Open Access

ARTICLE

Effects of Spark Energy on Spark Plug Fault Recognition in a Spark Ignition Engine

by A. A. Azrin^1,*, I. M. Yusri^1,2, M. H. Mat Yasin³, A. Zainal⁴

1 Faculty of Mechanical and Automotive Engineering Technology, Universiti Malaysia Pahang, Pekan, 26600, Malaysia
2 Centre for Automotive Engineering, Universiti Malaysia Pahang, Pekan, 26600, Malaysia
3 Automotive Technology Center (ATeC), Politeknik Sultan Mizan Zainal Abidin, Dungun, 23000, Malaysia
4 Automotive Engineering Department Education, Engineering Faculty, Universitas Negeri Yogyakarta, Yogyakarta, 55281, Indonesia

* Corresponding Author: A. A. Azrin. Email:

Energy Engineering 2022, 119(1), 189-199. https://doi.org/10.32604/EE.2022.017843

Received 10 June 2021; Accepted 09 August 2021; Issue published 22 November 2021

Abstract

The increasing demands for fuel economy and emission reduction have led to the development of lean/diluted combustion strategies for modern Spark Ignition (SI) engines. The new generation of SI engines requires higher spark energy and a longer discharge duration to improve efficiency and reduce the backpressure. However, the increased spark energy gives negative impacts on the ignition system which results in deterioration of the spark plug. Therefore, a numerical model was used to estimate the spark energy of the ignition system based on the breakdown voltage. The trend of spark energy is then recognized by implementing the classification method. Significant features were identified from the Information Gain (IG) scoring of the statistical analysis. -Nearest Neighbor (KNN), Artificial Neural Network (ANN), and Support Vector Machine (SVM) models were studied to identify the best classifier for the classification stage. For all classifiers, the entire featured dataset was randomly divided into standardized parameter values of training and testing data sets with the ratio of 70–30 for each class. It was shown in the study that the KNN classifier acquired the highest Classification Accuracy (CA) of 94.1% compared to ANN and SVM that score 77.3% and 87.9% on the test data, respectively.

---

Keywords

---