Self-diagnosis And Condition Assessment Of Long-bridge Health Monitoring System

With the rapid development of our national society and economy,the health monitoring system of long span bridges(BHMS)has become an important measure to asses and manage the conditions of the bridges.At present,a variety of monitoring data has been accumulated via BHMS.However,there would be lots of abnormal data due to its characteristics that are massive,multiple dimensional,random,susceptible,etc.The existence of abnormal data will not only affect the efficiency of subsequent data analysis,but also affect the scientific evaluation of bridge state assessment.Therefore,self-diagnosis technology is imperative for BHMS to ensure the availability,safety,reliability and efficiency of the system.In this paper,the self-diagnosis technology and the state evaluation methods of BHMS are mainly studied.At the same time,based on the health monitoring system,the state assessment of bridge structure is carried out according to bridge management and maintenance requirements.As a reference,many national BHMS played an important role in this study.The main contents are as follows:(1)The probability statistics of time domain,frequency domain,time-frequency domain are analyzed and summarized.Based on the traditional data self-diagnosis method,a self-diagnosis method based on generalized 3-deta is proposed.The method uses the moving interval to get statistical eigenvalue combined with the historical data statistics to set threshold in order to find and remove the abnormal data.At the same time,the data self-diagnosis method based on spectral kurtosis is proposed,which uses the power spectral density method to obtain the frequency domain characteristics of the vibration data.The vibration abnormal data is identified and removed according to the spectral kurtosis analysis of the frequency domain characteristic section.(2)The main factors influencing data correlation and relevance in BHMS were studied.A more convergent correlation can be obtained when eliminating the effect of data delay effect based on Fourier Least Squares Dependency Self-diagnosis method.A self-diagnosis method based on multiple regression analysis is proposed which optimizing multivariate variables using principal component analysis.Finally,a stable multivariate nonlinear model of variables is established to perform data anomaly diagnosis.At the same time,using the advantages of machine learning algorithm and massive monitoring data,a non-linear model were created for high-precision prediction of missing data and the removal of the outliers.(3)This paper focused on the current method of bridge state evaluation,and put forward the safety state evaluation method based on vehicle-bridge resonance response.Based on the wavelet packet decomposition method,the vibration acceleration amplitude and the speed correlation center value were extracted.In the meantime,the warning threshold setting method based on the interval estimation theory is used to evaluate the bridges' vibration data.The method has been successfully applied in Nanjing DaShengGuan BHMS.Based on the principle of string theory and frequency method,the frequency parameters of the cable were extracted by using the optimized parameter power spectrum function.The frequency after eliminating the temperature's impact can be used to evaluate the state of the cable based on the mean control chart model.