Suffusion potential assessment by self-filtration criteria

Abstract

The majority of civil engineering structures is either placed on or built with soils. Therefore, the stability of soil is of vital importance for the stability of structures. This thesis is focused on suffusion, which is a major type of soil internal instability. Suffusion is a hydro-mechanical phenomenon where fine soil particles are loosened, mobilised, and transported by seepage flow through a series of pore constrictions formed by the soil’s porous structure. This removal leads to a higher soil porosity, which intensifies the seepage flow. At some critical hydraulic gradient, the flow can restructure the soil on a micro scale to enhance the suffusion of bigger particles and/or initiate other types of internal instability such as piping or backward erosion. Therefore, suffusion is also acknowledged as an early step in some other types of internal instability of soils. The biggest challenge in suffusion assessment is the handling of the broad variation of spatial distribution of particles within the soil packing. With different particle arrangements, a gap-graded soil can be either a multi-layered filter or a suffusive material. The first contribution of this thesis is a new particle packing method, which facilitates the study of particle arrangements. Basing on the trilateration equations, this new packing method provides good control of particle position. Besides, this semi-analytical method allows variation of the porosity of the packing specimen. The second contribution is based on the packing method and allows determination of the pore constriction size distribution for given particle size distributions and predefined porosities. Another major difficulty in suffusion assessment is the definition of loose, transportable particles and the soil primary fabric, which forms the soil porous structure capable of transferring effective stress. Although there is no method available to determine those two fractions of soils, it is feasible to estimate their proportion via numerical simulation on the particle scale. The third contribution is the determination of the soil primary fabric by virtual oedometer tests under zero-gravity conditions. Soil particles are kept firmly within the soil primary fabric by a high effective stress. Meanwhile, loose particles float in pores because they do not transfer stress and are not affected by gravity. Thus, the identification of primary fabric can be dramatically simplified by a threshold of contacts. A quick suffusion assessment method based on the bending level of the particle size distribution is the fourth contribution. Experimental data show that suffusive, continuously graded soils are often characterised by a bended particle size distribution with a steep inclination in the coarse fraction and a long tail in the fine fraction. The bending level is represented by the farthest distance a particle size distribution is from its characteristic straight line connecting the starting and end point of the distribution. The correlation of the bending parameter with general slope gradations provides a good coherence with the empirical data. The fifth contribution is an approximate method to separate soil particles into primary fabric and loose particles. From the self-filtration criteria standpoint, soil primary fabric works as a filter, which prevents loose particles from penetrating through. Therefore, appropriate filter criteria can be applied on the separated particle size distributions to assess the suffusion potential of soils. The separation is based on a discovery from the numerical simulation mentioned above that there is often an overlapping interval of particle size between those two fractions for continuously graded soils. The sixth and last contribution is a new suffusion assessment method, which collates the particle size distribution of loose particles with the constriction size distribution formed by the primary fabric. The probability of a particle being transported to the next pore is calculated by the proportion of constrictions larger than the particle size. The suffusion potential is represented by the total probability of all loose particles. If this probability is larger than 50%, i.e. loose particles seem to be transported rather than kept in pores, the soil is considered to be suffusive. If the probability is between 20% and 50%, the suffusion potential of soils might depend on other factors, apart from geometrical resistance. The core of the thesis consists of one conference paper and four journal papers, three of which are submitted to highly ranked, peer-reviewed international journals. The practical outcomes of the thesis are: (1) Comprehensive suffusion assessment methods based on self-filtration criteria; (2) Evidence of the overlapping size interval, which is often overlooked in other methods; (3) Numerical proof that soil behaviours are dominated by a small proportion of soil particles involved in the primary fabric; and (4) A sequential packing method which can be used by other researchers for investigating processes influenced by the structure of the packing. It will be the future task to take into account the presence of water to complete the picture of suffusion.