Fault Detection and Isolation for Manipulator Robot Using Optimal Unknown Input Observer

Fault detection and diagnosis is one of the important and challenging issues in the field of control engineering. The robotic mechanical systems used instead of human in industrial, unreachable and hazardous spaces are always exposed to different kinds of stress and are susceptible to different kinds of fault in their operators and sensors. Fault detection and diagnosis in shortest possible time after the occurrence of fault, fault isolation, and detection of faulty components may prevent serious damages and additional costs. This paper aims to detect and diagnose the faults of manipulator robot using unknown input observer. The proposed observer is able to estimate the virtual modes, generate proper residuals, diagnose and detect sensor faults, and make fault detection process robust respect to disturbance and noise. The challenge of this observer is to determine its parameters which are sometimes inconsistent with each other and therefore have to be determined in the order of priority based on fault detection goals. In this paper, we optimize and determine the parameters of unknown input observer using optimization genetic algorithm. The proposed observer combined with the comparative threshold designed in this paper minimizes the number of wrong alarms and fault detection failures. The simulation results and their comparison with the extended Kalman filter confirm the efficiency of the proposed observer in the robust fault detection and diagnosis for manipulator robot.