Bearing fault detection by using graph autoencoder and ensemble learning

Meng Wang,Jiong Yu,Hongyong Leng,Xusheng Du,Yiran Liu

DOI: https://doi.org/10.1038/s41598-024-55620-6

IF: 4.6

2024-03-04

Scientific Reports

Abstract:The research and application of bearing fault diagnosis techniques are crucial for enhancing equipment reliability, extending bearing lifespan, and reducing maintenance expenses. Nevertheless, most existing methods encounter challenges in discriminating between signals from machines operating under normal and faulty conditions, leading to unstable detection results. To tackle this issue, the present study proposes a novel approach for bearing fault detection based on graph neural networks and ensemble learning. Our key contribution is a novel stochasticity-based compositional method that transforms Euclidean-structured data into a graph format for processing by graph neural networks, with feature fusion and a newly proposed ensemble learning strategy for outlier detection specifically designed for bearing fault diagnosis. This approach marks a significant advancement in accurately identifying bearing faults, highlighting our study's pivotal role in enhancing diagnostic methodologies.

multidisciplinary sciences

What problem does this paper attempt to address?

The paper primarily addresses the issue of bearing fault detection by proposing a new solution. In industrial automation and production processes, bearing faults are very common problems that can lead to reduced equipment efficiency, increased maintenance costs, and even safety incidents. Existing bearing fault diagnosis techniques often struggle to accurately distinguish between normal operating conditions and fault conditions, resulting in unstable detection outcomes.

To solve the above issues, the paper proposes a novel bearing fault detection method (BFDGE) based on Graph Neural Network (GNN) and Ensemble Learning. Specifically, the main contributions of this method include:

1. **Graph Structure Data Transformation**: A stochastic combination method is used to convert Euclidean-structured data into graph format, enabling processing by Graph Neural Networks.
2. **Feature Fusion**: By combining Graph Neural Networks with Ensemble Learning strategies, feature fusion is achieved to enhance the robustness of the model and the stability of detection results.
3. **Anomaly Detection**: A new ensemble learning strategy is designed specifically for anomaly detection in bearing fault diagnosis.

The study uses publicly available datasets from Case Western Reserve University (CWRU) and Xi'an Jiaotong University (XJTU) to validate the effectiveness of the method. These datasets contain both normal and abnormal signals, allowing the evaluation of the new method's ability to identify complex signal patterns and early faults.

In short, this paper aims to improve the accuracy and reliability of bearing fault detection by introducing techniques from Graph Neural Networks and Ensemble Learning.