Event-triggered fault detection for discrete-time LPV systems with application to a laboratory tank system

Abstract

This paper investigates a new event-triggered fault detection methodology for discrete-time dynamic systems characterized by linear parameter-varying models. An event-based linear parameter-varying observer is presented as the fault detection module to generate the residual signal. Moreover, two event-triggering conditions are proposed to transmit the sensor and scheduling variable data to the fault detection module only when required. A mixed H−/H∞ formulation of the problem is presented to attenuate the effect of disturbances and the control input on the residual signal, and at the same time to maximize the residual signal sensitivity to fault. A set of linear matrix inequality conditions are obtained to find the solution of the proposed problem based on the concept of the input-to-state stability. The proposed methodology is experimentally tested and validated on a laboratory two-tank system