Self-Learning Digital Twin for Offshore Floating Platform Dynamics and Productivity

Abstract

As offshore energy transitions to greener operations, turret-moored floating platforms increasingly function as advanced cyber-physical systems (CPSs) exposed to harsh seas and cyber-physical risks. This paper targets productivity and safety for turret-moored platforms with thruster-assisted heading control under time-varying environmental loads and potential cyber–physical anomalies. We develop an operational and control oriented self-learning digital twin that fuses physics based hydrodynamic modeling with online neural approximation to capture modeling errors and slow drifts during platform operation. Prediction residuals from the twin are used for lightweight anomaly monitoring, which provides early warning of abnormal sensor or actuator behavior and supports supervisory mitigation. Building on these predictions, an autonomous decision-support module optimizes the platform heading via a multi-factor performance index while adaptive controller maintains robust heading stabilization under bounded prediction residuals and advisory mismatch. Comprehensive simulations across three extreme multi-directional sea states and scaled basin experiments demonstrate substantial reductions in roll motion and improved heading regulation; the monitoring layer further provides actionable early warning for abnormal operation. The results show a practical pathway to monitoring, decision-making, and control for maritime energy CPSs.