Aluminum alloy arc additive manufacturing process and performance evaluation based on deep learning

The lightweight design of rail vehicle bogies effectively reduces overall energy consumption and enhances dynamic performance. Aluminum alloys, due to its lightweight and high-strength properties, have been widely utilized in the design of critical components. Arc Additive Manufacturing, characterized by its low cost and high efficiency, enables customized fabrication of components; however, its excessive thermal efficiency often results in reduced forming accuracy. This study addresses this issue through additive manufacturing experiments on 5356 aluminum alloy using three welding techniques: MIG-AM, DP-MIG-AM, and CMT-AM. The research investigates the influence of different combinations of three process parameters—stick-out length, welding current, and welding speed—on the macroscopic morphology of single-pass single-layer weld beads. Based on the macroscopic geometric data of single-pass single-layer welds, qualitative and quantitative relationships among parameters, quality, and forming characteristics are established. Furthermore, a data-driven machine learning approach is employed to predict the geometric morphology of weld beads, and a Particle Swarm Optimization (PSO) algorithm is integrated to optimize the model, enhancing its iteration efficiency and prediction accuracy. Finally, optimized process parameters for each welding technique are selected for single-pass multi-layer experiments, with evaluations conducted on deposition efficiency and mechanical properties.