Allan M. Zarembski

Some examples of big data in railroad engineering

The railroad industry is an infrastructure intensive industry that relies on significant amounts of information and data to operate and maintain each railroad. Using the US railroad industry as a model, this data collection encompasses the full range of railroad activities from tracking of goods shipments and car locations to managing train crews to inspecting and maintaining the infrastructure. This paper will look at this last area, inspection and maintaining the infrastructure and in particular the 330,000 km (200,000 miles) of railroad track in active use in the US. Using a broad range of inspection vehicle to collect data and a new generation of maintenance management software systems to analyze and interpret this data, railroads represent an industry that is starting to make extensive use of its “big data” to optimize its capital infrastructure and safely manage its operations while keeping costs under control. This paper presents examples of collection, storage and use of “big data” in the railroad engineering environment.

Predictive modeling of the rail grinding process using a distributed cutting grain approach

High-density railway lines experience a high rate of deterioration on the running surface of the rails; it can be addressed by rail grinding in order to reduce the frequency of rail replacement. Rail grinding includes additional complex features beyond what is usually considered in conventional grinding. Although extensive empirical experience exists to describe rail grinding, it can still be considered to be an emerging field that is in need of predictive theoretical guidance. This paper presents a newly developed modeling approach that is intended to provide a theoretical understanding of the rail grinding process and allow the prediction of rail grinding behavior and performance. The modeling is a bottom-up approach that starts from individual cutting grains and builds up to the rail grinding train level. First, grain distribution modeling is used to build a uniform template for the grinding simulation, based on the assumption of spherical grains with normally distributed sizes. Second, one representative slice is extracted as a grinding surface with stable grains. Protrusion heights and spacing distances of the cutting grains are analyzed to obtain the features of the grinding surface. Then the spherical grains are transformed into decahedrons with arbitrary poses, so as to closely approximate the actual surface of a grinding wheel. Third, the interactions of the cutting grains are combined into a model of a single grinding wheel and compared to test results from a single-wheel test. This allows for the connection between the utilized grinding parameters and the grinding results to be isolated, which is validated with supporting experiments performed on a single wheel. The individual wheel relationships can be combined into a full multi-wheel grinding pattern for estimating the simulation results of a multi-wheel grinding train. Eventually, the comparison between the simulations and grinding tests is used to show the effectiveness of the predictive rail grinding modeling at the level of a rail grinding train.

Analysis of Wheel-Rail Contact Stresses Through a Turnout

The turnout represents a complex component of the track structure that generates high levels of vertical and lateral dynamic forces. This in turn results in high levels of wheel/rail contact stress and corresponding high rates of track degradation, significantly greater than in conventional track. Furthermore, the location of these high contact stresses vary as a vehicle negotiates the turnout. This paper presents the results of a series of analyses and computer simulations developed to examine the wheel/rail contact behavior as a vehicle negotiates a turnout and to determine the location and magnitude of the associated wheel/rail contact stresses. These analyses include a procedure for aligning the wheelset profile to the rail profile pair, based on the actual rail profiles, as it varies through the turnout. Using the point-by-point alignment, the analyses then determine the location of the wheel/rail contact patch and then calculate the magnitude of the contact stress profile in that patch. The result is a map of the wheel/rail contact stress as a vehicle moves through the turnout.Copyright

Performance of a Track Geometry Car-Based Real-Time Dynamics Simulator Using Multiple Vehicle Types

A real-time dynamic simulation system designed to identify sections of track geometry that are likely to cause unsafe rail vehicle response is discussed. Known as TrackSafe, this system operates onboard a track geometry vehicle where the geometry measurements are passed as inputs to the dynamic model of one or more rail vehicle types. In order to comprehensively analyze the effect of the existing geometry on rail vehicle behavior, the system is capable of simultaneously simulating the response of several vehicle models, each over a range of traveling speeds. The resulting response predictions for each modeled vehicle and each simulated traveling speed are used to assess the track geometry condition and to identify locations leading to potentially unsafe response. This paper presents the latest work in the development of TrackSafe, specifically, the development and testing of eight new vehicle models is presented. The new car types modeled include a box car, flat car, and both a long and short tank car. Each can be simulated in a fully loaded or empty condition. Accuracy of the models is discussed in detail.Copyright

Risk-based scheduling methodology for audit inspections of curves on high-speed mainline tracks:

Agencies with safety oversight responsibilities of railroad tracks often perform walking audit inspections of tracks (also referred to as quality audits) to complement and oversee the regular inspections performed by the railway operator or maintenance manager. Traditionally, these audit inspections are scheduled based on the qualitative evaluation of the rail line by the inspectors, together with the available schedule of the inspector(s). This paper presents an approach to replace the current qualitative decision-making process for determining when and where to conduct audit inspections with a quantitative decision-making process. This quantitative process first establishes an acceptable level of risk in a given territory, and then taking into account the defect history and real-time track conditions, it schedules audit inspections based on those conditions. This risk-based scheduling methodology of audit inspections can be used by the safety oversight agencies and inspectors to monitor and “spot” check track conditions and provide oversight over the normal inspection process. The audit inspection’s frequency algorithm, presented in this paper, establishes the acceptable level of risk based on six years of Federal Railway Administration safety audit inspections data of the Amtrak North East Corridor. This methodology takes into account the track conditions in terms of the curve defect rate and optimizes the scheduling of audit inspections of mainline curves based on this defect condition. The risk-based curve audit inspections interval methodology outputs the required maximum curve audit inspections interval (time until next audit inspection or reinspection) while maintaining an accepted level of risk in the presence of real-time curve defect rates.

Towards a Better Understanding of the Rail Grinding Mechanism

Rail grinding continues to be one of the most effective techniques for extending rail life, improving wheel/rail contact behavior, and reducing the overall cost of track maintenance. While the ability to more effectively implement improved rail grinding programs continues to expand, the understanding of the grinding mechanism itself has not kept pace with the improved implementation. Thus, while railroad engineering and maintenance personnel have learned to better develop grinding patterns and profiles through empirical testing and field evaluation, the fundamental theoretical bases for the improved grinding performance have not kept pace. One such fundamental area of understanding is the modeling of the rail grinding process itself, both individually, as a function of a single grinding motor on the head of the rail, and in the more complex configuration of multiple grinding motors in a range of patterns. This paper presents the results of research directly aimed at better understanding these mechanisms and then utilizing this better understanding to develop a detailed rail grinding model that allows for the accurate analysis of not only an individual grinding motor but also a full grinding train application, as a function of pattern and speed. In the case of the single grinding motor on the head of the rail, this research looks at the fundamental mechanism associated with each cutting abrasive grinding grain in the grinding stone, and then expands that mechanism to a full 10 inch diameter grinding wheel as it cuts into the rail head at a defined angle and speed. Using actual rail profile data and grinding data, a theoretical grinding wheel model is developed and then calibrated with wheel test data and actual grinding (field) data. This single motor model is then expanded into a full grinding train model, such as for a 96 stone grinding train with 48 motors per rail, where it is able to analyze the full sequence of 48 motors as each motor individually and sequentially removes metal from the rail head. The resulting analysis is sensitive to such key factors as grinding speed, and the key pattern parameters of motor angles, sequence and power. The model is then calibrated to and compared with actual full scale rail grinding metal removal data from a major Class 1 railroad. Such an analysis tool allows railroads to analyze the performance of different grinding patterns in a real world operating setting, to improve their rail grinding practices and take further advantage of new technologies in rail grinding to better manage the grinding process and improve planning of grinding activities.