Scholarlink

<p>Brushless DC (BLDC) motors are widely used in industrial and automotive systems due to their high efficiency, low maintenance requirements, and compact structure. However, they are vulnerable to various electrical and mechanical faults, such as bearing wear and rotor imbalance, which can lead to unexpected downtimes. To address this issue, this study proposes ETNeXt, a lightweight, self-organizing fault detection framework based on acoustic signal analysis. The method applies a 7-level Multilevel Discrete Wavelet Transform (MDWT) with the ‘sym4’ wavelet to extract frequency-domain features, followed by triadic histogram feature generation using signum, upper ternary, and lower ternary functions. A hybrid feature selection process based on Neighborhood Component Analysis (NCA) and Chi-square (Chi2) methods identifies the most discriminative features. Classification is performed using Fine k-NN and Cubic SVM with tenfold cross-validation. The proposed ETNeXt model achieved up to 100% accuracy with Cubic SVM and 99.80% with kNN on a benchmark dataset, and maintained 99.95% accuracy on a separate test dataset, demonstrating strong generalizability. Compared to deep learning models, ETNeXt offers significantly reduced computational complexity, making it highly suitable for real-time, edge-based deployment thanks to its lightweight design.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

ETNeXt: integrated feature engineering and classification framework for BLDC motor fault detection

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月3日

查看全文
  • Burak Celik,
  • Ezgi Taskin,
  • Ayhan Akbal,
  • Mehmet Ozdemir

摘要

Brushless DC (BLDC) motors are widely used in industrial and automotive systems due to their high efficiency, low maintenance requirements, and compact structure. However, they are vulnerable to various electrical and mechanical faults, such as bearing wear and rotor imbalance, which can lead to unexpected downtimes. To address this issue, this study proposes ETNeXt, a lightweight, self-organizing fault detection framework based on acoustic signal analysis. The method applies a 7-level Multilevel Discrete Wavelet Transform (MDWT) with the ‘sym4’ wavelet to extract frequency-domain features, followed by triadic histogram feature generation using signum, upper ternary, and lower ternary functions. A hybrid feature selection process based on Neighborhood Component Analysis (NCA) and Chi-square (Chi2) methods identifies the most discriminative features. Classification is performed using Fine k-NN and Cubic SVM with tenfold cross-validation. The proposed ETNeXt model achieved up to 100% accuracy with Cubic SVM and 99.80% with kNN on a benchmark dataset, and maintained 99.95% accuracy on a separate test dataset, demonstrating strong generalizability. Compared to deep learning models, ETNeXt offers significantly reduced computational complexity, making it highly suitable for real-time, edge-based deployment thanks to its lightweight design.