Scholarlink

Purpose <p>This research introduces a One-Dimensional Convolutional Neural Network (1D-CNN) model aimed at predicting ten different health conditions of Drive-End (DE) bearings in a 1-hp induction motor equipped with a 6202 2RS wheel bearing. The objective is to enhance fault diagnosis accuracy and provide a robust solution for predictive maintenance in rotating machinery, specifically addressing mechanical bearing failures, which are responsible for a significant portion of motor faults.</p> Methods <p>Acoustic data from the induction motor, captured through an R15a physical sound sensor, was processed and fed into the 1D-CNN model. This model was designed to automatically extract relevant features from raw acoustic signals, and a fully connected neural network was employed to classify the different bearing fault conditions. The model's performance was benchmarked against the FaultNet model using both the Case Western Reserve University (CWRU) dataset and a proprietary acoustic dataset. Key evaluation metrics included classification accuracy and robustness across both datasets.</p> Results <p>The 1D-CNN model demonstrated high accuracy in categorizing all ten DE bearing health conditions across both datasets. The results indicated superior performance compared to FaultNet, highlighting the model's ability to predict bearing failures effectively. These findings emphasize the potential of the proposed model for real-time fault diagnosis in industrial applications.</p> Conclusions <p>The new developed 1D-CNN architecture offers a promising solution for predictive maintenance of rotating machines, enabling timely detection of bearing faults and reducing the risk of unforeseen breakdowns. The successful application of this model in both laboratory and real-world datasets supports its effectiveness and relevance within the framework of Industry 4.0.</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

Predictive Health Monitoring of Induction Motors Using 1D Convolutional Neural Network

  • 期刊论文

  • 发表日期：

    2025年1月9日

查看全文
  • D. Suganya,
  • R. Rajavel,
  • A. K. Lakshminarayanan

摘要

Purpose

This research introduces a One-Dimensional Convolutional Neural Network (1D-CNN) model aimed at predicting ten different health conditions of Drive-End (DE) bearings in a 1-hp induction motor equipped with a 6202 2RS wheel bearing. The objective is to enhance fault diagnosis accuracy and provide a robust solution for predictive maintenance in rotating machinery, specifically addressing mechanical bearing failures, which are responsible for a significant portion of motor faults.

Methods

Acoustic data from the induction motor, captured through an R15a physical sound sensor, was processed and fed into the 1D-CNN model. This model was designed to automatically extract relevant features from raw acoustic signals, and a fully connected neural network was employed to classify the different bearing fault conditions. The model's performance was benchmarked against the FaultNet model using both the Case Western Reserve University (CWRU) dataset and a proprietary acoustic dataset. Key evaluation metrics included classification accuracy and robustness across both datasets.

Results

The 1D-CNN model demonstrated high accuracy in categorizing all ten DE bearing health conditions across both datasets. The results indicated superior performance compared to FaultNet, highlighting the model's ability to predict bearing failures effectively. These findings emphasize the potential of the proposed model for real-time fault diagnosis in industrial applications.

Conclusions

The new developed 1D-CNN architecture offers a promising solution for predictive maintenance of rotating machines, enabling timely detection of bearing faults and reducing the risk of unforeseen breakdowns. The successful application of this model in both laboratory and real-world datasets supports its effectiveness and relevance within the framework of Industry 4.0.