This study contributes to Phase 2 of the Task Group 1 round robin in the NeT European Network. To obtain better prediction results, in the thermal analysis, two significant changes are used. The welding efficiency, η, is fixed at 75%, and the weld bead fusion boundary profiles are based upon macrographs taken from welded specimens, which have been destructively examined. In the subsequent mechanical simulation, a non‐linear kinematic or mixed isotropic–kinematic hardening model should be employed, and a progressive annealing scheme or explicit consideration of visco‐plastic or creep effects should be implemented to handle high‐temperature inelastic strains and reduce stress discontinuities. In this study, an uncoupled 3D thermal and mechanical analysis was carried out using the software code SYSWELD. In the thermal simulation, a two‐offset‐double‐ellipsoid heat source model was developed, and the parameters were fitted using the heat source fitting tool. Power intensity was applied to simulate 1‐s dwelling time at the weld start end. Offset distances between two double ellipsoids were adjusted to obtain the weld bead transverse fusion boundary profiles at different positions. Predicted temperatures were compared with the measured data by thermocouples on the test pieces. In the mechanical analysis, a new material constitutive model, non‐linear mixed hardening model, was developed. Tension–compression cyclic tests were simulated at different temperatures using three different material hardening models (isotropic hardening model, kinematic hardening model and non‐linear mixed isotropic–kinematic hardening model), and the predicted cyclic stress–strain curves were compared with the measured data. Effects of three different hardening models on the welding residual stresses were studied. Compared with the measured data, the optimum material hardening model was confirmed.