Research On Structural Damage Identification Method Of Overhead Vertical Wharf Based On Optimized Dynamic Kriging Model

Due to the structural type and stress characteristics of structural members,the damage and deformation characteristics of the overhead vertical wharf and the calculation process of repeatedly using the finite element model have certain complexity.In the past research on damage identification,particle swarm optimization and artificial fish swarm optimization have excellent performance in solving constrained optimization problems.However,on the basis of this advantage,this paper takes Henry gas solubility optimization(HGSO)algorithm,a meta heuristic optimization algorithm based on a certain physical principle,introduces the related theories of wavelet transform and Kriging surrogate model,and uses the idea of multi peak search to output multi parameters of EI plus point criterion function,and proposes a two parameter optimization algorithm based on Optimized Dynamic Kriging surrogate model Taking Chongqing Guoyuan Port Phase II project as an example,the damage identification method of stage structure is verified,and the application of this method in the damage identification of overhead vertical frame wharf structure is realized.The main research results are as follows:(1)Compared with static and dynamic particle swarm optimization algorithm,Henry gas solubility optimization algorithm has obvious advantages such as fast convergence speed,high convergence accuracy and stable performance.Compared with wpso algorithm,hgso algorithm can identify single damage,symmetric and asymmetric damage cases under the same initial parameters,and the relative error of identification result is less than 1.7%;At the same time,it shows that the number of "suspected" damage elements affects the computational efficiency of the optimization process.A dimension reduction method based on wavelet transform is proposed to analyze the location of singular points.(2)The first-order displacement modal difference can be used as the signal function of wavelet transform analysis;The greater the damage degree is,the greater the modulus maximum is,and the more obvious the singular signal is.In the second stage,the relative error of the damage degree identification result is within 5%,which realizes the application of two-stage structural damage identification based on wavelet hgso algorithm in the setting of damage condition of simply supported beam structure.(3)The optimal Ladin hypercube distribution experimental design has spatial uniformity and projection uniformity in the design space;The adaptive elite multi peak search algorithm(AE-HGSO)proposed in this paper is verified by a two-dimensional test function.The results show that the optimization process of the multi peak search algorithm finally converges to the peak point.This method can be used to dynamically add points to the multi peak parallel multi-point adding strategy.It is feasible to optimize Kriging model by dynamically adding points,The basic implementation process of Kriging agent model dynamic optimization based on multi peak parallel multi-point adding strategy is summarized.(4)The application of two-stage damage identification method based on Optimized Dynamic Kriging model in structural damage identification of overhead vertical wharf is studied.The results show that the method is effective in the case of single damage of pile and column,the relative error of the damage degree identification results is less than 2% after reaching the convergence condition,and the minimum fitness value is1.5162e-6 in the iterative process of each set condition.For the fitting accuracy of Kriging model optimized by dynamic adding points to other test sample points in pile damage condition,the fitting accuracy is poor because there are too few training points around the damage degree below 10%.Considering the sparsity of training points in the process of dynamic adding points,the Kriging model after iteration can replace the real response of modal frequency and modal confidence of more than 10% damage degree test sample points within the allowable error range,and the accuracy of identification results is more than 80%;The prediction accuracy is the highest at the corresponding real damage degree.