Structural Damage Identification Based On Time Series Model And Principal Component Analysis

Structural damage detection based on the time series analysis has been continuallyperfected and developed in recent years as an entirely new research method. Howeverthere are potential problems on its application. Its application for simple structure such assimply supported beam has been investigated deeply, while the application for complexstructure such as framework and space truss is only the beginning. Time series approachwas always used in state prediction while it was seldomly used in damage location. Thelinear stationary time series were widely used for damage detection while the use ofnonlinear time series was less. Aiming at above problems, according to the feature offramework and space truss, a further research on the application of time series analysis inthe damage detection will be discussed, and the main contents of this thesis are as follows.(1) The state-of-art and development trend of structural damage detection werereviewed in this paper. The advantages and disadvantages of different kind of methodswere summarized firstly. This paper focused on the basic principle of time series analysisin structural damage detection according to the internal relations between time seriesmodel and the structural system.(2) The modeling process of time series was investigated deeply. The main contentsare the analysis of excitation system, structure response data collection and preprocessing,estimation of model parameters, determination of the model order, and so on. The basicprinciple and application background of principal component analysis in structural damagedetection were explained in detail.(3) Structural damage detection based on time series and principal componentanalysis method was proposed in this paper. The parameters and RMS errors of time seriesmodel were used as damage sensitive feature for structural damage detection respectively.The damage detection matrix has numerous dimensions and contains a lot of overlappinginformation. The principal component analysis is used to reduce the dimension of thedamage detection matrix in order to simplify the calculation and to explain the effective information. According to the limitations of AR and MA models, based on the series ofthe experimental data obtained from framework structure of the Los Alamos NationalLaboratory, using the ARMA model to verify the availability of the method based on timeseries and principal component analysis.(4) The numerical simulation was performed for a double-layer space truss in damagedetection using time series and principal component analysis. The structural finite elementmodel was built with ANSYS procedures, and the acceleration data of each node under thewhite noise excitation was obtained. Then the ARMA model was built using MATLABLanguage, considering the change of acceleration response under different level influencesof noise. The study of the research shows that the damage sensitive feature in damagedunit nodes is much greater than the undamaged ones. In addition, to further validate thefeasibility and availability of the method.(5) Based on the series of the experimental data from framework structure of theLos Alamos National Laboratory, the GARCH model is presented to solve somelimitations which display in structural nonlinear damage detection using ARMA model.Comparing with this two methods, the GARCH model showed its superiority in nonlineardamage detection for structure damage location and resistance to environmental factorinfluence. The calculation process is convenient and with higher damage detectionaccuracy.In brief, the damage detection method based on time series and principal componentanalysis is suitable for complex structure such as framework and space truss. It issignificant for improving the effect of structural damage location in practice.