The quality of the spectral data collected by radiological survey systems depends on many factors including the survey environment, configuration of the system and its detectors, and the radionuclides in question. Algorithms in the field of machine learning have the potential to classify data that would be difficult and time-intensive for a human to analyze. Depleted and natural uranium spectra are of particular interest due to known contamination at domestic sites and world-wide. Several machine learning classifiers were developed with data collected from laboratory experiments. This thesis demonstrates the potential of machine learning algorithms to discriminate gamma-ray emitting sources using sparse, or low-count statistic, data. Effectiveness has been demonstrated for discriminating chemical forms of uranium, mixtures with differing uranium isotope distributions, and predicting source masses given certain detector geometries and a known target distribution. All activity has been supported by the U.S. Army Engineering Research and Development Center (ERDC).
| Identifer | oai:union.ndltd.org:MSSTATE/oai:scholarsjunction.msstate.edu:td-1679 |
| Date | 01 May 2020 |
| Creators | Finney, Austin |
| Publisher | Scholars Junction |
| Source Sets | Mississippi State University |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Theses and Dissertations |
Page generated in 0.0105 seconds