Attention‐Guided Coati‐Optimized GAF‐DenseNet169 for Fault Diagnosis in Electric Vehicle Drive Systems
International Journal of Energy Research · 2026-01 · 2026 (1)
DOI: 10.1155/er/1434794 ↗Kurz gefasst
Fachartikel in der Zeitschrift International Journal of Energy Research, online erschienen am 07.08.2026. Verfasst von Emrah Aslan. Die Arbeit wurde bisher 1-mal zitiert (Crossref) und nennt 38 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Reliable fault diagnosis of electric vehicle (EV) drive systems is essential for ensuring operational safety, predictive maintenance, and propulsion system reliability. Induction motors used in EV applications may be exposed to several electrical and loading abnormalities, including overload fault (OLF), overvoltage fault (OVF), undervoltage fault (UVF), phase‐to‐ground fault (PTGF), and phase‐to‐phase fault (PTPF). In this study, an attention‐guided fault‐diagnosis framework named COA‐CBAM‐GAF‐DenseNet169 is proposed by integrating Gramian angular field (GAF) signal imaging, DenseNet169‐based transfer learning, convolutional block attention module (CBAM), and the coati optimization algorithm (COA). First, raw motor signals are separated using a leakage‐free sequence‐level protocol and converted into Gramian angular summation field (GASF) images to preserve temporal dependencies in a two‐dimensional representation. Then, a pretrained DenseNet169 backbone extracts deep feature maps from the generated GAF images. To enhance fault‐related feature discrimination, a CBAM attention block is inserted after the DenseNet169 feature extractor, enabling both channel‐wise and spatial attention refinement. COA is used to optimize the learning rate, dropout rate, dense layer size, batch size, fine‐tuning depth, CBAM reduction ratio, CBAM spatial kernel size, and feature‐weighting coefficients. Experimental results show that the proposed COA‐CBAM‐GAF‐DenseNet169 model achieves 98.30% accuracy, 98.42% precision, 98.30% recall, 98.28% F1‐score, and 99.21% macro‐AUC, outperforming standalone transfer learning models, XGBoost‐based hybridization, attention‐free DenseNet169, and alternative optimization strategies. Ablation analysis, repeated‐run statistical validation, robustness testing, computational burden evaluation, and Grad‐CAM visualization further confirm the contribution and reliability of the proposed framework. The results indicate that the proposed attention‐guided and coati‐optimized model provides an accurate, interpretable, and near‐real‐time capable diagnostic framework for EV induction motor drive systems.
Zitieren
APA
Aslan, E. (2026). Attention‐Guided Coati‐Optimized GAF‐DenseNet169 for Fault Diagnosis in Electric Vehicle Drive Systems. International Journal of Energy Research, 2026(1). https://doi.org/10.1155/er/1434794
Vancouver
Aslan E. Attention‐Guided Coati‐Optimized GAF‐DenseNet169 for Fault Diagnosis in Electric Vehicle Drive Systems. International Journal of Energy Research. 2026;2026(1). doi:10.1155/er/1434794
BibTeX
@article{Aslan2026,
author = {Aslan, Emrah},
title = {Attention‐Guided Coati‐Optimized GAF‐DenseNet169 for Fault Diagnosis in Electric Vehicle Drive Systems},
journal = {International Journal of Energy Research},
year = {2026},
volume = {2026},
number = {1},
doi = {10.1155/er/1434794}
}
EUVIMED