Misfire Detection in an IC Engine Using Vibration Signal and Bagging Classifier

Out of all the problems in an automobile industry, misfire in an IC engine remains a difficulty which has received high attention. However, this problem prevails with deteriorating effects like drop in fuel efficiency, rise in power loss and harmful emissions containing substantial amount of unspent hydrocarbons. When misfire occurs, a discernible pattern attributed to a particular cylinder is observed. In the present study, these vibration signals have been used to extract statistical features conducive to detect misfire. Among all the extracted features, those with maximum contribution in determining classification accuracy were selected using J48 decision tree algorithm. Using these selected features with bagging classifier utilizing functional tree algorithm, the misfire and normal conditions were classified appropriately. Then, the viability of using this technique as an onboard system is discussed based on its classification accuracy and confusion matrices.