OBD Using Statistical Pattern Recognition 2000-01-3104
A new OBD approach using statistical pattern recognition, applied to the monitoring of the efficiency of catalytic converters, is introduced. Usually, because of the cost of engine electronic control units, automotive On-Board Diagnosis is achieved using mono-dimensional decision techniques. Decision thresholds are then tuned manually. Finally, diagnosis is accomplished under very restrictive operating conditions. The paper shows how statistical pattern recognition enables multi-dimensional data treatment by using the system current operating point in order to make a final decision. In the meantime, automatic calibration is possible through pattern recognition systems learning skills.