To solve the problem of low disassembly efficiency of complex products, an asynchronous parallel disassembly sequence planning method of multi-manipulator based on improved shuffled frog leaping algorithm was proposed. Depending on the characteristics of this method, the disassembly manipulators are taken as the research object, and the minimum completion time of disassembly is taken as the optimization objective. Based on the basic disassembly time of parts, the disassembly preparation time caused by the change of disassembly tools and directions is considered, so that the objective function evaluation model is more realistic. Then, the mathematical model of asynchronous parallel disassembly sequence planning is constructed and the disassembly information modeling method is optimized. We proposed a new method based on the fastener constraint matrix and the part constraint matrix to build the information model of product disassembly. Based on this model, the detachable conditions are derived and the feasible disassembly sequence is obtained. In order to improve the performance of the proposed algorithm, a double-link encoding method and an efficient decoding operator is proposed. The adjustment orders which based on adjustment position and the substitution formulas are constructed to carry out the new segmental local evolution of frog individuals, and the update formula of frog position is reconstructed. Furthermore, the two-point mutation operator is added to the global information exchange strategy to enhance the search ability of the algorithm. The proposed algorithm is applied to a screw lifting mechanism and four products of varying complexity, and the results are compared with two algorithms. The results show that the proposed method can solve disassembly problems of different scales and is especially effective in solving large-scale disassembly problems.