Engineering Studies in Support of the Development of High-Speed Track Geometry Specifications

Abstract

The Federal Railroad Administration has been directing engineering studies to support the development of high speed track geometry standards. These standards are intended to cover train operating speeds from 110 mph to 200 mph. The studies conducted include evaluation of the use of measuring track geometry with offsets from several chord lengths, computer simulations of vehicle response to track surface and alignment variations, application of the proposed specifications to previously measured track geometry, and comparison of proposed specifications to foreign practice. The proposed standards use multiple chords to control surface and alignment geometry. Single isolated geometry variations are allowed greater amplitudes than three or more repeated geometry variations. The results of the engineering studies indicate that use of multiple chords is effective in controlling a wide range of geometry variation wavelengths, from less than 30 feet to greater than 250 feet. The computer simulation studies show that at high speed, wheel/rail interaction dominates vehicle response to short wavelength (less than ∼100 feet) alignment variations, while carbody motions dominate vehicle response to long wavelength variations. Derailment and carbody accelerations are the principal concerns in vehicle response to track geometry variations. Application of the proposed specifications to previous measurements of high speed track on the Northeast Corridor indicates a relatively modest number of exception locations (∼1 location every 3 miles). Comparison of the proposed specification with foreign practice indicates that the proposed specification provides a generally similar level of control of track geometry.

Prediction of high-speed train wheel-rail relationship based on maximum information coefficient and LSTM

“…Ahmed [4] studied the dynamic interaction between wheel and rail in the presence of unsupported sleepers by an effective numerical method based on the central finite difference theory. Magdy [5] used ANN technology to model the wheel-rail vertical force of simplified standard urban rail train, but it has certain requirements for simplified degrees of freedom, and there is a problem of slow model response; Pang [6] used the improved large artificial neural network to predict the wheelrail force. The above research lacks the screening of model input, and there is a large redundancy in the input, which leads to low accuracy of subsequent prediction, and relies on a large number of experimental data, which brings certain challenges to the acquisition of data and the inspection and elimination of abnormal data.…”