基于代理模型技术的江海直达船体结构可靠性优化设计

Reliability optimization design of river-sea-going ship hull structure based on agent model technology

为提高船体结构可靠性优化设计这类多参数、高维度、高非线性优化问题的效率,弥补传统船体结构确定性优化设计对影响结构安全不确定性因素考虑的欠缺,本文以一艘江海直达船为研究对象,采用BP(Back Propagation)神经网络代理模型技术,结合SMOTE(Synthetic Minority Oversampling Technique)过采样算法增加极限状态失效面附近样本点数量,实现以较少的样本点获得较高精度的船体舱段结构极限状态代理模型,进而利用蒙特卡洛抽样模拟方法编制船体结构可靠度计算程序,在结构可靠性分析的基础上以降低船体舱段结构重量为目标,采用模拟退火优化算法开展船体结构可靠性优化设计。本文建立了一套完整有效的、基于代理模型技术的船体结构可靠性优化设计系统,有效提高了优化设计的效率,对江海直达船体结构可靠性优化设计具有指导意义。

In order to improve the efficiency of multi-parameter,high dimensional and high nonlinear optimi⁃zation of ship structure reliability optimization design and make up the lack of uncertainty factors affecting structural safety in traditional deterministic optimization design,a river-sea-going ship was taken as the re⁃search object.BP(Back Propagation)neural network agent model technique and SMOTE(Synthetic Minority Oversampling Technique)algorithm were used to increase the number of sample points near the failure sur⁃face,in order to obtain a high-precision limit state agent model of ship structure with fewer sample points.Combined with Monte Carlo simulation method,the reliability calculation program of hull structure was de⁃veloped.Structural reliability optimization analysis was performed adopting the simulated annealing optimiza⁃tion algorithm in order to reduce the structural weight.A set of complete and effective reliability optimization design system based on agent model technology was established to improve the efficiency of reliability optimi⁃zation design,which has guiding significance to the reliability optimization design of river-sea-going ship structures.