Finite-element prediction of distortion during gas metal arc welding using the shrinkage volume approach

Abstract

Distortion is a potential problem with all welded fabrications. To a large extent, industrial control of weld induced distortion is achieved by reliance on past experience, simple empirical formulae or rectification procedures. Rectification can be costly, whilst in large complex structures, empirical formulae are rarely applicable. Classical approaches to the modelling of welding distortion and residual stress, whilst accurate, have not been readily useable within industry. The time and cost associated with running such models appear to be the main reasons contributing to this situation. Nevertheless, the use of computer simulative techniques has the potential to significantly reduce the cost of welded fabrications by allowing for predictions to be made long before a single weld bead is put down on the workshop floor. Therefore, computer models that are aimed at predicting welding phenomena not only need to be accurate, but must also be affordable and capable of making predictions within industrial time frames if they are to be used by fabricators. This paper presents one such strategy. The Shrinkage Volume Method is a linear elastic finite-element modelling technique that has been developed to predict post-weld distortion. By assuming that the linear thermal contraction of a nominal shrinkage volume is the main driving force for distortion, the need to determine the transient temperature field and microstructural changes is eliminated. In so doing, the model solution times are reduced significantly and the use of linear elastic finite-element methods permits large, highly complex welded structures to be modelled within a reasonable time frame. Verification of the modelled results was carried out by an experimental program that investigated the distortion of plain carbon steel plates having differing vee-butt preparations. The initial models, which had assumed the edge preparation to be representative of the overall shrinkage volume, were in reasonable agreement with the experimentally determined distortion values. Further improvements to these results were made by using a thermal model to define better the effective weld shrinkage volume.