Modeling geotechnical uncertainty by bootstrap resampling
Abstract
The main objective of this dissertation is to demonstrate and extend the application of bootstrap resampling, a non-parametric simulation technique, to modeling the effect of uncertainty associated with geotechnical data on engineering designs. This technique is presented as an alternative to traditional statistical data analysis and Monte Carlo algorithms currently employed for these purposes. A real geotechnical database obtained from Heber Road, Imperial Valley, Southern California, is utilized throughout this study to perform traditional statistical analyses, Monte Carlo simulation, and bootstrap simulation. The database consists of many cone penetration and flat dilatometer tests performed at the Heber Road site between 1983 and 1985 by Purdue University in collaboration with the U.S. Geological Survey. The problems of correlation between two variables and autocorrelation of one variable over space, are analyzed using the bootstrap technique. The convergence and accuracy of bootstrap estimates are studied and discussed. Typical problems of bearing capacity and settlement of shallow foundations in granular materials are also analyzed using bootstrap resampling. The variability of safety factors with respect to bearing capacity failure, and the resulting probability of failure obtained from bootstrap resampling, are compared with the results obtained using Monte Carlo simulation. The uncertainty of sandy soil settlements is assessed by applying the bootstrap technique, in conjunction with the Schmertmann method, to cone penetration and flat dilatometer data. The results obtained using data from both testing techniques are compared and discussed. The benefits and limitations of bootstrap resampling to model geotechnical uncertainty are discussed and conclusions are presented. Some recommendations for future research are given.
Degree
Ph.D.
Advisors
Bourdeau, Purdue University.
Subject Area
Civil engineering|Geotechnology|Statistics
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.