Speaker
Description
Rapid and reliable localization of radioactive sources is a critical requirement in nuclear security, border protection, radiological emergency response, and the prevention of illicit trafficking of radioactive materials. In our previous work, we investigated radioactive source position estimation using a coded aperture device combined with machine learning, demonstrating the potential of data-driven methods for directional radiation sensing [1]. Although coded-aperture systems can provide strong localization capability, they are generally more complex in design and reconstruction and may involve higher implementation cost. In the present work, we propose a simpler and more compact directional detector architecture intended for security applications, while retaining high angular localization performance.
The proposed system consists of three small-form-factor CZT detectors, each with a crystal volume of 0.5 cm³, arranged azimuthally at 120° with respect to each other. To enhance the directionality of the system, each detector is asymmetrically shielded using a lead cone placed around the crystal, together with a cylindrical lead endcap at the bottom (Figure 1). This shielding configuration modifies the angular response of each detector and enables the extraction of source-direction information from the relative count-rate pattern of the three-detector array. An important practical advantage of the system is that it operates without any moving parts, making it more robust, simpler, and potentially more suitable for field deployment in operational security scenarios.
The system was studied using GEANT4 [2] simulations. A 1mCi 137Cs source was placed at a distance of 50 m from the center of the detector system and simulated at azimuthal angles from 0° to 355° in steps of 5°. For each source position, the total counts recorded by each detector over a fixed acquisition time interval were used as the input features to the machine-learning model. A gradient-boosted decision tree model, configured with 100 trees, was developed within the KNIME analytics platform [3] and trained on the simulated dataset and subsequently evaluated on an independent validation dataset.
The results demonstrate that the proposed approach can determine the source direction with an angular accuracy of 2°, showing strong potential for practical directional radiation sensing. Owing to its compact design, lower complexity relative to coded-aperture approaches, and lack of moving components, the proposed system is particularly attractive for fixed-site monitoring, mobile security platforms, and rapid deployment in nuclear security applications. Experimental validation of the detector system is currently in progress.
[1]K. Karafasoulis, I. Kaissas, C. Papadimitropoulos, K. Potiriadis, C.P. Lambropoulos, “A machine learning approach in the estimation of a radioactive source position using a coded aperture device,” Journal of Instrumentation (JINST), 18, C01062, 2023.
[2]GEANT4 collaboration, GEANT4 — A Simulation Toolkit, Nucl. Instrum. Meth. A 506 (2003) 250
[3]M.R. Berthold, N. Cebron, F. Dill, T.R. Gabriel, T. Kötter, T. Meinl, P. Ohl, C. Sieb, K. Thiel, B. Wiswedel, “KNIME: The Konstanz Information Miner,” Studies in Classification, Data Analysis, and Knowledge Organization (GfKL 2007), Springer, 2007.