A comparison of longitudinal adhesion models and identification of the dominant adhesion parameter

ABSTRACT

Accurately estimating the coefficient of adhesion (CoA) is important to the safety, maintenance, operation and wheel–rail dynamics of railway systems. Studies have reported the influence of various parameters on the CoA and developed subsequent models, but parameter contribution or model estimation accuracy has not been systematically examined. Therefore, this study aims to quantitatively compare the accuracy of Heuristic, Polach and Dynamic estimation models with statistical measures and to investigate the contribution of various parameters on the CoA. Experimental CoA data was collected using a twin-disc rig at 57 and 85 km/h. The percentage root-mean-square error (PRMSE) of each model was then calculated for the rolling, sliding and overall contact regimes at each speed and multiple regression analysis (MRA) was applied to examine the contribution of each parameter. The Dynamic model was found to most accurately estimate the CoA in the sliding contact regime at both speeds due to its ability to capture dynamic variance. The Polach model also provided accurate estimations, but could not capture dynamic variance at the sliding contact regime. Using MRA, the CoA was found to be most dependent on sliding velocity. Overall, this work contributes to increase the prediction accuracy of wheel–rail contact dynamics.

KEYWORDS:

• Estimation accuracy
• adhesion model
• dynamic variance
• parameter contribution
• coefficient of adhesion
• multiple regression analysis

Acknowledgements

This study was supported by a grant from the R&D Program of the Korea Railroad Research Institute, Republic of Korea.

Disclosure statement

No potential conflict of interest was reported by the author.