Dynamic online nonlinear robust detection and accommodation of incipient component faults for nonlinear dissipative distributed processes

This paper proposes a new nonlinear detection observer for the detection of component faults and their subsequent accommodation for a class of nonlinear distributed processes. Specifically, the proposed results constitute a significant extension to previous precursory work that utilized a linear observer for both detection and diagnosis of component faults for the same class of nonlinear distributed processes. An added feature of the proposed nonlinear observer under consideration is not simply the inclusion of nonlinear dynamics terms in the detection and diagnostic observers but the inclusion of such additional information in the expression for the dynamic residual threshold. The resulting improved version of a dynamic threshold contributes to the robustness of the detection scheme and minimizes the fault detection time. Two different dynamic threshold approaches are employed within the same nonlinear detection observer leading to improved detection time. A combination of the two thresholds is also considered to arrive at a detection observer that reaps the beneficial attributes of each of the thresholds, thus resulting in significant reduction of the fault detection time. Finally, an adaptive diagnostic observer is proposed that is subsequently utilized in an automated control reconfiguration scheme that accommodates the component faults. Copyright (C) 2011 John Wiley & Sons, Ltd.