Estimating Dynamic Characteristics of Gravel-Tire Chips Mixtures Using Artificial Intelligence Techniques

Abstract

Scrap Tire Derived Geo-Materials (TDGM) mixed with soil are often being used as environmentally friendly granular materials in sustainable construction of civil engineering projects for reducing dynamic loads acting on geo-structures and soil liquefaction remediation purposes. Predicting dynamic properties of TDGM-soil mixture is a complicated task due to the numbers of factor involved in soil- TDGM mixture. This study presents application of artificial intelligence technique in estimating dynamic characteristics of granular mixture of Gravel and Tire chips (GTCM). Support Vector Regression (SVR) and Artificial Neural Networks (ANN) were used for predicting shear modulus and damping ratio of GTCM. Shear modulus and damping ratio models were developed using ANN and AVR techniques. The models were trained and tested using a database that included results from a set of laboratory tests on the GTCM. Stress controlled cyclic triaxial tests were conducted on specimens of gravel and tire chips with different volumetric portions of gravel in mixture (GF). The tests were performed on GTCM specimens at an initial relative density of 50% under different initial effective confining pressures. Test results have shown that shear modulus and damping ratio of the granular mixtures are remarkably influenced by volumetric fraction of gravel in GTCM. Furthermore, shear modulus was found to increase with the mean effective confining pressure and gravel fraction in the mixture. It was found that a feed-forward multilayer perceptron model with back-propagation training algorithm have better performance in predicting complex dynamic characteristics of granular mixture than SVR one.