Neural network and second-order regression analysis in prediction of residual stress for high manganese steel after shot peening

Shot peening (SP) generates compressive residual stress (CRS) on component surfaces, a critical factor in enhancing fatigue performance. Recent research efforts have focused on optimizing CRS magnitude through parametric control of the SP process. We focused on the application of back propagation artificial neural networks(BPANN) and second-order regression analysis for predicting surface residual stress for high manganese steel components shot by means of different SP parameters, so multiple 3D FE dynamic simulations were conducted using an orthogonal experimental design. Research results show that the proposed two models could predict CRS of high manganese steel after different SP processes with reasonable accuracy, and the predictable ability of BPANN model is better than the predictable ability of second-order regression model, so the BPANN model can be applied to predict CRS efficiently for high manganese steel components shot using different parameters.