Merging models and data: Predictive modeling for guided wave

Abstract

New simulation tools for nondestructive evaluation (NDE) and structural health monitoring (SHM) are enabling predictive power that will quicken inspection accuracies, minimize inspection costs, reduce the need for data, and create pathways to new automatous inspection and monitoring methods. Simulations can be especially powerful tools for analyzing complex geometries and modern anisotropic materials, such as carbon-fiber reinforced composites, where NDE and SHM theory and practice is still developing. Yet, truly predictive simulations are yet to be realized. Most simulations rarely match with experimental data unless the simulation is meticulously tuned. In this paper, we describe a framework for merging existing models with experimental data to create better predictive simulations of guided waves and other complex acoustic environments. We leverage tools and theory from compressive sensing, sparse inversion, and convex optimization. We focus on guided waves due to their significant complexity and their wide use in SHM and other acoustic applications. Our results achieve prediction accuracies of greater than 90% for guided waves in both isotropic and anisotropic environments. We also demonstrate how predictive models can be used for a variety of applications, including time of arrival estimation and temperature compensation.