Abstract: The accurate identification of structural damage is an important part of structural health monitoring. The identified model usually needs to be modified in real-time, which costs substantial computational effort because of the factorization of global stiffness matrix. In this paper, a new method of state identification for truss structures is presented based on the theory of inelasticity-separated method, in which the total strain of the cross section is decomposed into elastic strain and damage strain, and the material damage is evaluated by stiffness degeneration. The concept of the proposed method is that the stiffness matrix related to the damage of the model is segregated from the global stiffness matrix, and only the stiffness matrix with small scale is updated and factorized for local damage problems in the whole computation process. Thus, the calculation efficiency can be improved greatly for damage identification of truss structures. Finally, an actual steel truss bridge was taken as example, and the damage identification of the main truss strings were studied by numerical simulation. Two conditions of static loads were designed to verify the proposed method. The results show that the method can accurately and efficiently identify the location and extent when of the structural damage when it occurs in local domains.