This paper studies the application of independent component analysis (1CA) to structural health monitoring (SHM). The 1CA technique is used to extract features from measurement sensors of a smart composite structure. To enhance information about damage, we propose a new damage detection methodology, based on the analysis of a symmetric matrix obtained from the mixing matrix. The subspaces associated to the symmetric matrix are used to define a damage index (Dl). Furthermore, we propose a bound associated to this Dl and decision on the state of the structure is taken. Experimentation on a composite smart structure shows the performance of the proposed approach.