Prediction modeling for yaw motion of deep-sea mining vehicle during deployment and recovery: A physics informed neural network (PINN) approach

This paper presents a physics informed neural network (PINN) method for constructing a yaw motion hydrodynamic model of the deep-sea mining vehicle during the deployment and recovery processes. Initially, by incorporating the motion equations of the underwater vehicle as part of the loss function, the synchronous construction and optimization of parametric and non-parametric hydrodynamic models are achieved. Subsequently, focusing on the mining vehicle "Lushan", the deployment and recovery processes of deep-sea mining vehicles are simulated using computational fluid dynamics (CFD) methods. The CFD simulation results are utilized as driving data for the mining vehicle hydrodynamic modeling, employing both the novel neural network approach and the conventional neural network (NN) method. A comparison case study reveals that the newly proposed neural network method not only enables synchronous identification of parametric and non-parametric models, but also exhibits resistance to NN overfitting and enhanced generalization capabilities.