Strength deterioration and predictive modeling of coarse-grained soil under penetrating erosion effects

Abstract

This investigation aims to explore the strength deterioration characteristics of coarse-grained soils (CGSs) under penetrating erosion. Unconsolidated undrained (UU) triaxial tests were conducted on untreated specimens and those subjected to four amplitudes of penetrating erosion force (p₀), to examine the impact of penetrating erosion on the shear performance of CGSs with varying mass ratios of mudstone fines (η). Furthermore, a strength prediction model, K_M-C(η, p₀), was developed based on the Mohr-Coulomb (M−C) strength criterion, incorporating both η and p₀ as influencing factors. The results indicate that untreated specimens predominantly exhibit bulging deformation upon failure, whereas treated specimens can be divided into distinct bulging and non-bulging zones, with significant transverse cracks emerging in the transition region between these zones. Moreover, treated CGS specimens demonstrate notable strain-softening behavior, as the deviator stress peak [(σ₁–σ₃)_max] decreases linearly with increasing penetrating erosion force p₀ under the same confining pressure (σ₃). Furthermore, minor variations in confining pressure have little effect on the strength weakening induced by penetrating erosion, and at low confining pressures, both untreated and treated specimens conform to the M−C failure criterion. Additionally, there exists a critical fine content threshold η_critical; when the fine content is below this threshold, increasing it enhances the stability of the CGS skeleton and its resistance to seepage erosion. Conversely, when the fines content exceeds η_critical, the filling effect of the fines weakens the stability of the soil skeleton, thereby reducing its resistance to penetrating erosion. In this investigation, η_critical is determined to be 9.15%. Lastly, the predictive model developed in this study accurately captures the influence of η and p₀ on the strength of CGSs, with the relative error between the measured and predicted strength values ranging from 1.1% to 14.9%.