SVR-based identification of nonlinear roll motion equation for FPSOs in regular waves

A novel system identification method, Support Vector Regression (SVR), is proposed for identifying the nonlinear roll motion equation of a FPSO vessel in regular waves. Firstly, the roll motion of a vessel is simulated, and the simulated data are used to identify the parameters in the roll motion equation. Then the roll motion is predicted by using the identified parameters, and the prediction results are compared with the simulated data to verify the identification method. Secondly, model test data of a FPSO vessel are used to identify the parameters in the roll motion equation. The roll motion is predicted by using the identified parameters and compared with the model test data. In addition, by using the model test data, the time histories of the nonlinear damping and restoring moments in the non-parametric roll motion equation are identified and the identified results are used to predict roll motion. Comparison of the prediction results with the model test data shows the validity of the identification method in identifying the non-parametric roll motion equation. It is shown that the SVR-based identification method can be effectively applied to parametric and non-parametric identification of the nonlinear roll motion equation.