Ballast Fouling Identification Through Statistical Pattern Recognition Techniques on Ballast Particle Movement

Ballast fouling is one of the most common undesirable conditions in tracks that adversely impact ballast performance. Poor performing ballast can cause rough track geometry and accelerate the deterioration rate of other track components such as rail, tie, and fasteners. Real-time monitoring of ballast condition can assist in providing responsive maintenance planning and safe train operation. Several studies have been done till date with the aim of providing an automatic and continuous monitoring of ballast. SmartRock is a wireless sensor that has proven capable of serving as a continuous monitoring system for ballast condition. This sensor closely resembles the ballast particle, and while embedded in the ballast layer, it can provide information regarding ballast particle movement under the load of passing trains in real time. In this study, a field experiment was conducted on a clean and a mud spot section with the same traffic and weather conditions. Four SmartRocks were placed in each of these sections, and data recorded was analyzed using statistical pattern recognition technique. Linear discriminant analysis (LDA) is the algorithm deployed in this study to predict fouling of ballast through SmartRock data. The results of this study are encouraging toward the use of the SmartRock system together with LDA algorithm as a monitoring tool on the state of track ballast.