Permeability evolution of granular soils in an internal erosion context – PERSEE

France relies on a significant stock of hydraulic structures with more than 9 000 km of protection against flooding, 8 000 km of dikes for navigation canals and 1 000 km of hydroelectric canals. The number of small embankment dams is around several tens of thousands, while the number of large dams approaches 600. An important aspect of this French hydraulic asset is its age: while most dams are older than half a century, most dikes are more than 100 years old. Hence, the maintenance of such a wide and old patrimony requires a costly upkeep and calls for scientific progress on the mechanism of internal erosion of geomaterials. In addition, the probable consequences of climate change on the sea level and continental hydrology will lead to increasing solicitations on coastal and fluvial structures, which will reinforce the need for their surveillance and maintenance.

Hydraulic earth structures can suffer from instabilities induced by internal erosion processes, which are responsible for 46% of all disorders. The risk management related to volumetric erosion, named suffusion, calls for the numerical modelling of these structures. Such modelling requires the development of a new relationship that can describe the evolution of the permeability during the suffusion process, i.e. including the evolutions of the grain size distribution (GSD) and the constriction size distribution that both describe the soil’s microstructure.

Overall, five numerical methods and several experimental tests will be used to adapt the concept of “controlling constriction size” to soils susceptible to suffusion. Our project is organized in four steps:

(i) First, numerical specimens will be studied to better understand the physical links between pore space characteristics extracted from granular specimens by using the discrete element method and the permeability that can be computed thanks to a numerical full field homogenization technique. The idea is to work on numerical specimens constituted of spherical grains and simplified GSDs, with respect to that of in-situ soils.

(ii) Second, a physically-based relationship relating the key microstructure characteristics, identified previously, to the permeability will be sought with a semi-analytical homogenization method. This method provides an estimate of the permeability from a simplified and implicit representation of the microstructure, such as spherical grains, cylindrical pores, etc. This first relationship will be validated against the numerical simulations realised in step (i), by focusing on simplified GSDs and spherical grains. Next, this relationship will be extended to in-situ soil GSDs by using a probabilistic approach to obtain the key microstructure characteristics. Throughout steps (i) and (ii), the obtained results will be challenged for intact specimens and for suffusion-induced heterogeneous ones.

(iii) The validation of this extended physically-based relationship will be realised against several permeation and suffusion tests. In addition, multiple grain size distributions (initial, post-suffusion, eroded grains) will be realised to further validate the numerical approaches developed in step (ii) and (iv).

(iv) Finally, the extended physically-based relationship and the probabilistic approach will be implemented in our domestic finite element code which features a hydro-mechanical continuous model extended to suffusion. The numerical predictions of the permeability will be validated against the suffusion tests, performed in step (iii), and subsequently on a reduced physical model of dike.

The final product of this project, in addition to the deliverables of each step, will be a poro-mechanical simulation code that accounts for suffusion-induced fine grain loss and permeability changes. The model implemented inside this code will have been carefully validated against experimental data to allow its practical use by dam and dike stakeholders.