Research On Inversion Method Of Arch Dam Material Parameters Considering Uncertainty Of Material Parameters And Monitoring Information

Concrete arch dams have been widely used in hydropower development in western China due to their high economic efficiency,good seismic performance,and strong overload capacity.Due to the importance of dam engineering safety,it is of great significance to study the operational and safety behavior analysis methods for concrete arch dams.The parameters of dam concrete and bedrock materials,as indispensable parameters in the finite element calculation of arch dams and important indicators for determining their operational status,play an important role in accurately assessing the safety status of arch dams.Therefore,this paper takes a concrete arch dam as the research object and adopts research methods such as prototype observation,numerical simulation,sensitivity analysis,and probability statistics to study the inversion method of arch dam material parameters.The main contents are as follows:(1)To solve the problem of reasonable selection of monitoring points in the inversion of arch dam material parameters,the finite element method and sensitivity analysis are combined to accurately analyse the material parameter sensitivity of deformation at different parts of the arch dam.A numerical calculation model for arch dam deformation analysis is established based on the finite element method.The Latin hypercube sampling method is used for experimental design.The sensitivity of the material parameters of arch dam deformation is analysed using the Grey Relational Grade.Determining the material parameter sensitivity and its spatial distribution of dam deformation provides a theoretical basis for the selection of parameters in material parameter inversion and the optimization of displacement monitoring points.(2)Aiming at the lack of consideration of various uncertainties in traditional inversion of arch dam material parameters,a Bayesian inference-based inversion method for arch dam material parameters is proposed.To improve the computational efficiency of material parameter inversion,the Latin Hypercube Sampling method is used to generate a set of material parameter samples,and the dam displacement corresponding to different parameter combinations is calculated based on finite element forwards analysis.Then,the particle swarm optimization method is used to optimize the Long Short-Term Memory neural network to establish a surrogate model for the relationship between material parameters and arch dam displacement.On this basis,DiffeRential Evolution Adaptive Metropolis is used to achieve random inversion of arch dam material parameters based on Bayesian inference theory.The updated inversion ability of Bayesian inference is used to track the changes in arch dam material parameters during operation.(3)Because single measurement point inversion results cannot effectively represent the overall parameters of the arch dam,based on the results of the material parameter sensitivity analysis,high sensitivity measurement points are selected as the research object.Particle Swarm Optimization is used to optimize the hyperparameter of the long-and short-term memory neural network,improve the calculation accuracy of the multi-input multioutput surrogate model,establish a joint likelihood function for multiple measurement points,and perform inverse analysis of material parameters under the monitoring information of multiple measurement points.It has been proven that using only sensitive monitoring points can also achieve accurate inversion analysis of material parameters and better than single point.(4)Considering the influence of the assumed prior distribution of material parameters and the artificially determined likelihood function forms on the inversion results in Bayesian inference inversion,this paper summarizes the commonly used prior distribution and likelihood function forms of parameters in various geotechnical engineering problems.Taking an arch dam as an example,the influence of the prior distribution and likelihood function forms of material parameters on the posterior distribution of material parameters is systematically studied,which provides a reference for the selection of prior distribution and likelihood function forms of material parameters in arch dam inversion problems.