Research On Damage Identification Method Of Wharf Frame Structure Based On Support Vector Machine

China’s port construction industry is booming,but after decades of operation,a large part of the built port buildings have been damaged,which has varying degrees of impact on the structural safety,so the exploration of port buildings damage identification is a hot spot in the field of engineering.Support vector machine(SVM)is a kind of model based on traditional machine learning algorithm,which is often used to solve dichotomy problems.In this context,a wharf frame structure model is established in this paper.Combined with the damage identification method of support vector machine,the damage location and damage degree of wharf frame structure are identified.The main research work is as follows:(1)The finite element model of 2-storey 2 × 2 span wharf frame structure was established by using ANSYS finite element software,and two kinds of finite element models were set,including the non-destructive finite element model of wharf frame structure and the lossy finite element model of wharf frame structure.Firstly,modal analysis of wharf frame structure is carried out to extract dynamic characteristic parameters.Then the transient dynamic response of the structure is analyzed to extract the dynamic response value of acceleration in the force direction.The natural frequency values and period values under respective damage conditions were compared,and the dynamic response characteristics of wharf frame structures under lossy and nondestructive conditions,as well as under different damage degrees were compared.(2)Based on the theory of wavelet packet analysis,the extracted acceleration dynamic response values were decomposed into three layers of wavelet packet,and the damage index was constructed.Considering the damage of beam and pile positions,the damage index--energy change rate under different damage degrees was calculated.By comparing and studying the same damage sites of damaged structures and nondestructive structures,different damage sites under the same damage degree,and different damage degrees at the same damage site,the numerical law of the change rate of nondestructive energy and damage energy in these cases is found to reflect the internal damage law of structures.It is feasible to input the damage index into the support vector machine model.(3)After extracting the energy change rates under various damage degrees and taking them as sample data pretreatment,the SVM models on the beam and the pile are respectively established by using the SVM theory,which are SVM-Z(support vector machine on the pile)and SVM-L(support vector machine on the beam),for damage identification of wharf frame structure.Study compared in three different kernel functions: polynomial kernel function,linear kernel function,RBF kernel function is under the effect of damage identification based on support vector machine,in order to determine the optimal kernel function,combined with suitable to identify the damage parameters,identify the damage when using support vector machine(SVM)classification model,framework of wharf structure damage location identification of duan;Support vector machine(SVM)regression model is used to identify the damage degree of wharf frame structure.(4)When identifying the location and extent of structural damage of wharf frame structure,it is inevitable that there will be various errors,such as environmental noise and model error.In real life,noise is inevitable and directly affects the accuracy of damage identification results.By adding different degrees of white Gaussian noise,the damage identification effect of support vector machine considering noise is discussed.