On-the-fly Segmentation Approaches for X-ray Diffraction Datasets for Metallic Glasses

TL;DR

This paper introduces three real-time segmentation methods for X-ray diffraction data from metallic glasses, enabling faster analysis by grouping similar spectra and reducing manual effort without losing scientific detail.

Contribution

It presents novel on-the-fly segmentation approaches—attribute extraction, nearest-neighbor distance, and cluster analysis—for XRD data, improving efficiency over traditional manual methods.

Findings

Methods enable rapid grouping of XRD spectra

Significantly reduces analysis time for metallic glasses

Maintains scientific insights with fewer spectra to examine

Abstract

Investment in brighter sources and larger detectors has resulted in an explosive rise in the data collected at synchrotron facilities. Currently, human experts extract scientific information from these data, but they cannot keep pace with the rate of data collection. Here, we present three on-the-fly approaches - attribute extraction, nearest-neighbor distance, and cluster analysis - to quickly segment x-ray diffraction (XRD) data into groups with similar XRD profiles. An expert can then analyze representative spectra from each group in detail with much reduced time, but without loss of scientific insights. On-the-fly segmentation would, therefore, result in accelerated scientific productivity.