Multi-scale sensing for damage identification

Damage identification is important for the lifetime prediction of any structure. In a composite structure, damage can occur at several material scales from micro-cracking to global buckling or delamination. This makes the identification of damage difficult with a single sensing device. In this paper, we propose to monitor a structural volume with an embedded optical fiber sensor network measuring strain, integrated strain, and strain gradients. Two methods are also compared for data fusion of the multi-scale data in order to determine damage parameters. The first calculates strain maps directly from the data; the second method uses a neural network. As an example, an isotropic, homogeneous structural volume with a localized crack is modeled. The results demonstrate that (a) the multi-scale sensing approach improves damage identification and (b) the neural network is a method well adapted for the multi-scale data fusion and significantly improves the damage identification capability.