Fast Prediction of Welding Distortion Using ISM and i-ISM with Experimental Validation

Welding mechanics plays an important role in the design of welding joint to minimize the residual distortion and stress. In the past decades, the TEP-FEM (thermal elastic-plastic finite element method) has shown its excellent capability in tracing the stress and strain during welding. Nevertheless, the computation expense in aspect of CPU time and computer memory becomes a big issue when the finite element model is large. For practical welded structures, the elements required for a transient thermal mechanical analysis can easily reach several millions, which make the computation difficult to run on modern computers or cost too much time to complete. In order to improve the computing speed while maintaining the accuracy, the ISM (iterative substructure method) was developed about ten years ago. According to the temperature distribution, the FE model (Region A) was divided into strongly nonlinear region B (near heat source) and weakly nonlinear region A-B (rest part excluding B). The A-B and B regions are solved separately, and balance on the boundary between the regions is realized by iterative calculations. Recently, the i-ISM (inherent strain based ISM) was proposed by introducing the concept of inherent strain into the ISM scheme. The region C which covers a part large enough around the heat source can be solved by fixing its boundary for a time interval and then it is released to solve the global model. This approach is based on the fact that strain in the rest part beyond C region will not change appreciably. To validate the two schemes ISM and i-ISM, experiment on a fillet welding model was carried out and three deformation components namely longitudinal shrinkage, transverse shrinkage and angular distortion were measured after welding. Analysis by conventional FEM, and ISM were then performed on the same model. Furthermore, a large scale model with nearly two millions of DOF (degree of freedom) was simulated by ISM and i-ISM. Finally, the accuracy and efficiency of the proposed methods have been confirmed.