A Support Vector Machine-Based Approach for Bolt Loosening Monitoring in Industrial Customized Vehicles

Area of Science:

• Engineering
• Machine Learning
• Condition Monitoring

Background:

• Machine learning enhances machine condition monitoring for superior fault diagnosis.
• Industrial environments with custom equipment challenge traditional monitoring methods.
• Bolted joints are critical for structural integrity, yet bolt loosening detection is under-researched.

Purpose of the Study:

• To investigate vibration-based detection of bolt loosening in a custom sewer cleaning vehicle's rotating joint.
• To evaluate the effectiveness of support vector machines (SVM) for this specific application.
• To determine optimal accelerometer configuration and model strategy for reliable fault detection.

Main Methods:

• Utilized vibration data from a custom sewer cleaning vehicle transmission.
• Employed support vector machines (SVM) for classification of bolt loosening.
• Analyzed the impact of accelerometer number and placement (upstream/downstream).
• Compared single SVM models versus condition-specific models.

Main Results:

• A single SVM model utilizing data from four accelerometers achieved 92.4% overall accuracy.
• Mounting accelerometers both upstream and downstream of the joint improved detection reliability.
• The study identified an effective approach for monitoring rotating joint health.

Conclusions:

• Vibration-based detection using SVM is a reliable method for identifying bolt loosening in rotating joints.
• A generalized SVM model with strategically placed accelerometers offers robust fault detection.
• This research contributes to maintaining the structural integrity of critical industrial components.