The importance of initial particle position on filter performance in granular soils

This paper conducted an extensive set of gravity-driven filtration simulations using the discrete element method (DEM). The simulations were used to statistically analyse the inherent variability in filter performance due to variations in initial particle position using one-way ANOVA for repeated measures tests. The simulations accounted for group effects associated with the interaction of base particles within the voids of the filter particles, and considered a size ratio that captured the complete range of filtration behaviour. The simulations were performed in a highly controlled manner, such that the sole difference between the base-filter realisations was initial particle position. Four filtration characteristics, based on filter efficiency and filling degree, were considered to define filter performance. One-way ANOVA for repeated measures tests conclusively showed that initial filter particle position had a statistically significant effect on filter performance for all four characteristics. Significant variability was observed with normalised ranges up to 16.0%. Increasing the domain width did not eradicate the observed statistically significant differences in filter performance. These observations highlighted the importance of considering multiple stochastically generated base-filter realisations in DEM studies before drawing conclusions about the performance of filters due to the inherent variability of the particle-scale mechanisms underlying the filtration process.