Have a Question?
Ask the Graduate
College at our new
Doctoral Dissertation Announcement
Candidate: Rennie Bwalya Kaunda
Doctor of Philosophy
Title: Some Applications of Gaussian Quadrature and Neural Network Modeling in Earth Flows and Other Slow Moving Landslides in Cohesive Slope Materials
Dr. Ronald Chase, Chair
Dr. Alan Kehew
Dr. William Sauck
Dr. Karlis Kaugars
Dr. James Selegean
Date: Friday, May 4, 2007 1:00 p.m. – 3:00 p.m.
1122 Rood Hall
Geometrical changes and progressive displacements in earth flows and other slow moving landslides triggered by climatic changes may be addressed by digital modeling. Gaussian quadrature, a numerical integration technique through fixed points, is employed to compute geometrical areas defined by stratigraphic (soil or rock layering) units, vertical pole projections and a slip surface, based on kinematic admissibility. An example from the Lake Michigan coast shows that the total internal geometrical area is found to be preserved during the course of the progressive deformation. Displacement monitoring of the slope shows that it became less stable over a period of eleven years due to progressive failure. The Gaussian quadrature technique allows representation and manipulation of geometrical models in a digital format amiable to the display of volumetric changes. Three different types of neural network models are also developed based on the back propagation algorithm for landslide problems in Michigan, England and the French Alps. The first artificial neural network model predicts slip surface positions based on measured surface displacements and soil types. The second neural network model predicts slope displacement rates from temperature and groundwater level data. The third model predicts ground water levels based on temperature data. The fourth model predicts displacements from precipitation records. The predicted slip surface positions using artificial neural networks closely match well the measured positions of slip surfaces at all three sites. Also, the neural network models are able to predict ground water levels and displacements from climate data. The digital exactness of Gaussian quadrature and neural network modeling allows for applications that are in a usable, quantifiable format for engineers and other mitigation planners. This digital format can be applied to a wide variety of slope stability problems of concern.