Due to the complexity of hull structures, traditional optimization methods are prone to fall into the local optimum and have slow solution speeds.

For this reason, based on an adaptive mutation particle swarm optimization (AMPSO) algorithm, BP neural network and genetic algorithm (GA), combined with orthogonal experiments designed by Isight/Nastran, an AMPSO-BP-GA structural optimization method is proposed. Subsequently, the optimizations of cross-bar truss and gangboard structures are used as examples to verify the accuracy and feasibility of the algorithm.

The calculation results show that under the same constraints, the weight of a cross-bar truss structure optimized by the AMPSO-BP-GA method is 2 272.1 kg, which is lighter than structures optimized by other methods; and using the AMPSO-BP-GA method, the weight of a gangboard is reduced by 33.3% compared with the 25.4% weight reduction of the GA-BP-GA method and 17.9% weight reduction of the PSO-BP-GA method, demonstrating that the AMPSO-BP-GA method has superior optimization results.

Compared with the three methods of BP-GA, PSO-BP-GA and GA-BP-GA, the AMPSO-BP-GA method has a better effect in the optimization of lightweight structure, and can provide references for the optimization design of hull structures.