Improved Kriging Model And Its Application For Stability Reliability Analysis Of Tunnel Surroundings

As for the reliability calculations of practical engineering problems, in most cases,the explicit relationship between the state function and the basic random variables do not exist or function relationship is more complex, then the numerical simulation method can be applied to calculate failure probability, but the workload is often enormous simulation. To solve this difficulty, the response model has benn put forward to approximate the state function of the structure in the mathematical theory,By the 90s of 20th century, CC Bucher, who has proposed a method by using quadratic polynomial to approximate the limit state equation which has been called the response surface method, the basic idea of this method assumed a analytical expression between the unknown parameters of a state function and the basic random variables, and then use interpolation methods to determine the unknown parameters in the expression. But the expression form of the response surface method and the location of interpolation points,which have tremendous impact on the accuracy and have become the main issues for scholars of various disciplines.Currently, the main response model method are as follows response surface method, Taylor series method, support vector machines.for a long time, the role of kriging model in the structural reliability calculation has not been enough attention, which consists of a parameter process and a non-parametric random process and has a strong advantage in the prediction.On this paper, firstly, analyzing the exact existing problems by using JC method and response surface method to calculate the reliability index through various numerical examples; secondly, study the various existing design methods, and followed use best uniform point set sunch as Hammers experimental design to construct the training sample points,then study the accuracy of error of several major experimental design on Kriging model and RSM;thirdly as for the limitations for determining model parameters of the kriging method, genetic algorithm have been proposed to determine the relevant model parameters,which could obtain the optimal estimation model,in this case, the Kriging model is applied to approach the limit function of the reliability index calculations.As for stability analysis of underground structure, firstly analyzing the main features of design calculations and stability of surrounding rock, and then researching the results of stability analysis of cavity design calculations based on the generalized Hoke-Brown strength criterion proposed by C.Carranza-Torres and others,whose format have some big weakpoints,so in this paper developed format based on Matlab language,which can be widely used in tunnel design calculations, are proposed; Secondly, study the current treatment defect of the statistical methods for geotechnical parameters, take the fuzzy random variable weight theory processing some experimental samples, as for example a specific project, the improved kriging model is applied to calculate reliability index of underground structures,and develop a corresponding computer program.