Statistics Based Detection and Isolation of UEGO Sensor Faults

Stringent emission regulations mandated by California air regulation board (CARB) require monitoring the upstream exhaust gas oxygen (UEGO) sensor for any possible malfunction causing the vehicle emissions to exceed certain thresholds. Six faults have been identified to potentially cause the UEGO sensor performance to deteriorate and potentially lead to instability of the air-fuel ratio (AFR) control loop. These malfunctions are either due to an additional delay or an additional lag in the transition of the sensor response from lean to rich or rich to lean. Current technology detects the faults the same way (approximated by a delay type fault) and does not distinguish between the different faults. In the current paper, a statistics based approach is developed to diagnose these faults. Specifically, the characteristics of a non-normal distribution function are estimated based on the UEGO sensor output and used to detect and isolate the faults. When symmetric operation is detected, a system identification process is employed to estimate the parameters of the dynamic system and determine the type of operation. The proposed algorithm has been demonstrated on real data obtained from both Ford F150 and Mustang V6 vehicles.© 2013 ASME