[Oil atomic spectrometric feature selection by Parzen window based vague sets theory]

Area of Science:

• Tribology and Spectrometry
• Machine Condition Monitoring

Context:

• Oil atomic spectrometry generates vast, ambiguous data, hindering accurate gearbox fault diagnosis.
• Traditional feature selection methods struggle with the complexity of spectrometric data.

Purpose:

• To develop a novel method for selecting fewer, more effective spectrometric features for fault diagnosis.
• To enhance the efficiency and accuracy of gearbox fault diagnosis using oil analysis.

Summary:

• A new method defines Mean Vague Sensibility (MVS) to quantify feature sensitivity to wear states.
• Vague set theory and Bayesian formulas are used to calculate membership degrees, reducing reliance on human experience.
• The approach effectively identifies high fault-sensitive features from extensive spectrometric data.

Impact:

• Enables more accurate and efficient gearbox fault diagnosis through optimized feature selection.
• Reduces data processing complexity in condition monitoring applications.
• Provides a robust framework for analyzing complex spectrometric data in wear analysis.