Deformation Monitoring And Safety Warning Of Concrete Dam

In this paper,the signal decomposition theory,clustering analysis method,time series model,extreme value theory etc.are used to study the deformation monitoring methods of concrete dams.A concrete dam deformation evolution trend analysis method is proposed,a deformation monitoring model and a deformation warning index for multiple measuring points in concrete dams space are established.And a safety assessment of the deformation state of a concrete dam is carried out based on an engineering example.The main research contents are as follows:(1)The nonlinear and nonstationary deformation monitoring sequence of concrete dams is decomposed into several subsequences with different frequencies by the variational mode decomposition(VMD).The physical characteristics of subsequences are represented by the dispersion entropy(DE)and maximum mutual information coefficient(MIC).Analysing the development trend and spatial-temporal distribution law of the vertical displacement of a concrete dam during the long-term deformation evolution process.A concrete dam deformation prediction model based on long shortterm memory(LSTM)network,and the deformation state of the dam is diagnosed.(2)In view of the correlation between different measuring points of concrete dams deformation,a composite prediction model of concrete dams deformation based on hierarchical clustering and improved panel data is established.By using the agglomerative hierarchical clustering(AHC)method based on Ward,the measuring points with similar change law are divided into one deformation area,and different panel data models are established for different deformation areas.According to the idea of residual correction,the VMD-ARIMA method is used to deal with the residual series of the panel data model.The prediction value of reconstructed residual sequence is superimposed with panel data model to construct a composite prediction model.(3)Based on considering the correlation of the measuring points of concrete dams spatial deformation field,combined with the ellipse fitting method and the extreme value theory(EVT),a method for formulating the deformation early warning index of concrete dams space with multiple measuring points is proposed.Through principal component analysis(PCA),the measured values of several deformation related points are mapped into two-dimensional space.The peaks over threshold(POT)model in the EVT is used to determine the early warning point.The initial elliptic equation of the data points distribution is fitted by the least square(LS)method.And the early warning point is used to modify the initial elliptic equation to obtain the warning boundary of spatial deformation of the concrete dam.