Fault identification for rolling bearing based on ITD-ILBP-Hankel matrix

Area of Science:

• Mechanical Engineering
• Signal Processing
• Fault Diagnosis

Background:

• Bearing failures exhibit non-stationarity and nonlinearity, complicating fault feature extraction.
• Traditional 1D local binary pattern (1D-LBP) is sensitive to noise, leading to inaccurate quantization and diagnosis.
• Existing methods struggle to effectively isolate and analyze fault signatures in complex vibration signals.

Purpose of the Study:

• To enhance the accuracy of bearing fault diagnosis by addressing noise interference and imprecise quantization.
• To propose a novel combined method integrating improved 1D-LBP with intrinsic time-scale decomposition (ITD) and Hankel matrix analysis.
• To develop a robust signal processing strategy for reliable identification of bearing fault features.

Main Methods:

• A signal preprocessing strategy involving first-order difference and ITD decomposition to obtain proper rotation components (PRCs).
• An improved 1D-LBP quantization criterion using root mean square (RMS) as the threshold to mitigate noise and extreme value influence.
• Hankel matrix construction and signal reconstruction in a low-dimension space to reduce interference and extract hidden features.

Main Results:

• The proposed ITD-ILBP-Hankel method effectively reduces noise and improves the accuracy of 1D-LBP quantization.
• Component-wise signal processing and reconstruction successfully highlight bearing fault features.
• Fault feature frequencies are accurately extracted, enabling reliable fault type judgment.

Conclusions:

• The developed ITD-ILBP-Hankel method offers a significant improvement for bearing fault diagnosis compared to classical techniques.
• The novel quantization criterion and signal processing strategy enhance the robustness and effectiveness of fault feature extraction.
• This approach provides a reliable tool for diagnosing bearing failures in complex industrial environments.