Ching-Ting Pi

The Observer-based Approach of Model-based FDI

Abstract

In recent years, fault detection and isolation (FDI) problem has been receiving a great deal of attention in a wide variety of industries. In this thesis, we focus on the observer-based approach of model-based FDI. The methodology is based upon use of an observer, known as Proportional Integral (PI) Observer, which will estimate the states of the system as well as the instrument fault vectors in case there are some actuator and/or sensor faults within the system. In addition, it is possible to identify the shape and magnitude of the fault with this approach. Unknown input observer (UIO) is another estimator which can achieve the same goal under restrictive assumptions. With proportional integral observer (PIO), we can estimate the fault directly without the limitations imposed by UIO. A systematic design procedure is developed for PIO which allows for fast and accurate fault detection while accounting for the structural uncertainties and parameter variation in the system's dynamical model. The reduced-order PIO will also be introduced in this thesis. The overall performance of PIO is verified through simulation on VTOL Helicopter example, DC-DC converter example, and an IC engine example.

Advisor: Bahram Shafai