| The underground structure represented by subway station is an important part of urban lifeline project.During the construction of underground structures in cities,the site selection line of underground structures will inevitably pass through the liquefiable soil.Under the strong earthquake,the liquefaction of soil layer will cause great damage to the underground structure,including the floating of underground structure caused by foundation liquefaction and the uneven settlement of foundation after earthquake.At present,most of the existing prediction methods are about predicting earthquake liquefaction,and the applicability of the existing methods is limited,the accuracy of prediction is not high.In addition,there are few studies on the prediction of site seismic liquefaction disaster,especially on the problem of site containing underground structures.Method based on bayesian network,this paper firstly established a free field seismic liquefaction bayesian network model for prediction of subsidence disasters by finite element numerical model,then analyzes the influence with underground structure site seismic liquefaction hazard factors,established the site seismic liquefaction of underground structures in bayesian networks liquefaction disaster risk assessment model.Aiming at the problem of liquefaction and buoyancy of underground structures in earthquake,the improvement measures of anti-liquefaction and buoyancy are analyzed.The main research of this paper includes the following aspects:(1)Based on the Bayesian network method,a Bayesian network model of assessing seismic liquefaction-induced settlement is constructed,in which 12 significant factors including the earthquake parameters,soil parameters and field conditions combining with liquefaction potential and liquefaction potential index are considered.Through some cases study,the Bayesian network model has obvious advantages in assessment performance,comparing with RBF(Radial Basis Function)neural network method and I&Y(Ishihara&Yoshimine)simplified calculation method.The Bayesian network model not only has better assessment accuracy and reliability,but can also perform reverse causal reasoning.In the analysis of sensitive factors on the two machine learning models,ground peak acceleration,duration of the earthquake and standard penetration test blow count are more sensitive among 12 factors,of which I&Y simplified calculation method also use identical factors.(2)Seven factors influencing seismic liquefaction and structural buoyancy of underground structures represented by subway stations were selected:peak ground acceleration,duration,relative density of sandy soil,thickness of liquefiable soil beneath subway stations,buried depth of subway stations,width to height ratio of subway stations,and depth of underground water table.Through the establishment of finite element numerical model,the above influencing factors were analyzed respectively,and the seven influencing factors were used to establish the bayesian network evaluation model applicable to the seismic liquefaction hazard of the site containing underground structures.Performance indexes such as overall accuracy(OA),accuracy(Pre),recall rate(Rec)and F1 value were selected to verify the effectiveness and accuracy of the bayesian network model,and analyze the sensitivity of various influencing factors in the model.(3)For represented by subway station of the underground structures in the process of encounter earthquake effect,may cause the earthquake liquefaction buoyancy problem,analyzes the setting steel sheet pile wall around the subway station,drainage layer of gravel soil and improve cut off wall and the improvement of the floatation response of the seismic measures of gravel drainage layer,compared the various measures of buoyancy effect,and analyzes its working mechanism.The analysis results show that both the gravel soil layer and the truncated wall around the subway station can inhibit the floating of the subway station to a certain extent,and the improved gravel soil layer and the truncated wall have better anti-floating effect than other similar methods. |
