Assessment of Fuzzy Logic Radioisotopic Pattern Identifier on Gamma-Ray Signals with Application to Security

Abstract

Analysis of acquired nuclear detector gamma-ray signals for recognition of present radioisotopic signatures is crucial to national security and security applications. Identification algorithms must be accurate and rapid. Artificial intelligence is a scientific field with a variety of tools suitable to implement automated processing of nuclear signals. The use of low resolution portable detectors to measure gamma-ray signals has found a wide use in security and safeguards applications. In this paper, the fuzzy logic based analysis methodology that has been previously developed is applied and assessed on a variety of nuclear signals obtained with a low resolution scintillation detector, and more particularly a sodium iodide (NaI) detector. Various types of fuzzy membership functions are employed and their performance is assessed with regard to the number of positive detections, misses, and false alarms. Furthermore, recorded results from the set of low resolution gamma ray signals are used to estimate the detection sensitivity for each membership function. Results demonstrate the overall effectiveness of the fuzzy logic based identifier, and consist of the main course for the assessment of each membership function. Furthermore, comparison of results designates the triangular membership function as the best membership shape for this type of detector signals.

Nuclear Power Plants With Artificial Intelligence in Industry 4.0 Era: Top-Level Design and Current Applications—A Systemic Review

“…McCoy et al [91] put forward a conceptual model of NPP operation safety and monitoring, which uses interactive component calibration software to comprehensively monitor NPP operation activities. The analysis method based on fuzzy logic is applied to deal with various nuclear signals obtained by lowresolution scintillation detectors, especially sodium iodide (NaI) detectors [92]. Lagari et al [93] designed four differ- ent ANNs to automatically determine the nuclear resonance characteristics according to the experimental cross-section data in a fast and reliable way, which made the modeling and simulation more reliable, thus reducing the degree of experimental testing.…”

Survey of Machine Learning Approaches in Radiation Data Analytics Pertained to Nuclear Security

“…Inevitably, the field of nuclear security was also influenced by machine learning, with several methods been developed or are under development. Though these methods have different aims, their ultimate goal is to support the analysis of data to identify patterns of interest that are correlated with the use of nuclear materials [9].…”

Application of Fireworks Algorithm in Gamma-Ray Spectrum Fitting for Radioisotope Identification

“…Further, application of B-spline functions for least square fitting is presented in (Zhou et al, 2009), while spectrum fitting using Levenberg-Marquardt optimization is discussed in (Fatah & Ahmed 2011). Genetic algorithms are adopted for searching for an optimal combination of template spectra in (Carlevaro et al, 2008), and fuzzy logic based approaches in (Alamaniotis et al, November 2009) and (Alamaniotis et al, January 2014), while a hybrid evolutionary-fuzzy in (Huang et al 2011). Other proposed methodologies employed Bayesian Statistics such as the work described in (Stinnett & Sullivan, 2013), likelihood ratios (Forsberg et al, November 2009), Local-Global graphs (Bourbakis et al, 2008), and neural networks (Yoshida et al, 2003).…”