Experimental Investigation of Global Backward Erosion and Suffusion

Many embankment dams constructed with a core of non-plastic or low plasticity silt-sand-gravel, frequently of glacial origin but sometimes fluvioglacial or alluvial origin, have experienced internal erosion. This has often expressed itself with the development of sinkholes, or with intermittent episodes of increased leakage which then reduces, probably as the filter or transition system seals with eroding soil. The internal erosion mechanisms involved in these incidents are not well understood.In this investigation, the aim was to test a wide range of particle size distributions (PSDs) from those similar to those of Sherard (1979), to those testing as clearly suffusive as in the Wan and Fell (2004, 2008) testing. Tests were permitted to progress for as long as there was evidence of on-going internal movement of soil particles as evidenced by erosion, or changes in seepage flow rates and piezometer pressures within the sample. 37 tests on 22 soil samples were performed in a custom built 450mm diameter, 500mm high cell. Test durations ranged from 15 minutes to 223 days with gradients commencing at 1 and rising up to a maximum of 10 if required.All 22 soils tested were internally unstable as evidenced by PSDs after testing showing finer and coarser PSD layers and manometer readings indicating lower and higher permeability layers. It was concluded that to equate the term internal instability only to suffusion as proposed by ICOLD (2014), and Fell et al (2015) is too restrictive and the term should apply to all mechanisms of internal movement of particles of soil within the soil matrix including suffusion and global backward erosion.The internal erosion process was very rapid for suffusive soils, typically occurring in minutes in the tests, and at a gradient of 1, but for soils subject to GBE, and no-erosion soils the internal movement of particles continued for weeks and months in tests, and re-activated when the overall gradient was increased. For GBE the erosion process occurred over a range of gradients.The behaviour of the soils varied to some extent with the size of the base mesh with soils subject to GBE potentially significantly more erodible for a base filter/ mesh significantly coarser than a continuing erosion filter. Suffusion appears to be not so dependent on the mesh size.A new method for predicting the amount of erosion and the erosion mechanism based on the particle size distribution of the soil has been developed which is related to the ability of the soil to self-filter. This method predicts the internal erosion mechanism and amount of erosion quite well and is recommended for use. However, for important decisions it is recommended that laboratory permeameter tests modelling the core and filter / transition or rockfill downstream of the core are carried out