Observer-based fault detection for magnetic coupling underwater thrusters with applications in jiaolong HOV.

This study proposed an observer-based fault detection method for magnetic coupling underwater thrusters. To improve the accuracy of a thruster system model, a modeling identification method based on local recurrent neural networks was proposed, which can be described using state space equation. The algorithm for selecting model parameters was obtained by constructing a nonlinear constrained optimization model. Based on an identification model, a sliding mode observer was developed and employed for online fault reconstruction. Compared with traditional analytical-model-based thruster fault diagnosis methods, the proposed method can determine the fault cause to improve the submarine safety. The proposed method was validated based on the data of Jiaolong human occupied vehicle (HOV). • A modeling identification method based on local recurrent neural networks is proposed. • The parameters selecting algorithm is obtained by constructing an optimization model. • A sliding mode observer is developed and employed for online fault reconstruction. [ABSTRACT FROM AUTHOR]