Session: 06-01: Machine Learning and Statistical Methods in NDE I

Paper Number: 147668

147668 - Anisotropic Guided Wave Dispersion Curves for Physics-Informed Learning 

Abstract: 

One challenge in guided wave structural health monitoring is learning dispersion curves (i.e., the guided wave’s wavenumber or velocity as a function frequency) from measurements. This information can be used to build predictive guided wave models to improve flaw detection, localization, and characterization.  It can also be used directly for diagnostics. For example, the local wavefield dispersion curves can identify delaminations and porosity in composites or identify thickness losses in metallic structures. However, the anisotropic relationships within the many material systems, such as fiber-reinforced composites, are difficult to decipher and even more difficult to learn/estimate. For highly anisotropic and heterogeneous materials, there is generally no known mathematical expression for describing the dispersion curves. As a result, the dispersion curves are often considered to simply vary as a function of angle with no additional assumptions, making their extraction even more difficult.

This presentation discusses a fundamentally different mathematical approach to understanding and learning anisotropic guided wave dispersion curves. We start by discussing the anisotropic wave equation, which is not commonly found outside of fundamental mathematical studies on partial differential equations. We demonstrate that the solutions to the anisotropic wave equation allow the production of non-elliptical wave propagation from a single point source. We discuss the tunable parameters of these solutions and how they ultimately modify our waves. With that knowledge, we demonstrate how to mathematically express anisotropic dispersion curves, which do not need to be explicitly written as a function of the propagation angle. Furthermore, we show how these dispersion curves can be estimated as part of a physics-informed learning paradigm.

We learn anisotropic dispersion curves from a wavefield traveling in a unidirectional carbon fiber-reinforced polymer matrix composite material. The waves in the material exhibit two types of anisotropy. The first type of anisotropy is characterized by one direction of propagation being preferred. The second type of anisotropy is characterized by the wavefront not exhibiting an elliptical pattern. Both characteristics of the waves are learned, and their anisotropic dispersion curves are shown. Based on these anisotropic dispersion curves, we demonstrate the ability to emulate the wave propagation and use differences between our emulation and the measured wavefield to identify regions of damage.

Presenting Author: Joel Harley University of Florida

Presenting Author Biography: Joel B. Harley is an Associate Professor in the Department of Electrical and Computer Engineering at the University of Florida, Gainesville, FL. His interests include integrating numerical and analytical models with machine learning, and data science for smart infrastructure monitoring and material characterization. Dr. Harley is a recipient of 2021 Achenbach Medal from the International Workshop on Structural Health Monitoring, a 2020 IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society Star Ambassador Award, a 2020 and 2018 Air Force Summer Faculty Fellowship, a 2017 Air Force Young Investigator Award, and a 2014 Carnegie Mellon A. G. Jordan Award. He has published more than 170 technical journal and conference papers, including four best student papers. He is also an Associate Editor and member of the editorial board for Structural Health Monitoring: An International Journal, a member of the editorial board for Ultrasonics, a member of the IEEE Ultrasonics, Ferroelectrics, and Frequency Control Society Technical Program Committee, a member of the IEEE Signal Processing Society, and a member of the Acoustical Society of America.

Authors:

Joel Harley University of Florida
Charlie Tran University of Florida
Woohyun Eum University of Florida
Amanda Beck University of Florida
Michael Macisaac University of Florida
Matthew Stormant University of Florida
Ghatu Subhash University of Florida