An important subject in the research on the behaviour of railway structures is the validation of numerical models by means of in-track measurements. Many forms of track measurements have taken place all over the world, mainly in the sense of strain and acceleration measurements, but a more unexplored area is that of the direct measurement of stresses in the ballast.
The aim of this research is to find the best way to equip a railway sleeper with measuring instruments in order to use it for validation of computational models. The sleeper must be able to measure the the vertical velocity of the sleeper as well as the normal stresses on the ballast-sleeper interface over time.

In order to predict the quantities to be measured, a finite element model was build for this research with the Kratos open source software package. The model comprises one single sleeper on top of a volume of ballast, both in solid elements. A dynamic load was applied vertically on top of the sleeper to simulate the passing of a double axle bogie. Linear elastic springs were fixed on top of the sleeper to represent force of the rail pulling the sleeper back to its initial place.
In order to analyse the effect of hanging sleepers, interface elements were used in between the ballast and the sleeper to simulate hanging gaps. The idea is that these elements transfer no stresses when the surfaces are physically separated, and adopt Mohr-Coulomb criteria as soon as the surfaces approach each other. This method turned out to be very difficult to implement, despite numerous discussion with the software engineer it was not possible to script this in the FE software in such a way that all calculations converged without any errors.
During the modelling of the sleeper, it was found that the forces on the sleeper are highly dependent on the resistance of the rail. The rails have such a high bending stiffness that they hardly show any curvature over the length of only a few meters, so the sag of the rail is mainly dependent on the behavior of the surrounding sleepers. It was concluded that, to get a good picture of the behaviour, it is not sufficient to model only one individual sleeper, but it is necessary to look at a larger part of the railway track in order to involve the coupling between the sleepers.
The results from the computational model were used to determine the magnitude range and the frequency domain of the values to be measured, in order to be able to select the right measuring instruments. The results were also used to select the right positions for the sensors on the sleeper. Due to the constraints described above, a certain margin of uncertainty has been adopted.

The vertical speed of the sleeper is best measured with accelerometers, the large amount of previous applications of this type of measurement has shown that this measurement works well and will not lead to problems.
It was considered to use pressure cells for the measurement of interface stresses, though this device turned out to be difficult to fix at the sleeper and it gives very little information on the distribution of stresses. It was concluded that the best way to measure the ballast-sleeper interface stresses is by using the matrix based tensile surface sensor (MBTSS). This instrument is able to capture the stress distribution on a surface over time, by measuring the electrical current flow through a matrix of conductive lines, the current is resisted when certain forces are applied on the surface. Tekscan is a manufacturer that offers a wide range of MBTSS types, including sensors previously used in railway measurements. A certain protection will be needed to protect the sensor from being damaged by the ballast particles.
Previous applications of MBTSS in railway structures showed the calibration of the sensor to be very difficult. This presumably has to do with the way the stresses are distributed over the sensor surface. Since the stresses of the ballast are often compressed to very small contact points, it should be made sure the stresses will not get lost between the sensels, that are the points on the surface where the stress magnitudes are measured. The advice is to cover the MBTSS with an under sleeper pad, this is a relatively stiff mat that, in addition to offering protection to the sensors, also spreads the ballast forces. Laboratory tests will have to be performed in advance of the field measurement to test whether a correct calibration is possible. It can not be made sure that a fully successful calibration method will be found and therefor additional measurements are recommended for a validation for the stress magnitudes. A first validation can be found by equating the sum of the vertical forces with the mass times the acceleration of the sleeper. To find the sum of the vertical forces, the load on top of the sleeper at the rail-sleeper connection, has to be measured as well. This can be realized by placing MBTSS between the baseplate of the rail and the sleeper. Similar measurements described in literature proved this type of measurement to be feasible. Furthermore, the result of the acceleration measurement will have to be used here as well. A second validation can be performed by determining the moment distribution over the length of the sleeper, this can be conformed to the magnitude and distribution of the vertical stresses on the sleeper. The most reliable way to perform this validation is by using fiber measurements to find the moment distribution over the whole of the length. The most important positions over the length of the sleeper are under the rails and in the middle of the sleeper, because the peaks in the moment distribution are expected here based on the computational model calculations.