Blind identification of incident waves and response transfer functions of a marine vessel based on measured responses

Estimating incident waves onboard is crucial for ensuring safe and efficient operation of a marine vessel. This paper is concerned with an in-situ response-based wave estimation method, in which incoming waves are inversely estimated via measured wave-induced responses and corresponding transfer functions (TRFs); a method widely referred to as the ’Wave Buoy Analogy (WBA)’. Specifically, a new phase-resolved WBA technique is studied. In a transfer function-dependent technique, the accuracy of estimated waves via the WBA is in principle contingent upon the accuracy of TRFs. However, it can be challenging to provide precise TRFs onboard, as the vessel's weight distribution and the operational conditions may vary from voyage to voyage. To address this issue, this paper proposes a novel approach to simultaneously attain estimations of response TRFs and incoming wave profiles based on measured wave-induced responses. Specifically, parametrized TRFs (P-TRFs) are introduced using a well-trained Artificial Neural Network (ANN), and the parameters characterizing the P-TRFs are identified through optimization, based on pseudo responses under reconstructed phase-resolved incident waves. A linear strip theory, the so-called New Strip Method (NSM), is utilized to train the ANN. Bayesian Optimization (BO) is employed to identify the parameters in the P-TRF. A numerical investigation using synthetic measurements generated via known TRFs is first made. Following this, the proposed approach is validated against experimental campaigns using scaled models of a bulk carrier and container ship, respectively, in long-crested irregular waves. It is unveiled that the estimation results of TRFs and incoming waves demonstrate a high degree of correlation with the experimental measurements. The computational cost of the presented approach is low, thus making the approach practically feasible for real-time applications.