Drag reduction of SUBOFF model with deep-learning-optimized distributed partial-slip boundaries

Partial-slip boundaries have been demonstrated to be effective at drag reduction of underwater vehicles. However, in practical applications, covering the entire body surface with slip materials is cost-inefficient. To address this issue, a deep learning-based approach is developed to optimize the spatial distribution of slip boundaries on the SUBOFF bare-hull model under prescribed engineering constraints. First, a convolutional neural network surrogate model is trained to predict drag efficiently, enabling rapid evaluation of different slip-boundary configurations. The optimal distribution is then determined through systematic exploration of the parameter space. Numerical simulations of the optimized configurations reveal two key findings: (1) slip boundaries significantly influence the skin-friction drag, whereas their effect on pressure drag remains negligible and (2) a well-designed distributed partial-slip boundary can effectively regulate the development of near-wall turbulence, leading to enhanced the overall drag reduction performance. This study provides a promising framework for the intelligent design of drag-reducing surfaces for underwater vehicles.