A new estimation/decoupling approach for robust observer-based fault reconstruction in nonlinear systems affected by simultaneous time varying actuator and sensor faults

Abstract

This paper presents an estimation/decoupling approach to design a fault estimation observer for nonlinear systems affected by a simultaneous actuator and sensor faults. In this approach, the sensor fault is treated as unknown inputs while the state and the actuator faults are estimated by recalling the extended state observer (ESO) idea. A Takagi-Sugeno Multiple-Integral Unknown Input Observer (TSMIUIO) is used to achieve this approach. First, the TSMIUIO estimates an extended state consists of the original system state and the actuator faults by using the ESO idea. On the other hand, the sensor faults are decoupled from the estimation of the extended state by treating such faults as unknown inputs. Then, the TSMIUIO permits an implicit estimation of sensor faults by using the output signals. The benefits earned by such an approach are: (1) the approach does not presume constraints on the time behaviour of the sensor faults. (2) By decoupling the sensor faults, the approach eliminates the bi-directional interaction (BDI) between the estimated signals. (3) It offers an implicit estimation of sensor faults by using the output signals. The design algorithm is given in a linear matrix inequality (LMI) form. The single-link flexible joint robot and the FAST 5MW benchmark wind turbine have been used to show the effectiveness of the proposed method.