In order to accurately evaluate the structural health condition, the improved particle swarm optimization（IPSO）algorithm and BP algorithm integrated organically is applied to train the artificial neural network and is used for structural damage identification. A benchmark problem proposed by the International Association for Structural Control and the American Society of Civil Engineering （IASC-ASCE） Task Group on Structural Monito- ring is investigated. The damage location for four different cases is estimated. Under the influence of the model error, measurement noise etc. , the damage identification results show that the presented method is satisfactory.