非可行解驱动进化算法和多元分析技术在船型参数优化中的应用

Application of infeasibility driven evolutionary algorithm and multivariate analysis in shipform-parameters research

引入非可行解驱动进化算法(infeasibility driven evolutionary algorithm,IDEA)和多元分析技术开展船型参数优化和设计模型分析.针对一艘散装货舱在概念设计阶段的船型参数设计,应用IDEA算法进行多目标优化,然后采用距离理想解最近的方法对Pareto解集进行量化评价,选取一个满意的设计方案,最后应用多元分析技术分析Pareto解集获取船舶设计变量之间特性,即采用层次聚类方法得到样本或者变量之间的相互距离关系和等距特征映射(Isomap)的降维方法,得到变量在二维平面上的映射图,采用最小二乘法得到Pareto解集上变量之间的拟合关系式.数值结果表明:IDEA运算速度快,Pareto解集分散性良好.基于多元分析技术的数据挖掘应用能够获得对模型更多的认识,揭示模型内在关系.

Infeasibility driven evolutionary algorithm（ IDEA） is employed to find the approximate Pareto setsof ship form parameters optimization. and multivariate analysis is used to analyze ship design model. A bulk carrier concept design problem is studied in the analysis. The method,which ranks the distance of each alternative solution to the utopian solution,is used to obtain comprehensive evaluation of the Pareto solutions. The inherent characteristics of model can be investigated from Pareto sets by data mining skills. The relations between variables can be studied from hierarchical clustering method. The dendrogram can provide group information about variables. The distance among all variables can be investigated from Isomap,which uses the graph distance as an approximation of the geodesic distance. The multiple nonlinear regressions are also examined for finding the relations between design variables and objective functions. The numerical results show that the IDEA algorithm can find the Pareto sets effectively,and the multivariate technique can be utilized for data mining in ship design.