Slope Engineering Safety Monitoring Method Based On Bayesian Theory And Spatio-Temporal Clustering

With the large-scale construction of water conservancy and hydropower projects in China,many high slope projects of water conservancy hubs have emerged.Slope safety monitoring can reflect the real safety state of slopes during construction and has become an important method to judge the safety and stability of slopes.It is urgent to explore and study the safety monitoring method of the high slope construction period,provide timely and accurate prediction information,and avoid the loss of life and property caused by slope disasters.Due to the small amount of deformation monitoring data during the slope construction period in the traditional modelling process,there are no environmental monitoring data,and the spatiotemporal information of the monitoring data is not deeply excavated.Therefore,this paper takes the safety monitoring data of a slope during the construction period as an example,establishes a safety monitoring model of slope deformation and prestress during the construction period according to Bayesian theory and the spatiotemporal clustering method,and develops safety monitoring data analysis software for slopes during the construction period on this basis.The main research contents and results are as follows:(1)The characteristics of displacement and prestress changes during slope construction are comprehensively analysed from the aspects of time and space variation,eigenvalue statistics and deformation magnitude,and the main influencing factors of slope deformation and prestress changes are analysed in combination with the construction process.The analysis of monitoring data shows that excavation disturbance and weak structural planes are the main factors leading to slope deformation and prestress change.Slope deformation is mainly controlled by large structural planes.With the gradual implementation of slope reinforcement measures,the slope has basically stabilized.(2)Taking slope displacement monitoring data as an example,a slope deformation safety monitoring model based on the Bayesian vector autoregressive(BVAR)algorithm is proposed.The Minnesota prior distribution is used to solve the problem of too many model parameters,and the joint posterior distribution of the model is obtained by Gibbs sampling.Finally,the model prediction value and confidence interval are obtained.The study of engineering examples shows that the residual sum of squares,mean absolute error and root mean square error of the BVAR model are smaller than those of the vector autoregressive(VAR)model,indicating that the BVAR model has higher prediction accuracy and a better interval prediction effect.(3)Taking the prestressed monitoring data of a slope as an example,a prestressed safety monitoring model based on the Ward clustering method and Bayesian panel vector autoregressive(BPVAR)algorithm is proposed.The Ward clustering method is used to divide the prestressed time series and spatial measuring points.The BPVAR model is used to fit and predict the prestressed panel data.The research on the engineering example shows that the prestressed monitoring sequence is divided into two periods,and the spatial measuring points are divided into four categories.The multiple correlation coefficient between the fitting value and the measured value of the BPVAR model is above 0.90,and the prediction error is less than that of the vector autoregressive model,the autoregressive moving average model and the long-term and short-term memory neural network model.The measured prestress values are all within the 95%confidence interval,which can provide a reference for the safety state discrimination of slope engineering.(4)Using the MATLAB App Designer design tool,combining the Bayesian vector autoregressive model,Ward clustering method and Bayesian panel vector autoregressive model proposed in this paper,the safety monitoring data analysis software of the slope construction period is developed.Research on engineering examples shows that the software can improve the efficiency of slope safety monitoring,make the analysis of monitoring data more convenient and intelligent,and has strong portability and applicability.