Reliability analysis for settlement prediction of a strip footing.

摘要

In traditional footing design methodologies, the prediction of settlement usually uses a deterministic value of elastic modulus which is estimated by testing the soil at a limited number of locations. The reliability of the design is controlled by using a factor of safety. However, in practice, the soil tests may not represent the actual soil properties due to many sources of uncertainty. Also the factor of safety will not adequately predict the reliability of the design since it does not take variability of the parameters into account.;An investigation into the reliability of a traditional footing design methodology is performed in this thesis. The soil mass is modeled as a spatially varying and random elastic modulus field and the practical traditional design procedures are realized by simulation. The required footing width is decided using traditional methodologies after virtually sampling a number of locations in the simulated soil mass. The resulting required footing is placed on the simulated soil mass and the actual settlement computed using the finite element method. The procedure is repeated many times so that the reliability of the traditional design method, based on a random sample, can be estimated.;The main purpose of this thesis is to study the reliability of an existing foundation design method for the purpose of developing a reliability-based code. In particularly, a semi-analytical method to assess the reliability of a modified Janbu design method is developed as a function of the soil's variance and scale of fluctuation. The results of the semi-analytical method are compared with those from simulation and the agreement appears reasonably good. It is concluded that the method is a good estimate of the unconditional probability distribution of the settlements predicted by the modified Janbu Approximation.