Fault Diagnosis of a Rotor and Ball-Bearing System Using DWT Integrated with SVM, GRNN, and Visual Dot Patterns

Area of Science:

• Engineering
• Mechanical Engineering
• Signal Processing

Background:

• Motor bearing failures are a significant cause of unexpected downtime.
• Real-time monitoring is crucial for preventing catastrophic failures and optimizing maintenance schedules.

Purpose of the Study:

• To develop and validate a real-time system for inspecting working motor ball-bearings.
• To differentiate between normal and abnormal bearing operation accurately.

Main Methods:

• Vibration signal acquisition using accelerometers.
• Signal decomposition and feature extraction via discrete wavelet transform (DWT).
• Classification using support vector machines (SVM), general regression neural networks (GRNN), and convolutional neural networks (CNN).

Main Results:

• Successful real-time diagnosis of motor bearing conditions.
• Demonstrated effectiveness of DWT, SVM, GRNN, and CNN in malfunction classification.
• Established a unique experimental process for motor diagnostics.

Conclusions:

• The proposed system provides a reliable method for real-time motor bearing inspection.
• The findings support the development of an integrated diagnostic and early warning system for motor malfunctions.