Real Time Detection Filters for Onboard Diagnosis of Incipient Failures
This paper presents the real time implementation of detection filters for the diagnosis of incipient failures in electronically controlled internal combustion (IC) engines. The detection filters are implemented in a production vehicle. Recent results  have demonstrated the feasibility of a model-based failure detection and isolation (FDI) methodology for detecting partially failed components in electronically controlled vehicle subsystems. The present paper describes the real time application of the FDI concept to the detection of faults in sensors associated with the engine/controller In a detection filter, the performance of the engine/controller system is continuously compared to a prediction based on sensor measurements and an analytical model (typically a control model) of the system. Any discrepancy between actual and predicted performance is analyzed to identify the unique failure signatures related to specific system components.