Detection of Engine Misfire Events using an Artificial Neural Network

New environmental legislation places increasing demands on automobile emission controls, requiring new approaches to engine management and diagnostics systems. This paper demonstrates the use of an Artificial Neural Network (ANN) solution for misfire detection in spark ignition engines. The solution is based on a truly parallel hardware implementation of an ANN. The network is developed by a data-driven training process, using data with known incidences of misfires. No analytical or algorithmic methods need to be developed in order to train or use the ANN for misfire detection. There is minimal processing overhead on the main processor of the engine management unit, freeing resources for alternative engine management tasks or enabling the use of less costly processor solutions.