Model Prediction of Soil Parameters via Experimental Analysis for the Geotechnical Design

Abstract

The geotechnical properties of soils play a crucial role in civil engineering projects. This study uses experimental design techniques to develop predictive models for soil properties such as cohesion, friction angle, and bulk density. The problem addressed is the need for reliable, data-driven models to predict these properties based on easily measurable soil characteristics, thus facilitating better infrastructure planning and construction. This research aims to apply the Taguchi method, a robust design of experiments (DOE) approach, to identify the significant factors affecting soil properties and develop accurate predictive models. Soil samples from various locations in Malaysia were analysed for different percentages of clay, silt, sand, resistivity, and moisture content. These variables were chosen as factors in Taguchi's design, with four factors at two levels. The predictive formulas derived offer a practical means to estimate soil properties based on measurable parameters. Formulas were developed for the cohesion, angle of friction and bulk density, demonstrating the systematic influence of the factors studied. Results show that the sand's percentage and 1D resistivity had the most significant effects, with moisture content contributing to variations in the cohesion.