Research On Structural Damage Identification Based On Global And Local Timing Analysis

With the development of economy and society,super high-rise and long-span structures are gradually increasing,which makes structural condition assessment and damage identification technology more and more important.Combined with structural health monitoring and damage identification technology,the changes in physical parameters and mechanical behavior of the structure during service can be detected in time,and early warning will play an important role in the operation and maintenance of the structure.With the vigorous development of sensor technology,computer science and other fields,the structural health monitoring system has collected a large amount of structural data information.How to use massive monitoring data for structural damage identification and safety assessment has become a research hotspot in this field.Based on the time series analysis method,this paper starts from the overall and local characteristics of the time series data,and carries out the relevant research on the structural vibration response data.The main research contents and results are as follows:(1)Based on the overall time series analysis,this paper uses a linear autoregressive moving average(ARMA)model and a nonlinear generalized autoregressive conditional heteroscedasticity(GARCH)model to identify relevant structural damage.Research,establish the corresponding linear damage index(Standard Deviation of Residuals,SDR),nonlinear damage index(Standard Deviation of Innovation,SDI)through simply supported beam finite element model and wooden truss The bridge laboratory model verifies the feasibility of the holistic time series analysis method.The research results show that: for the simply supported beam model,the SDR index based on the ARMA model can identify the larger damage of the structure,and the SDI index based on the GARCH model can identify and locate the larger damage;while for the small damage case,the two damage indexes show not good.For the wood truss bridge model,both the SDR and SDI indexes can identify large structural damage and locate the damage location;however,the two indexes are not sensitive to small structural damage.(2)Based on the local time series analysis,in view of the insufficient ability of the ARMA model and the GARCH model to identify small damage,the Shapelet Transform algorithm was used to extract the local features of the time series data,and the local time analysis of the simply supported beam model and the wooden truss bridge model was carried out.Sequence Analysis Research.The research results show that:for the simply supported beam model,when using Shapelet Transform method to extract Shapelets,the highest information gain(Information Gain)value is 0.92.The classification accuracy of damage identification is 96.67%,and the identification accuracy of small damage is 100%.For wood truss bridge structures,the highest IG value is 0.63 when Shapelets are extracted using the Shapelet Transform method.The classification accuracy of damage identification is 100%,and the identification accuracy of small damage is 100%.It can be seen that the local time series analysis method based on Shapelet Transform can effectively extract local features representing the structural state information and accurately identify small structural damage.(3)In view of the efficiency of Shapelet Transform extraction,this paper combines the Largest Triangle Three Buckets(LTTB)dimensionality reduction method to first perform LTTB dimensionality reduction processing on the time series data of the simply supported beam model and the wooden truss bridge model.Then perform Shapelet Transform analysis.The research results show that for the simply supported beam model,the highest IG value is 0.66 when filtering the Shapelets of the time series data after LTTB dimension reduction.The classification accuracy of damage identification is 93.3%,and the accuracy of small damage identification is 100%.The computational cost of Shapelet Transform to extract Shapelets is reduced to 25.3% of the original.For the wooden truss bridge model,when filtering the Shapelets of the time series data after LTTB dimension reduction,the highest IG value is 0.41.The classification accuracy was 99.2%.The computational cost of Shapelet Transform to extract Shapelets is reduced to 22.9% of the original.It can be seen that the Shapelet Transform method based on LTTB dimensionality reduction can effectively ensure the accuracy and reduce the computational cost.