Prediction of geotechnical properties of rice husk ash-stabilized soil systems

Abstract

Rice Husk Ash (RHA) is one attractive alternative that is used as a full/partial replacement of cement/lime in problematic soil stabilization. This paper introduces statistical models; multiple regression analysis (MRA) and artificial neural network (ANN) for the prediction of Unconfined Compressive Strength (UCS), Soaked California Bearing Ratio (S-CBR), Maximum Dry Density (MDD), Optimum Moisture Content (OMC), and Plasticity Index (PI) of RHA-stabilized clayey soil. S-CBR and MDD of RHA-stabilized soil can be predicted with linear and non-linear MRA and UCS, OMC, and PI can be predicted with ANN models with prediction accuracy > 95%. In the validation process, all the proposed models express prediction errors around ±25%. A Parametric Analysis (PA) and a Sensitivity Analysis (SA) were performed to evaluate the variation of UCS with the influencing input parameters. In general, the analysis suggests 6-12% RHA with a very little amount of cement (4-8%) or lime (4-9%) as the optimum mix proportion for soft soil stabilization.