Rolling Bearing Fault Diagnosis Using Multi-Sensor Data Fusion Based on 1D-CNN Model.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Rolling bearings are critical components in rotating machinery.
• Effective end-to-end fault diagnosis is essential for operational reliability and safety.
• Existing methods may face challenges in handling complex multi-sensor data for fault detection.

Purpose of the Study:

• To propose a novel hybrid model for end-to-end fault diagnosis of rolling bearings.
• To enhance the accuracy and generalization ability of fault diagnosis systems.
• To effectively integrate multi-sensor data for improved diagnostic performance.

Main Methods:

• Utilizing the Bald Eagle Search (BAS) algorithm to optimize Sparse Wavelet Decomposition (SWD) parameters, creating BAS-SWD.
• Applying BAS-SWD for signal preprocessing and extraction of sensitive orthogonal components (OCs) with high spectrum kurtosis.
• Developing an improved 1D-Convolutional Neural Network (1D-CNN) model, incorporating VGG-16 architecture, for feature extraction and fusion.

Main Results:

• The BAS-SWD effectively preprocesses raw sensor signals and extracts relevant features.
• The hybrid 1D-CNN model successfully fuses features from decomposed multi-sensor data.
• Comparative experiments demonstrate the proposed model's high accuracy, effectiveness, and good generalization ability across different datasets.

Conclusions:

• The proposed hybrid model, integrating BAS-SWD and 1D-CNN with multi-sensor data fusion, provides a robust solution for rolling bearing fault diagnosis.
• The method achieves superior performance compared to existing approaches.
• This approach offers a promising direction for intelligent fault diagnosis in industrial applications.