Bearing-Fault Diagnosis with Signal-to-RGB Image Mapping and Multichannel Multiscale Convolutional Neural Network.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Traditional bearing fault diagnosis methods struggle with complex conditions and limited data.
• Existing deep learning models like CNNs have limitations in feature extraction and handling small datasets.

Purpose of the Study:

• To propose a novel intelligent bearing fault diagnosis method combining signal-to-RGB image mapping (STRIM) and multichannel multiscale CNN (MCMS-CNN).
• To address the limitations of traditional and existing deep learning approaches in bearing fault diagnosis.

Main Methods:

• Signals from multiple sensors are converted into RGB images using the STRIM method for feature fusion.
• A multichannel multiscale CNN (MCMS-CNN) is developed to extract multi-scale features from RGB images.
• Network architecture is optimized (wider, shallower) to prevent overfitting on small datasets.

Main Results:

• The proposed STRIM and MCMS-CNN method effectively fuses multi-sensor data.
• The MCMS-CNN architecture successfully learns complementary and abundant features at different scales.
• The method demonstrates superior fault classification performance compared to other deep learning approaches on the CWRU dataset.

Conclusions:

• The novel STRIM and MCMS-CNN approach offers an effective solution for intelligent bearing fault diagnosis.
• This method overcomes limitations of traditional techniques and existing deep learning models, particularly with limited fault samples.
• The proposed technique provides a robust and high-performing solution for bearing fault diagnosis in industrial settings.