High-speed underwater vehicles can experience a dramatic drag reduction when a natural supercavity forms: a vapor-filled envelope that surrounds the body, minimizing contact with water. The shape of this cavity is critical to maintaining stability and performance, but predicting it accurately is challenging due to the complex multiphase physics involved.
The paper introduces a physics-guided multi-fidelity surrogate modeling approach to predict the natural supercavity geometry behind an axisymmetric conical cavitator. By combining low-fidelity and high-fidelity data with physical constraints, the model aims to improve accuracy while reducing computational cost compared to purely high-fidelity simulations.
The authors focus specifically on the conical cavitator geometry, a common configuration for generating supercavities. The results suggest that this hybrid approach can effectively capture the cavity shape, offering a practical tool for designing high-speed underwater vehicles where drag reduction is a key performance goal.