Nuclear Analytics: Enabling New Multimodal Approaches to Nuclear Threat Detection

Jonathan Dreyer | 20-ERD-045

Project Overview

This project investigated new approaches to noninvasive nuclear monitoring that provide useful contextual information with more flexible and sensitive capabilities for early warning of nuclear proliferation. Current radiation detectors do not provide contextual information and must be monitored continuously for signatures of materials of concern (fissile materials, radiological hazards, etc.) that exceed a fixed threshold. Radiation sensing also has limited range and does not always provide adequate information for a conclusive assessment. The effectiveness is further limited by the weak signatures of most materials of concern, the ease of reducing those signatures with shielding, and the confusion with legitimate radiation sources and background radiation. Non-invasive, non-radiation sensors provide additional contextual information, and provide the user and subsequent analyses with relevant information about the vehicle or container (such as estimates of the type of object and its size, mass, velocity, time, and distance of closest approach). To exploit multimodal measurements and extend our capabilities beyond what is available from the deterministic methods currently used for analysis of radiation sensors, we designed, procured, and deployed a custom measurement testbed, developed physics-informed features from contextual sensors, and created a software infrastructure to investigate machine-learning approaches to fusing data from contextual sensors. This project established a unique capability for evaluating a variety of multimodal measurement system configurations and produced datasets to foster further growth of nuclear analytics.

Mission Impact

The proliferation of nuclear weapons is a top U.S. priority for 21st century national security, and the ability to prevent proliferation of nuclear weapons faces increasing challenges in the shifting global environment. Critical to this effort is the ability to provide early warning of proliferation activities. Radiation measurements are a key component of current early-warning detection of nuclear proliferation activities. Multimodal sensor systems (measurement systems with multiple types of sensors) can improve detection sensitivity and enable new deployment options for monitoring nuclear activity. This work has established a unique capability for evaluating a variety of multimodal measurement system configurations and produced datasets to foster further growth of nuclear analytics.

Publications, Presentations, and Patents

Bonnie E Canion, 2021. "Nuclear Material Movement ERC." Livermore, CA. LLNL-PRES-826223.