改进的交叉熵法在船型优化中的应用

Application of an improved cross-entropy method in hull form optimization

针对常规智能优化算法在处理基于集成仿真的船型优化问题时易陷入局部最优解且优化效率低的缺点,本文采用具有处理高维(high-dimensional,HEB)问题潜力的交叉熵(cross-entropy,CE)法求解带约束的船型优化问题。本文研究了在CE法中嵌入"接受-拒绝"法,并对初始方差的选取方法进行自适应设计。通过数值函数的测试结果表明:改进的CE法较常规智能优化算法具有更好的优化性能。将改进的CE法应用于5100TEU集装箱船兴波阻力性能优化,根据优化结果可知:改进的CE法鲁棒性好,对多变量、多约束的船型优化问题能以较快的速度收敛到全局最优解。

Hull form optimization based on integrated simulation is easy to fall into local optimal solution and is of low efficiency when using general intelligent optimization algorithms. A cross-entropy( CE) method recently developed which has potential advantages in dealing with HBE problems was applied to hull form optimization with constraints. And an accept-reject method was embedded into the CE method,where an adaptive initial variance selection method was introduced. Result of tests on numerical functions shows that the improved CE method has better optimization performance than the conventional intelligent optimization algorithm. The improved CE method was implied to wave-making resistance performance optimization of a 5100 TEU container. Optimization results show that the improved CE method has better robustness,and is faster to converge to global optimum for multivariate multiconstraint hull form optimization issues.