Adaptive fuzzy-neural-based multiple models for fault diagnosis of a pneumatic actuator

Due to the inherent nonlinearity and uncertainty, fault diagnosis in pneumatic actuators is a very difficult task. Developing the models of nonlinear systems with adaptive network-based fuzzy inference systems (ANFISs) has recently received attention. Modeling that are built upon ANFISs overcome the disadvantages of ordinary fuzzy modeling and can be very suitable for generalized modeling of nonlinear plants. In this paper, we setup a group of models, which are relatively common in practice, corresponding to various situations of a pneumatic actuator, including normal, low and high supply pressure. We construct a multiple models-based fault diagnosis system to generate residual signals and detect fault occurrence using the novel concept of minimum index of sum of the absolute values of the residual errors. The trade-off between the robustness and the sensitivity of the developed scheme is considered to isolate faults by employing a fault index. The effectiveness of the proposed fault isolation scheme is demonstrated via experiments.