.-L., and Nieh, I.-F., 2007,
Units in millimeter. (b) 6s for non-nominal conditions after weld-
“Modification of a Neural Network Utilizing Hybrid Filters for the Compensa-
ing. Units in millimeter. tion of Thermal Deformation in Machine Tools,” Int. J. Mach. Tools Manuf.,
47, pp. 376–387.
[8] Söderberg, R., Lindkvist, L., and Carlson, J., 2006, “Virtual Geometry Assur-
illustrated using color coding. In this figure, it is not possible to ance for Effective Product Realization,” 1st Nordic Conference on Product
show the variation of welding for nominal conditions, since there Lifecycle Management—NordPLM’06, Göteborg, Sweden.
is only one nominal assembly and consequently no variation [9] Gao, J., Chase, K., and Magleby, S., 2009, “Comparision of Assembly Toler-
ance Analysis by the Direct Linearization and Modified Monte Carlo Simula-
between different assemblies. All welding conditions besides the tion Methods,” Proceedings of ASME Design Engineering Technical
tolerances are constant during this study. It can though be seen Conference, Boston, USA.
that the amount of variation increases from Fig. 10(a), where 6s [10] Shen, Z., Ameta, G., Shah, J., and Davidson, J., 2005, ” A Comparative Study
for all non-nominal conditions before welding is shown, to of Tolerance Analysis Methods,” ASME J. Comput. Inf. Sci. Eng., 5(3), pp.
247–256.
Fig. 10(b), where 6s for non-nominal conditions after welding is [11] Söderberg, R., and Lindkvist, L., 1999, “Computer Aided Assembly Robustness
illustrated. It is accordingly possible to establish that the welding Evaluation,” J. Eng. Des., 10(2), pp. 165–181.
procedure increases the amount of variation and also that the [12] Wärmefjord, K., Söderberg, R., and Lindkvist, L., 2008, “Tolerance Simulation
added variation is not distributed evenly all over the assembly. of Compliant Sheet Metal Assemblies Using Automatic Node-Based Contact
Detection,” Proceedings of IMECE2008, Boston, USA.
For example, the values in the upper left corners in Figs. 10(a) [13] RD&T Technology, 2009, RD&T Manual, Mölndal.
and 10(b) are almost the same, while the variation has increased [14] Goldak, J., and Springer., M. A., 2005, Computational Welding Mechanics,
significantly in the area that is red in Fig. 10(b). Springer, New York.
Figure 10 does show the importance of including the effect of [15] Lindgren, L., 2001, “Finite Element Modeling and Simulation of Welding. Part
1: Increased Complexity,” J. Therm. Stresses, 24(2), pp 141–192.
welding in variation simulations, since it leads to a significant [16] Lindgren, L.-E., 2007, Computational Welding Mechanics: Thermomechanical
increase of the level of variation in the final assembly. and Microstructructural Simulations, CRC Press, Boca Raton, FL.
[17] Lindgren, L., 2001, “Finite Element Modeling and Simulation of Welding. Part
3: Efficiency and Integration,” J. Therm. Stresses, 24(4), pp. 305–334.
4 Conclusions [18] Pahkamaa, A., Karlsson, L., Pavasson, J., Karlberg, M., Näsström, M., and
Goldak, J., 2010, “A Method to Improve Efficiency in Welding Simulations for
The geometrical variations on parts and the influence of heating Simulation Driven Design,” Procedings of ASME 2010 International Design
and cooling due to the welding process both strongly affect the Engineering Technical Conference and Computers and Information in Engi-
final geometrical outcome of an assembly. Those two aspects are, neering Conference, Montreal, Canada.
however, usually simulated separately, which may give mislead- [19] Karlsson, L., Pahkamaa, A., Karlberg, M., Löfstrand, M., Goldak, J., Pavasson,
J., 2011, “Mechanics of Materials and Structures—A Simulation-Driven Design
ing results. In this work, variation simulation and welding simula- Approach,” J. Mech. Mater. Struct., 6, pp. 277–301.
tion are combined. [20] 20 Goldak Technologies Inc., 2010, www.goldaktec.com.
The method is applied to a case study and the results are ana- [21] Goldak, J., 2008., “Distortion and Residual Stress in Welds: The Next Gener-
lyzed with respect to both variation and deviations from nominal ation,” Proceedings of 8th International Conference on Trends in Welding
Research, Pine Mountain, Georgia.
values. [22] Corben, H. C., and Stehle, P., 1994, Classical Mechanics, Dover Publications,
For the deviation, there are evident differences in deviation New York.
from nominal when the welding simulations are applied to parts [23] Deo, M., and Michaleris, P., 2002, “Experimental Verification of Distortion
under nominal conditions compared to when they are applied to Analysis of Welded Stiffeners,” J. Ship Prod., 4, pp. 216–225. Available at
http://www.ingentaconnect.com/content/sname/jsp/2002/00000018/00000004/
parts under non-nominal conditions (due to tolerances added in art00005.
the variation simulation). This motivate why it is not enough to [24] Deo, M., Michaleris, P., and Sun, J., 2003, “Prediction of Buckling Distortion
carry out welding simulations on nominal parts. of Welded Structures,” Sci. Technol. Weld. Joining, 8(1), pp. 55–61.