| dc.description.abstract | This paper proposes a strain-based performance warning method for bridge main girder monitoring using the long-term data obtained for varying load conditions, i.e., the variations in temperature, wind, and traffic load. Because the temperature field variation of the main girder is easy to measure, a correlation model between temperature and strain of the main girder was first established through a novel representative temperature, namely, the canonically correlated temperature. Based on this model, temperature effects on the main girder strain can be accurately estimated and eliminated. However, the influence of wind and traffic load on the main girder strain was difficult to quantify. Thus, principal component analysis was then employed to model the main girder strain after eliminating temperature effects, and the first two principal components were extracted to represent the effects of wind and traffic load, which could subsequently be eliminated. The remaining principal components were then used to reconstruct the model errors, which were minimally influenced by the varying operating conditions. After this analysis, two warning indexes (i.e., the Euclidean distance and the Mahalanobis distance) were defined for the model errors and the remaining principal components to detect potential performance degradations. In addition, a location index was deduced based on contribution analysis to indicate where the performance degradation occurs. Finally, an engineering application to a cable-stayed bridge was carried out to verify the capability and validity of the proposed method. | |
