A novel deep learning framework for rolling bearing fault diagnosis enhancement using VAE-augmented CNN model

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Rolling bearings are critical components in industrial machinery, and their operational integrity directly impacts overall equipment performance.
• Effective fault diagnosis is essential for continuous monitoring and maintenance of rolling bearings in industrial settings.
• Traditional fault detection methods struggle with noisy, complex vibration data, leading to reduced accuracy and reliability.

Purpose of the Study:

• To develop an innovative fault diagnosis methodology for rolling bearings using deep learning.
• To enhance the accuracy and robustness of fault detection in mechanical equipment by addressing challenges in vibration signal analysis.
• To precisely identify and classify rolling bearing faults by extracting detailed vibration signal features.

Main Methods:

• A deep learning approach combining Variational Autoencoders (VAEs) for noise robustness and Convolutional Neural Networks (CNNs) for feature expressiveness.
• Utilizing the reparameterization trick for unsupervised learning of latent features.
• Incorporating adaptive threshold methods, the "3/5" strategy, and Dropout for model optimization.

Main Results:

• The VAE-CNN model achieved over 90% diagnosis accuracy for various fault types across different rotational speeds.
• Experimental validation demonstrated superior performance compared to traditional deep neural network models without VAE augmentation.
• The model showed significant improvements in accuracy and robustness for rolling bearing fault diagnosis.

Conclusions:

• The proposed VAE-CNN model offers a powerful and reliable solution for rolling bearing fault diagnosis in industrial environments.
• Deep learning, particularly the VAE-CNN architecture, effectively addresses the complexities of vibration signal analysis for machinery health monitoring.
• This methodology significantly enhances the precision and dependability of fault detection, contributing to improved industrial equipment maintenance.