Application Of Statistical Pattern Recognition On Structural Damage Identification Based On The Time Series Model Method

In recent years,the study on probabilistic damage identification methods arouses relevant scholars' concern increasingly,both at home and abroad.In the methods mentioned above,the damage identification method based on probabilistic statistic theory is rather promising,which is worthy to be further studied.This dissertation concentrates on the structural damage identification through applying statistical pattern recognition and time series model,based on probabilistic statistic analysis.In the framework of statistical pattern recognition,extract structural acceleration data and establish the time-series model,and then use parameters of time series model to construct damage sensitive features and damage extent indice.At last,identify damage of structure through statistic discriminance.In this dissertation,a numerical example and standard damge test are used to validate the feasibility of the proposed method and indicators.The results imply that the method and indicators have the ability to detect damage,locate damage and quantify damge,and have a certain ability to resist noise.The main content of paper can be concluded as follows:(1)Discuss the steps to build time series model,select the appropriate model and methods about parameter estimation,determining model order and test residual error of the model.Some Studies show that the parameter of model is closely related to characteristics of structure and can indicate the change of structural inherent characteristics,based on which two damage sensitive features are proposed.Besides,damge extent indice are construct based on the features to quantify damage extent.(2)Combining with the actual time series data,discuss detailedly the selection of the key indicators in the process of structural damage identification,such as the selection of model,the determination of model order,sample length and the overlap rate of data samples.In addition,the impact of these factors on the final recognition result is also discussed.Secondly,assess the ability of damage sensitive features and damage extent indice to identify damage in the case of simulation conditions.Besides,test the robustness of the features to noise.(3)Through standard damage test,examine the ability of damage sensitive features to detect damage and assess the recognition effect of the indices to quantify and locate damage.Meanwhile,assess the ability of these indicators to identify the damage of different forms and the damge occurs at two places at a time,so as to lay a foundation for the application in the engineering practice.