Efficient Gearbox Fault Diagnosis Based on Improved Multi-Scale CNN with Lightweight Convolutional Attention.

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Gearbox operation is critical for industrial machinery reliability and lifespan.
• Accurate fault diagnosis is essential for preventing catastrophic failures and ensuring operational continuity.

Purpose of the Study:

• To develop an intelligent fault diagnosis framework for gearboxes.
• To enhance diagnostic accuracy and robustness, especially under complex working conditions.

Main Methods:

• Proposed an intelligent diagnosis framework combining Empirical Mode Decomposition and multimodal feature co-optimization.
• Developed a fault diagnosis model fusing a multi-scale convolutional neural network (MS-CNN) and a lightweight convolutional attention (LCA) model.
• Utilized dynamic convolutional kernel generation and grouped convolution for efficient local-global feature modeling.

Main Results:

• The proposed model achieved diagnostic accuracy exceeding 98.9%.
• Demonstrated superior performance compared to existing methods under complex conditions like variable speeds and strong noise.
• The lightweight convolutional attention mechanism reduced computational complexity while maintaining high accuracy.

Conclusions:

• The developed intelligent diagnosis framework offers highly efficient and accurate gearbox fault detection.
• The model's robustness makes it suitable for real-world industrial applications with challenging operating environments.
• The fusion of MS-CNN and LCA presents a promising approach for advanced machinery health monitoring.