| In recent years,Structural Health Monitoring（SHM）technology has been increasingly used in the conservation of wooden structures of ancient buildings as a noninvasive preventive conservation tool.The massive monitoring data of structural health monitoring system can provide important data support for the conservation and repair of wooden structures of ancient buildings,and make effective diagnosis of the structure.In this paper,the long-term monitoring data of the structural health monitoring system of a Tibetan ancient building timber structure in the past eight years as the research object,and the following studies are conducted for the identification of abnormalities in its longterm monitoring data in the past 8 years,and the main research contents and conclusions are as follows.（1）This paper analyzes the characteristics of the temperature data and strain data of Tibetan ancient building wood structures,and uses the correlation coefficient between the two to conclude that both temperature and strain change periodically on a daily and yearly basis.Combination and power spectrum analysis show that the correlation between the two is extremely high.Moreover,the time-average value is used as the research unit,and the effect of strain lagging temperature effect can be ignored.That is,in the timber structure of Tibetan ancient buildings,the changes of strain and temperature can be regarded as occurring synchronously.（2）The preprocessing of sensor distortion data is a necessary means to ensure the accuracy and reliability of data.This paper proposes the Sobel-CNN model to identify sensor distortion data.The actual structure of the monitoring data was verified,and the model was successfully used to classify and identify three data distortion modes（ie,data duplication,data missing,and outliers）and normal modes.The accuracy of the model was 95.4%.（3）Based on the high correlation between strain and temperature,a regression model between strain and temperature data is established based on SSA-PR and VMD-PR methods,and strain is separated into two parts,temperature strain and residual strain.And it is concluded that compared to the SSA-PR model,the advantage of the VMD-PR model is that it can quickly process a large amount of monitoring data,and both models can effectively separate the strain caused by the temperature effect.For the separated temperature strain,the Prophet model is used to predict the separated temperature strain for two years,and the actual measurement data is used to verify the prediction accuracy of the model.The promotion of the model can realize the five-year prediction of the temperature strain.The future trend of temperature strain.Regarding the residual strain after separation,it is regarded as an indicator of structural damage,and the Raida’s rule is used to identify the abnormal change trend in the residual strain.It is concluded that the residual strain will change abnormally in July and August,which is assumed to be environmental Among the factors,the influence of humidity and the influence of the increase of tourists.（4）Based on the research based on the high correlation of strain data between homologous multi-sensors in the same substructure,the traditional K-means clustering algorithm is improved,and the improved algorithm is used to compare the correlation between homologous multi-sensors in the same structure.The coefficient matrix is calculated to realize the positioning of abnormal measuring points.63 diagrams,13 charts,57 references... |
PDF Full Text Request