Two-step identification approach for effective bridge health monitoring is proposed to alleviate the issues associated with many unknown parameters faced in the real structures and to improve the accuracy in the estimate results. It is suitable for on-line monitoring scheme, since the rigorous damage assessment is not always needed to be performed whereas the alarming for potential damage occurrence is to be continuously carried out. In this study, two-step identification approach is incorporated. In the first step for screening potential damaged members, three different methods were utilized: (1) Damage Indicator Method based on the Modal Strain Energy (DIM-MSE), (2) Probabilistic Neural Networks (PNNs), and (3) Neural Networks using Grouping technique (NNs-Gr). Then, in the second step, the conventional neural networks technique is utilized for damage assessment on the screened members. The proposed methods are verified through a field test on the northern-most span of old Hannam Grand Bridge over Han River in Seoul, Korea. The issues on measurement noise, modeling errors and multiple damages are addressed.