A Unified Fault Diagnosis Approach Utilizing Filtering and Adaptive Approximation for Process and Sensor Faults in a Class of Continuous-Time Nonlinear Systems

摘要

This paper develops an integrated filtering and\udadaptive approximation-based approach for fault diagnosis of\udprocess and sensor faults in a class of continuous-time nonlinear\udsystems with modeling uncertainties and measurement noise. The\udproposed approach integrates learning with filtering techniques\udto derive tight detection thresholds, which is accomplished in two\udways: 1) by learning the modeling uncertainty through adaptive\udapproximation methods and 2) by using filtering for dampening\udmeasurement noise. Upon the detection of a fault, two estimation\udmodels, one for process and the other for sensor faults, are\udinitiated in order to identify the type of fault. Each estimation\udmodel utilizes learning to estimate the potential fault that has\udoccurred, and adaptive isolation thresholds for each estimation\udmodel are designed. The fault type is deduced based on an\udexclusion-based logic, and fault detectability and identification\udconditions are rigorously derived, characterizing quantitatively\udthe class of faults that can be detected and identified by\udthe proposed scheme. Finally, simulation results are used to\uddemonstrate the effectiveness of the proposed approach.