Self-propulsion performance predictions of AUV based on response surface methodology

The self-propulsion performance of an autonomous underwater vehicle (AUV) is predicted based on computational fluid dynamics (CFD) and response surface methodology (RSM). The innovations are that a data-driven predictive framework is proposed and a model of coupled rotation speed, self-propulsion velocity and power is developed. In the present study, the discretized propeller method and multiple reference frames (MRF) model are adopted. The design of experiment (DOE) and numerical simulations are executed with inlet velocity and propeller rotation speed as inputs and drag, thrust and torque as outputs. Using the simulation results as sample data, the prediction accuracy of multiple models is compared. The polynomial model is applied to fit the response points, describing the relationship between inputs and outputs. The goodness of fit and predictive ability is assessed using the adjusted coefficient of determination (R2adj), root mean square error (RMSE) and relative error (RE). Finally, the predicted results are further validated by experiments, and all relative errors are below 6%. A linear function of self-propulsion velocity versus rotation speed and a specific cubic function of propeller input power versus self-propulsion velocity is obtained. The prediction method and achievements are beneficial for the navigation, control and safety of AUVs.