With the development of the field of ocean engineering, the impact of human activities on the ocean environment is increasing. Among them, marine structures such as ships and submarines inevitably generate various noises in the ocean environment during navigation and operations. Noise control of underwater structures and prediction of the acoustic field in the ocean environment is an active research field. This paper proposes a hybrid solver based on the finite element method and kernel method to analyze vibration and acoustic propagation of the underwater structure in the ocean environment. In the proposed hybrid solver, the finite element method is employed to obtain the acoustic-vibration coupling response of the underwater structure, the normal wave model is used to obtain the underwater acoustic Green's function for far-field acoustic propagation. Based on near-field acoustic information and Green's function, the machine learning technique is used to predict the far-field acoustic propagation response in the ocean environment. The benchmark example is carried out to demonstrate the accuracy and efficiency of the proposed hybrid solver in comparison with the COMSOL simulation.