Learning acoustic emission signatures from a nanoindentation-based lithography process: Towards rapid microstructure characterization

摘要

We present an approach for rapid identification of the salient microstructural phases present on a metallic workpiece surface via nanoindentation-based lithography process. We employ a machine learning approach to connect the time-frequency patterns of the corresponding acoustic emission (AE) signals with the underlying microstructural phases. Results show that the AE frequencies in the range of 0.3-1 kHz and 30-50 kHz can discriminate between the microdynamics of the lithography process arising from different microstructural compositions and thereby predict these microstructural phases with accuracies exceeding 95%. We also draw physical interpretations of our "black-box" machine learning model and demonstrate that the physical insights into the underlying AE signals allow us to identify novel patterns and possible microstructural anomalies.