Large-scale three-dimensional complex components are important components of ships, marine equipment and containers. There are many kinds of weld joints and complex spatial distribution in the components, and some of the weld joints have special welding process constraints such as collaborative synchronous arc welding and welding direction, so the existing collaborative welding path planning method is difficult to be applied to such complex scenarios. A multi-stage optimization method for dual robot collaborative welding path planning combining global and local was proposed. The three-dimensional space welding joint model of complex components and the dual welding robot collaborative model were established. According to the number of collaborative synchronous welds, the components were divided into corresponding number of small blocks, and the approximate optimal solution of local path planning for each weld block was solved by ant colony algorithm. Global path planning between multiple blocks was carried out by genetic particle swarm optimization algorithm. The results show that the method of multi-stage optimization based on block division can effectively solve the problem of dual robot synchronous welding path planning for large-scale three-dimensional complex components under special welding process constraints.