摘要
Finite element (FE) modelling methods were implemented to perform a weakly-coupled weld simulation activity on a series of simple plate welds, to determine the effects of altering the frequency of saving the thermal time-step result upon the mechanical results. By definition, the thermal results will be unaffected, but the mechanical results are calculated from the saved thermal results, hence can be changed when the frequency of saving thermal time-steps is altered. By default, most weakly-coupled thermal-mechanical solvers will save every single thermal time-step, for accuracy. Results indicated that during the welding operation, the thermal time-steps could be reduced to saving 1-in-every-2 thermal time-steps with minimal loss in mechanical accuracy. However, during the cooling operation, every time-step was required to be saved. Whilst this seems almost counter-intuitive that the time-step during the cooling operation is in some way more critical than during welding, it must be stated that the FE software employed for this exercise has a setting allowing the time-steps to become progressively large during cooling, when thermal gradients are much lower and as such both thermal and mechanical calculations are easier to converge.
Finite element (FE) modelling methods were implemented to perform a weakly-coupled weld simulation activity on a series of simple plate welds, to determine the effects of altering the frequency of saving the thermal time-step result upon the mechanical results. By definition, the thermal results will be unaffected, but the mechanical results are calculated from the saved thermal results, hence can be changed when the frequency of saving thermal time-steps is altered. By default, most weakly-coupled thermal-mechanical solvers will save every single thermal time-step, for accuracy. Results indicated that during the welding operation, the thermal time-steps could be reduced to saving 1-in-every-2 thermal time-steps with minimal loss in mechanical accuracy. However, during the cooling operation, every time-step was required to be saved. Whilst this seems almost counter-intuitive that the time-step during the cooling operation is in some way more critical than during welding, it must be stated that the FE software employed for this exercise has a setting allowing the time-steps to become progressively large during cooling, when thermal gradients are much lower and as such both thermal and mechanical calculations are easier to converge.
作者
Richard P. Turner
R. Mark Ward
Richard P. Turner;R. Mark Ward(PRISM2 Research Group, School of Metallurgy & Materials, The University of Birmingham, Edgbaston, Birmingham, UK)
