Explainable Deep Learning for Predictive Maintenance and Energy-Efficient Optimization of Intelligent Electric Vehicle Powertrain Systems
Computer and Decision Making: An International Journal · 16.11.2026 · 4 (1) · 512-522
DOI: 10.1922/comdem.v4i1.19200 ↗Kurz gefasst
Fachartikel in der Zeitschrift Computer and Decision Making: An International Journal, online erschienen am 16.11.2026. Verfasst von Dr. Raghavendra Sheddi und 5 weiteren Autor:innen. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Electric vehicle powertrain systems require reliable condition monitoring, predictive maintenance, and efficient energy management to support safe, sustainable, and intelligent operation. This study evaluated battery, motor, maintenance, operational, and energy-related characteristics using a quantitative secondary-data approach based on 1,000 cleaned observations from the publicly available EVIoT-PredictiveMaint dataset. Descriptive statistics, independent-samples t-tests, chi-square analysis, and Pearson correlation were used to examine statistical patterns. A feed-forward deep neural network was developed for failure prediction, while SHAP analysis was applied to interpret the contribution of individual predictors. A neural-network regression model was also used to estimate power consumption, followed by exploratory multi-objective optimization. The findings showed that motor temperature and remaining useful life differed significantly between failure and non-failure conditions, whereas most other variables showed weak or non-significant relationships. The deep-learning classifier demonstrated limited ability to detect failure cases despite high overall accuracy, indicating the influence of class imbalance and weak predictive patterns. SHAP analysis identified several mechanical, battery, road, and operational variables as relatively influential, although their overall effects were modest. Energy-consumption prediction was also limited, while the Pareto-based analysis identified operating conditions associated with lower energy use and acceptable component health. Overall, the study demonstrates the exploratory value of explainable deep learning and optimization for intelligent EV powertrain analysis, while emphasizing the need for larger, balanced, longitudinal datasets with verified failure events and richer sensor information.
Zitieren
APA
Sheddi, D. R., A, D. M. B., M., M., Vyas, M. K., shaikh, S., & Tak, D. R. (2026). Explainable Deep Learning for Predictive Maintenance and Energy-Efficient Optimization of Intelligent Electric Vehicle Powertrain Systems. Computer and Decision Making: An International Journal, 4(1), 512-522. https://doi.org/10.1922/comdem.v4i1.19200
Vancouver
Sheddi DR, A DMB, M. M, Vyas MK, shaikh S, Tak DR. Explainable Deep Learning for Predictive Maintenance and Energy-Efficient Optimization of Intelligent Electric Vehicle Powertrain Systems. Computer and Decision Making: An International Journal. 2026;4(1):512-522. doi:10.1922/comdem.v4i1.19200
BibTeX
@article{Sheddi2026,
author = {Sheddi, Dr. Raghavendra and A, Dr Manjunatha B and M., Manu and Vyas, Monika K and shaikh, Shakila and Tak, Dr. Ritu},
title = {Explainable Deep Learning for Predictive Maintenance and Energy-Efficient Optimization of Intelligent Electric Vehicle Powertrain Systems},
journal = {Computer and Decision Making: An International Journal},
year = {2026},
volume = {4},
number = {1},
pages = {512--522},
doi = {10.1922/comdem.v4i1.19200}
}
EUVIMED