Robustness in quantitative model-based fault diagnosis

Quantitative model-based fault diagnosis has become a popular issue in safety-critical systems, e.g., aircraft, spacecraft, chemical processes and nuclear plants. The use of dynamical system model information has been widely recognized as an important approach to fault detection and isolation for the case when there are no repeated hardware units. A prerequisite for feasibility of quantitative model-based fault diagnosis is satisfactory robust performance with respect to uncertainties. The paper discusses the different problems in robustness and surveys the state of the art in robust solutions for quantitative model-based fault diagnosis. The state observer with disturbance de-coupling design is recommended as a good solution for robustness in fault diagnosis. Further research topics in robust fault diagnosis are outlined