Research On Damage Identification Of Civil Structure Based On PSO-SVM

With the rapid development of modern building technology,how to monitor the health and safety state of civil engineering structures and ensure their safety and reliability has become a concern.Although the traditional damage identification method based on vibration test develops rapidly,it is difficult to measure the structural excitation information accurately and the error is large.Therefore,from the point of view of solving the problem of learning based on samples,this paper puts forward a method of structural damage identification based on PSO-SVM combined with big data intelligent algorithm,and the main research work is as follows:Firstly,this paper introduces the present development,basic theory and methods of structural damage identification at home and abroad.Secondly,a damage identification method based on PSO-SVM is proposed,which solves the uncertainty of classification and regression performance when traditional support vector machine(SVM)selects parameters manually,and adopts particle swarm optimization(PSO)instead of manual selection of optimal SVM parameters to achieve its best classification and regression performance.Thirdly,a damage identification method based on PSO-SVM and curvature mode is proposed,and the theory of structural curvature mode difference damage identification is deduced.Using the numerical simulation analysis of simply supported beam,the curvature mode difference(sample)between single damage and multi-position damage is obtained,and the sample set(test set and training set)of the learning problem is formed.The unoptimized SVM and PSO-SVM are applied to the sample set respectively to predict the damage location and degree of simply supported beam.By comparing the results of the two damage identification methods,it is found that the method based on PSO-SVM curvature mode difference has a better accuracy.Fourthly,a damage identification method using the natural frequency change rate as the SVM feature vector is proposed,and the damage identification theory of natural frequency change rate is derived.A four-story frame structure is simulated and analyzed to obtain the natural frequency change rate(sample)of the structure under typical damage conditions,and the sample set(test set and training set)of the learning problem is formed.The unoptimized SVM and PSO-SVM methods are applied to the sample set respectively to predict the location and degree of damage of frame structure.By comparing the results of the two damage identification methods,the superiority of the damage identification method based on PSO-SVM and curvature modal difference is verified.Finally,the main work and research results of this paper are summarized: Based on the traditional SVM damage identification,the method of using PSO to replace the manual selection of SVM parameters is proposed,and combined with the vibration parameters of civil structure,it is proved that this method is more efficient and accurate than the traditional SVM method.