Fuzzy-Empowered Decision Making Integrated with DDDAS-Matrix Profile Framework for Anomaly Detection in Radiation Measurements

This paper explores the field of nuclear security by employing advanced data analytics to detect unusual patterns in radiation measurements linked to illicit nuclear activities. A significant challenge in this analysis is the pervasive presence of background radiation, which often masks signals from illegal radioactive sources. Therefore, accurately estimating and subtracting background radiation from measurements is crucial for improving the detection of hidden nuclear activities. In this study, a fuzzy empowered decision-making module is integrated with the synergistic framework of the Dynamic Data Driven Applications Systems (DDDAS) paradigm with the Matrix Profile method (MP) for anomaly detection in radiation spectra. The Matrix Profile method creates a background model from a single radiation measurement obtained from a sensor, whereas DDDAS provides a way for updating background models by assimilating new incoming measurements. The proposed fuzzy logic empowered module enhances decision-making by taking into consideration several parameters and eliminating the need for a hard decision threshold. Results from a comprehensive dataset containing both pristine background and source-influenced measurements show a significant improvement of the DDDAS-MP framework in anomaly detection accuracy through the utilization of the fuzzy logic decision making.