Efficient Reliability-based Design For Slopes Using Probabilistic Sufficiency Factor

The traditional deterministic design method uses a safety factor to determine a reasonable design by the way of trial and error.Obviously,it is difficult to reflect whether the design results are reliable due to a large number of uncertainties,such as the variability of geotechnical materials and model uncertainty,existing in the process of slope analysis.Reliability-based design(RBD)in geotechnical engineering is attracting more attention because of its ability to effectively consider a variety of uncertainties.In the current stage,a variety of design codes adopt the simplified RBD,such as load and resistance factor design.However,the engineers do not know these assumption before obtaining these partical coefficients and can not guarantee the reliability of design after using these partial coefficients.In the design process,geotechnical engineers not only need to consider the reliability of the project,but also consider to save design costs.It is an important problem for engineers to select a satisfying design scenario that meets both the requirements of reliability and cost.In this connection,full probabilistic design method is very attractive,but the implementation process is not easy for a large number of candidate design scenarios,which serversly limits the practicability and applicability of full probabilistic design.It is an urgent problem to improve the computational efficiency of full probabilistic design.Therefore,this thesis proposes an efficient full probabilistic design method using probabilistic sufficiency factor.The main works and conclusions are as follows:(1)Propose an efficient full probabilistic design method using probabilistic sufficiency factor,which significantly improves the efficiency of full probabilistic design.The probabilistic sufficiency factor is introduced to construct the design response surface.Design response surface replaces the process of reliability evaluation,which avoids repeated evaluation of the reliability of each candidate design scenario.And it has been proved that the second order design response surface can meet the precision of design requirement.(2)Probabilistic sufficiency factor(PSF)is redefined from the perspective of cumulative distribution function so that it can be widely applied to various RBD methods.PSF can reasonably evaluate whether the design has reached the target reliability and greatly improves the accuracy of the design response surface.(3)Propose an efficiency method of calculating PSF based on the generalized subset simulation,which avoids re-executing the random simulation process for different design scenatios and improves the efficiency of obtaining PSF.(4)The validity,accuracy and efficiency of the proposed method are verified by the examples of slope,which facilitates the application of full probabilistic design in RBD of slope.