Fractal analysis of geometry data for railway track condition assessment

This dissertation presents the results of the study that was conducted to examine the use of fractal analysis to quantify track geometry data meaningfully so that the geometry data could then become a much greater part of the railroad decision-making process. The report begins with a background discussion on railway track geometry and measurement techniques. The theory behind fractals is presented along with the divider fractal dimensioning technique. The divider method is discussed in detail to provide an edifying example of the fractal analysis concept, and to show how this method is used to quantify railroad track geometry data. The concept of “true” versus “natural” fractals, multi-fractals, and long-range dependency was explored and examples supporting the discussion were presented.; The dissertation presents the working procedures for fractal analysis of track geometry data based on the divider fractal dimensioning technique. These procedures were used as the basis for two computer programs (FTEval and FTAuto) for fractal analysis of large volumes of railway track geometry data. This study focused primarily on vertical profile since it is the geometry parameter most affected by substructure condition.; Fractal analysis was performed on a large amount of track geometry data and edifying examples are presented that show how fractal analysis effectively characterizes the functional condition of the track. The different orders of roughness within geometry data are also discussed. The benefit of fractal analysis for examining different lengths of track is shown. Examples of specific applications of fractal analysis of geometry data are given, including categorizing track, waveform characterization and quantification of the overall track quality.; The results of a study to examine the benefits of fractal analysis of geometry data for maintenance planning are given. A discussion of the results of an empirical study is also included. The empirical study attempted to correlate known substructure conditions (based on previous studies) with the results of fractal analysis performed on a history of geometry data.