Exclusion tendency-based observer design framework for active fault diagnosis

This paper proposes a new closed-loop observer-based active fault diagnosis (AFD) framework using a bank of set-valued observers (SVOs). Each SVO is designed for one healthy or faulty system mode to obtain the corresponding output estimation set (OES). The principle of fault diagnosis is to check whether the considered OES includes the system output. Compared with the existing works that mainly make use of SVOs to obtain accurate estimation results, our proposed method optimizes the observer gains to deform OESs such that the exclusion tendency of the system output in each OES is maximized. With this idea, inconsistent system modes can be removed rapidly, and faults can be diagnosed. The design logic is formulated as a bi-level max-min problem, which is transformed into a linear problem by Lagrangian duality and solved efficiently. At the end of this paper, a quadruple-tank example is used to compare the results of our proposed method with an existing fault diagnosis method to illustrate the effectiveness of the proposed method. (c) 2023 Elsevier Ltd. All rights reserved.