摘要
船舶多学科设计优化是典型的多目标优化问题。采用基于物理规划的多目标优化算法可以得到均匀分布的Pareto前沿,但对于复杂的船舶多目标优化问题,其优化质量有待改进。文章首先对物理规划进行改进,改善其偏好函数的光顺性及对偏好区间的适应性,以改进其优化效果。然后采用缩减搜索域和转动搜索域技术,将矩形的伪偏好结构转化为菱形,从各个方向对目标空间进行搜索,在此基础上进一步提出转动伪偏好结构技术以扩展搜索范围,由此可以降低Pareto前沿对伪偏好结构的要求,提高其分布质量和范围。文章以数值算例和船舶概念设计优化为例进行验证,并与多目标进化算法进行对比,证明了改进后的多目标优化算法的有效性。
With Multi-objective optimization algorithm based on physical programming an even distribution of Pareto front is able to be obtained, but the optimization quality need to be improved for complex multi-objective optimization problem in ship design. In this paper, the applicable effect of physical programming is improved with improving the smoothness of preference function and the adaptability of preference region. Then technology related to the shrink and rotation of search domain is applied to convert the pseudo-preference structure from rectangle to rhombus, while the search of objective space is performed in every direction. On this basis technology in rotating pseudo-preference structure is proposed to further extend search range. Comparing with multi-objective evolutionary algorithm by a numerical example and a conceptual ship design optimization, the effectiveness of the improved multi-objective optimization algorithm has been verified.
出处
《中国造船》
EI
CSCD
北大核心
2014年第3期53-63,共11页
Shipbuilding of China
基金
国家自然科学基金重点项目(51039006)
国家自然科学基金项目(51279147)
国家自然科学基金项目(51179143)
国家自然科学基金(51479150)
关键词
多目标优化
物理规划
PARETO前沿
均匀分布
multi-objective optimization
physical programming
Pareto front
even distribution
