A reduced order data-driven method for resistance prediction and shape optimization of hull vane

Hull Vane (HV) is an energy-saving appendage introduced by BV company to reduce total ship resistance. Shapewise, HV a hydrofoil wing transversely fixed at the transom bottom of hull. In this paper, data-driven shape optimization method proposed for HV. To avoid time-consuming resistance evaluation designs via viscous flow solver, we develop Machine-Learning (ML) based model that predicts hull's in presence For purpose, Principal-Component Analysis (PCA) first implemented dimensionality problem, and then prediction trained with most influential Principal Components (PCs). Given these PCs capture maximum geometric variance original design space, higher accuracy can be achieved expense few training samples. After phase, integrated optimizer, which searches dimensionally-reduced space optimal The obtained results 70% reduction aid PCA approximately 98% predicting Compared reference HV, optimized one reduced 1.2%. • A parametric on NACA used Vane. component analysis reduces accelerate optimization. Surrogate principal components infer stern zone pressure distribution due difference between suction sided. impact generated lift force vessel's trim sinkage.