Abstract
Trailing suction hopper dredgers (TSHDs) are characterized by their bulky hull forms. Traditional mathematical models based on Taylor series expansions lack sufficient predictive accuracy when estimating the hydrodynamic forces and maneuvering responses of such large vessels under highly nonlinear operating conditions, primarily due to the ambiguous physical interpretation of higher-order terms. To address this issue, this paper proposes a physics-informed neural network (PINN) hydrodynamic identification method that incorporates the cross-flow drag mechanism. A hybrid hydrodynamic model architecture combining linear potential flow with nonlinear cross-flow effects is developed. For longitudinal resistance, a polynomial model consistent with physical symmetry is retained; for lateral force and yaw moment, a cross-flow integral model based on strip theory is introduced. Utilizing a PINN network structure, the draft distribution function H(x) of the model ship’s hull geometry is explicitly embedded within the physical computation layer. Experimental data from large-amplitude oscillation tests conducted using a planar motion mechanism (PMM) were successfully employed to invert the cross-flow drag coefficient distribution along the ship’s length. The proposed model significantly enhances generalization ability and physical interpretability for large-amplitude roll and yaw motions, thereby providing a more reliable and physically consistent framework for predicting TSHD maneuvering behavior in complex sea conditions.