A Study on Wheel Member Condition Recognition Using 1D-CNN.

Area of Science:

• Railway Engineering
• Condition Monitoring
• Machine Learning

Background:

• Current railway wheel inspection relies on manual, periodic checks, missing early defect detection.
• This reactive approach risks operational disruptions, safety hazards, and increased maintenance costs.
• A predictive maintenance strategy is crucial for enhancing railway safety and efficiency.

Purpose of the Study:

• To develop a real-time tire wear classification system for light-rail rubber tires.
• To improve the predictive maintenance capabilities for railway vehicle wheel conditions.
• To reduce operational costs, enhance safety, and prevent service delays.

Main Methods:

• Collected operational data (temperature, pressure, acceleration) from light-rail vehicles.
• Processed and analyzed sensor data to generate a comprehensive training dataset.
• Utilized a 1D Convolutional Neural Network (1D-CNN) model for tire condition classification.

Main Results:

• The 1D-CNN model achieved a highly accurate tire wear classification rate.
• Demonstrated an exceptional performance with 99.4% accuracy in classifying tire conditions.
• The system enables real-time monitoring and prediction of tire wear.

Conclusions:

• The proposed real-time classification system effectively predicts railway tire wear.
• This predictive approach significantly enhances operational safety and reduces maintenance-related delays.
• The technology offers a viable solution for proactive railway wheel condition management.