Investigation of hysteresis behavior and critical parameters in ventilated cavitation flows using experiments, simulations, and machine learning

One of the important and complex phenomena in ventilation supercavitation is the existence of hysteresis variations, which involves the study of the coefficients of ventilation formation and collapse. In this article, this phenomenon using a multifaceted approach, employing experimental, numerical methods, and also a machine learning model has investigated. Initially, the hysteresis curve was studied on the disc-shaped cavitation at different Froude numbers using a combination of experimental and numerical methods. Subsequently, Bayesian optimization algorithm was employed to fine-tune the hyperparameters of the XGBoost model, resulting in the BOA–XGB model. The outcomes of this optimized model agree well with the both experimental and numerical results. The results indicate that, with an increase in the Froude number, the values of the ventilation formation and collapse coefficients initially exhibit an increasing trend, followed by a decreasing trend. Also, the results show that the collapse ventilation coefficient has a linear relationship with the formation ventilation coefficient. The results obtained from the BOA–XGB model demonstrate that as the cavitation diameter increases, the maximum value of the formation ventilation coefficient linearly moves towards smaller Froude numbers.