Deformation Prediction And Reliability Analysis Of Earthworks In Permafrost Regions

The frozen soil in China is mainly distributed in the northern provinces and the Qinghai Tibet plateau.In the process of construction of highway and railway in permafrost regions,the disease of pennafrost earthworks occurs frequently.Therefore,it is of great important to study the deformation and reliability of earthworks.At the same time,it is of great significance to reduce earthworks diseases and risk assessment.In this paper,the deformation and reliability of permafrost earthworks in Qinghai-Tibet Railway are analyzed.The specific research contents and conclusions are as follows:(1)A BP neural network model is established,which input is the average ground temperature,the height of the earthwork and the depth of permafrost tables,the output is the deformation of the earthwork.In addition,the program compiled by MATLAB is used to predict the deformation of the earthworks within 30 years and then verify it.The results show that under the condition of negative temperature,the lower the temperature is,the larger the frost heave is.Similarly,under the condition of positive temperature,the higher the temperature,the greater the thaw settlement.(2)The deformation analysis model based on the hydrothermal field of permafrost earthwork is established.And the COMSOL numerical simulation software is applied to the permafrost earthwork.The hydrothermal field and deformation field of the earthworks within 30 years are obtained,and the results were compared with the results of BP neural network to verify its accuracy.(3)The change rule of the temperature field standard deviation which obtained by the stochastic finite element approach is analyzed.The results show that the standard deviation of temperature field decreases gradually with the increase of depth.The relationship between the temperature standard deviation and the variation coefficient and time are linear and logarithmic respectively;(4)A new improved neural network response surface approach(named INNRS)is proposed and the program is compiled by MATLAB.This approach is combined with BP neural network and the improved response surface method.Only one response surface need to be formed to obtain the reliable index accurately.The application of the program to the permafrost earthwork project obtains that the reliable index and failure probability of deformation within 30 years.In this paper,the deformation and reliability of earthworks in permafrost regions are studied in order to provide theoretical basis for the research of construction,performance and risk assessment.