Collision detection and fault diagnosis with DMAO-GRU for flow regulating valve

Abstract Fault diagnosis and condition monitoring are critical to the real-time and accurate flow regulation of the flow regulating valve. For the blockage fault, one of the main types of flow control valve faults, this paper proposes a collision detection and fault diagnosis method based on gated recurrent unit (GRU) and dynamic model-based adaptive external torque observer (DMAO). Firstly, combined with the generalized momentum, the DMAO based on multi-layer perceptron and system dynamic model is constructed. Compared with the dynamic model with fixed parameters, the estimation error is reduced by 25%. Secondly, this paper proposes a collision detection method based on DMAO-GRU. In the absence of an accurate dynamic model, the optimal performance of collision detection is achieved by parameter adaptive estimation and time series signal modeling technology, compared with other detection methods. Moreover, according to the estimation of DMAO and collision detection results, this paper proposes a method to detect the degree of valve port blockage and realizes condition monitoring and fault-tolerant control.