Misfire Detection Using a Neural Network Based Pattern Recognition Technique

Abstract

This contribution investigates the practical application of artificial neural networks to misfire detection in gasoline engines.The problem of misfire detection is formulated as a pattern recognition problem. A feed-forward multiple-layer neural network is used for the classification of firing and misfiring events. Emphasis is given on the trade-off between performance, computational cost and implementabilitv of the technique on a production electronic control unit (ECL). The developed technique is applied to a six cylinder gasoline engine to detect misfire events over the whole range of operation defined by official on board diagnosis (OBD) regulations. Experimental results on a passenger car are presented.