Scholarlink

<p>Learning better feature representations instinctively through Convolutional Neural Networks (CNN) has inspired solutions to the unresolved issues in stator current-based multi-class fault diagnosis of induction motor drives. The current envelope of stator current, acquired using the Hilbert transform, has proven to be an effective pre-processing method for handling the complex data patterns of motor current and revealing the masked defect information. The self-synthesized quality features through deep convolutional layers outperform traditional methods, achieving unmatched accuracy compared to feature engineering schemes. The feature engineering method is also developed using DHT-DWT-based feature extraction, with a novel approach to selecting the appropriate mother wavelet. The most noffachievement of this research is addressing the unique advantages of hybridizing the signal processing technique with the CNN model, where the enrichment in feature quality is achieved by unveiling the buried fault information close to the dominating supply frequency. The proposed method is reliable for analyzing multi-class motor fault detection and has a strong generalization approach. The compact design of the hybrid CNN-envelope approach, which works with very low-resolution stator current sampled at 1.28&#xa0;kHz, reduces computational intricacies significantly, making it a suitable candidate for real-time applications.</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

Signal processing and machine learning techniques based hybrid approaches for decent fault classification of induction motor

  • 期刊论文

  • 发表日期：

    2026年3月18日

查看全文
  • Parth Sarathi Panigrahy,
  • Paramita Chattopadhyay

摘要

Learning better feature representations instinctively through Convolutional Neural Networks (CNN) has inspired solutions to the unresolved issues in stator current-based multi-class fault diagnosis of induction motor drives. The current envelope of stator current, acquired using the Hilbert transform, has proven to be an effective pre-processing method for handling the complex data patterns of motor current and revealing the masked defect information. The self-synthesized quality features through deep convolutional layers outperform traditional methods, achieving unmatched accuracy compared to feature engineering schemes. The feature engineering method is also developed using DHT-DWT-based feature extraction, with a novel approach to selecting the appropriate mother wavelet. The most noffachievement of this research is addressing the unique advantages of hybridizing the signal processing technique with the CNN model, where the enrichment in feature quality is achieved by unveiling the buried fault information close to the dominating supply frequency. The proposed method is reliable for analyzing multi-class motor fault detection and has a strong generalization approach. The compact design of the hybrid CNN-envelope approach, which works with very low-resolution stator current sampled at 1.28 kHz, reduces computational intricacies significantly, making it a suitable candidate for real-time applications.