強化学習を用いた溶接残留応力低減のための溶接順序最適化システムの開発
概要
Currently, various welding is performed in the process of manufacturing all kinds of structures such as ships and bridges, and metal 3Dprinters are attracting attention in recent years. However, there are several problems in these manufacturing method. Residual stress occurs inside the products due to thermal processing. It is thought that defects such as cracks during manufacturing, deformation and fatigue fracture are caused by the tensile stress on the surface. Therefore, it is necessary to optimize the construction method as welding orders and to reduce the residual stress on the product surface. However, it is impossible to analyze all welding sequences of multilayer welding because of physical and temporal constraints. On the other hand, research on artificial intelligence has developed recently, and AI which exceeds human ability has appeared. In this research, we aim to construct a system that automatically obtains the optimum the welding sequence in which the residual stress on the product surface decreases by using AI for searching the welding sequence. In this research, we study the learning method for optimizing of the weld sequence based on reinforcement learning and show the possibility of applying AI to optimizing of the weld sequence.
