<p dir="ltr">Guided wave structural health monitoring applications produce large amounts of complex, high-dimensional datasets that are prime targets for the growing number of machine learning strategies. Yet, machine learning is not widespread in guided wave structural health monitoring. This is due to two important challenges that need to be overcome. First, guided wave experimental data is often highly redundant. That is, we rarely have data from 1000s of different structures. Instead, data is often taken from a small number of structures over time. As a result, machine learning methods for guided waves cannot easily generalize across different structures and different geometric configurations. Second, purely data-driven machine learning approaches offer few engineering assurances (physically and mathematically) when compared with physics-based techniques. This is important and necessary for trust in safety-critical systems.</p><p dir="ltr">To address these two challenges, this presentation explores the integration of the wave equation into machine learning models. We refer to this as wave physics-informed learning. This approach is effectively achieved by penalizing the cost function of a machine learning algorithm. This forces an underlying wave structure in how we represent data within the machine learning system. We discuss a couple of different algorithms of wave physics-informed learning for guided wave structural health monitoring applications. Specifically, we integrate a wave-informed penalty into unsupervised representation learning frameworks. We show that our framework can both generalize across structures and provide physical knowledge by demonstrating how it can extract dispersion curves (velocity as a function of frequency) from arbitrary geometries. The results point to the future possible benefits of such learning architectures.</p>

Wave Physics-Informed Velocity Learning With Guided Waves

Paper Proceedings: https://asmedigitalcollection.asme.org/qnde/QNDE2021/volume/85529

ASME 2021 48th Annual Review of Progress in Quantitative Nondestructive Evaluation (QNDE2021)

Supporting the need among researchers worldwide for additional online and multimedia content, ASME now offers the ASME Conference Video Collection. This platform will collect and provide long-term access to conference presentation videos, encompassing all technology areas within ASME's technical conference and events program. About 30,000-40,000 engineers across all mechanical engineering and related disciplines participate in ASME conferences each year. This includes authors of approximately 6,000 to 8,000 technical papers published by ASME, the majority of whom also submit video presentations for our virtual conferences. The video collection will be available as an add-on to existing product packages or as a separate subscription.