A Novel Rolling Bearing Fault Diagnosis Method Based on the NEITD-ADTL-JS Algorithm.

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Bearing faults are critical in rotating machinery, leading to equipment failure.
• Traditional fault diagnosis methods struggle with accuracy and robustness, especially with complex signals.
• Transfer learning offers potential but requires effective signal preprocessing and parameter optimization.

Purpose of the Study:

• To develop an innovative bearing fault diagnosis method enhancing transfer learning accuracy.
• To introduce a novel signal preprocessing technique for improved signal quality.
• To optimize fault diagnosis parameters using advanced algorithms for superior performance.

Main Methods:

• A Noise Eliminated Intrinsic Time-Scale Decomposition (NEITD) algorithm for adaptive signal decomposition and feature extraction.
• An improved adaptive deep transfer learning (ADTL) network for enhanced diagnostic capabilities.
• Integration of the Jellyfish Search (JS) algorithm for adaptive optimization of fault diagnosis parameters.

Main Results:

• The NEITD algorithm demonstrated superior signal decomposition performance compared to traditional methods.
• The combined NEITD-ADTL-JS method achieved a 5.29% improvement in fault diagnosis accuracy.
• The proposed method showed enhanced sensitivity and recognition capabilities across various fault types.

Conclusions:

• The innovative NEITD-ADTL-JS method significantly enhances bearing fault diagnosis accuracy and robustness.
• Adaptive signal decomposition and deep learning optimization are crucial for effective fault diagnosis.
• This approach offers a promising solution for reliable condition monitoring in rotating machinery.