Traditional analyses of ship roll damping and added moment of inertia rely on free roll decay and forced roll tests, but acquiring linear (small angles) and nonlinear (large angles) relationships demands extensive computational cases and parameter fitting, limiting efficiency. To address this, this study couples Support Vector Machine (SVM) and Neural Ordinary Differential Equation (NODE) networks: SVM solves for added moment of inertia, linear damping, and nonlinear damping, while NODE constructs a complete model for the roll motion equation. Using the DTMB5415 hull form, Computational Fluid Dynamics (CFD) simulations of forced roll build a “time-angle-moment” sample space, and the coupled model learns and predicts free roll decay under different initial angles. The results show that SVM effectively determines roll damping and added moment of inertia from constant-amplitude variable-frequency and constant-frequency variable-amplitude data, reducing required cases significantly. NODE’s simulation of free roll decay validates coefficient accuracy. Within a certain angle range, the SVM-NODE model meets rapid roll motion analysis needs, providing an innovative method for ship roll research and engineering.
To improve unified modeling of steady two-dimensional lid-driven cavity flow across a wide range of Reynolds numbers, this study proposes a Vorticity-Enhanced Physics-Informed Neural Network (VE-PINN). The method augments a standard velocity-pressure PINN with a vorticity-transport residual and uses a logarithmic Reynolds-number embedding, , for multi-regime training. Using CFD benchmark data as supervision and evaluation, we conduct systematic ablation studies on network architecture, loss weighting, sampling density, input embedding, and physical constraint over , together with out-of-range extrapolation tests. The results show that the logarithmic Reynolds-number embedding improves cross-regime training stability and reduces the multi-regime mean relative error, while the vorticity-transport constraint improves the reconstruction of velocity fields and secondary vortical structures with only a modest increase in training cost. Further comparisons based on contour fields, centerline velocity profiles, vortex-core locations, and vorticity intensity indicate that VE-PINN provides improved accuracy, physical consistency, and generalization relative to the baseline PINN in the present benchmark. These findings suggest that, for the steady cavity-flow problem considered here, combining logarithmic parameter embedding with derivative-level physical constraint is a practical and effective strategy for parametric PINN modeling.