A framework for final drive simultaneous failure diagnosis based on fuzzy entropy and sparse bayesian extreme...

Area of Science:

• Mechanical Engineering
• Reliability Engineering
• Signal Processing

Background:

• Accurate diagnosis of simultaneous failures in mechanical systems is critical for preventing catastrophic events.
• Existing diagnostic approaches often struggle with the complexity and overlapping features of multiple simultaneous failure modes.

Purpose of the Study:

• To develop a novel framework for the simultaneous failure diagnosis of final drive systems.
• To improve the accuracy and efficiency of diagnosing multiple, concurrent failure modes.

Main Methods:

• Feature extraction using wavelet packet transform and fuzzy entropy to reduce noise and identify failure characteristics.
• Training probability classifiers with paired sparse Bayesian extreme learning machine (ELM) using single failure modes.
• Employing a Grid search method with mixed single and simultaneous failure samples to establish an optimal decision threshold.

Main Results:

• The proposed framework effectively extracts representative features and reduces noise interference.
• Paired sparse Bayesian ELM demonstrates high generalization and sparsity for classifier training.
• The Grid search method provides superior global optimization for decision threshold generation.
• Experimental results, validated by F1-measure, show superior diagnostic accuracy and efficiency compared to Support Vector Machine (SVM) and Probability Neural Networks (PNN) based methods.

Conclusions:

• The developed framework offers a robust and efficient solution for simultaneous final drive failure diagnosis.
• The combination of advanced feature extraction, sparse Bayesian learning, and optimized thresholding significantly outperforms traditional diagnostic techniques.