PI: D. Vaughan Griffiths Institution: Colorado School of Mines "Probabilistic Analysis of Geomaterials with Randomly Distributed Shear Strength"
Although sophisticated techniques are available for risk assessment, and although such techniques are important for analyses involving soil materials that are inherently variable, the civil engineering profession has been slow to embrace risk approaches in practice. Major reasons for this reluctance include the lack of knowledge about probabilistic methods, and the lack of confidence in the parameters which must be used as "input" to these methods.
This research will address these limitations with particular reference to the probabilistic analysis of slope stability, including the assessment of stability of existing dams and embankments, ands especially for the particular case when the soil is weakened by earthquake loading. The analytical part of this research will bring together two powerful tools of engineering analysis and modeling, namely the theory of random fields and the finite element method. By assigning each soil element soil properties based on a fixed underlying statistical distribution, soil properties will be accounted for in a systematic way. Nonlinear finite element elasto-plastic analysis will then be used to assess the stability of slopes made up of soils with randomly distributed properties.
The choice of soil properties is crucial, and this research will also focus on the interpretation of available data and presumptive values in the literature. Data sources will include the U. S. National Geotechnical Experimentation Sites (NGES) data bases and information from the Norwegian Geotechnical Institute (NGI). The results of the analyses will be presented where possible in the form of charts or semi-empirical relationships relating probability of failure to soil variability and spatial correlation. It is believed that studies of this type will provide guidance to geotechnical engineers in the design process and improve confidence in the use and interpretation of probabilistic methods.
