Abstract: Aimed at the problem that ship structures reliability-based design optimization has poor efficiency and difficult convergence due to high nonlinearity, a reliability-based design optimization of ship structures based on SMOTE (synthetic minority oversampling technique) algorithm is proposed. The innovation is to establish an improved BP (back propagation) neural network by SMOTE algorithm and the limit state function with high approximation could be set up with fewer samples, which overcomes the shortcoming that old model cannot satisfy both accuracy and efficiency requirements. Moreover, the feasibility of establishing BP neural network model by SMOTE algorithm has been proved by a mathematical instance. Then put the improved BP neural network and simulate anneal arithmetic into single loop approach optimized strategy to solve the issue of ship structure reliability-based design optimization. The result confirms the efficiency and accuracy of the model. In addition, the proposed model can provide an idea for large engineering structures reliability-based design optimization.

Key words: reliability-based design optimization, limit state function, single loop approach, SMOTE (synthetic minority oversampling technique) algorithm