Keywords: railway engineering, track geometry, system identification, dynamic systems, neural networks, simulations.

This paper introduces a new approach to retrieve both reference track geometry data and track irregularities from the measured track data of urban light-rail systems. Track maintenance, forecast of irregularities and track/vehicle interaction requires knowledge of the reference track geometry. If a reference track has not been defined beforehand the development of a reference track geometry identification process becomes necessary.

The investigation of track quality is a major requirement for optimal maintenance of tracks. Obviously, measurements of the track geometry as well as the reference track geometry are necessary to evaluate the track irregularities. In the case of heavy rail systems reference track geometry is specified during the design process and is composed of standardised geometric forms such as straight lines, clothoids and curve segments. In contrast, light-rail systems are constructed with regard to the historical topology of roads and other traffic zones. Consequently, the requirements of modern railway systems with respect to geometric track design are often strongly compromised.

The basic idea to retrieve the reference track from measured data is to first apply a two-step filter procedure to the data (first in the spatial and then in the frequency domain) and subsequently describe the reference track by the radius of the local curvature. Track irregularities are consequently defined as measured deviations from this reference track.

It is obvious, that track quality conditions have a strong influence on vehicle reactions. Besides classical approaches to directly assess the track quality e.g. by spectra a dynamic simulation of a rail vehicle may be utilised to predict passenger ride comfort. Therefore, both vehicle manufacturers and transport operators have a great interest to accurately predict the ride behaviour of the vehicle for a given track.

The achieved track description is used for black-box modelling of a light-rail vehicle where the input data are separated into reference track geometry and irregularities. As a result of a non-linear input transformation of the reference track geometry and irregularities, the results for models with linear system dynamics show comparable performance to non-linear models identified with artificial neural networks. The whole procedure is demonstrated using measured track and dynamic vehicle data from a light urban rail system. Validation results for the lateral dynamics of the rail vehicle show excellent results for the proposed method.

purchase the full-text of this paper (price £20)

go to the previous paper
go to the next paper
return to the table of contents
return to the book description
purchase this book (price £145 +P&P)