Abstract: The collaborative optimization involved in multidisciplinary engineering system design is complicated because of the presence of variable couplings. With regard to the present problems of the existing solving strategies, a fast solving method based on classification and topology decoupling is proposed, in which the copy surrogate is adopted for the couplings among different clusters whereas the topology unidirectional discipline feasible strategy is used for the couplings in the same cluster. Firstly, the disciplines are clustered using the sensitivity analysis; Then the clusters are made independent by adding some constraints at the system level; At last, a tree structure without loop is obtained by performing the topology decoupling operation. With our method, the solving process is decreased to a great extent and the system solving process is accelerated. Two analytical multidisciplinary optimization problems are solved and the results are compared, which demonstrates the effectiveness of the proposed method.