A probabilistic model for predicting suffusion-induced fines loss

Accurate prediction of suffusion is essential for evaluating internal erosion risks and ensuring the stability of hydraulic structures. However, most existing suffusion laws are developed through curve fitting of experimental results and rely heavily on empirical parameters. Moreover, they generally fail to account for the effects of particle gradation changes and suffusion boundary effects. To address these limitations, this study proposes a new probabilistic model for suffusion development. The model constructs a fine particle transport mechanism using a network model to quantify the suffusion probability. Based on this probability, the cumulative mass of eroded fines and their spatiotemporal distribution are calculated using a flux-based approach. The proposed model was validated against four experiments conducted under both constant and multi-stage hydraulic gradients. The results show good agreement with experimental results, confirming the capability of the model to effectively predict the progression of suffusion. In addition, a parametric analysis was conducted to investigate the influence of initial fines content and distribution, relative density, permeability coefficient, and sample length on suffusion behavior. This work advances the understanding of suffusion mechanisms and provides new insights for geotechnical risk assessment.