Prediction of Structural Steel Strength with Lap Joint Welds Using Artificial Neural Networks and Finite Element Simulation

Welding is a critical and sensitive technique for joining parts in industry. It has numerous applications and, in some cases, is irreplaceable. Therefore, it's essential to examine components produced by welding to determine the tolerable strength of welded sections for design and manufacturing purposes.This study utilized thermo-elasto-plastic finite element analysis to examine the thermomechanical behavior that emerges from the arc welding of overlapping 37St steel sheets. The welding process was simulated in two stages using ANSYS software. The weld strength was measured through two independent analyses: thermal and mechanical. To simulate the heat source during welding, the element "birth and death" feature was utilized. Due to the high temperature gradient in the weld zone, the material's thermophysical and mechanical properties were considered temperature-dependent.The calculation of weld strength is of great importance. Given that experimental methods are time-consuming and costly, the provision of fast and economical methods is significant. In this article, weld strength has been calculated using two methods: simulation with ANSYS software and artificial neural networks with MATLAB software. Finally, the results obtained from these two methods (numerical simulation and artificial neural networks) have been compared with experimental results.