Track buckling due to excessive rail temperature may cause derailments with serious consequences. To minimize the risk of derailments, slow orders are typically issued on sections of track in areas where an elevated rail temperature is expected and risk of track buckling is increased. While slow orders are an important preventive safety measure, they are costly as they disrupt timetables and can affect time-sensitive shipments. Optimizing the slow order management process would result in significant cost saving for the railroads. The Federal Railroad Administration’s (FRA’s) Office of Research and Development has sponsored the development of a model for predicting rail temperatures using real time weather forecast data and predefined track parameters and a web-based system for providing resulting information to operators. In cooperation with CSX Transportation (CSX) and FRA, ENSCO Inc. conducted a comprehensive model verification study by comparing actual rail temperatures measured by wayside sensors installed at 23 measurement sites located across the CSX network with the rail temperatures predicted by the model based on weather forecast data over the course of spring and summer 2012. In addition to the correlation analysis, detection theory was used to evaluate the model’s ability to correctly identify instances when rail temperatures are elevated above a wide range of thresholds. Detection theory provides a good way of comparing the performance of the model to the performance of the current industry practice of estimating rail temperature based on constant offsets above predicted daily peak ambient air temperatures. As a next step in order to quantify the impact of implementation of the model on CSX operations, heat slow orders issued by CSX in 2012 on 10 selected subdivisions were compared to theoretical heat slow orders generated by the model. The paper outlines the analysis approach together with correlation, detection theory and slow order comparison results. The analysis results along with investigation of past heat related track buckle derailments indicate that the railroad would benefit from adopting the rail temperature prediction model along with flexible rail temperature thresholds. The implementation of the model will have a positive impact on safety by allowing for issuing of advance heat slow orders in more accurate, effective and targeted way.

The railroad industry uses slow orders, sometimes referred to as speed restrictions, in areas where an elevated rail temperature is expected in order to minimize the risk and consequence of derailment caused by track buckling due to excessive rail temperature. Traditionally, rail temperature has been approximated by adding a constant offset, most often 30°F, to a peak ambient air temperature. When this approximated maximum rail temperature exceeds a given risk threshold, slow orders are usually issued for a predefined period of the day. This “one size fits all” approach, however, is not effective and suitable in all situations. On very warm days, the difference between rail temperature and ambient air temperature can exceed railroad-employed offsets and remain elevated for extended periods of time. A given temperature offset may be well suited for certain regions and track buckling risk-related rail temperature thresholds but less accurate for others. Almost 160,000 hours of rail temperature measurements collected in 2012 across the eastern United States by two Class I railroads and predicted ambient air temperatures based on the National Weather Service’s National Centers for Environmental Prediction (NCEP) data were analyzed using detection theory in order to establish optimal values of offsets between air and rail temperatures as well as times when slow orders should be in place based on geographical location and the track buckling risk rail temperature threshold. This paper presents the results of the analysis and describes an improved procedure to manage heat-related slow orders based on ambient air temperatures.