A regression-based damage detection method for structures subjected to changing environmental and operational conditions

Damage detection of civil engineering structures during the past decade has focused on eliminating the effects of the changing environmental and operational conditions, from the effects of damage. In the literature, a regression analysis has been adopted to construct a model between the vibration properties of structures, and the environmental and operational parameters to represent the undamaged state of the structures, for damage detection. However, using the environmental and operational parameters in the analysis has several limitations. For example, these parameters are not always available which may affect the performances of the damage detection methods. Regression between the vibration properties only has also been proposed in the literature where multivariate statistical tools have been adopted to extract the relationships among the properties. However, these methods have the problem that it is more difficult to detect damage in the multivariate situations and a regression target is usually needed, which is difficult to determine. Therefore, a damage detection method which uses the simple regression analysis, is developed in this paper. The vibration properties of structures are used as both the independent and dependent variables in the developed method. This has the advantages that the environmental and operational conditions are not needed and the multivariate statistical tools are not required for data processing. The developed method is applied to a beam structure model and the Z24 Bridge, in Switzerland, and the results obtained demonstrate that the method can successfully classify between undamaged and damaged states. The traditional regression analysis method is also applied to the two structures and it was found that better results are obtained using the method developed in this paper.