| Description |
The underlying principle to locate active or passive sources in many fields start with removing baseline information. In underwater acoustics, researchers use localization algorithms to locate sea mammals. In speech, these algorithms locate and track active speakers. In seismology, localization is used to determine the source of a seismic event. In semiconductor manufacturing, these methods are used to detect and find random defects on wafers. In guided wave structural health monitoring, finding damage is typically accomplished by comparing measurements before damage (i.e., baseline data) and after damage (i.e., test data). Yet, in practical scenarios, baseline data are either uncertain or unavailable. Data from surrogate structures (structures similar to the test structure) could replace baseline data, but due to small differences in material properties, thickness, temperature, and other effects, these data are unreliable. This dissertation proposes a dictionary learning framework that overcomes the two challenges of baseline data: uncertainty and unavailability. The framework takes inspiration from transfer learning to reconstruct accurate baseline data. The framework learns and combines wave propagation characteristics of a test structure with geometric information from surrogate structures to create a synthetic damage-free baseline data. The synthetic damage-free baseline data are then used to solve three important problems in the field of guided wave structural health monitoring. Specifically, we address damage wavefield isolation, damage localization, and damage detection on geometrically complex structures with multipath reflections. Furthermore, simple numerical simulations are also investigated to replace surrogate structures. |
