改进的MODSA算法在船型优化中的应用

针对传统智能优化算法处理高维优化问题时易陷入局部最优解且优化效率低的问题,文章采用近几年提出的基于模型的动态抽样分配(MODSA)算法作为研究对象,该算法具有处理高维优化问题的潜力,但对某些复杂高维问题很难搜索到全局最优解。为避免MODSA算法陷入局部最优解,采用多元正态分布作为抽样分布并推导相应参数更新式;为进一步提升该算法的优化效率,采用均匀设计确定初始抽样分布的期望值并通过Sigma管理水平自适应确定初始方差。通过数值函数测试结果表明:改进的MODSA算法具有更好的优化性能。最后,将改进的MODSA算法应用于5100TEU集装箱船兴波阻力性能优化。

An improved MODSA and its application in hull form optimization

Traditional intelligent optimization algorithms are easy to fall into local optimal solution and ineffi⁃cient in solving high-dimensional optimization problem.Model-based Optimization with Dynamic Sample Al⁃location(MODSA)algorithm proposed in recent years has potential in dealing with high-dimensional issues.But when the high-dimensional issues are complex,it is hard to find the global optimal solution.To avoid the MODSA algorithm falling into local optimal solution,multivariate normal distribution is introduced for sam⁃pling,and the corresponding expressions of the parameter update are deduced.To further improve the optimi⁃zation efficiency of the MODSA algorithm,the uniform design is adopted to determine the expectations of the initial sampling distributions.And Sigma management level is used to adaptively determine the initial vari⁃ance.The test results of the numerical functions show that the improved MODSA has a better optimization performance.Finally,the improved MODSA is applied to wave-making resistance optimization of a 5100TEU container ship.