Flow characteristics and advanced forecasting method for ventilated cavitating flow around an axisymmetric body with different ventilation modes

Experiments are conducted to investigate ventilated cavitating flow around an axisymmetric body under different ventilation modes, with a systematic analysis of key characteristics in ventilated cavity morphology. Building on Akaike information criterion (AIC), an AIC-driven automated prophet model is proposed to achieve advanced forecasting of ventilated local cavity dynamics across varying ventilation conditions. The study identifies two dominant characteristics in ventilated cavity morphology: ventilation-induced characteristics and shedding characteristics, accompanied by significant lag effects. To address these characteristics, the proposed model incorporates: (1) an additional regressor component for ventilation-induced characteristic; (2) a multi-period seasonality component for shedding characteristic, and (3) a phase lag compensation integrated into the regressor to account for lag effects. Hyperparameters are automatically optimized using simulated historical forecasts and the AIC to enhance predictive accuracy. The results show that by coupling the cavity variation characteristics proposed in this study, the model successfully forecasts cavity length under both steady and sinusoidal ventilation modes, capturing both shedding characteristics and ventilation-induced characteristics. The errors in root mean square error, mean absolute percentage error, and shedding frequency remain within 12%. Moreover, this study further validates the model's exceptional long-term forecasting capability, as the forecast accuracy remains stable across different ventilation modes over time spans of 0.4, 0.8, 1.2, and 1.6 s.