Full-field, high-spatial-resolution detection of local structural damage from low-resolution random strain field measurements

Structural damage is typically a local phenomenon that initiates and propagates within a limited area. As such high spatial resolution measurement and monitoring is often needed for accurate damage detection. This requires either significantly increased costs from denser sensor deployment in the case of global simultaneous/parallel measurements, or increased measurement time and labor in the case of global sequential measurements. This study explores the feasibility of an alternative approach to this problem: a computational solution in which a limited set of randomly positioned, low-resolution global strain measurements are used to reconstruct the full-field, high-spatial-resolution, two-dimensional (2D) strain field and rapidly detect local damage. The proposed approach exploits the implicit low-rank and sparse data structure of the 2D strain field: it is highly correlated without many edges and hence has a low-rank structure, unless damage-manifesting itself as sparse local irregularity-is present and alters such a low-rank structure slightly. Therefore, reconstruction of the full-field, high-spatial-resolution strain field from a limited set of randomly positioned low-resolution global measurements is modeled as a low-rank matrix completion framework and damage detection as a sparse decomposition formulation, enabled by emerging convex optimization techniques. Numerical simulations on a plate structure are conducted for validation. The results are discussed and a practical iterative global/local procedure is recommended. This new computational approach should enable the efficient detection of local damage using limited sets of strain measurements.