关键词: 装配单元, 亲密度, Dijkstra算法改进, 树结构层级化

Abstract: A method of clustering component assembly units with hierarchical branching for the trunks of tree structure was proposed aiming at the problems of low parallelism of assembly work and difficult analysis of feature deviation in the component assembly processes. Firstly， the component tree structure was constructed according to the assembly association among the parts， and the assembly feature association index among the branches was analyzed. The index system and the application rules were established to express the closeness between the branches， and the closeness matrix was constructed to complete the representation of the trunks of the component tree. Secondly， with the closeness between branches as the initial input， an improved Dijkstra path planning algorithm was introduced， and the optimal solution of the local shortest path between the trunk and multiple branches was obtained by using the advantages of the algorithm itself. The stem-branch hierarchy matrix was constructed to realize the stem-branch hierarchy of the component tree structure， and then the assembly unit clustering was completed by merging the branches. Finally， virtual assembly was carried out with the results of unit partition， and interference was analyzed in the physical workshops to optimize the final partition result. The feasibility and effectiveness of the proposed method were verified by an example of assembly unit clustering of a vehicle differential assembly. 

Key words: assembly unit, closeness, improved Dijkstra algorithm, hierarchical tree structure