The rail network operated by DB Netz AG is a core element of Germany’s transport infrastructure and is expected to accommodate significantly increased traffic volumes by 2030. Achieving this goal requires comprehensive upgrades to the existing infrastructure to support higher speeds and axle loads.

However, such upgrades often necessitate a reassessment of embankment stability based on current design standards. This frequently leads to a paradox: even embankments that have performed safely and reliably for decades may no longer meet today's analytical verification criteria – sometimes even under current loading conditions. If an upgrade is pursued nonetheless, the embankments must be brought into full compliance with modern standards. This typically entails extensive and costly rehabilitation, often resulting in considerable financial and material resources being spent unnecessarily.

Project Background and Objectives

The root of this discrepancy often lies in overly simplified calculation models. Traditional approaches typically neglect key aspects, one of which is the natural spatial variability of geotechnical properties like friction angle and cohesion.

The aim of this project was therefore to investigate the application of the Random Finite Element Method (RFEM), an advanced numerical approach that explicitly accounts for this inherent spatial variability in geotechnical parameters.

Methodology and Implementation

The study was based on a representative cross-section of a railway embankment from a real-world DB Netz AG construction project. Statistical distributions of the relevant soil parameters were derived from field investigation data. These distributions were then transformed into spatially variable random fields to reflect natural soil heterogeneity as realistically as possible.

To generate these complex random fields, we developed a out own software solution at the Institute of Geotechnics. This tool accounts not only for spatial variability but also for cross-correlation effects – such as the interdependence between the friction angle and cohesion. Figure 1 illustrates six sample random fields for the friction angle, all generated from the same statistical input data.

The resulting, more “realistic” site models were then imported into the finite element software Optum G2, and the stability of the embankments were evaluated using the strength reduction method. A typical failure mechanism observed in these analyses is illustrated in Figure 2.

Since single random field simulations cannot provide probabilistic information, Monte Carlo simulations were conducted. These enabled the derivation of a probability density function for the factor of safety (FoS). Based on this distribution, a probabilistic assessment of the embankment’s stability could be carried out.

Results and Outlook

The results demonstrate that the RFEM enables a more realistic, probabilistic assessment of stability. This approach has the potential to significantly increase efficiency in the management of existing infrastructure by reducing unnecessary and costly rehabilitation measures. With the planned incorporation of probabilistic design methods into the upcoming generation of the Eurocode, RFEM is expected to gain further practical relevance and broader acceptance.