Computational methods and experimental validation of welding distortion models

Abstract

Multiply-stiffened, thin plate, welded fabrications are used in a wide variety of transport fields, however the resulting out-of-plane distortion associated with welding exacts a severe design penalty. Depending on the information required, the size of the structure under investigation and the computer power at hand, three computational strategies may be considered to predict welding distortion. If prediction of the localized residual stresses from welding is of major importance, then a full transient, uncoupled thermo-elastoplastic analysis is preferred. This method is not readily applicable to predict welding distortions in industrial-scale welded structures. More computationally efficient models are required and two other models are suggested in the current study. A series of experimental tests of a realistic nature were performed to validate the proposed computational strategies. Computational and experimental study of butt and fillet welding of small and industrial size fabrications is considered.