Identification of bridge structures using genetic algorithms

Abstract

In this study, a structural identification method is developed for bridges using genetic algorithms, and it is applied to the finite element model updating and damage estimation of a plate girder bridge. Structural identification has increasingly become an important research topic in conjunction with damage assessment and safety evaluation of existing structures. The proposed method using genetic algorithms uses only the payoff information, not derivatives or other auxiliary knowledge as in the conventional methods. Moreover, the present method may give the better estimate, which is not related to the local minimum of the error criteria function. In this study, the averaged normalized frequency response function modulus is used for estimating the natural frequencies from the acceleration data measured at many different locations. The numerical simulation study on a grid structure indicates that the proposed method using genetic algorithms can identify the damage locations and severities very well. The grouping approach used to reduce the number of the unknowns in each identification process is founded to be very effective. The results of the experimental study on a plate girder bridge indicate that the present method can provide a better finite element model than the initial model constructed from the design drawings and the locations of the damaged members with relatively large inflicted damages can be reasonably identified.