5. Conclusions
In this article, a stochastic finite element procedure based on project-specific data was implemented to assess the slope stability analysis of a case study of an earth dam. This procedure involved an evaluation of the input parameters required in an SFEM model based on field data. Furthermore, this evaluation was probabilistic concerning parameters that have a strong influence on the FoS. This paper performs a probabilistic stability analysis by SFEM for an existing earthfill dam, considering the spatial variability of soil properties based on field data. Indeed, the localization of the compaction control measurements performed during the earth dams’ construction is put to good use in this approach thanks to the geostatistical analysis that allows modeling the spatial variability of soil properties by random fields. Finally, MCS were performed on the deterministic FEM model to give a probabilistic distribution of FoS, and thus a reliability evaluation from the reliability index. Such an approach could be used as part of a global dam safety assessment combined with a risk analysis like that described in [45]. Based on the results obtained on the case study, the following conclusions and statements can be presented:
(1) The probabilistic analysis performed in this article can be easily applied to limited datasets generally available for earth dams. It allows overcoming the small number of in-situ or laboratory geotechnical tests by taking advantage of available data.
(2) Geostatistics provide a powerful tool for assessing the spatial variability of soil properties, even in the particular case of embankment dams. It is interesting to take this variability into account because of its significant influence on FoS (and thus on the reliability index) as it tends to reduce failure probability. The use of geostatistical methods nonetheless requires a large sample of geolocalized data, as with the compaction control measurements in the case study. Nowadays, this geolocalization can be performed easily with GPS technology. However, geostatistical analysis is in most cases only possible on compaction control measurements, as in the case studied with dry density.
