Optimization method for hierarchical space reduction method and its application in hull form optimization

Abstract

Hull form optimization can be considered a typical “black-box” problem owing to the large number of design variables controlling the hull form, the time required to evaluate the hull form performance, and the unknown hull form performance space structure. Therefore, problems such as low optimization efficiency and difficulty in obtaining the optimal solution persist. Therefore, in this study, a novel optimization method that uses SOM (Self-Organizing Maps) and MARS (Model-based Annealing Random Search) to stratify the design space is proposed. In the proposed method, the design space is divided into an initial global space, a potential subspace, and a significant subspace during the optimization process. The SOM and MARS techniques are used for data mining to reduce the search domain of the subspaces and determine the global optimal solution. The proposed hierarchical space reduction method (HSRM) was validated using multiple functions as examples and one hull form optimization design problem. The proposed method could improve the optimization efficiency and accuracy of the solution.