Online Seakeeping Forecast Coupling System Identification With Nonlinear Pseudo Spectral Potential-Flow
This paper presents a grey-box framework for a robust deterministic, real-time prediction of coupled surge-heave-pitch motion in waves. The approach combines a physics-based online system identification with a nonlinear pseudo-spectral potential flow solver. Rather than prescribing a complete hydrodynamic model from captive model tests or relying on a black-box surrogate, three coupled compact, dynamics equations are identified with constrained ridge-regression of a short time-history of free-running body motion data in waves. The PSPF solver is then used to provide the future incident wave at the body's location. The PSPF and dynamic system models are coupled for continuous real-time future motion prediction in a receding horizon architecture, while the model coefficients are re-identified online to enable robust prediction to changing wave-conditions. This novel framework is demonstrated based on CFD data that has been validated for a freely floating two-dimensional rectangular box in waves. The results demonstrate a feasible framework for accurate motion forecasting up to around eight wavelengths into the future utilising data that can be gathered onboard a ship in real operations. Furthermore, it is shown that the framework is robust to changing sea-states and agnostic to underlying model geometry.