Segmentation of Online Ferrograph Images with Strong Interference Based on Uniform Discrete Curvelet Transformation.

Area of Science:

• Tribology
• Mechanical Engineering
• Image Processing

Background:

• Online visual ferrography (OLVF) monitors equipment wear by analyzing lube oil debris.
• Bubbles in lube oil create interference shadows, leading to misidentification of wear debris by traditional algorithms.
• This misidentification results in significant errors in wear debris characteristic extraction.

Purpose of the Study:

• To propose a jam-proof method for binarizing wear debris images obtained from OLVF.
• To enhance the accuracy of wear debris analysis in the presence of bubble interference.
• To improve the reliability of online equipment wear monitoring.

Main Methods:

• A uniform discrete curvelet transformation (UDCT)-based method was developed for image binarization.
• Multiscale analysis of OLVF ferrograms was performed using UDCT.
• Nonlinear transformation of UDCT coefficients was applied for low-frequency suppression and high-frequency denoising.
• The Otsu algorithm was employed for final image binarization under interference.

Main Results:

• The proposed UDCT-based method effectively suppresses interference from bubbles in OLVF ferrograms.
• The technique achieves accurate binarization of wear debris images even with strong interference.
• The method significantly reduces errors in wear debris characteristic extraction.

Conclusions:

• The jam-proof UDCT-based binarization method enhances the accuracy and reliability of online wear debris analysis.
• This approach offers a robust solution for overcoming bubble interference in OLVF systems.
• The improved monitoring capability contributes to more effective equipment maintenance and prediction.