Machine learning in analytical spectroscopy for nuclear diagnostics [Invited]

Abstract

Analytical spectroscopy methods have shown many possible uses for nuclear material diagnostics and measurements in recent studies. In particular, the application potential for various atomic spectroscopy techniques is uniquely diverse and generates interest across a wide range of nuclear science areas. Over the last decade, techniques such as laser-induced breakdown spectroscopy, Raman spectroscopy, and x-ray fluorescence spectroscopy have yielded considerable improvements in the diagnostic analysis of nuclear materials, especially with machine learning implementations. These techniques have been applied for analytical solutions to problems concerning nuclear forensics, nuclear fuel manufacturing, nuclear fuel quality control, and general diagnostic analysis of nuclear materials. The data yielded from atomic spectroscopy methods provide innovative solutions to problems surrounding the characterization of nuclear materials, particularly for compounds with complex chemistry. Implementing these optical spectroscopy techniques can provide comprehensive new insights into the chemical analysis of nuclear materials. In particular, recent advances coupling machine learning methods to the processing of atomic emission spectra have yielded novel, robust solutions for nuclear material characterization. This review paper will provide a summation of several of these recent advances and will discuss key experimental studies that have advanced the use of analytical atomic spectroscopy techniques as active tools for nuclear diagnostic measurements.

Laser Ablation Plasmas and Spectroscopy for Nuclear Applications

“…ANNs are becoming increasingly popular within the scientific community; spanning across disciplines including neuroscience, astrophysics, and analytical chemistry; as scientists seek to implement machine learning techniques into the analysis of their data. [418][419][420][421] However, as many experts cautioned against using PLS-DA without understanding the underlying mathematics and assumptions of the technique, we will make that same argument here concerning neural networks. While neural networks have become very appealing in the last few decades due to their impressive performance on complex data sets, they require very large training data sets that are composed of at least a few thousand samples; 395 Microsoft has developed a database for large-scale face recognition that contains millions of images.…”

Machine Learning-Assisted Determination of C6H14 Mole Fraction From Molecular Emissions of Laser-Induced Hexane–Air Plasmas

“…These results parallel the performance of similar models trained to perform chemical analysis of complex actinide metals. 30 The versatility of the ANN structure lends it to be more…”

“…These results parallel the performance of similar models trained to perform chemical analysis of complex actinide metals. 30 The versatility of the ANN structure lends it to be more flexible in training on spectral data sets with features from multiple elements or molecular species, which in turn enables better predictive performance. The quantitative metrics evaluating the performance of all five trained and tested models are summarized in Table III, with the best achieved RMSEP and LOD in bold.…”

“…27,28 ML regression methods based on decision trees and neural networks have gained popularity for nuclear material analysis, providing calibration models with superior precision and sensitivity for quantifying alloying metals in plutonium. 29,30 Recently, Li et al 31 implemented data fusion with Raman and near-infrared spectral features of gasoline for high-precision quantification of methanol. The dynamic complexity of molecular emission features seen in plasmas containing hydrocarbons makes ML constructs an ideal option for developing high-fidelity analytical models.…”