An Ensemble Convolutional Neural Networks for Bearing Fault Diagnosis Using Multi-Sensor Data

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Multi-sensor data fusion is crucial for accurate fault diagnosis in rotating machinery.
• Existing fusion methods often overlook information loss, impacting diagnostic performance.
• Complex operating conditions exacerbate challenges in machinery fault detection.

Purpose of the Study:

• To propose an ensemble convolutional neural network (CNN) model to address information loss in multi-sensor data fusion for bearing fault diagnosis.
• To enhance the accuracy, robustness, and generalization of fault diagnosis systems.
• To collect comprehensive fault information by integrating multi-sensor and single-sensor data features.

Main Methods:

• Developed an ensemble CNN model with three branches: one multi-channel fusion CNN and two 1-D CNNs.
• The multi-channel branch extracts coupling features from multi-sensor data.
• The 1-D branches extract inherent features from single-sensor data, minimizing information loss.
• Employed a support vector machine ensemble strategy to fuse results from individual branches.

Main Results:

• The proposed ensemble CNN model effectively reduces information loss during data fusion.
• The model demonstrates superior accuracy and robustness in bearing fault diagnosis compared to other methods.
• The integration of multi-channel and single-channel feature extraction captures comprehensive fault information.

Conclusions:

• The proposed ensemble CNN model offers an effective solution for bearing fault diagnosis by mitigating information loss in multi-sensor fusion.
• The approach enhances diagnostic performance, providing more reliable results for rotating machinery.
• The study highlights the importance of addressing information loss for robust fault detection systems.