Hierarchical space reduction method based on self-organizing maps and K-means clustering for hull form optimization

ObjectiveDue to its high-dimensional, computationally expensive, and 'black-box' characteristics, hull form optimization based on CFD usually leads to low efficiency and poor quality. To solve the above problems, this paper proposes a hierarchical space reduction method (HSRM) based on the self-organizing maps (SOM) method and K-means clustering. MethodThe visualization results of SOM are clustered and the regions of interest are extracted. In this way, data mining is used to extract the knowledge implicit in the sample simulation data which can then be used to guide hull form optimization and improve its efficiency and quality. The proposed method is applied to the hull form optimization of a 7500-ton bulk carrier.ResultsThe results show that the total drag of the optimal hull form obtained using traditional particle swarm optimization (PSO) and HSRM is reduced by 1.854% and 2.266% respectively, with HSRM leading to a higher-quality optimized solution. ConclusionThe proposed method can guide the optimization algorithm to search in the direction of the optimal solution, effectively improving the efficiency and quality of optimization.