Rapid imaging of special nuclear materials for nuclear nonproliferation and terrorism prevention

We introduce a neutron-gamma emission tomography (NGET) technique for rapid detection, three-dimensional imaging, and characterization of special nuclear materials like weapons-grade plutonium and uranium. The technique is adapted from fundamental nuclear physics research and represents a previously unexplored approach to the detection and imaging of small quantities of these materials. The method is demonstrated on a radiation portal monitor prototype system based on fast organic scintillators, measuring the characteristic fast time and energy correlations between particles emitted in nuclear fission processes. The use of these correlations in real time in conjunction with modern machine learning techniques provides unprecedented imaging efficiency and high spatial resolution. This imaging modality addresses global security threats from terrorism and the proliferation of nuclear weapons. It also provides enhanced capabilities for addressing different nuclear accident scenarios and for environmental radiological surveying.

MACHINE-LEARNING ALGORITHM FOR SHIELDED SPECIAL NUCLEAR MATERIALS DETECTION

A machine-learning algorithm has been implemented by use of a neural network as a preliminary study on the applicability of this method to special nuclear materials detection. The algorithm predicts the presence of the 238U isotope when learning from a gamma spectrum data measured with a high-purity germanium detector from a sample of depleted uranium. In this work, both a fully connected neural network and a convolutional neural network have been implemented, and the performance of different configurations of the network has been studied. The use of convolutional network showed better performance over the fully connected network, with cost function and success rate values supporting a better prediction while avoiding overfitting. Furthermore, implemented network features such as filtering, max-pooling, dropout regularization, and momentum optimization also showed improved prediction performance.