Large-Scale Triaxial Testing on Suffusion Characteristics and Permeability Evolution of Gap-Graded Soils under Different Controlling Constriction Sizes

Abstract

Internal erosion is an important concern in coarse-grained soil engineering, yet large-scale triaxial experimental studies on this subject remain extremely scarce. As a prerequisite for seepage erosion test design, the selection criteria for leakage opening size lack a reliable experimental basis or quantitative guidelines. A self-developed large-scale triaxial testing system with coupled seepage and stress was used to study the suffusion process involved in internal erosion. The controlling constriction size, D C * , was introduced as a characterization parameter for the leakage opening size to systematically analyze the evolution of particle loss, gradation evolution, and hydraulic characteristics under different D C * conditions. Under stepwise loading, an increase in D C * lowers the critical hydraulic gradient for particle migration and increases the cumulative loss rate of fine particles. Under the tested material and loading conditions, a marked transition in fine-particle loss was observed between D C * / P min = 1.5 and 2.0, with the fine-particle loss rate at D C * ≥ 2 P min reaching 20 to 25 times that at D C * ≤ 1.5 P min , where P min denotes the minimum particle size missing from the gap-graded soil. When D C * ≤ 1.5 P min , fine particles clog the seepage path and accumulate at the bottom of the specimen; when D C * ≥ 2 P min , the gradation curves of all sections shift toward the coarser direction. The postsuffusion void ratio e n , calculated by comprehensively accounting for the fine-particle loss rate and volumetric strain, is introduced. The results indicate that D C * controls the evolution of hydraulic conductivity by regulating fine-particle loss and the associated postsuffusion void ratio e n . A positive correlation between e n and hydraulic conductivity k was identified, and an empirical prediction formula was established based on the experimental data under different D C * values.