A theoretical model to predict suffusion‐induced particle movement in cohesionless soil under seepage flow

Abstract

AbstractSuffusion is a grave threat leading to accidents in embankment dams across the world. The process has been widely studied experimentally and numerically with gap‐grading cohesionless soil. One of the direct observations in the studies is the finer particle loss of the soil. However, the movement of the finer particles in the numerical models is commonly predicted with an empirical law. This paper proposes a theoretical particle movement model based on the force balance of soil particles and a probability distribution of the soil pores under one‐dimensional upward seepage flow. In the proposed model, the velocity and the critical movement state of soil particles were deduced, and a method to predict the critical hydraulic gradient was established. The finer particle movement rate was estimated theoretically using the particle movement velocity and probability. The model was verified with several experiments and showed a good consistency in the critical hydraulic gradient, seepage velocity, finer particle loss and hydraulic conductivity. Finally, the changes in the soil parameter during the suffusion process, the effects of the hydraulic gradient and initial porosity and the model applicability are discussed quantitatively. The results indicate that the modelled soil parameter changes match well with the movement characteristics of the finer particles observed in the experiment. The particle movement process is very sensitive to changes in the hydraulic gradient and soil initial porosity. The theoretical model provides a new method to assess the suffusion process for internal unstable cohesionless sand soil.Highlights A evaluation model of the suffusion‐induced soil particle movement was proposed The model was established based on force equilibrium of soil particles The model was credible in varied factors based on experiment data The model evaluated suffusion process, variables affects and model applicability were discussed