Dr. Firas Khasawneh (ME Dept., Michigan State University) 

12/8/2022  3:10pm

Abstract: 

Quantifying patterns in visual or tactile textures provides important information about the process or phenomena that generated these patterns. In manufacturing, these patterns can be intentionally introduced as a design feature, or they can be a byproduct of a specific process. Since surface texture has significant impact on the mechanical properties and the longevity of the workpiece, it is important to develop tools for quantifying surface patterns and, when applicable, comparing them to their nominal counterparts. While existing tools may be able to indicate the existence of a pattern, they typically do not provide more information about the pattern structure, or how much it deviates from a nominal pattern. Further, prior works do not provide automatic or algorithmic approaches for quantifying other pattern characteristics such as depths' consistency, and variations in the pattern motifs at different level sets. This talk leverages persistent homology from Topological Data Analysis (TDA) to derive noise-robust scores for quantifying the type of the pattern, e.g., pattern centers forming rectangular or hexagonal grids, as well as motifs' depth and roundness in a pattern. Specifically, point cloud persistence is used to quantify the type of the pattern after estimating the motifs’ centers, while sublevel persistence is used to quantify the consistency of indentation depths at any level set. Moreover, we combine sublevel persistence with the distance transform to quantify the consistency of the indentation radii, and to compare them with the nominal ones. We show the effectiveness of our methods using digital scans of surfaces subjected to Piezo Vibration Striking Treatment (PVST). Although the tool in our PVST experiments had a semi-spherical profile, we present a generalization of our approach to tools/motifs of arbitrary shapes thus making our method applicable to other pattern-generating manufacturing processes.


Bio: Dr. Firas Khasawneh is an Assistant Professor of Mechanical Engineering at Michigan State University. He received his PhD in 2010 from Duke University, and his Masters in 2007 from the University of Missouri-Columbia. His areas of expertise include nonlinear dynamics and vibrations, topological signal processing, and time series analysis. Dr. Khasawneh’s research has been funded by the National Science Foundation (NSF), and the Airforce Office for Scientific Research (AFOSR).