Numerical simulation and ANN algorithm-based predictive modeling in self-piercing riveting of aluminum and high-strength steel sheets

This study contributes to examine the impact of design variations on self-piercing riveting (SPR) in dissimilar metal joints, focusing on DP780, DP980, 1180MS steel, and Al6061 aluminum sheets with total thicknesses ranging from 2 to 4 mm. The study aligns with the target for aluminum content in light vehicles, which is set at 570 net pounds per vehicle by 2030. On the other hand, dual-phase steel combinations are crucial in the automotive industry due to their superior strength-to-weight ratio and excellent formability, which enhance vehicle safety and fuel efficiency. To address these factors, the study presents several key innovations: (1) the development of a novel artificial neural network (ANN) model for predicting riveting quality and mechanical properties, capturing parameters not covered by simulations alone; (2) the use of a contour graph method to optimize sheet thickness and die depth, introducing a new approach for achieving optimal SPR results; and (3) the establishment of a new correlation between process chain quality and shear test evaluations for self-piercing rivets in dissimilar metals. Results show that a die depth of 2.25 mm is most effective for joining 1-mm (1180MS) and 2-mm (Al6061) materials, achieving a maximum tensile force of 9.26 kN and absorbing up to 36.02 J of energy. The ANN model demonstrated high prediction accuracy with MAPEs ranging from 7.56 to 15.8%, highlighting its potential for integration into industrial applications. By allowing precise prediction of joint performance, the ANN model and the contour graph offer a transformative tool to optimize the SPR process, minimize development costs, and improve production efficiency in automotive manufacturing.