Machine Learning-Enhanced Rapid Design of Hydrodynamic Shape for Underwater Vehicles Pedigrees

Currently, underwater vehicles are playing an increasingly important role in ocean observation, and rapid vehicle design is required for the increasingly diversified missions. Hydrodynamic shape is a vital discipline in underwater vehicle design, which consists of many mutually coupled design parameters. In this article, a machine learning-enhanced rapid design method is proposed to design the hydrodynamic shapes for the establishment of underwater vehicle pedigrees. The framework of the method is introduced, and a case study is detailed for the shape design of the underwater glider (UG). The shape of UG is optimized following various criteria to validate the feasibility of replacing the traditional computational fluid dynamics method with the machine learning method. Finally, a circulating water channel test is performed to validate the accuracy of the hydrodynamic data, while sea trial data are adopted to verify the correctness of the dynamic model for motion performance evaluation. The design method proposed can largely shorten the design cycle of underwater vehicles, and the broad design space explored in the case study helps build the hydrodynamic database of traditional UG hulls, thereby promoting industrialization.