Design of Built-In Tests for Robust Active Fault Detection and Isolation of Discrete Faults in Uncertain Systems

Abstract

This work deals with the design of health maintenance tests for systems operating under uncertainty. Faults masked by uncertainty or uncertainty interpreted as a system fault often lead to subsequent safety and performance complications instantiated as false alarms, no fault founds, and non-detections. Here, a model-based active fault detection and isolation algorithm is employed in the form of a semi-infinite, worst-case scenario design program. The objective of this program is to maximize the fault detection and isolation capability at the worst-case realization of uncertainty, by manipulating the admissible system inputs. Fault detection and isolation are improved by increasing the distance between sensed outputs of the fault-free and faulty systems. The proposed method is demonstrated with application to the benchmark three-tank system with uncertain pipe flow coefficients, along with a faulty pump actuator and an unknown tank leak. The corresponding optimization problem is solved locally and globally using different worst-case design algorithms.