A comparison of sliding mode and unknown input observers for fault reconstruction

This paper compares the use of a recently proposed sliding mode fault detection and isolation scheme with a linear scheme based on an unknown input observer. Both methods seek to reconstruct actuator and sensor fault signals for a class of uncertain systems. Although the explicit details of the two approaches appear quite different, an underlying link between them is exposed and investigated. The methods are compared on data collected from a laboratory scale crane rig on which (known) faults were deliberately introduced.