Prediction of train wheel diameter based on Gaussian process regression optimized using a fast simulated annealing...

Area of Science:

• Mechanical Engineering
• Artificial Intelligence
• Railway Engineering

Background:

• Train wheel diameter dynamically decreases with mileage, impacting speed and location accuracy.
• Traditional Gaussian Process Regression (GPR) faces hyper-parameter challenges with standard optimization methods.
• Accurate real-time wheel diameter monitoring is crucial for train operational efficiency.

Purpose of the Study:

• To propose and evaluate a Fast Simulated Annealing-Gaussian Process Regression (FSA-GPR) algorithm for predicting train wheel diameter.
• To address the limitations of traditional GPR optimization techniques.
• To improve the accuracy of train speed and location measurements.

Main Methods:

• Development of an FSA-GPR algorithm for wheel diameter prediction.
• Comparative analysis against traditional GPR, ABC-GPR, and GA-GPR algorithms.
• Validation using real-world data from a DF11 train during a major repair period, assessed by RMSE, MAE, R2, and residual values.

Main Results:

• The FSA-GPR algorithm demonstrated superior prediction accuracy compared to other evaluated methods.
• Predictions from FSA-GPR closely matched the actual measured wheel diameter data.
• Significant improvements in accuracy metrics (RMSE, MAE, R2) were observed with FSA-GPR.

Conclusions:

• The proposed FSA-GPR algorithm effectively predicts train wheel diameter, mitigating issues from wheel wear.
• Integration into vehicle-mounted systems can automate wheel diameter updates, reducing manual labor.
• Enhanced accuracy in speed and position measurement contributes to improved train operational effectiveness.