Skip navigation
Please use this identifier to cite or link to this item: http://repository.iitr.ac.in/handle/123456789/18085
Title: Support vector machine classifier for engine misfire detection using exhaust sound quality
Authors: Singh S.
Potala S.
Mohanty A.R.
Gibbs B.
Published in: Proceedings of 24th International Congress on Sound and Vibration, ICSV 2017
Abstract: This paper proposes a novel non-contact-based technique to detect engine misfiring us-ing sound quality metrics of the sound waves measured near the exhaust to train a sup-port vector machine (SVM) classifier. This method was tested on a four-stroke, four-cylinder SI engine run on a wide range of load torques, 20 to 50 Nm, and wide range of speeds, 1260 to 3340 rpm, where at every test condition a cylinder was misfired intermittently. 52 sound signals were measured near the exhaust, containing 26 pairs of no misfiring con-dition and its corresponding one cylinder misfiring condition. The key sound quality metrics namely, Zwicker Loudness, Roughness and Fluctuation Strength of the exhaust sounds were used to train and test an SVM classifier. The algorithm correctly classified misfiring signals and correct signals with 95.2% training accuracy, 90% test accuracy, and 0.01 s compu-tation time. Thus, exhaust sound quality metrics can successfully predict misfiring of an SI engine using SVM. The proposed technique could be advantageous over existing misfire detection techniques, as this method does not require an in-cylinder, engine-attached or exhaust-attached measurement, thus eliminating the need for costly, high maintenance and less durable sensors. Further, the presented method is computationally faster and robust over wider torque and speed range than most existing techniques.
Citation: Proceedings of 24th International Congress on Sound and Vibration, ICSV 2017, (2017)
URI: http://repository.iitr.ac.in/handle/123456789/18085
Issue Date: 2017
Publisher: International Institute of Acoustics and Vibration, IIAV
Keywords: Cylinder
Fluctuation Strength
Load
Loudness
Misfire
Roughness
Speed
Author Scopus IDs: 56215217100
57195637368
7102044449
Author Affiliations: Singh, S., Indian Institute of Technology Kharagpur, Department of Mechanical Engineering, Kharagpur, West Bengal, India
Potala, S., Indian Institute of Technology Kharagpur, Department of Mechanical Engineering, Kharagpur, West Bengal, India
Mohanty, A.R., Indian Institute of Technology Kharagpur, Department of Mechanical Engineering, Kharagpur, West Bengal, India
Appears in Collections:Conference Publications [ME]

Files in This Item:
There are no files associated with this item.
Show full item record


Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.