OBD of Diesel EGR Using Artificial Neural Networks

To detect malfunctions of the EGR system of a passenger car diesel engine, a neural network approach was selected using Self Organizing Maps (SOM). Self Organizing Maps are self-learning technologies that can be used to retrieve typical data patterns in large data sets. This technology is very efficient for identifying if patterns from a new, modified or changed system are similar to already existing patterns. The SOM outputs a measure of similarity to ‘typical system behavior patterns’. As an OBD function, this value is a measure for system anomaly detection.

Performing dynamic tests using standard driving cycles, not only was the occurrence of a malfunction within the EGR system detected by the neural network, the cause of the malfunction could also be identified.