A thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology

Thesis (M.Eng.) - Asian Institute of Technology, 1985

This thesis is mainly concerned with the application of statistical and probabilistic methods in analyzing the soil. data and soil structure performance . The trial embankment in Nong Ngoo Hao was used ±or case study. Both 3 dimensional model considering the and effect along the axis of embankment and 2-aimensional model of plane strain slope stability analysis using Simplified Bishop method were used in the study. The only variable considered is undrained shear strength. Prom the investigation, there exists a trend in mean value of undrained. shear strength which increase with depth, this was eliminated before application. The variance of undrained shear strength was verified to be stationary with depth using run test. The correlation distance in horizontal direction was found to be about 20 m while in vertical direction was about O. 2 m after detrending. The PDF used in 2-dimensiona1 monte -Carlo simulation model was normal distribution. The curve of probability of failure versus plane strain safety factor was constructed from 2-D and 3-D model for design purposes. The results show that a design safety factor of 1.12 is recommended for economical design for 3-D model analysis, and 1.28 is recommended tor conservative design for 2-D model analysis if the design criteria of probability of failure equal to 10-a is selected. However 2-D model and J-D model were proved to produce probability of failure within acceptable agreement under the same conditions. For Nong Ngoo Hao site, the BJERRUM 1 s correction fact or was successfully applied to explain the failure conditions. The corresponding probability of failure was 15% for 3-D model and 50% for 2-D model. In 3-D model analysis it was indicated that spatial variability of soil properties contributes very little to the probability of slope stability failures, it was significantly overshadowed by model error. The sensitivity analysis o f statistical parameter was also performed to see the influence of the change of soil characteristics. Computer program PROBISH developed by Utah State University was employed in 3-D model and program combination of Nearest-Neighbor realization generated technique and simplified Bishop circular method was used in 2-D model. Furthermore, simple Nearest-Neighbor method was used to present the simulation soil profile from data available.