Research Of Arch Dam Structural Damage Diagnosis Using Statistical Pattern

In order to study structural damage diagnosis method for arch dam and realize non-destructive online health detection and safety evaluation,this paper proposes a arch dam structural damage diagnosis method based on statistical pattern recognition.The method is based on the statistical pattern recognition techniques,combined with the ARMA time series model,feature extraction model of principal component analysis method and work modal parameter identification method based on HHT techniques,to detect arch dam structure damage.This method is essentially the status identification of structure,to reduce differences and highlight is to distinguish differences between different state structure damage.Its process includes data acquisition,setting up time series model of ARMA based on acquisited vibration data,extracting the characteristic parameters of the model by using principal component analysis method,keeping enough features of the corresponding data at the same time compressing it to the first steps of principal component,at last selecting several proper principal component and diagnosing the damage by establishing statistical indicators,most of all,the latter three steps are focus in diagnosing method.Based on this,the model order,model validation,model parameter estimation,the characteristic parameters reduction and statistical indicators selection is deeply studied,which is tested by Damage simulation calculation of finite element numerical simulation model and measured dates of damaged plaster arch model test,it shows that this method can effectively and accurately diagnose the damage of arch structure,besides identify injury severity and location of the structure sensitively.For this damage detection method,data analysis process is simple,the measured data have strong noise immunity to the environment,above all it has important theoretical and practical engineering value.The main work of this study of damage detection methods are as follows:1.Select ARMA time series model as the basic model to do modeling,determine the order of ARMA model by AIC criterion and verify the validity of the model by compared with other structural modal parameters identification method.2 Take part AR model parameters as the model feature extraction,extract and compress it by principal component analysis.In addition,take several principal components which most of the information for study,which is based on principal component contribution rate after dimensionality reduction.3 Finite element simulation,conduct acceleration response data of derived measuring point by diagnostic methods described in this paper,furthermore,structural damage condition can be diagnosed.4 Conduct plaster model damage vibration test under ambient excitation conditions,and analyse acceleration response data of measuring point by this method,furthermore,damage information of arch structure can be diagnosed.