Skip to Main Content
This paper addresses the robust (in the disturbance de-coupling sense) design problem of fault detection filters for uncertain sampled data systems. The only available signal is the output variable. A novel reduced-order unknown-input observer is proposed such that both the estimation error and the state vector vanish asymptotically, robustly with respect to disturbances affecting the continuous-time system. A test-function reveals the occurrence of actuator faults in the plant. An application to a three tank system is displayed.