J019: Reinforcement learning-based optimal fault-tolerant control for offshore platforms скачать в хорошем качестве

J019: Reinforcement learning-based optimal fault-tolerant control for offshore platforms

Ziaei A.*, Kharrati H.*, Salim M., Rahimi A. (2022, Mar.) Reinforcement Learning-Based Optimal Fault-Tolerant Control for Offshore Platforms, Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control Engineering, from SAGE Publications, Volume 236, Issue 6, pp. 1187-1196, doi: 10.1177/09596518221080323 This academic article presents a novel fault-tolerant control strategy for steel-jacket offshore platforms using reinforcement learning. The researchers address the challenge of maintaining structural stability against external wave forces even when actuator faults occur. By utilizing a fault observer to estimate malfunctions and integrating those estimates into a cost function, the system adapts its control law online through policy iteration. This method is particularly significant because it allows the platform to function effectively without requiring complete knowledge of the system's internal dynamics. Comparative simulations demonstrate that this approach utilizes less control force and exhibits greater robustness to mass variations than traditional offline methods. Ultimately, the study highlights how adaptive dynamic programming can enhance the reliability and performance of vital marine infrastructure.