Composite propeller design optimization for cavitation minimization using deep learning-based objective parameter prediction model

Abstract

Recently, composite propellers have attracted attention as a means of reducing cavitation. To maximize cavitation reduction when using a composite propeller, the design of a composite propeller must be optimized. In this study, deep learning-based prediction models for composite propeller design optimization and design optimization procedures based on these models are proposed. The prediction models are trained using a training dataset consisting of the training input data obtained from a data scan grid, and the training output data include cavitation volume, adaptive deformation, and failure index. To minimize cavitation on a composite propeller, a composite propeller design optimization procedure for propeller geometry and composite lay-up sequence based on the developed prediction models is established. By following the proposed procedure, an optimized composite propeller design that minimizes cavitation volume and adaptive deformation is obtained. The cavitation of the optimized composite propeller is approximately half that of the original propeller. This result verifies the effectiveness of the proposed design optimization procedure based on the developed prediction models.