In-life Bridge State Evaluation Based On Statistical Pattern Recognition Technique For Monitoring Information

With the bridge structural health monitoring technology is widely used,the existing bridge structure system response signal analysis and structure behavior evaluation based on statistical pattern recognition theory has been paid more and more attention.In this paper,based on statistical pattern recognition theory,to explore the internal relations between existing bridge structure response signal and the structure of the system,analyze the behavior patterns characteristic of bridge structure system and the mechanism of production,to get the characteristic parameters of characterize the structural state of the evolution by modeling the system,building damage sensitive indicators to get the differences between the feature vectors in order to obtain evolution patterns of the structure state,combined with laboratory scale model and ASCE Benchmark structural model to make experimental verification,the formation of a state based on the assessment of existing bridges theory technology system and algorithm flow in statistical pattern recognition.The main contents include:(1)Analysis statistical characteristics of bridge monitoring information.Combined with information,statistics theory and technology,analysis the statistical characteristics of the bridge structure dynamic real-time response information from the perspective of signal and system,the results show that the structure acceleration information with high-frequency has obvious characteristic features of strong background noise,strong randomness and nonstationary,nonlinearity.(2)Analysis bridge monitoring information statistical pattern and its feature recognition method.Introducing timing analysis method to establish the system model with bridge output response information,statistical models was constructed by the use of the ARMA model and the regression coefficient was extracted as feature vector to evaluate the structural state.The ARMA model combined with structural dynamic system and the effectiveness of the model parameters and structural state of theoretical arguments timing models to characterize the structure of the system evolution of internal relations;binding long autoregressive model method and generalized least squares analysis model order and parameter estimation accuracy by the regression function from the applicability of the model characterization of the real system to verify and improve the model to characterize the accuracy of information evolution model;according to the parameters of the model identification for characterization of the evolution of structure state eigenvector,implements the monitoring data to the feature vector information together.(3)The correlation analysis between bridge monitoring information statistical models feature and structural states.Introducing the theory of statistical pattern recognition on the basis of constructing the index of the characteristics which characterization the structure state,by building damage sensitivity index for different stages of the differences between the characteristics of characterization of structure state evolution pattern.On the one hand,the structural damage sensitive indicators was constructed based on the Mahalanobis distance,by setting the confidence interval for structure condition of threshold,overcome the complex calculation defects of traditional threshold method,through the sensitivity to damage index between control threshold value to obtain the distribution of structure state evolution pattern;on the other hand,introducing the statistical process control theory mean control chart and construction T statistics,through multiple hypothesis testing to construct a confidence interval as the structural condition of threshold,using statistical mode difference index distribution both inside and outside the threshold to identifying the state of the structure evolution.Finally the scale model structure and ASCE Benchmark model experiment was carried out for experimental analysis under ambient excitation to provide experimental verification of sensitivity of the index and the algorithm program,the experimental results verify the ability of the damage sensitive indicator for structural damage recognition is stronger and sensitive for minor injury and has certain sensitivity position.