Feasibility of Locomotive-Mounted Broken Rail Detection

This research examined the feasibility of applying Time Domain Reflectometry to the task of broken rail detection. The key innovation is to regard the track as a two-wire electrical transmission line, and then apply the same electrical pulse/echo technique that is commonly used to locate breaks and shorts in electrical cables. The aim of the proposed locomotive-mounted system is to allow each train to independently test the track ahead for broken rails and track occupation, up to a given detection range. This approach is designed to work in conjunction with Communication-Based Train Control to replace the present wayside signaling system and eliminate the associated per-mile maintenance costs. A theoretical framework for the proposed test system was developed to explore the likely benefits and problems, and a MATHCAD signal path model was developed to estimate the detection range. Anticipated problems that were considered included track-shunting structures such as turnouts; interoperability with the existing track circuits; RF safety and licensing issues; and the complex mechanical and electrical requirements for coupling signals between the train and the track. The signal path model incorporated factors such as the estimated transmission line characteristics of railroad tracks with wood or concrete ties and wet or dry ballast conditions; initial pulse characteristics; signal coupling losses to and from the rail; electrical reflection models for various configurations of broken rail or track occupation; and signal processing techniques that could be applied to recover highly attenuated echo signals. To estimate transmission line characteristics, previous European research on railroad track electrical properties was extrapolated to the required test frequencies. Railroad track was found to have a high attenuation rate because of ballast conductance, and the attenuation rate was predicted to increase significantly at the higher frequencies needed for pulse/echo testing. Even so, the model predicted that a range of one mile for wet ballast and two miles for dry ballast could be possible by using correlation processing to enhance the echo signals, and a variety of initial pulse characteristics to cover different segments of the required search range for various track conditions. The given range might be suitable for some train categories but not for others, depending on the full-service braking distance for each category of train. It is recommended that field measurements be made of the track attenuation rate over a broad range of frequencies for various U.S. track conditions. The aim would be to validate and fine-tune the model, and then reassess the predicted detection range compared to the safe stopping distance of each category of train.