Sensor Selection and Placement for Track Switch Condition Monitoring through Validated Structural Digital Twin Models of Train-Track Interactions

Due to increased passenger traffic and more frequent load exertion on the railway infrastructure, the rate of deterioration for track systems has increased. Switches and Crossings (S&Cs), which are the track components that help trains change direction are highly susceptible to degradation and would thus benefit from continuous condition monitoring using physical sensors. However, the critical locations for the placement of sensors need to be determined to obtain the most useful data for condition monitoring. To this end, Multi-Body Simulations (MBS) and Finite Element (FE) Analysis have been carried out for the interaction between a train and a railway switch. From the outputs of the MBS simulations, damage prediction models were used to determine the locations with the highest risk of damage along the length of the switch. Parametric simulations for different train speeds, wheel-rail friction conditions, axle loads, and fault introduction were carried out through the detailed 3D solid element FE model of the local region of the switch. It was observed from previous field experimentation that a more linear relationship is obtained between the wheel-rail contact force and sensor measurements for strain gauges than accelerometers in the presence of surface crossing faults. Therefore, stress/strain signals obtained from the simulations were post-processed to determine strain sensor placement. Ideal sensor placement locations have been determined through the algorithm that has been developed whilst considering the material fatigue life as well as the mutual information between the fault occurrence and the sensor placement locations. The virtual strain measurements from these validated simulation models, which can be turned into live Digital Twins can also be used for sensor selection for condition monitoring.