Detection and Isolation of Multiple Consecutive Faults in Nonlinear Uncertain Systems

This paper presents a robust dynamic fault detection and isolation scheme for nonlinear uncertain systems subject to multiple consecutive actuator and process faults. A model-based monitoring framework consisting of multiple submodules is developed, with the number of submodules dynamically adjusted based on the number of detected faults. Each newly designed submodule is activated to: (i) approximate the cumulative sum of all detected faults, (ii) isolate the last detected fault, either partially or fully, and (iii) detect the next fault occurrence. A key feature of the proposed approach is the utilization of the online learning-based approximation of the cumulative sum of previously detected faults to enable detection and isolation of subsequent consecutive faults in a computationally efficient manner. The boundedness of the system variables and the robustness of the proposed scheme are analytically shown. Finally, a numerical simulation validates the presented approach and demonstrates its effectiveness and applicability.