Title: Misfire detection in I.C. engine through ARMA features using machine learning approach
Authors: Sumedh Mulay; V. Sugumaran; S. Babu Devasenapati
Addresses: School of Mechanical and Building Sciences, Vellore Institute of Technology, Chennai Campus, Vandalur-Kelambakkam Road, Chennai 600127, India ' School of Mechanical and Building Sciences, Vellore Institute of Technology, Chennai Campus, Vandalur-Kelambakkam Road, Chennai 600127, India ' Sri Guru Institute of Technology, Coimbatore 641110, India
Abstract: One of the prime problems engines are facing is Misfire, as it leads to the power loss along with the exhaust gas containing air pollutants like CO and NOx. This paper proposes a predictive model for misfire detection using machine learning approach. For the present study, vibration signals acquired using the piezoelectric accelerometer were taken into consideration as a pattern of a misfire for each cylinder is specific in nature. Then, ARMA features were extracted from acquired vibration signals followed by Feature selection using J48 decision tree algorithm. For feature classification, the functional tree classifier was used. In this study, the classification accuracy of 92.2% was achieved. The proposed model was tested on the engine test rig wherein every cylinder misfire tests were conducted. This work can be improved by using different classifier algorithms for more accurate misfire detection.
Keywords: machine learning approach; ARMA features; feature extraction; feature selection; functional tree classifier.
Progress in Industrial Ecology, An International Journal, 2018 Vol.12 No.1/2, pp.93 - 111
Received: 17 Oct 2017
Accepted: 02 May 2018
Published online: 25 Oct 2018 *