Unsteady cavitating hydrofoil surface pressure prediction based on long short-term memory network

Abstract

<p indent="0mm">As a typical nonlinear dynamic problem, the evolution law and hydrodynamic characteristics of an unsteady cavitating flow are affected by noise and computer accuracy, thus, rendering it difficult to be predicted for a long time. In this paper, a long short-term memory network is used to develop a multivariable pressure prediction model, wherein the input is the velocity value and the vapor volume fraction, and the output is the time series of the pressure coefficient. The large eddy simulation (LES) data of the hydrofoil is predicted via extrapolation. The results show that the prediction results agree with LES simulation results well, demonstrating good generalization and the ability to predict more future results.