Diagnosis of Multiple Faults in Rotating Machinery Using Ensemble Learning

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Signal Processing

Background:

• Rotating machine fault diagnosis is crucial for preventing downtime and enabling condition-based maintenance (CBM).
• Deep learning models offer automated feature extraction but struggle with diverse fault types, operating conditions, and component scales.
• Existing methods face challenges in handling single and multiple faults across different rotating components like gearboxes, bearings, and shafts.

Purpose of the Study:

• To propose a comprehensive learning approach for diagnosing single and multiple faults in diverse rotating machine components.
• To address the limitations of current deep learning models in handling variations in scale, operating speed, and load conditions.
• To develop an optimized signal processing and ensemble learning framework for robust fault diagnosis.

Main Methods:

• Utilized optimized signal processing transforms, including bicoherence, spectral kurtosis, and cyclic spectral coherence, for feature extraction.
• Employed deep blending ensemble learning for enhanced fault diagnosis capabilities.
• Integrated a compound dataset from multiple public repositories for training and validation.

Main Results:

• The proposed approach demonstrated superior performance in diagnosing single and multiple faults across different rotating machine components.
• Achieved improved diagnostic accuracy compared to state-of-the-art methods on a combined dataset.
• Showcased the effectiveness of the framework with minimal retraining for new fault scenarios.

Conclusions:

• The developed comprehensive learning approach effectively diagnoses faults in rotating machinery, outperforming existing methods.
• Optimized signal processing and ensemble learning provide a robust solution for complex fault diagnosis scenarios.
• The framework supports reliable condition-based maintenance (CBM) decision-making through accurate and verifiable fault detection.