Rapid data-driven individualised shape design of AUVs based on CFD and machine learning

Abstract

ABSTRACT The shape design of autonomous underwater vehicles (AUVs) has an essential influence on their hydrodynamic performance. In this study, a data-driven approach is proposed based on computational fluid dynamics (CFD) and machine learning to efficiently design the hydrodynamic shape of AUVs with different sailing velocities, angles of attack (AOAs) and volumes. Based on the data-driven approach, a shape optimisation design for AUVs is established. Moreover, the variation of optimal shape with the design constraints of AOA, velocity and volume is also explored, and the results prove that a relatively larger body with a sharper head is more beneficial for the Myring shape AUVs sailing with a higher velocity and a larger AOA. The results of CFD calculations and those of the data-driven approach are compared, and a circulating flume test is carried out to prove the reliability of CFD method. The comparison verifies the reliability of the optimisation result.