Dynamic Analysis And Prediction Of Dynamic Response Of Earth-Rock Dam Considering Spatial Variability Of Overburden Material

The uncertainty of the material parameters of the overburden foundation often significantly affects the seismic response of the asphalt concrete core dam above it.This paper combines the spatial random field simulation technology with the finite element method to realize the random field simulation and"non-intrusive"random finite element calculation of the spatial variability and correlation of the covering material based on the Monte Carlo method.Taking the vertical permanent deformation of the dam crest as the index of seismic damage classification,the seismic vulnerability analysis of the asphalt concrete core earth-rock dam was carried out,and the influence of the spatial variability of the material parameters of the overburden foundation on the probability of the dam settlement classification failure was summarized.Finally,based on the XGBoost machine learning algorithm,the non-linear mapping relationship between static and dynamic material parameters of the overburden foundation,ground motion parameters and the vertical permanent deformation of the dam body was established,and the coefficient of variation of multiple material parameters of the foundation and any combination of seismic waves were realized.Permanent deformation prediction.The details are as follows:(1)Based on the orthogonal decomposition experiment,the sensitivity of each parameter of the equivalent linear viscoelastic constitutive model to the response of the dam body acceleration and permanent deformation is analyzed,and the random parameters of the dynamic constitutive model in this paper are selected.The normal or lognormal distribution function is used to consider the spatial variability of the material parameters of the overburden,and the Gaussian autocorrelation function is used to consider the spatial correlation,and a Gaussian spatial random field discrete method based on the Cholesky covariance decomposition method is established.The secondary development of ABAQUS based on python language assigns the random field to the finite element model,and realizes the random field simulation and"non-intrusive"random fi nite element calculation of the spatial variability and correlation of the cover material based on the Monte Carlo method.(2)Three different seismic waves were selected and 7-level amplitude modulation was carried out according to the peak acceleration,considering the random static parameters of the overburden foundation material,the random dynamic parameters and the static and dynamic parameters at the same time,and the three working conditions were randomly selected from each working condition.Random parameters of the materials were grouped and discretized in space,3150 times of non-linear seismic response analysis of the overburden-asphalt core earth-rock dam was carried out.The probability of exceeding the standard,probability density distribution and graded failure probability of the vertical deformation of the dam under different seismic waves and working conditions are statistically analyzed.Taking the vertical permanent deformation of the dam crest as the seismic damage classification index,based on the analysis of the average vertical deformation and the coefficient of variation,the seismic vulnerability of the asphalt concrete core earth-rock dam considering the spatial variability of the overburden foundation material parameters is drawn.Sex curve.(3)According to the 7 random parameters in the static and dynamic constitutive model of the overburden,three factor levels are selected for each parameter,and 27 sets of parameter variation combination conditions are designed.Each set of conditions is randomly sampled 10 times,combined 3 different seismic waves and 7-level peak acceleration amplitude modulation,a total of 5670 times of overburden-asphalt core earth-rock dam nonlinear seismic response finite element analysis was carried out.The mean value and coefficient of variation of the vertical deformation of the dam body obtained in each calculation are counted as the XGBoost model training,learning and testing samples.The XGboost model is constructed and the model parameters are optimized and adjusted.Based on the weight analysis of the influence of various random parameters on the vertical deformation,the non-linear mapping relationship between static and dynamic material parameters of the overburden foundation,ground motion parameters and the vertical permanent deformation of the dam was established,and multiple material parameters of the foundation were realized.Permanent deformation prediction under any combination of mutation and seismic waves.