Detection and isolation of process faults from actuator faults and sensor faults for a typical MIMO dynamic system

Abstract

For a typical MIMO (Multiple-Input Multiple-Output) nonlinear dynamic system, fault detection and isolation usually aim at process faults with an assumption that actuator faults and sensor faults do not occur at the same time, which is not always the case. This paper uses Extended State Observer for real-time process fault detection and fuzzy inference for fault isolation. It then investigates the coupling relationship among process faults, actuator faults and sensor faults, and presents how a combination of different types of faults could lead to undetected faults or false fault detection and isolation. Finally, a method to isolate actuator faults from process faults is presented. A three-tank MIMO nonlinear system is used to help illustrate the presented fault detection and isolation techniques.