Scholarlink

Purpose <p>Accurate fault detection in mechanical components is essential to minimize equipment replacement costs, avoid production downtime, and reduce the risk of human injury. However, traditional fault detection methods often fall short in achieving reliable and precise identification of motor faults in rotating machines. This study aims to develop an intelligent classification model for the effective detection of roller bearing faults.</p> Methods <p>The proposed approach integrates Ensemble Empirical Mode Decomposition (EEMD), Principal Component Analysis (PCA), and a Bi-directional Long Short-Term Memory (Bi-LSTM) neural network. EEMD is first applied to decompose noisy vibration signals into Intrinsic Mode Functions (IMFs), and relevant IMFs are selected based on their correlation with the original signal. PCA is then employed for dimensionality reduction, retaining only essential features from five different fault types. Finally, a Bi-LSTM model is trained on the reduced features to learn dynamic patterns for fault classification.</p> Results <p>The proposed model effectively distinguishes between five types of roller bearing faults and demonstrates superior performance compared to models trained on raw features and several state-of-the-art methods. The evaluation based on Precision, Recall, F1 Score, and Accuracy confirms that the ensemble approach enhances classification accuracy and robustness.</p> Conclusion <p>Combining EEMD, PCA, and Bi-LSTM provides an efficient and reliable framework for roller bearing fault classification. The intelligent system outperforms conventional models in accuracy and adaptability, offering a promising solution for real-time fault diagnosis in rotating machinery.</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

Enhanced Fault Diagnosis of Rolling Bearings with Noise Filtering and Neural Networks

  • 期刊论文

  • 发表日期：

    2025年7月5日

查看全文
  • Sujit Kumar,
  • Bam Bahadur Sinha

摘要

Purpose

Accurate fault detection in mechanical components is essential to minimize equipment replacement costs, avoid production downtime, and reduce the risk of human injury. However, traditional fault detection methods often fall short in achieving reliable and precise identification of motor faults in rotating machines. This study aims to develop an intelligent classification model for the effective detection of roller bearing faults.

Methods

The proposed approach integrates Ensemble Empirical Mode Decomposition (EEMD), Principal Component Analysis (PCA), and a Bi-directional Long Short-Term Memory (Bi-LSTM) neural network. EEMD is first applied to decompose noisy vibration signals into Intrinsic Mode Functions (IMFs), and relevant IMFs are selected based on their correlation with the original signal. PCA is then employed for dimensionality reduction, retaining only essential features from five different fault types. Finally, a Bi-LSTM model is trained on the reduced features to learn dynamic patterns for fault classification.

Results

The proposed model effectively distinguishes between five types of roller bearing faults and demonstrates superior performance compared to models trained on raw features and several state-of-the-art methods. The evaluation based on Precision, Recall, F1 Score, and Accuracy confirms that the ensemble approach enhances classification accuracy and robustness.

Conclusion

Combining EEMD, PCA, and Bi-LSTM provides an efficient and reliable framework for roller bearing fault classification. The intelligent system outperforms conventional models in accuracy and adaptability, offering a promising solution for real-time fault diagnosis in rotating machinery.