Wake field prediction and optimization of submarine with fillet based on data-driven deep learning model

The fillet on the submarine is a rounded structure designed based on the body-stern appendages, which effectively weakens the horseshoe vortex at the junction between the rudder and the submarine hull, thereby improving the propeller's inflow quality. To investigate the impact of stern fillet shape on the wake flow, this research develops a data-driven steady wake field prediction model for submarines based on the U-Net architecture. By comparing the computational fluid dynamics (CFD) simulation results with the wake field prediction model, it is demonstrated that the prediction efficiency of flow field is significantly improved, and the prediction accuracy can be maintained simultaneously. Furthermore, the effects of stern fillets on the wake field at different radii are analyzed based on the prediction model, and the optimal stern fillet shape parameters are identified. Compared to the original model, the optimized fillet design reduces the velocity non-uniformity of wake flow by 20%.