Statistical Modeling for the Prediction of Undrained Shear Strength from Index Properties of Cohesive Soils found in Agaro Town

Abstract

Construction of infrastructure projects are booming in developing countries in recent years. For these conditions, geotechnical engineering properties of soils, especially strength properties, are vital for design purposes. The laboratory equipment's and field instruments are not available in all areas to get these engineering properties of soil as well as they require more effort, time, and money. Thus, Geotechnical engineers usually endeavor to develop statistical models that best fit a specific area and soil type, which is used especially for analysis and design purposes. In this study, Statistical Modeling for the Prediction of Undrained Shear Strength from Index Properties of Cohesive Soils found in Agaro Town is studied. Index properties and undrained shear strength behavior of these soils were performed in the laboratory of Jimma Institute of Technology (JiT) and Jimma University college of Agriculture and Veterinary Medicine(JUCAVM) soil department. In the present work,undisturbed and disturbed soil samples from fifteen test pits at 1.5 m and 3.0 m depth are collected. Thirty representative samples were taken from test pits. For all test procedures, American Society for Testing & Material (ASTM) standard was used. Combining selected variables, single linear regression (SLR) and multiple linear regression (MLR) models were developed for the prediction of undrained shear strength parameter. To develop the intended statistical models for a study, SAS JMP Pro 13, Statistical Package for Social Science Software (SPSS v22) and Microsoft Excel-2013 soft wares are used. The study shows that undrained shear strength parameter (Cu) was significantly correlated with liquid limit, plastic limit (PL),bulk density, dry density,natural moisture content and plasticity index (PI) whereas it was not significantly correlated with specific gravity and liquidity index of study area soil. From the study, the best Model is obtained from multiple linear regression (MLR) analysis and given by:Cu=224.032-2.272*PL-2.485*PI; coefficient of determination (R 2 )=0.806,p value=0.00,Tolerance=0.923 and Variance inflation factors (VIF)=1.084,Durbin Watson=2.791. Using the developed model, undrained shear strength parameter can be figured as well as it is expected to have wide application in the construction to minimize the cost, effort, and time for laboratory tests of undrained shear strength of the study area.