Abstract: In this paper, a damage detection algorithm is proposed for large size structures under limited input and output measurements. A large size structure is decomposed into small size substructures based on its finite element formulation. The interconnections between adjacent substructures are treated as ‘additional unknown inputs’ to substructures. By sequentially utilizing the extended Kalman estimator for the extended state vector and a least squares estimation for the unmeasured external inputs, the algorithm can identify element dynamic parameters in each substructure. Structural local damage in the large size structure can be detected by tracking the changes in the identified values of dynamic parameters at an element level, e.g., the degrading of stiffness parameters. The numerical example of detecting structural damages in a relatively large size plane truss bridge validates the feasibility of the proposed algorithm. Compared with other approaches, the proposed algorithm requests fewer input and output measurements.