Application and optimization of an artificial neural network to forecast thermo-mechanical behaviour of HLE steel under weld spot effect

The reliability of assembled structures is significantly influenced by the applied thermomechanical stresses and the robustness degree of the simulation numerical methods. The utilization of classical numerical methods such as the finite element method (FEM), extended finite element method XFEM, and mean weighted residuals method are computational costs due to the complexity of the materials behaviour laws, physicals mathematical model and laboratory apparatus cost. To ensure accurate investigation techniques, it should be performed a numerical model used for resolving welding physical equations governed. The main objective of this study is to architect and optimize an intelligent model based on an artificial neural network to resolve a complex model of the calculation effect of spot welding on the behaviour of HLE steel. The ANN model gives a strong correlation between the dataset as numeric input and the target. The artificial neuron network gives a proxy model approach to exploit input data and results extracted by simulation of weld spot using finite element method FEM. The performance evaluation of the ANN model was carried out using mean square error and regression analysis. As a result, the present model ANN gives with minimum computational cost a good match of temperature estimating, equivalent stress and strain along the contact area of two thin plates of steel studied assembled by weld spot with a comparison between classical models using FEM.