Nonlinear fluid-structure interactions and wake dynamic of flexible Antarctic krill trawls with fish pump: high-fidelity numerical simulation and VMD-enhanced deep learning architecture

Optimizing the design of Antarctic krill beam trawls requires accurate predictions of the coupled nonlinear dynamic and hydrodynamic behaviors, a challenge complicated by unsteady turbulent flows, large deformations, and operational constraints. This study proposes an integrated framework that couples high-fidelity fluid-structure interaction modeling with a Variational Mode Decomposition (VMD)-enhanced deep learning architecture to realize robust prediction of trawl nonlinear dynamic responses. A scaled physical model was tested in a flume tank, while numerical simulations employed unsteady Reynolds-Averaged Navier–Stokes approach paired with the Shear Stress Transport k-ω turbulence model, coupled to a nonlinear finite element structural model. To balance computational efficiency with accuracy, a grouping mesh technique based on a modified Tauti’s law was implemented, achieving experimental validation with maximum relative errors of 8.37% in drag and 9.09% in net mouth height. The nonlinear dynamic responses from FSI simulation were decomposed via VMD into Intrinsic Mode Functions (IMFs), whose spatial features were processed by a Convolutional Neural Network (CNN) and temporal sequences captured by a Long Short-Term Memory (LSTM) network. Comparative evaluation demonstrated that the hybrid VMD–CNN–LSTM model significantly outperformed LSTM and CNN–LSTM architectures, achieving R² values up to 0.996 and reducing prediction errors by up to 45%. Key findings show that a fish pump increase drag by 1.59–2.44%, expands the net mouth opening by 14%, and induces symmetrical large-scale wake vortices. Structural deformation rises by up to 43%, attributed to flow obstruction and low-frequency oscillations (0.05–0.1 Hz) synchronized with drag fluctuations. Furthermore, higher mesh grouping factors embedded in the framework produce amplified downstream velocity deficits and more symmetric wake fields, which correspond to improved hydrodynamic efficiency and operational stability of the trawl gear. The established multi-module, physics-informed integrated framework delivers a scalable, complete numerical analysis workflow for the parametric design of low-energy, stable, ecologically sustainable Antarctic krill fishing trawls.