Railway track stiffness is an important track property which can help with the identification of maintenance related problems. Railway track stiffness can currently be measured using stationary equipment or specialised low-speed vehicles. The concept of using trains in regular service to measure track stiffness, has the potential to provide inexpensive daily 'drive-by' track monitoring to complement data collected by less-frequent monitoring techniques. A method is proposed in this paper for the detection of track stiffness variation through an analysis of vehicle accelerations resulting from the vehicle-track dynamic interaction (VTI). The Cross Entropy optimisation technique is applied to determine the track stiffness profile that generates a vehicle response that best fits the measured vertical accelerations of a railway carriage bogie. Numerical validation of the concept is achieved by using a 2-dimensional half-bogie dynamic model, representing a railway vehicle, to infer a global track stiffness profile along a track. The Track Stiffness Measurement Algorithm (TSMA) is implemented in Matlab. This paper reports the results of the numerical simulations. The proposed method gives good estimates of the track stiffness. To the authors' knowledge this is the first time an optimisation technique has been applied to the determination of railway track stiffness.

The use of sensors fixed to in-service trains has the potential to provide real-time track condition monitoring to inform maintenance planning. An Irish Rail intercity train was instrumented for a period of 1 month so that a numerical method developed to find track longitudinal profile from measured vehicle inertial responses could be experimentally tested. A bogie-mounted accelerometer and gyrometer measured vertical acceleration and angular velocity as the train made regular service operations between Dublin and Belfast on the island of Ireland. Cross entropy optimisation is used to find a track longitudinal profile that generates a numerical inertial response that best fits the measured response. Tolerance limits are used to inject variance where required to ensure a good match between measured and modelled signals. A section of track with known track settlement history is selected as a case study. A level survey was undertaken during the measurement campaign to characterise the longitudinal profile through the test section. Bandpass filters are used to compare inferred profiles and the surveyed profile. Good agreement is found between the two profiles although improvements in accuracy and reproducibility are required before conformance with current standards is achieved.

ABSTRACT:The longitudinal profile of a railway track excites a dynamic response in a train which can potentially be used to determine that profile. A method is proposed in this paper for the determination of the longitudinal profile through an analysis of vehicle accelerations resulting from the train/track dynamic interaction.The Cross Entropy optimisation technique is applied to determine the railway track profile elevations that best fit the measured accelerations of a railway carriage bogie. Numerical validation of the concept is achieved by using a 2-dimensional quarter-car dynamic model, representing a railway carriage and bogie, traversing an infinitely stiff profile. The concept is further tested by the introduction of a 2-dimensional car dynamic vehicle model and a 3 layer track model to infer the track profiles in the longitudinal direction. Both interaction models are implemented in Matlab. Various grades of rail irregularity are generated which excite the vehicle inducing a dynamic response. Ten vertical elevations are found at a time which give a least squares fit of theoretical to measured accelerations. In each time step, half of these elevations are retained and a new optimisation is used to determine the next ten elevations along the length of the track. The optimised elevations are collated to determine the overall rail longitudinal profile over a finite length of railway track.