LAUSR

Abstract Online tuning of vessel models based on onboard measurement data can reduce the uncertainties of vessel motion prediction, and therefore potentially increase the safety and cost efficiency for marine operations. Among the uncertain vessel parameters, the roll damping coefficient is very important and highly nonlinear. In reality, roll damping depends on the sea state and vessel condition. This paper proposes two different procedures for tuning the sea state dependent roll damping coefficient together with other uncertain vessel parameters, i.e., 1-step tuning and 2-step tuning procedures. In addition, a roll damping prediction model based on Gaussian process regression is also proposed to predict the roll damping for future sea states based on historical data. The tuning procedure together with the proposed prediction model form an iterative closed loop of continuously improving the knowledge about the roll damping online, also estimating the model uncertainty based on prior knowledge, sampling uncertainties, and the applied kernel. Case studies are presented to demonstrate the procedures.