Fault estimation observer design for descriptor-LPV systems with unmeasurable gain scheduling functions

Abstract

This paper addresses a design of a sensor Fault Estimation Observer (FEO) for Descriptor-Linear Parameter Varying (DLPV) systems. In contrast with conventional problem, the Gain Scheduling Functions (GSF) are considered to be unmeasurable. In order to decouple faults, an augmented system is constructed by considering faults as an auxiliary state vector. Then, a FEO is designed to estimate simultaneously the original system states and faults. The influence of the error provided by the unmeasurable GSF is minimized using the H∞ theory, and the stability is guaranteed based on a Lyapunov equation. Sufficient conditions for the existence of the FEO are set in terms of Linear Matrix Inequalities. The resulting formulation provides a FEO robust against the unmeasurable scheduling gains. The effectiveness of the FEO is illustrated by means of a numerical simulation example.