Probabilistic study of slope stability under rainfall condition
by Zhang Lulu
Ph.D. Civil Engineering
xxvi, 354 leaves : ill. ; 30 cm
Rainfall-induced landslides pose significant hazards in many parts of the world especially in tropical mountainous areas. A prediction model can be established to evaluate the performance of a soil slope under rainfall condition. Geotechnical physical model tests such as field tests and centrifuge tests can also be used to investigate the performance of a slope....[ Read more ]
Rainfall-induced landslides pose significant hazards in many parts of the world especially in tropical mountainous areas. A prediction model can be established to evaluate the performance of a soil slope under rainfall condition. Geotechnical physical model tests such as field tests and centrifuge tests can also be used to investigate the performance of a slope.
A great number of uncertainties are involved in the problem of slope stability under rainfall condition. For example, in situ soil properties are subject to inherent variability. The predicted performance of a slope from a prediction model is expected to deviate from reality due to uncertainties of input parameters and systematic errors of the prediction model. When the prediction model is calibrated using physical model tests, uncertainties of the tests are introduced into the estimated model error.
The primary objective of this research is to investigate slope stability under rainfall infiltration within a probabilistic framework. Using probabilistic approach, different types of uncertainties can be accounted for in a systematic and quantitative way. The stability of a slope can be represented by not only a safety factor but also a reliability index, which represents the probability of satisfactory performance of the slope. The information from various physical model tests can be utilized systematically to evaluate errors of prediction models by the Bayesian updating approach. In addition, reliability-based design can be conducted for the slopes.
Based on a nonlinear elastic coupled formulation for seepage and deformation of unsaturated soils, a coupled hydro-mechanical analysis program is developed. Numerical solutions to the coupled formulation are obtained using a finite element partial differential equation solver, FlexPDE. A finite element based slope stability program FESSA is also developed to calculate the safety factor of the slope based on stresses and pore-water pressure distributions obtained from the coupled numerical analysis. Coupled numerical modeling and slope stability analysis are conducted for two slopes. The first is the Sau Mau Ping slope, where a disastrous landslide occurred in 1976 after approximately 30 hours of heavy rainstorm. The second is the Kadoorie test slope where a field test was conducted to study the response of the instrumented nailed loose fill slope under artificial rainfall condition. Comparisons between the field measurements and the calculated results are made.
Essential soil properties that influence seepage and deformation in unsaturated soil slopes are identified. Uncertainties of soil porosity θs, saturated permeability ks, the parameters a and n in the Fredlund and Xing SWCC model and the shear strength parameters Mcs, Mcol, Τ and λ are investigated based on measurements from field and laboratory tests. Reliability analyses are conducted for the Sau Mau Ping slope using a stratified sampling technique to generate random samples of input random variables. Uncertainties of initial conditions, boundary conditions and rainfall characteristics are not considered.
A Bayesian framework is proposed to evaluate the model error of a prediction model using information from physical model tests. The procedures of model calibration are illustrated using the information from the field test in the Kadoorie slope and two centrifuge tests to calibrate the coupled numerical modeling program and FESSA developed in this study. The results illustrate that if the preparation of a centrifuge model is carefully conducted and the variability of soil properties in the centrifuge model is very low, the calibration can be very efficient. It is also found that when modeling the same prototype, if the quality of sample preparation is maintained the same in preparing different sizes of centrifuge models, a centrifuge model with a larger size is better for calibration than a centrifuge model with a smaller size.
As a special topic of slope stability under rainfall condition, the reliability of nailed soil slopes in Hong Kong is investigated. A spreadsheet for slope stability and reliability analysis of nailed slopes is programmed. Slope stability analysis and reliability analysis are conducted for three example nailed loose fill slopes. The effects of soil nails in loose fill are illustrated and discussed. Finally, the reliability-based design for nailed loose fill slopes is illustrated.