Efficient slope reliability analysis and risk assessment based on multiple Kriging metamodels

Abstract

Kriging metamodel is an effective method for slope reliability analysis because of its flexibility and accuracy in data interpolation. However, the accuracy of traditional Kriging metamodel needs to be improved when facing complex limit state functions in slope stability problems. In addition, traditional Kriging metamodel cannot provide the key component—slope failure consequence—for risk assessment of slope failure, which greatly hinders the application of the Kriging method. This paper develops a multiple Kriging (MK) metamodels method to address the above problems within the framework of limit equilibrium method. The proposed MK metamodels are developed by simultaneously establishing a Kriging metamodel for each potential slip surface of a slope, with a global optimization search for the correlation parameters by the genetic algorithm. Slope stability analysis is then efficiently performed on the MK metamodels to obtain the associated factor of safety (FS) and locate the critical slip surface (CSS). Since the FS and the associated CSS are available, slope reliability analysis and risk assessment can be conducted with ease, thereby circumvent the abovementioned issues. For illustration and validation, the proposed model is applied to two examples, including a two-layered cohesive soil slope and a practical case of the Congress Street cut slope in Chicago. Both the influence of regression models and correlation models on the accuracy of the MK metamodels are discussed. The results show that the proposed model is more accurate than the conventional model in estimating the slope failure probability. More importantly, it is shown that the proposed model can effectively assess the risk of slope failure, which greatly extends the application scope of the Kriging based method.