ANALYSIS AND DESIGN OF EMBANKMENT DAM SLOPES: A PROBABILISTIC APPROACH (SOIL MECHANICS, INDIANA)
Abstract
The purpose of this research was to develop probabilistic procedures to improve present methods of analysis and design for embankment dams. Presently, the safety of an embankment slope is assessed by calculating a factor of safety which relates soil strength and loading conditions. Strengths are represented by single values chosen from scattered test results; both the "strengths" and the factor of safety are uncertain quantities. In the proposed procedure, uncertainty was accommodated by treating strength as a random variable. Procedures were developed to characterize soil strength by a beta distribution fitted to three probabilistic moments. Strength variability for the Clarence Cannon Dam was estimated using a large number of tests obtained during embankment construction. The use of a probabilistic design method necessitates the prediction of the expected mean and variance of embankment strength. Data from Cannon Dam was used to develop a recommended prediction procedure. A review of existing probabilistic slope stability methods revealed several limitations regarding the general analysis and design of dam embankments: these relate to the ability to accommodate embankment zonation, loading conditions, possible failure surface shapes, and the shape of the strength probability distributions. An analysis procedure was developed that accommodates skewed strength distributions, the correlation between the (phi) and c strength parameters, multiple soils, and failure surfaces of a general shape. Results are expressed in terms of a "probability of failure" or reliability index. Using the developed method, a series of parametric studies were performed for Cannon Dam to evaluate the relationship between probability of failure and loading conditions, parameter skewness, parameter correlation, reservoir level, and factor of safety. Different loading conditions having similar factors of safety were found to have greatly different probabilities of failure. The probability of failure was found to be strongly dependent on strength parameter correlation and somewhat dependent on parameter skewness. The method was also used to study Shelbyville Dam, where a slide had occurred, and to evaluate the increase in reliability due to its repair. Recommendations are offered for the implementation of the proposed procedures in design practice.
Degree
Ph.D.
Subject Area
Civil engineering
Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server.