Application of Machine Learning to Investigate Grinding Parameters of a Novel Railway Grinding Machine

Rail surface defects develop through repeated wheel–rail interactions and require timely grinding intervention to maintain rail integrity, enhance safety, and avoid costly rail replacement. Although prior studies have optimized individual grinding parameters such as rotational speed, stone granularity, and feed rate, the combined influence of multiple interacting parameters on metal removal rate and surface quality remains insufficiently understood, hindering development of optimized grinding patterns. This project develops an uncertainty-aware machine learning framework tailored to the limited size of laboratory-generated railway grinding datasets. A Gaussian Process Regression (GPR) model will predict key performance indicators — Metal Removal Rate and Surface Roughness- from inputs such as rotational speed and grinding wheel orientation while quantifying predictive uncertainty. Leave-one-out cross-validation maximizes data utilization, GridSearch optimization identifies hyperparameters, and a fixed random state ensures reproducibility; performance is evaluated with the coefficient of determination and Root Mean Squared Error. A complementary Support Vector Classifier is developed to classify grinding burn conditions, with data-imbalance mitigation via class weighting and Random Minority Oversampling, evaluated by accuracy and F1-score. Interpretability is central to the effort: SHAP analyses will rank feature importance so railway management can identify the most influential grinding parameters, and Partial Dependence Plots will clarify individual parameter effects. Expected outcomes include unified grinding patterns that achieve target rail profiles while maximizing material removal efficiency and surface quality, supporting the industry shift from corrective to preventive grinding, extending rail service life, reducing replacement costs, and strengthening railway operational safety.

Language

• English

Project

• Status: Active
• Funding: $150,000.00
• Contract Numbers:

  69A3552348323

• Sponsor Organizations:

  Office of the Assistant Secretary for Research and Technology

  University Transportation Centers Program
  Department of Transportation
  Washington, DC  United States  20590
• Managing Organizations:

  Howard University

  2400 6th Street, NW
  Washington, DC  United States  20059
• Project Managers:

  Bruner, Britain

• Performing Organizations:

  University of Maryland, College Park

  Department of Civil and Environmental Engineering
  College Park, MD  United States  20742
• Principal Investigators:

  Attoh-Okine, Nii

• Start Date: 20260803
• Expected Completion Date: 20270504
• Actual Completion Date: 0
• USDOT Program: University Transportation Centers

Subject/Index Terms

• TRT Terms: Machine learning; Maintenance of way; Rail grinding; Regression analysis; Uncertainty
• Subject Areas: Maintenance and Preservation; Planning and Forecasting; Railroads;

Filing Info

• Accession Number: 01999932
• Record Type: Research project
• Source Agency: Research and Education for Promoting Safety (REPS) University Transportation Center
• Contract Numbers: 69A3552348323
• Files: UTC, RIP
• Created Date: Aug 19 2026 4:05PM