Detecting and diagnosing saturation faults

This paper presents a simple method for the detection and diagnosis of saturation faults in actuators for closed loop control systems. These faults are regarded as unexpected changes of the saturation level of the actuators. Assuming that there is a unknown nonlinear function which represents the relationship between the saturation and closed loop system performance such as overshoot and rise time, etc, a neural network is applied to approximate the nonlinear function. This neural network accepts the inputs as the performance parameters of the closed loop system and generates outputs as the estimated saturation faults. In the paper, the status of the actuator is represented by a parameter f. As a result, the estimated f (namely, the set of all f) produces the results for the fault detection and diagnosis of actuators. The status of the actuator is defined as 'healthy' if the set of all f = 1. Otherwise, a saturation fault in the actuator has occurred. A simulated example is included to demonstrate the use of the method and desired results have been obtained.