Research On Data Reconstruction Method For Wireless Health Monitoring Of Large-span Steel Structure Of Network Security Center

The large-span spatial steel structure is widely used in large public buildings with its beautiful shape and good load-bearing performance.Due to the huge size of the large-span space steel structure and the crisscrossing of components of different specifications,the overall structure needs to undergo multiple stress changes from the beginning of construction to the formation of the structure and the period of service.It is very important to monitor the health of the space steel structure throughout its life cycle.However,due to the impact of the durability of the monitoring equipment,the temperature field and the uncertain factors of the outdoor environment,the monitoring data of the steel structure is missing,which affects the effectiveness of the structural health assessment.Therefore,The research background in this article is the large-span steel structure from the ‘National Cyber Security Talent and Innovation Base Exhibition Center’(‘Network Security Center’ for short).Our goal is to study the reconstruction method of steel structure health monitoring data missing,and its practicality and effectiveness are analyzed.It provides a basis for the monitoring of steel structures during the construction and service phases.The main research work of the thesis is as follows:(1)Analyze the configuration plan of stress and strain monitoring points for large-span steel structures based on design calculations,building monitoring specifications,and actual engineering status.Based on the characteristics of the large-span steel structure of the Network Security Center,Health monitoring modules are built with wireless monitoring method.(2)Analyze the sources and types of missing monitoring data for large-span steel structures.Based on the improved linear regression model,the method of reconstruction of stress missing data for temperature correlation and measurement point correlation is studied,and the data reconstruction effect of different types of missing data is analyzed.The results show that the temperature and stress correlation coefficient of the monitoring point is above 0.9,and the average error between the reconstructed stress missing data and the measured data is less than 5%.Meanwhile,the data missing rate of the temperature-dependent reconstruction should not exceed 30 %.When the correlation coefficient between single measurement point and multiple measurement points is above 0.9,the average error between the reconstructed stress missing data and the measured data is close to 5%.The data missing rate of the monitoring point correlation reconstruction should be controlled within 30%.For a higher missing rate,the use of multiple regression can effectively improve the accuracy of reconstruction.For different types of continuous and discrete stress missing data,the reconstruction error of the two types of missing stress data is mostly within 5%.Overall,the reconstruction error of stress missing data is within the acceptable accuracy range of engineering monitoring.(3)On the basis of using linear regression method to reconstruct the missing stress data of steel structure,the reconstruction value of the local stress data has a large deviation from the true value in view of the non-linear relationship between the local stress monitoring point and the whole.Adopting improved neural network method to study the reconstruction method of missing stress data of steel structure.The results show that the accuracy of reconstructed stress missing data is higher than that of the linear regression method,and the applicable data missing rate range of this method cannot exceed 20%.