Abstract -- This project is a sub-project of the main project “Collection and Analysis of Bridge Responses to Earthquakes” leaded by C. H. Loh, director of National Center of Research on Earthquake Engineering. The aim of this work is to apply a wavelet neural network to detect the damge of bridges from their project is a sub-project of the main project “Collection and Analysis of Bridge Responses to Earthquakes” leaded by C. H. Loh, director of National Center of Research on Earthquake Engineering. The aim of this work is to apply a wavelet neural network to detect the damge of bridges from their dynamic responses to earthquakes. The detecting results can be used for bridge’s health monitoring in long term. This study establishes the procedure of structure damage detection using a wavelet neural network. A record of dynamic responses of a structure to small earthquake is first employed to train a wavelet neural network; the structutal parameters of the structure, therefore, are derived and stored in the network. By comparing the differences of the dynamic response of the structure between masured in site and computed via the wavelet neural network, the damage of the structure can be decteded. The methodology is first verified in successfully processing the measured responses of a five-story steel frame under shaking table tests untaken by National Center for Research on Earthquake Engineering in Taiwan. The proposed procedure is finally applied to process the measured earthquake responses of a unit of elevated bridge in the east-to-west express high way in Taipei.