Integrated optimization of multi-station multi-robot assembly welding line: Application for automotive industry

• Published in: Expert systems with applications 2025-04, Vol.267, p.126116, Article 126116
• Main Authors: Wang, Ye, Wang, Xuewu, Chen, Sanyan, Gu, Xingsheng
• Format: Article
• Language: English
• Subjects: Assembly welding line; Industrial robot; Multi-objective optimization; Scheduling and allocation; Welding sequence planning
• ISSN: 0957-4174
• DOI: 10.1016/j.eswa.2024.126116

In automotive industry, the optimization of components mass manufacturing process is crucial for enhancing production efficiency. Automatic assembly welding line is the ideal choice for modern mass production. Existing research primarily focuses on the limited coupling of sub-problems for simple production line. This paper emphasizes the integrated optimization of complex coupling problems of multi-station multi-robot production line. Highly coupled sub-problems such as robot allocation among stations, task allocation among stations, and welding sequence planning are integrated optimized, while meeting a large number of production line composition and parts assembly constraints. At the same time, the robot jumping stations work and man-robot synchronous work are also studied. Large coupling problems with numerous constraints are integrated into a unified and novel multi-station multi-robot allocation and sequence optimization framework without considering collision. Moreover, a set of innovative and complete system optimization model and algorithm are proposed to address the intricate issues. Firstly, the optimization model based on the framework is established considering constraints including robots stations accessibility, robots welding accessibility, welding integrity and process feasibility. Then, a three-layer chromosome is designed to represent the decision space of the coupled integrated optimization problem. To ensure the rationality and validity of chromosomes, a chromosome correction method, which eliminates the robot inaccessibility and welding incompleteness, is devised for population iteration. Meanwhile, a similarity taboo non-dominated sorting rule is proposed and applied to enhance NSGA-III, resulting in the STNSGA-III algorithm. Finally, the cases of actual and mimetic automotive rear floor welding tasks are studied to verify the effectiveness of the STNSGA-III. Compared with three state-of-the-art multi-objective evolutionary algorithms and the artificial optimization result, the comprehensive performance of STNSGA-III results is superior to the comparison algorithms, which shows the effectiveness and feasibility of the proposed model and algorithm for optimizing the high-coupling assembly welding line.