Evaluation of Machine Learning Algorithms for Nuclear Monitoring Applications
Author(s)
Pace, Joshua
Advisor(s)
Editor(s)
Collections
Supplementary to:
Permanent Link
Abstract
Measuring radioactive signatures from decaying elements is crucial for international nuclear safety and monitoring. For instance, sparse amounts of radioactive xenon detected within the atmosphere serve as evidence of nuclear fission reactions. 131m Xe, 133 Xe, 133m Xe and 135 Xe are key xenon gaseous isotopes which international monitors use to recognize and categorize nuclear events. When these gaseous elements decay, unique beta-gamma energy spectra are produced; however, the different spectra from these elements can overlap. The work of the present study is to use machine learning methods, in particular the kmeans clustering algorithm, to classify and differentiate various radioactive spectra. Several methods of data organizing are implemented to investigate the algorithm’s precision in identifying and grouping distinct energy spectra. Overall, this thesis argues that machine learning can be used to categorize distinct energy spectra and provide accurate isotope identification.
Sponsor
Date
2025-07-28
Extent
Resource Type
Text
Resource Subtype
Thesis