Reliability-based design optimization of ship structure using SMOTE algorithm and dynamic surrogate model

Abstract

Objectives Aiming at the problem that it is difficult to ensure global approximation accuracy and computational efficiency in the traditional reliability-based design optimization of ship structures, a reliability optimization strategy based on a dynamic surrogate is proposed.Methods The BP nueral network surrogate is constructed with initial sampling points generated by the optimal Latin hypercube design method. The global optimization algorithm and single loop method (SLA) of the reliability optimization design are used to find the current global optimal solution. The sample points around the optimal solution are then added using the synthetic minority over-sampling technique (SMOTE), and the surrogate is updated to improve its accuracy near the global optimal solution until the optimization iteration converges.Results The SMOTE algorithm can synthesize the sample points located near the failure surface so that the surrogate fits the limit state function efficiently; the SLA decouples the reliability optimization problem into a deterministic optimization problem, improving calculation efficiency while maintaining calculation accuracy.Conclusions This optimization method is validated using a mathematical problem and ship structure reliability optimization. The optimization results show that the method can effectively reduce the calculation cost while obtaining the global optimal solution of the analysis model.

Keywords

• ship structure
• reliability-based design optimization
• dynamic surrogate model
• synthetic minority over-sampling technique (smote)
• bp neural network model
• single loop algorithm (sla)