Multi-Filter Clustering Fusion for Feature Selection in Rotating Machinery Fault Classification

Area of Science:

• Machine Learning
• Signal Processing
• Engineering

Background:

• Traditional filter methods for feature selection in fault classification lack clear guidelines on feature quantity and necessity.
• Efficient and effective feature selection is crucial for developing accurate fault classification models.

Purpose of the Study:

• To develop a novel Multi-Filter Clustering Fusion (MFCF) technique for automated and efficient feature selection.
• To improve the accuracy, efficiency, and robustness of fault classification models for rotating machinery.

Main Methods:

• A multi-filter approach is employed for feature clustering, followed by automatic selection of key features.
• The union of key features is identified, and an exhaustive search determines the optimal feature combination.
• Fault classification models are developed using the MFCF technique for rotating machinery.

Main Results:

• The MFCF technique effectively and efficiently selects features for fault classification.
• Classification models utilizing MFCF demonstrated good accuracy in distinguishing normal and abnormal conditions in rotating machinery.
• The proposed method shows robustness in the fault classification of rotating machinery.

Conclusions:

• The MFCF technique offers a significant advancement in feature selection for fault classification tasks.
• MFCF provides a reliable approach to enhance the performance of classification models in rotating machinery diagnostics.
• The study highlights the potential of integrated filter methods for improved machine condition monitoring.