Real-time classification of rotating shaft loading conditions using artificial neural networks

Area of Science:

• Mechanical Engineering
• Signal Processing
• Artificial Intelligence

Background:

• Rotating machinery health monitoring is crucial for preventing failures.
• Intermittent faults in rotating shafts necessitate continuous, real-time analysis.
• Traditional vibration analysis methods may miss transient fault conditions.

Purpose of the Study:

• To evaluate artificial neural networks (ANNs) for classifying rotating shaft conditions.
• To compare ANN performance against other discriminant analysis techniques.
• To explore the utility of time-series moments as input features for fault detection.

Main Methods:

• Utilized orthogonal vibration data represented as a two-dimensional vector time series.
• Calculated time-series moments as input features for rapid computation.
• Applied signal processing techniques to enhance feature discrimination.
• Compared artificial neural networks with other discriminant analysis methods.

Main Results:

• Artificial neural networks demonstrated effectiveness in classifying shaft conditions.
• Time-series moments proved to be efficient input features for ANNs.
• The study compared time-domain feature analysis with frequency-domain analysis.

Conclusions:

• Artificial neural networks provide a viable method for real-time condition classification of rotating shafts.
• Time-series moment analysis is a computationally efficient approach for fault detection.
• Continuous monitoring with ANNs enhances the detection of intermittent machinery faults.