Fault Diagnosis Method for Imbalanced Data Based on Multi-Signal Fusion and Improved Deep Convolution Generative...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Data Science

Background:

• Accurate fault diagnosis is vital for machine operation.
• Deep learning excels at feature extraction but requires sufficient training data.
• Imbalanced fault data, common in engineering, significantly degrades deep learning model accuracy.

Purpose of the Study:

• To develop an intelligent fault diagnosis method robust to imbalanced datasets.
• To enhance the accuracy of fault diagnosis in mechanical systems.
• To address the challenge of insufficient fault data in practical applications.

Main Methods:

• Wavelet transform for signal processing and feature enhancement.
• Pooling and splicing for data squeezing and fusion.
• Improved adversarial networks for data augmentation and synthetic sample generation.
• An enhanced residual network with a convolutional block attention module for improved diagnosis.

Main Results:

• The proposed method effectively generates high-quality synthetic data samples.
• Demonstrated significant improvements in diagnosis accuracy for both single-class and multi-class imbalanced datasets.
• Validated effectiveness and superiority using two distinct bearing datasets.

Conclusions:

• The developed method offers a promising solution for imbalanced fault diagnosis.
• Successfully enhances machine fault diagnosis accuracy under data scarcity.
• Highlights the potential of advanced deep learning techniques in addressing real-world engineering challenges.