ABSTRACT
ABSTRACT Life cycle performance and risk management are often mentioned as critical tasks for infrastructure managers. However, without proper data collection and analytics these tasks cannot be executed. This paper discusses lessons learned from a case where a data analytics approach was deployed when an unexpected phenomenon occurred on the Dutch High Speed Line (HSL-Zuid). In November 2014, it was found that large sections of the HSL-Zuid were affected by a severe type of rolling contact fatigue (RCF). The RCF resulted in deep cracks on top of the rail. These damages were unexpected as the rails were only 5 years in operation and these rails were expected to last about 20–25 years with proper maintenance. In this case, resulting in about 20 km of rail replacements and multiple additional grinding campaigns. As the causes of defects were unknown, the authors applied data analytics to evaluate the possible causes of the RCF. Several measurements of the infrastructure, maintenance and the rolling stock resulted in a set of parameters. Then, a bottom-up approach is proposed for evaluating the affected sections to find similar parameter values among these over the whole track. The idea was to look for parameter values which could explain why certain sections were affected by the defects while others were not. The outcomes of the analysis indicated that it is highly likely that one type of rolling stock was affecting the rails in the curves of the HSL-Zuid. As the track was designed at the high-speed sections for 220–300 km/h and this type of rolling stock was driving below design speed, different loading of the rails throughout the curves occurred. Lessons learned from this case do not only apply to the technical area of wheel/rail and vehicle/infrastructure interfacing, but also to the usage of data analytics itself and life cycle management. From this case study, it is discussed how data collection and analytics can be better embedded by (rail) infrastructure managers from an early stage of development and use of infrastructure. Further scientific development for infrastructure data analytics are also discussed.
