Data Analysis Of Structural Health Monitoring Based On Multidimensional Probability Distribution Models

Analyzing structural health monitoring data based on probability distribution model is one of the key research contents in the field.With the development of structural health monitoring sensing network,the dimension of monitoring data increases.Therefore,probabilistic modeling,probabilistic prediction and probabilistic warning based on multidimensional probability distribution are the core tasks of health monitoring data analysis.This paper conducts research on the above key issues,the main work includes:(1)The development of structural health monitoring and the probabilistic analysis of structural health monitoring data are reviewed.This paper summarizes the uncertainty and correlation of monitoring data,the probability analysis method of monitoring data,and the research status of monitoring data probability analysis based on multidimensional probability distribution model.(2)The theoretical framework of probability analysis of monitoring data based on multidimensional probability distribution model is proposed.The framework is based on two methods of Copula theory and Gaussian Mixture model to obtain a multidimensional probability distribution model;based on the Bayesian model selection to obtain the optimal multidimensional probability distribution model;based on the optimal model to conduct monte Carlo simulation,target variable probabilistic prediction,target probabilistic warning.Finally,the validity of the theoretical framework is verified based on the simulation study.(3)To conduct multidimensional temperature monitoring data probability analysis.Based on the temperature monitoring of the Dowling Hall Bridge at Tufts University,the proposed theoretical framework is applied to obtain the multidimensional air temperature and component temperature probability distribution model based on Copula theory and Gaussian mixture model respectively;and the optimal multidimensional probability distribution model is selected based on Bayesian method;and the probabilistic prediction of target temperature under complete and incomplete information is considered respectively.The results show that: there is a significant spatial correlation between Dowling Hall Bridge multidimensional air temperature and component temperature;the multi-peak Gaussian mixture model is the optimal multidimensional probability distribution model that describes the multidimensional temperature data of the bridge;the probability density function of the temperature of the target component can be improved by increasing the given information;and the probability density function of the temperature of the target component can be obtained given the air temperature only.(4)To conduct multidimensional environmental factors-modal parameters monitoring data probability analysis.Based on the monitoring of environmental factors(temperature and humidity)and modal parameters(first to third order modal frequency)of the East Asian Building of the University of Macau,the proposed theoretical framework is applied to obtain the multidimensional environmental factors and modal parameter probability distribution models based on Copula theory and Gaussian mixture models,respectively,and selects the optimal multidimensional probability distribution model based on Bayesian method,taking into account the target modal parameter under complete and incomplete information,respectively The probability prediction of modal parameters is finally combined with the abnormal probability discrimination method to provide an early warning of the overall health status of the structure.The results show that: there is a significant correlation between the first to third order mode frequencies of East Asian buildings and temperature and humidity;multi-peak Gaussian mixture model is the optimal multidimensional probability distribution model that describes the building’s multidimensional environmental factors and modal parameter data;the probability density function of the target modal frequency can be improved by increasing the given information;and the results of monitoring data analysis show that the overall health of the structure has no safety risks during the monitoring period.