Actuator and Sensor Fault Reconstruction Using an LPV Sliding Mode Observer

Abstract

This paper presents a new actuator and sensor fault reconstruction scheme for Linear Parameter Varying (LPV) systems, based on a sliding mode observer. Two separate observer schemes for actuator and sensor fault reconstruction are presented. For the design of the actuator fault reconstruction scheme, the varying input distribution matrix has been factorized into a flxed and a varying component. A virtual system comprising the system matrix and the flxed input distribution matrix is used for the design of the observer. The flxed input distribution matrix facilitates a coordinate transformation which deflnes the observer gains and ensures a stable closed-loop reduced order sliding motion. The ‘output error injection signals’ from the observer are used for reconstruction. For the sensor fault design, augmenting the LPV system with a flltered version of the faulty measurements allows the problem of sensor fault reconstruction design to be posed as an actuator fault reconstruction problem. Simulation tests based on an LPV and high-fldelity model of a large transport aircraft have been used to demonstrate the proposed actuator and sensor FDI scheme. The simulation results show good reconstruction of the faults for both actuator and sensors.