Research On Safety Monitoring Method Of Concrete Arch Dam Deformation Based On Bayesian Inference

To evaluate the true safety status of the structural deformation of a concrete arch dam during operation,this paper conducts research from three perspectives:dynamic monitoring model,spatiotemporal monitoring model,and parameter inversion.Bayesian inference theory is used to handle the uncertainty information in the establishment of the monitoring model and the parameter inversion process,to explore the spatiotemporal information in the deformation monitoring data,and to improve the rationality of the positive and inverse analysis results of the arch dam deformation monitoring data.Therefore,the basic theory of structural deformation safety monitoring during the operation period of concrete arch dams is developed,providing technical support for ensuring the long-term safe service of arch dams and scientific decisionmaking for dam safety management departments.The main work of this paper has the following aspects:(1)In order to address the time-varying and uncertain nature of concrete arch dam deformation,a dynamic safety monitoring model for arch dam deformation is proposed using the Bayesian dynamic linear model(BDLM)to leverage its dynamic modeling capabilities and interpretability advantages.This model can update in real-time and quantify model errors and prediction uncertainties.The proposed arch dam dynamic safety monitoring model is validated based on measured data from a specific arch dam,and the generalization ability of the model under different component expression forms is compared and analyzed within the frameworks of the hydrostatic-seasonal-time(HST)model,hydrostatic-seasonal-temperature-time(HSTT)model,and hydrostatic-seasonal-state(HSS)model.In order to diagnose the safety status of arch dam deformation,the non-convergence of the deformation ageing component separated from the dynamic safety monitoring model is proposed as a criterion for identifying the degree of convergence or divergence of the ageing deformation,and an engineering example is used to validate this approach.(2)Aiming at the strong multi-dimensional spatiotemporal evolution law and distribution characteristics of concrete arch dams,a spatiotemporal monitoring model of arch dam deformation that comprehensively considers spatiotemporal similarity and model optimization is proposed to accurately evaluate the deformation safety state of arch dam structure.The multiindex clustering method of panel data is used to divide and cluster the deformation measurement points at different positions of the arch dam body to distinguish the deformation law of different areas of the dam.The principal components analysis(PCA)method was used to extract the comprehensive displacement of multiple measurement points of the deformation behavior of each partition to extract the uniform deformation law of each measurement point in the partition.Then,considering the uncertainty of the model,Bayes model averaging(BMA)is averaged on the set of regression models.The results of engineering examples show that the Bayesian model can give robust and effective prediction results on the deformation of arch dams on average.The results of zonal deformation mechanism analysis show that the deformation of the arch dam body of the project conforms to the general law,and the contribution of the temperature component of the arch dam during operation is mainly reflected in the middle of the dam body,and shows a certain lag effect,and the contribution of the aging component is mainly reflected in both sides of the dam shoulder.(3)Aiming at the time-varying mechanical parameters of concrete arch DAMS,a mixed model is established by using BDLM framework to obtain the time-varying adjustment coefficients of water pressure components,and the actual time-varying parameters of DAMS are retrieved by using the time-varying adjustment coefficients and structural calculation parameters.Aiming at the spatial property of mechanical parameters of concrete arch dam,a method of subdivision parameter inversion based on PCK proxy model and Bayesian inference is proposed.The case study shows that the time-varying parameter inversion based on BDLM framework can well reflect the time-varying and spatial property of the dam elastic model,and the zonal inversion can better solve the influence of the disturbance of the measuring point position on the inversion results.