Our research project manager, Virginie Delavaud, carried out her PhD thesis for SNCF, the french railway company, with ENSTA ParisTech, about railway rolling noise.

Rolling noise is the main source of railway transportation noise for a wide speed range (between 50 and 320 km/h). Rolling noise occurs when a wheel moves on a rail in a straight line. On the same scope, the impact noise is due to discrete irregularities on either of the two structures, such as rail joints or wheelflats.


Client problem

To predict rolling noise, SNCF uses a simulation tool based on a frequency approach. Although very efficient in case of standard straight ballasted track and constant train speed, this approach is limited to wheels and rails without large amplitude defects. Thus, the wheel/rail interaction model is not adapted to the prediction of impact noise. A time dommain approach is required to extend the rolling noise model to include impact noise.

The main purpose of our work was to build a time domain simulation tool of the wheel/rail vertical interaction, in the scope of rolling and impact noise modelling. Knowing the wheel and rail rugosity, the final model gives the vibration leval of the rail in the time domain. In the future, the model will be inversed allowing the prediction of wheel surface state knowing the rail surface and the rail vibration level.

With the liberalisation of the european railway market, this tool could allow to detect the defects present on the train wheels which damage the railway tracks. This will allow the owner and operator of the railway infrastructure to target the trains that dammage the rails.

Solution proposed

A wheel/rail interaction model has been developped. A railway track is composed by rails supported by periodically distributed pad-sleeper systems, and ballast.

Based on the bibliography study of rolling noise modelling and for simplicity of the model, the rail is represented by a beam of finite length. To represent the "infinite length" characteristic of the rail, some innovative absorbing boundary conditions were developped to simulate the waves' infinite propagation. Moreover, some mass-spring-damper systems are periodically distributed under the rail to model the pad-sleepers-ballast systems. In order to predict the rail vibration level, it is sufficient to model the wheel with a simple 1 degree of freedom system. The whell/rail contact is modelled thanks to Hertz theory, modified to take into account the surface rugosity of both structures.

An experimental characterization of the wheel defects influence on the rail vibration levels and on the pass-by noise was conducted. A freight test train was specifically assembled. Some wheels of the test train were chosen for the presence of different kinds of defects. Moreover, a wheelflat was artificially made by grinding down one of the healthy wheel. The test track was chosen to include 2 types of links between rails : one part with long welded rails and one part with rail joints. Two test areas were thus equiped. The wheel and rail surfaces on oth experimental sites were precisely characterized. Acoustic and vibration measures were then recorded during the train pass-by.

This measurement campain enabled the validation of two objectives. First, the general specifications for the design and implementation of a wheel defect detection tool, based on acoustic and vibration measurement on the railtrack, were defined. These specifications introduced amongs others, a maximum acoustic and vibration level beyond which a wheel defect is detected. The second objective of the measurement campain was the use of the rugosity measurements on the wheel and rail as inputs for the numerical model in order to validate it. Although the model gives results in the time domain, all the validation analyses were conducted in the frequency domain. The accelerance curve of the railway track vibration is usually used to adjust the frequency models. Thus the measured accelerance curve wasused to validate the time domain model of the wheel/rail interaction.

Other missions

  • PhD
  • Data processing

Our CTO, Baptiste Coulange, did his PhD for CNES (the French Space Agency) and University of Paris Descartes on detecting of aliasing in satellite images.

Earth observation satellites are optimised to obtain the best image quality. The size of the sensors and the parameters of the optical chain are designed to obtain the best compromise between image resolution and acquisition artefacts. The higher the resolution the more artefacts are present.

Aliasing which is one of the artefacts in satellite images can lead to image misinterpretation. It is thus critical to detect it. By using the duality between spatial localisation in the image and aliasing relationships in the Fourier transform plane, it was possible to develop and validate a aliasing detection algorithm.