Mechanical equipment fault diagnosis method based on improved deep residual shrinkage network

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Bearing faults are critical in mechanical equipment, and their diagnosis is vital for safety and efficiency.
• Environmental noise often corrupts bearing vibration signals, hindering accurate fault diagnosis.
• Traditional methods struggle with noise, leading to reduced diagnostic accuracy and reliability.

Purpose of the Study:

• To develop a robust fault diagnosis method with strong noise resistance for mechanical equipment.
• To enhance the performance of deep residual shrinkage networks for bearing fault detection.
• To improve the accuracy and generalization capability of fault diagnosis systems.

Main Methods:

• Proposed a novel fault diagnosis method, CCSDRSN, enhancing traditional deep residual shrinkage networks.
• Incorporated an improved soft threshold function to maximize useful fault signal extraction.
• Developed a new activation function for multi-dimensional nonlinear transformation of time-frequency maps.
• Utilized depthwise separable convolutions for network optimization and parameter reduction.

Main Results:

• The CCSDRSN method demonstrated strong noise resistance in high-noise environments.
• Achieved high diagnostic accuracy for bearing faults.
• Validated effectiveness using Case Western Reserve University datasets.
• Showcased good generalization performance across different load conditions.

Conclusions:

• The proposed CCSDRSN method offers a significant advancement in fault diagnosis for mechanical equipment.
• The method's noise resistance and accuracy make it suitable for real-world applications with noisy vibration data.
• The architectural innovations contribute to an efficient and effective diagnostic tool.