Efficient positioning error compensation for robots in wire arc hybrid manufacturing systems

Wire arc additive manufacturing is a promising technology but is still limited by insufficient manufacturing accuracy. Despite numerous studies on process parameters to enhance manufacturing precision, the errors introduced by robot in hybrid manufacturing systems have not been effectively addressed. Unique on-site conditions such as varying robot poses and large working spaces have rendered many previous methods ineffective, making error compensation a challenging task. To solve this issue, an efficient compensation method for robots in wire arc hybrid manufacturing systems is proposed. A similarity-Radial Basis Function Neural Network is proposed to tackle pose variation issues that hinder error compensation methods, guaranteeing accuracy despite robot pose variations. However, the process of sampling to train neural networks is arduous. Arbitrarily reducing the number of sampling points is not feasible. Instead, optimizing the sampling process is a more effective approach. In this paper, we adopt the workspace Measurement and Positioning System and design a novel target based on circumferential constraints, presenting a comprehensive measurement optimization solution. This solution makes the arduous sampling process for training no longer difficult, significantly reducing the sampling time. Experimental verification shows that after using the proposed method for compensation, the positioning error decreased to 0.20 mm, and the compensation efficiency also significantly increased over 60%. To further validate the practical application of the method, real manufacturing tests are conducted in practical manufacturing scenarios. The results demonstrate good compensation effects, proving the feasibility of the compensation method.