Uncertainty Analysis Of Consolidation Theory

Nowadays,vacuum preloading combined with plastic vertical drain pipe(PVD)is the main consolidation treatment method for dredger fill in practical engineering.However,due to the high water content of dredger fill,many parameters in consolidation process cannot be accurately obtained(such as consolidation coefficient,compression coefficient,etc.).The prediction of engineering only by using the soil parameters obtained from field survey is always uncertain.Therefore,finding a fast and effective method to characterize soil parameters,which can accurately predict the consolidation degree of dredger fill in practical engineering,has great significance.In this paper,a Bayesian framework based on in-situ observation for back analysis and updating of soil parameters in consolidation process is proposed.The framework can update soil parameters through field observation in consolidation process of dredger fill,and the updated parameters can effectively and accurately predict the degree of soil consolidation in consolidation process.In the Bayesian framework,the uncertainty of soil parameters is calibrated and captured by probability distribution.And the MetropolisHastings(M-H)algorithm of Markov Chain Monte Carlo(MCMC)sampling method is used to obtain the posterior distribution of soil parameters in combination with the settlement data of engineering,and then the reasonable predicted consolidation degree is obtained.In this paper,two cases are used to verify the effectiveness of the proposed Bayesian framework.Case 1 is a vacuum preloading project in a test area before the start of largescale vacuum preloading project in Taizhou city,Zhejiang province,China and Case 2 is a vacuum preloading project in Shenzhen city,China.The results show that the accuracy of predicting the consolidation degree obtained by updating the soil parameters with only one month one region data in Case 1 is not enough,and the probability distribution of predicting the soil consolidation degree can be obtained accurately by using the field settlement data of two months one region in Case 1.If the two-month measured data cannot be obtained,more measured data can be obtained by increasing the number of measured regions,and then the uncertainty of updating parameters can be reduced by Bayesian updating,thus more accurate prediction can be obtained.However,the uncertainty of predicting consolidation degree distribution from the latter is slightly higher than that from the former.In addition,the influence of prior information of soil parameters on posterior information is analyzed by Bayesian framework.The results show that,as long as there are enough data,the influence of prior information on posterior information is small,and the uncertainty of soil parameters decreases with the increase of measured data.As long as two months' measured data can effectively reduce the uncertainty of parameters.And more data can reduce the uncertainty of parameters,but the decrease of parameter uncertainty is not very great.Because there is no prior information about the consolidation coefficient in Case 2,the non-information prior distribution is used for the consolidation coefficient.The uncertainty of soil parameters can be effectively reduced by only one month of measured data,and the accurate degree of soil consolidation can be obtained.The more accurate degree of soil consolidation can be predicted by using two-month data.In addition,combined with the results of Case 1,the influence of uncertainty of consolidation parameters on predicted consolidation degree is analyzed.The increase of variation coefficient of consolidation parameters will reduce the mean value of predicted consolidation degree and increase the uncertainty of predicted consolidation degree.