Development of rail temperature prediction model and software

The railway track buckling occurs all over the world due to inadequate rail stress adjustment, which is greatly influenced by the variation in weather induced rail temperature and the rigidity of the track structure. Climate change and the ever increase in extreme changes in temperatures have made buckling an ever more prevalent problem in the railway industry. The ultimate goal of any research in the area of track stability management is to comprehensively manage rail buckling and the subsequent procedures that follow after buckling. The first step to have a clear understanding of how the temperature change of the rail track is influenced by the environmental conditions. The second step is to have an accurate prediction of what the environmental conditions will be in the next day so that management procedure can be put into place.

This study aims to develop a model and software that is capable of predicting rail temperature 24 hours in advance that is as accurate for use in the rail buckling management. Two distinct and separate mathematical manipulations are performed to achieve this goal.

One method used weather forecasts from the Australian Bureau of Meteorology (BoM) and forecasts the weather for the location that the rail is situated. This involves using 3-dimensional cubic interpolation that is the weather parameters are interpolated in 2-dimensions geographically and then 1-dimensionally through time. An interactive software is written in MATLAB to convert the BoM raw data into a rail temperature forecast for this study. The result is a 15 -minute forecast for every 3.06 km. The second method used multivariate linear regression, to predict the rail temperatures 24 to 48 hours in advance.

To validate the rail temperature predications, 3 months field test spanning June, July and August 2010, is conducted on Queensland Rail's (QR) coal network, this involved erecting an automated weather station (AWS) and adhering temperature sensors on to a section of track. The guidelines of World Meteorological Organization's (WMO) were followed for implementation of the AWS on site (WMO 2008). The AWS model WXT520 , produced by Vaisala (Vaisala 2009) was used in this study which an off the shelf product that is similar to what some rail compaies are already using for continues monitoring of critical sites.

The temperature sensors (surface thermocouples) and an off the shelf product Salient system's rail -stress modules are used to measure rail surface temperatures on both rails of the track (Salient Systems Inc 2009). The sensors were attached to the surface of the rail track to directly measure temperature change of the rail profile throughout the diurnal cycle. Statistical correlations between the different measured points of the rail profile are evaluated in relation to the diurnal cycle to assess the accuracy of current rail temperature measuring practices.

Statistical evaluation of how well the BoM predictions compare with weather parameters at the field experimentation site are performed, so too is a statistical evaluation of the accuracy of the rail temperature model developed. The prediction model is compared with the existing empirical methods as found in the literature review and an assessment of track conditions. This is a flag ship study in Australia; the main purpose of this study is to prove in a test case scenario that a rail temperature forecast without use of weather instrumentation is possible and the accuracy of the prediction is as good if not better than the instrumentation calculation.