Data-driven approach to study the polygonization of high-speed railway train wheel-sets using field data of China’s HSR train

Abstract

Environmental factors, like seasonality, have been proved to exert significant impact on reliability of China high-speed rail train wheels in this article. Most studies on polygonization of train wheels are based on physical models, mathematical models or simulation systems. Normally, characteristics and mechanisms of wheel polygonization are studied under ideal conditions without considering the impact of the environment. However, in practical use, there are many irregular wear wheels and irregular wear cannot be explained by theoretical models with assumptions of ideal conditions. We look at two possible factors in polygonization: seasonality and proximity to engines. Our analysis of field data shows the environmental factor has more impact on wheel polygonization than the mechanical factor. Based on the Bayesian models, the mean time to failure of the wheels under different operation conditions is conducted. A case study of China’s HSR train wheels is conducted to confirm the finding. The case study shows the degree of polygonal wear is much more severe in summer than other seasons. The finding may give a totally new research perspective on polygonization of train wheels. We use Bayesian analysis because this method is useful for small and incomplete data sets. We propose three Bayesian data-driven models.