Guided wave reconstruction in complex geometries with a dictionary learning framework

Guided waves are an attractive tool for structural health monitoring (SHM) due to their ability to interrogate large areas of a structure. Yet, guided waves are characterized by multi-modal and frequency dispersive behavior and quickly grow in complexity with the structure. For example, guided wave reflections from plate edges, fasteners, or joints will lead to complex, difficult to analyze data. Most SHM algorithms try to remove or ignore these reflections. Yet, knowledge about the reflections and their acoustic behavior can significantly improve detection and localization algorithms. Hence, accurate knowledge about guided waves reflections is of a significant interest in SHM. In this paper, we reconstruct and predict guided wave measurements from geometric environments with reflections. We leverage dictionary learning and sparse recovery algorithms to achieve this goal. Dictionary learning is used to learn the building blocks of guided waves from simulation data. Sparse recovery algorithms are used to create predictive models of wave propagation based on experimental data. From simulation results, we show that our framework can successfully predict wave propagation, including reflections, across an aluminum plate with an accuracy of more than 95% from just 20 measurements. We demonstrate similar results with experimental data.

from #MedicinebyAlexandrosSfakianakis via xlomafota13 on Inoreader http://ift.tt/1O2JMlI
via IFTTT