Inversion Analysis Of High Concrete Dam Based On Swarm Intelligence-agent Model

China is a big country for building dams.The safety of dams is an important guarantee for the safety of people’s lives and property.With the increase of the number of high concrete dam in China,the inversion of the comprehensive mechanical parameters of the dam and foundation has become a research hotspot.In this paper,based on the direct method and three optimization algorithms and combined with Gaussian process regression response surface model,the elastic modulus of high concrete dam in static state is analyzed by inversion.The main research contents are as follows:(1)The research background and development status of dam inversion analysis are summarized.Jaya algorithm,Particle Swarm Optimization(PSO)and Gray Wolf Optimizer(GWO),as popular optimization algorithms,are widely used in engineering optimization problems due to their characteristics of less adjustment parameters and obvious optimization effect.Since the traditional inversion analysis needs to be combined with the finite element calculation,it will greatly increase the time required for the inversion.In this paper,Gaussian process regression response surface model(GPR-RSM)is adopted to replace the traditional finite element calculation.This method not only greatly reduces the calculation time,but also meets the requirements of computational accuracy.(2)The so-called GPR-RSM is to establish the mapping relationship between elastic modulus and displacement,the basic content of GPR,RSM as follows: the first to use Latin hypercube sampling in every increment of elastic modulus under regulation range respectively to extract elastic modulus as a certain number of training samples and test samples of the elastic modulus,the starting of the number of training samples and test samples of 15 D,and then extract the modulus of elasticity of input in the finite element to calculate the displacement of monitoring points value,the training sample and test sample is by the modulus of elasticity and its corresponding monitoring displacement values.Will training samples for training in the GPR,elastic modulus and displacement to establish the mapping relationship,then the training sample and test samples to test the accuracy in the GPR model,and because of the increase in the number of samples,the training sample and increasing the accuracy of the test sample,when sample increases to a certain number,its precision changed little basic can be ignored,the final will be the training samples of Gaussian process regression response surface model instead of the traditional finite element computing applications on the inversion analysis of high concrete dams.(3)Two high concrete dams,one roller compacted concrete gravity dam and one concrete arch dam are selected as examples.The three optimization algorithms and the Gaussian process regression response surface model were used to analyze the elastic modulus inversion of two calculation examples respectively in the static state.Finally,the performance and rationality of the three algorithms and the GPR-RSM model applied in the inversion analysis of high concrete dams were evaluated based on the elastic modulus inversion results.