Lightweight design on bridge crane exist the following problems by ANSYS: Mostly using static indicators as a design criterion,ignoring the dynamic performance; the whole process takes a long time; and the parameters optimized often should be rounded,weaken the effect. In this paper,taking the bridge crane with 20t- 22. 5m as an example,utilization of ANSYS and multidisciplinary design optimization software Isight,based on the theory of Latin hypercube design,BP neural network,Hooke-Jeeves algorithm,presenting a better and faster optimization method for main girder of bridge crane. By this method,optimization time shortened,and a decrease in mass of girder of bridge crane by 14. 84%.