Evaluating Performance of Conical Filter Systems Using Numerical and Laboratory Methods

Abstract

A significant contributor to retaining wall failure occurs due to inadequate drainage in the backfill. Studies showed 33% of retaining wall failures around the world occurred due to missing or inadequate drainage systems. Even though failure caused by drainage is high, very few United States Departments of Transportation are specific about the backfill material allowed. Traditional weep hole design makes use of pipes perpendicular or parallel to the wall to promote filtration; often covered with a geosynthetic for soil retaining purposes. This study seeks to determine the performance of a recent pore pressure mitigation system through the usage of conical geotextile filters and to investigate an alternative numerical method to effectively determine the type of geotextile in these filters.A numerical model based on a computational fluid dynamics and discrete element method (CFD-DEM) coupled approach was developed to simulate particle movement in the graded filter zone and piping through the geotextiles located in retaining wall backfills. The model was used for conventional as well as conical geotextile filter systems that use a series of woven and nonwoven geotextiles filtering backfill soils with varying fines contents. Poisson line processes and image processing techniques were used to study the pore structure of the nonwoven geotextiles. The results indicated that conical filter systems contribute to higher soil piping rates but provided higher permeability than conventional geotextile filtration system counterparts. The model predictions compared with the laboratory measurements indicated that the movement of particles (i.e., suffusion) influenced the soil-geotextile contact zone permeabilities and caused a decrease in system permeabilities. A retention ratio, αsl, successfully predicted piping rates for different types of woven and nonwoven geotextiles with a percent error range of 13-30%, and was converted into a performance chart. A machine learning algorithm was implemented to create woven and nonwoven zones within the performance curves. Overall, the model predictions were comparable to the laboratory results, suggesting the applicability of the model. Once validation was complete, a conversion retention ratio, αc, was developed for practical usage of the performance charts.