A Novel Image-Based Diagnosis Method Using Improved DCGAN for Rotating Machinery

Area of Science:

• Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Rotating machinery is critical in industrial systems, and its faults can lead to significant damage.
• Accurate fault diagnosis is essential for maintaining system health and preventing failures.

Purpose of the Study:

• To propose a novel image-based diagnosis method for rotating machinery fault classification.
• To enhance feature recognition and self-learning capabilities using improved deep convolutional generative adversarial networks (DCGAN).

Main Methods:

• Vibration signal data from rotating machinery was converted into time-frequency feature 2-D images using continuous wavelet transform.
• An adaptive deep convolution neural network (ADCNN) was integrated with generative adversarial networks (GANs) for improved feature self-learning.
• The proposed DCGAN method was evaluated for image feature classification performance.

Main Results:

• The developed image-based method demonstrated effective feature recognition for rotating machinery.
• The integration of ADCNN and GANs improved the self-learning ability of the diagnostic model.
• The proposed DCGAN approach outperformed other fault diagnosis methods in image feature classification.

Conclusions:

• The novel image-based diagnosis method using improved DCGAN offers superior performance for rotating machinery fault classification.
• This approach enhances the reliability and efficiency of fault diagnosis in industrial systems.
• The study highlights the potential of advanced deep learning techniques for machinery health monitoring.