Research On Algorithm Of Damage Diagnosis Model For Cable-stayed Bridges Based On Data Mining

The data collected during the health monitoring of the cable-stayed bridge contains many important information.The data mining technology can effectively mine these information,and the damage diagnosis of the cable-stayed bridge can be carried out accordingly.The identification effect of cable-stayed bridge damage diagnosis is closely related to the selected data mining algorithm.Based on the requirements of damage location and quantification for cable-stayed bridge damage diagnosis,this thesis selects a variety of data mining algorithms for algorithm research,trial calculation and comparison,in order to obtain more efficient data mining algorithms suitable for cable-stayed bridge damage diagnosis.The main research work of this thesis is as follows :(1)The calculation formula of curvature mode is sorted out and the calculation program is compiled by python language.The structural damage diagnosis process and damage index selection principle based on data mining are summarized,which provides a basis for the damage diagnosis method using curvature mode as the attribute of data mining algorithm.(2)The code of four algorithms of support vector machine(SVM),K-nearest neighbor,decision tree and multi-layer perceptron neural network is compiled by python language.Data mining experiments are carried out in classification and regression data sets.The performance of the algorithm is tested by experiments and the best performance algorithm is selected.(3)The mechanism of the Reptile Search Algorithm(RSA)is analyzed.In view of the shortcomings of the algorithm,such as too fast population aggregation in the late iteration,easy to lead to local optimization and low optimization efficiency.On the basis of RSA,the Bernoulli shift chaotic map is introduced to improve the population initialization formula,the iterative update formula is improved by using the timevarying nonlinear adaptive weight,and the Cauchy and Gaussian mutation strategies are added to realize the adaptive mutation of the population space.An Improved Reptile Search Algorithm(IRSA)is proposed.By comparing the convergence curves of IRSA and RSA on 13 benchmark functions,it can be seen that IRSA has stronger optimization ability.(4)Based on the above research,a damage diagnosis method of cable-stayed bridge based on IRSA optimized SVM is proposed.In this method,two damage indexes are defined according to the requirements of damage location and quantification.The influence parameters of SVM in damage location and quantification are optimized by IRSA,and the optimized SVM is used to complete the damage diagnosis of cablestayed bridge.The proposed method is applied and verified by a cable-stayed bridge example,and compared with three damage identification methods of SVM,RSA-SVM and SMA optimized SVM.By comparing the identification results of the four damage identification methods,it is found that the results of IRSA-SVM identification are closer to the actual damage.