S057-01
Application of Dynamic Mode Decomposition with Control (DMDc) to cluster, characterize and extract information from acoustic signals

Tuesday, 15 December 2020: 07:02
Virtual
Charles Fieseler, University of Vienna, Vienna, Austria, Chven A Mitchell, Stanford University, Stanford, CA, United States, Nathan Kutz, University of Washington Seattle Campus, Applied Mathematics, Seattle, WA, United States, Chaoyi Wang, Purdue University, West Lafayette, United States and Laura J Pyrak-Nolte, Purdue University, Department of Physics and Astronomy; Department of Earth, Atomospheric and Planetary Sciences; Lyles School of Civil Engineering, West Lafayette, IN, United States
Recorded Presentation
Abstract:
Monitoring damage evolution in rock requires identification of geophysical signatures that are linked to the physical properties of rock or the cracks that form during damage. We performed laboratory experiments to identify signatures of (1) damage evolution in acoustic emission (AE) signals recorded from clay-bearing samples and (2) synthetic fractures in laboratory-scale cased cement boreholes from elastic waves.

Machine learning techniques were employed to build model-based features that can be used for classification and analysis of the relevant dynamics. Features of intrinsic dynamics and external input were discovered in the recorded signals by extending the recently developed Dynamic Mode Decomposition with Control (DMDc) to incorporate sparsity regularization. With the DMDc method, intrinsic frequencies and external inputs are constructed to provide a model that can be integrated forward in time. Such a model can also be used to denoise signals by removing features that cannot be produced by the learned intrinsic dynamics.

The application of DMDc and clustering, which was performed using Gaussian Mixture models, provided robust methods for extracting and classifying features from the data, especially for extracting reflections recorded in experiment (2) that were not readily apparent with other methods. DMDc has the potential to illuminate hidden signatures of static and dynamic fractures, thus providing a unique opportunity to better understand geophysical signatures related to fracturing and wave propagation in natural systems.

Acknowledgments: A portion of the experimental data used in this work is supported by the U.S. Department of Energy, Office of Energy Efficiency and Renewable Energy (EERE), Office of Technology Development, Geothermal Technologies Office. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Basic Energy Sciences, Geosciences Research Program under Award Number (DE-FG02-09ER16022).

Previous Abstract | Next Abstract >>