Abstract: The displacement mode parameters and strain mode parameters are combined withneural networks techniques on the studies of damage detection of structures.In order to obtainmore sensitive signatures for structural damage detection, after considering the analysis of current signatures, six kinds of signatures for damage detection are studied and the sensitivity todamages is compared. The results show that six kinds of signatures can all be used to predict thestructural damage. It is also clearly seen that damage signatures based on strain data are generally more sensitive to damage than that on displacement data.