This study explores detecting rail defects (track ride quality exceptions) during train operations using real-time x, y and z acceleration data measured and collected on the side frame of a 315,000-lb gross rail load instrumented freight car. Different analysis tools were developed and employed to capture the peculiar characteristics of the data. This includes Harmonic analysis of the data, Wavelet analysis, energy density analysis, and correlation analysis. Based on the analysis results, different filtration and processing techniques were tried to identify the defects throughout the test data. A method that augments autocorrelation to wavelet based singularity detection showed promising results in capturing three types of exceptions: rail fractures, chipped rails, and broken concrete foundations. In addition, blind tests were conducted with several datasets and the algorithm proved to be 100% accurate in detecting the studied defects.

This content is only available via PDF.
You do not currently have access to this content.