Abstract: To investigate further the seepage erosion mechanism and predict the distribution of particles washed away in the filter system, this paper develops a filter void network model for the dense state of filling soil. Predictions of this model are analyzed and compared with those of two previous models in the literature in terms of the depth of base particles infiltrating into the filter, as well as with numerical simulations; And its feasibility and applicability are verified further in an engineering case. The results show the model of Locke and Indraratna gives the largest infiltration depth among the three; our new model predicts a depth smaller than that of Silveira’s model when the probability of base particles passing the constriction is low, while it predicts infiltration depths increasingly greater than Silveira’s as the passing probability increases. Variation trends in its maximum infiltration depths at the characteristic particle size ratio D15/d85 are the same as those in Zou’s simulations, but a greater discrepancy appears between the two calculations due to the limitation of Zou’s in soil sample size and infiltration time, especially at a size ratio close to or greater than 4 (the threshold value in the filter criteria). Our void model is a new approach for the filter design of real projects when the filter gradation design envelope is confirmed.

Key words: filter material, base soil particle, void model, infiltration depth