Railways could experience a significant set of problems associated with dynamic interaction of the train-track system. These problems mainly stem from wheel and track defects due to deterioration or lack of maintenance. In this paper, a feasible way of vibration-based monitoring is investigated with emphasis on wheel flat identification of train-track system. The major component of the monitoring system is a new rail pad sensor made of specialty plastic material that has several competitive advantages over conventional sensors. To identify wheel flat using the proposed rail pad sensor, two methods are proposed. The first method is predictive-corrective determination of impact load. After obtaining the force time history of the rail pad from the sensor, wheel-rail interaction force is predicted by using the steady state response from the measured pad force. Subsequent corrections are carried out to account for the impact effect caused by wheel flat. The second method is a nonclassical system identification method based on genetic algorithm (GA). This method can be used to identify the shape of the wheel flat, specifically the length and depth of wheel flat as well as the position where the impact happens. Thepredictive-corrective method is suitable for real-time monitoring which can process data in seconds and give quick estimation of the peak contact force with acceptable accuracy. The GAbased method is able to identify the shape of the wheel flat with relatively small error, but it requires several minutes of data analysis.
doi: 10.12783/SHM2015/267