Multiple model-based fault diagnosis using unknown input observers

Abstract

In this paper we have solved a problem that how to develop a method to establish a UIO to diagnose a liner system contains modeling uncertainty, actuators fault and external interference, and then built multiple models based on the UIO to diagnose the probably faults on the actuator. We use the technic of reduced-order UIO to separate the original system into two different subsystems. It not only reduce the amount of calculation but also robust for modeling uncertainty and the external interference through this method. After that we consider the case of multiple models, using the UIO to construct the different fault case respectively, and then the switch function will be used to select the optimal one from these models. Through that we should find the fault described by the relevant fault model since the switch function close to zero. Finally a numerical example was presented to illustration that the method with UIO can diagnose the fault effectively.