Efficient track maintenance: methodology for combined analysis of condition data

Measured condition data may contain lots of information that is not efficiently used. By using techniques from the field of pattern recognition, this article outlines a methodology for eliciting new information. The methodology is demonstrated on a large set of condition data originating from track geometry quality, dynamic stiffness, and ground penetrating radar. The case study gives evidence for the importance of dynamic stiffness measurements as to determine soil- and embankment-related track problems, whereas problems originating from the upper part or the track structure do not benefit significantly.