Damage identification of urban overpass based on modal frequency and genetic neural network

The finite element model of left auxiliary bridge of Qianjin Overpass is built and vulnerable sections of structure are chosen as research objects. In consideration of the asymmetry of the bridge, change rate of modal frequency is chosen as input parameter for genetic neural network, and identification ability of damage location and level is studied. The result shows that this method can successfully identify location of single damage and multi-damage; The error of damage level identification for test samples is less than 5% and the interpolation ability is better than the extrapolation ability. This indicates the method has good practice prospects.