Due to the complexity of the hull structure, traditional optimization methods are prone to fall into the local optimum and the solution speed is slow. Based on adaptive mutation particle swarm optimization (AMPSO), BP neural network, genetic algorithm (GA), combined with orthogonal experiments designed by Isight/Nastran, this paper proposes a structural optimization method—AMPSO-BP-GA. Subsequently, the optimization of the ten planar girders and gangboard structure were used as examples to verify the accuracy and feasibility of the optimization algorithm. The results show that the weight of the ten-bar truss optimized by the proposed AMPSO-BP-GA method is 2272.1kg under the same constraint conditions with the lightest weight as the objective function, which is lighter than the structure optimized by other methods; Using the AMPSO-BP-GA method to reduce the weight of the gangboard by 33.3%, the 25.4% weight loss of the GA-BP-GA method and the 17.9% weight loss of the PSO-BP-GA method are better optimized. This method can provide reference for the optimal design of ship structure.