Model-Based Condition Monitoring of Railway Switches and Crossings

Doctoral thesis, 2024

Railway switches and crossings, often known as S&C or turnouts, enable trains to change track. This operating feature comes at a cost because S&C have rail discontinuities which cause a significant contribution to the dynamic loading and higher degradation rates compared to ordinary plain line track. These higher rates of deterioration present a promising opportunity for implementing condition monitoring systems that have the potential to enhance maintenance decision-making and surpassing the capabilities of periodic inspections conducted by measurement cars or track engineers. This thesis is therefore focused on developing novel processing tools to increase the amount of condition information that can be extracted from sleeper mounted accelerometers using advanced multibody simulation (MBS) models.
The developed condition monitoring framework provides a robust model-based identification of railway ballast support conditions, wheel–rail impact forces, and crossing rail geometrical irregularities, as well as closed form condition indicators computed directly from measurement signals. The presented analysis algorithms demonstrate resilience in handling extensive datasets, with the total acceleration database used for this thesis consisting of around three years of remote field recordings for eight crossing panels.
An essential signal processing technique presented in this thesis is an innovative method for reconstructing sleeper displacement. It relies on integrating acceleration in the frequency domain and using band-pass-based functions for detrending baseline distortion. Using these track displacements, an approach is demonstrated for independently observing the sleeper support conditions and the geometry of the crossing rail from a single measurement source by separating measured displacement into dynamic and quasi-static domains based on two distinctly determined track response wavelength domains.
The MBS investigations carried out in this thesis demonstrate a robust link between the condition of the crossing rail geometry, the contact force between the wheel and rail, and the proposed condition indicators that have been established based on the dynamic track responses. The MBS
models serve as a basis for developing a procedure based on the Green's Kernel Function Method (GKFM) for an inverse identification of vertical wheel–rail contact forces and crossing rail geometry from measured sleeper accelerations. Additionally, this thesis demonstrates the possibility of determining ballast stiffness properties without prior knowledge of the crossing geometry, thus actually enabling the use of GKFM for inverse identification of wheel–rail contact force and crossing rail geometrical irregularity in a crossing panel equipped with a single sleeper mounted accelerometer.
In addition to measured sleeper accelerations, the geometry of the running surface of six crossings have been measured in situ using a laser scanner. These geometries have, together with measured wheel profiles, been implemented in the MBS models to account for a range of operating conditions.