Fault Detection and Isolation for Linear Dynamical Systems

Fault Detection and Isolation for Linear Dynamical Systems

Monteiro, D.

Oliveira, P. (mentor)
Rosa, P. (mentor)
Silvestre, C. (mentor)

Aerospace Engineering

Control and Operations

Control and Simulation

2015-11-18

Safety and reliability of a dynamical system is a concern that have always pursued designers in both academia and industry. Monitoring the health status of a system is even more relevant for safety critical applications, such as chemical and nuclear plants, medicine, transportation, and security systems. The occurrence of abnormal events on these processes may lead to malfunctions and disasters in ultimate fault conditions, as witnessed in the past. The paramount importance of the topic and the increasing interest in multiple-model approaches under the scope of on-line fault detection and isolation motivates this thesis. Initially, focus is given to classical multiple-model adaptive estimation (MMAE) in which an in-depth study is undertaken for the design of a scheme capable of determining the working regime of a system. This is done by identifying the region where the fault parameters lie under the associated uncertainty domain. The design procedure is built on a performance-based strategy, which ensures a well-defined level of state estimation performance despite the fault location. Due to the high computational complexity of the classical MMAE approach, in what follows we propose a novel bank design based on the combination of Kalman and robust H2 filters. This strategy leads to a substantial reduction on the number of estimators in the bank, while preserving the desired state estimation performance. In both approaches a prominent study on convergence properties is performed, so that robustness of the methods is guaranteed. Computational simulations based on a generic helicopter model are also executed to prove the potential of the strategies developed and provide a verification basis for the theoretical results achieved.

multiple-model adaptive estimation
model-based fault diagnosis
robust H2 filters
state estimation in uncertain systems

http://resolver.tudelft.nl/uuid:4ea07ffc-6749-4ceb-95b8-448f779d87d5

Student theses

master thesis

(c) 2015 Monteiro, D.