Uncertainty Characterization for Soil Cohesion in a Project Site in Nasiriyah Using Bayesian Methods

Abstract

High uncertainties arias through the characterization of soil parameters because of the lack of data obtained from geotechnical reports. Reducing these uncertainties may improve the characteristic values of soil parameters. This research aims to probabilistically characterize a soil's cohesion parameter in Nasiriyah. The Bayesian approach has been applied to soil data obtained through a project in Nasiriyah. The soil at the site is classified as lean clay, and the soil cohesion has been evaluated using two Bayesian methods: the ordinary, normal distribution method (OND) and the Marcove Chain Monte Carlo-based Bayesian approach (MCMC) method. The previous knowledge utilized in the Bayesian approach was based on 20 boreholes, and the subjective probability approach has functioned in the prior probability distribution. The OND method deduced a mean value of cohesion of (195.9 kPa) and a standard deviation of (14.68 kPa), (COV) 7.49%. It was noted that the probability distribution has a more significant effect than the previous distribution on the posterior distribution. The MCMC method summarized the probabilistic description of the soil characteristic, through which it reached the mean and the subsequent standard deviation (167.49) kPa (109.8) kPa, respectively, and the coefficient of Variation (COV) was 65.6%. It is considered the most appropriate and common method, especially in high-dimensional data when the results are not well known because it can provide a probabilistic value for the not well-known data.