Research on hydrodynamic forces prediction of underwater vehicle based on Co-Kriging model

Abstract

To enhance the accuracy of hydrodynamic forces predictions for underwater vehicle within certain resource constraints, this study integrates high and low-fidelity samples using the Co-Kriging method, combined with an expected improvement (EI) sequential infill criterion to construct a hydrodynamic forces prediction model. Different grid densities of computational fluid dynamics (CFD) calculations are used to distinguish high and low-fidelity samples. Taking the Joubert BB2 underwater vehicle as the research object, hydrodynamic forces predictions are conducted for various angles of attack and speeds. The effectiveness of the EI infill criterion in improving model prediction accuracy is validated. Furthermore, compared to the traditional Kriging model under the same computational resources, the Co-Kriging method, which integrates high and low-fidelity samples, significantly outperforms the Kriging model constructed solely from high or low-fidelity samples in overall prediction accuracy for underwater vehicle hydrodynamic forces.