Shear Strength Prediction of Soil based on Probabilistic Neural Network

Background: Shear strength parameter is an essential engineering property of soil which affects different aspects of soil such as bearing capacity of the soil, stability of slope, inclination of dam and retaining structures. Methods: In the following study, Probabilistic Neural Network (PNN) is applied as a promising tool for the estimation of shear strength. The input variables used for developing the model are index properties such as (water content (w), Plasticity Index (PI), Dry Density (DD), Gravel %(GP), Sand %(SP), Silt%(STP), and Clay%(CP) of soil and an attempt has been made to develop a neural model to predict the shear strength parameter of soil, viz, cohesion “c” and internal friction angle “φ”. Findings: The values of c and φ predicted by the model are comparable with the laboratory results. Trained data are validated to confirm the efficiency of PNN for determination of shear strength parameter. Application: Unlike other neural network approaches it uses the Bayesian estimation theory and gives time efficient results. These results can be utilized efficiently to determine the shear strength parameter using index properties of soil.