Identification of cavitation regimes using SVM: A combined numerical, experimental and machine learning approach

The phenomenon of artificial cavitation is a significant and practical aspect of reducing drag forces on devices interacting with water. Among various regimes, the supercavity regime, where the entire body or a substantial portion of it is enveloped by a cavity, plays a crucial role. This study investigates the cavitation phenomenon using a combination of numerical experimental methods and machine learning techniques. Specifically, the Support Vector Machine (SVM) classification model is employed to identify the type of artificial cavitation regime and to determine the occurrence of the supercavity regime based on input variables. The findings indicate that the Radial Basis Function (RBF) model outperforms other machine learning models in accurately detecting the cavity regime type, achieving an accuracy of over 95% in identifying the supercavity regime. Additionally, results from the RBF model demonstrate that the supercavity regime is observed at low cavitation numbers, exhibiting the longest cavity length.