Structural Damage Identification Based On Time Series Analysis

For the modern engineering structure,the requirement of structural safety exists in the whole structural life cycle.In order to ensure that the structure will not be damaged suddenly during the operation period,resulting in life danger and economic loss,the evaluation of engineering structure state is particularly important.In recent years,structural damage identification has been a research hotspot in structural state assessment.In order to ensure the applicability of the damage identification method for different models,as well as the accuracy of the identification.Based on the time series analysis method,this thesis introduces the concept of neural network and statistical analysis theory to study the damage identification methods for different models and problems,There are the main contents:(1)In order to verify the damage identification method of neural network,a five degree of freedom spring damping system without noise effect and a benchmark model with 5% noise effect are used.The time series method is used to parameterize the acceleration response signal of the structure,and the autoregressive coefficient matrix of the model is extracted as the input of neural network.The analysis results show that the method can accurately identify the damage location information of these two different models and low level noise interference have no obvious effect on the recognition results.The method is applicable to different models.(2)For complex structures that are difficult to model,it is impossible to construct known damage states in advance.For this kind of damage identification problem,based on ASCE benchmark model and Dongjin station model,time series analysis and probability statistics theory are combined.According to the modeling results of time series analysis of a large number of samples,the damage index based on the first three model coefficients is constructed,and the probability density function of the index is fitted.For the identification of different damage conditions,the probability estimation results are given.It is found that with the development of modern sensor technology and signal processing technology,using a large number of measured data,the method of structural damage identification with statistical method has a great development space.