Predicting Hydraulic Conductivity of Geosynthetic Clay Liners Using a Neural Network Algorithm
Yu Tan, Jiannan Chen, Craig H. Benson
Abstract
Hydraulic conductivity tests on geosynthetic clay liners (GCLs) to evaluate chemical compatibility can require months to years to reach equilibrium. There is a need for alternative methods to screen GCLs for chemical compatibility that are more expedient. In this study, a neural network machine learning (ML) algorithm was used to predict the hydraulic conductivity of Na-Bentonite (NaB) GCLs to leachate chemistry. Development of the ML predictive model (MLPM) included five steps: data collection, data cleaning and normalization, algorithm selection, parameters optimization, and model validation and evaluation. The MLPM is based on data collected from two decades of tests conducted on NaB GCLs with a broad range of leachates. Bentonite characteristics, permeant chemistry, and stress conditions are incorporated into the MLPM. Validation showed that the MLPM predicts hydraulic conductivity within one order of magnitude of the measured hydraulic conductivity in 85% of the cases.