Grey-box models for wave loading prediction

The quantification of wave loading on offshore structures and components is a crucial element in the assessment their useful remaining life. In many applications well-known Morison's equation employed to estimate forcing from waves with assumed particle velocities accelerations. This paper develops grey-box modelling approach improve predictions force structural members. A model intends exploit enhanced predictive capabilities data-based whilst retaining physical insight into behaviour system; context work carried out here, this can be considered as physics-informed machine learning. There are number possible approaches establish model. demonstrates two means combining physics (white box) (black components; one where simple summation components, second white-box prediction fed black box an additional input. Here used physics-based component combination Gaussian process NARX - dynamic variant more regression. Two key challenges employing GP-NARX formulation that addressed here selection appropriate lag terms proper treatment uncertainty propagation within GP. best performing model, residual GP-NARX, was able achieve 29.13\% 5.48\% relative reduction NMSE over Equation black-box respectively, alongside significant benefits extrapolative circumstances low dataset coverage.