Integrated Gradient-Based Continuous Wavelet Transform for Bearing Fault Diagnosis.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence

Background:

• Bearing fault diagnosis is critical for rotating machinery safety and efficiency.
• Deep learning (DL) models show promise for bearing fault diagnosis, often using Continuous Wavelet Transform (CWT) for data preprocessing.
• Traditional CWT generates time-frequency images where identifying important frequency components for DL models remains a challenge.

Purpose of the Study:

• To propose an Integrated Gradient-based continuous wavelet transform (IG-CWT) method for optimizing frequency component selection in CWT.
• To enhance the accuracy of deep learning-based bearing fault diagnosis by improving data preprocessing.

Main Methods:

• Developed the IG-CWT method to detect important frequency components and their ranges from sensor data.
• Applied IG-CWT for data preprocessing in bearing fault diagnosis tasks.
• Validated the IG-CWT method using four benchmark bearing datasets and three distinct deep learning models.

Main Results:

• The IG-CWT method effectively identifies critical frequency components for DL models.
• Experiments demonstrated that IG-CWT preprocessing leads to higher fault diagnosis accuracy compared to standard CWT.
• The proposed method showed superior performance across multiple datasets and DL architectures.

Conclusions:

• The IG-CWT method offers a significant advancement in preprocessing vibration data for bearing fault diagnosis.
• Optimizing frequency component selection through IG-CWT enhances the effectiveness of deep learning models.
• This technique contributes to more reliable and accurate machinery health monitoring.