Prediction of residual stresses in resistance spot weld

Purpose - The purpose of this paper is to predict residual stresses in resistance spot weld of 2 mm thick aluminum 6061-T6 sheets. The joint use of finite element analysis and artificial neural networks can eliminate the high costs of residual stresses measuring tests and significantly shorten the time it takes to arrive at a solution. Design/methodology/approach - Finite element method and artificial neural network have been used to predict the residual stresses. Different spot welding parameters such as the welding current, the welding time and the electrode force have been used for the simulation purposes in a thermal-electrical-structural coupled finite element model. To validate the numerical results, a series of experiments have been performed, and residual stresses have been measured. The results obtained from the finite element analysis have been used to build up a back-propagation artificial neural network model for residual stresses prediction. Findings - The results revealed that the neural network model created in this study can accurately predict residual stresses produced in resistance spot weld. Using a combination of these two developed models, the residual stresses can be predicted in terms of spot weld parameters with high speed and accuracy. Practical implications - The paper includes implication for aircraft and automobile industries to predict residual stresses. Residual stresses can lower the strength and fatigue life of the spot-welded joints and determine the performance quality of the structure. Originality/value - This paper presents an approach to reduce the high costs and long times of residual stresses measuring tests.