Supercavitating vehicles are difficult to guide through traditional underwater acoustic homing

making electromagnetic detection after water exit a potential solution to this issue. In order to obtain the trajectory parameters of the water exit process of a supercavitating vehicle under multi-factor coupling conditions

a numerical model of the water exit process of the supercavitating vehicle was established. Orthogonal experimental conditions with four factors and five levels were designed

and a dataset of the initial and terminal motion parameters of the supercavitating vehicle was obtained through simulation computation. A prediction model for the initial and terminal motion parameters of the supercavitating vehicle water exit process was established using feedforward neural networks combined with Bayesian optimization. The research results indicate that during the water exit process

the cavitation on the windward side begin to collapse first

and the collapse speed on the leeward side is faster; All test sets of the BO-FNN model have R2 values above 0.9

indicating strong generalization of the prediction model and more accurate predictions of velocity and angular velocity than pitch angle and angle of attack; Comparison with FNN model shows that BO-FNN model has better prediction results. Obtained research results can provide reference for the water exit trajectory design of supercavitating vehicles.