Advanced Computing, Mathematics and Data
Inverse Modeling of X-ray Imaging to Combat Nuclear Materials Trafficking
Algorithm could improve detection capabilities in current X-ray detectors
Simulated inspection of a layered, baggage-like object that contains a thin, shielded plutonium wedge (a) (not to scale). A single energy radiograph is shown in (b), along with the material estimations from the adaptive inverse algorithm that indicate an equivalent of color vision (by material) for a single-view radiograph using spectral X-ray detector data. Enlarge Image.
Results: As part of the ongoing effort to curtail illicit trafficking of nuclear materials, scientists from Pacific Northwest National Laboratory and the University of Texas at Austin studied inverse modeling of spectral X-ray radiography to determine the presence of nuclear materials in small containers or composite objects, such as baggage. In their examination, the researchers applied an inversion algorithm to synthetic radiographs of objects composed of layers of plutonium, cotton, steel, lead, aluminum, and copper to estimate the quantities of these materials within the layers. Their work showed that coupling the algorithm with current-generation commercial detectors has the potential to distinguish small quantities of nuclear materials from other high-Z materials, such as nuclear material stowed within airline passenger luggage.
Why it Matters: Extending existing imaging technology to detect small amounts of plutonium or highly enriched uranium hidden in composite objects could provide a major advance in combating the threat of nuclear materials trafficking.
Methods: Regularized inversion techniques, such as those Andy Gilbert (the paper’s lead author) applied to this problem, have proven effective in distinguishing between multiple components of a given signal in a variety of applications. Using a list of possible materials and associated X-ray attenuation profiles as basis functions that are coupled to Beer’s Law as a forward model to account for attenuation by photon absorption and scattering, the inversion breaks down the overall attenuated signal in each image pixel into components from the individual materials. These components are the coefficients in the forward model, representing areal densities of materials. Importantly, the algorithm has been developed to be adaptive, requiring minimal user input if deployed. Despite complicating factors associated with imaging high-density, high-Z objects, these techniques show promise for nuclear trafficking applications.
What’s Next? Building on the success of inverse modeling and using today's spectrally sensitive detectors to detect the presence of small amounts of plutonium in the simulated composite objects tested to date, the scientists are examining the model’s potential to adapt and expand to high-energy X-ray systems and multiple X-ray spectra, as well as its application to verification of arms control treaty-limited items.
Acknowledgments: This research was supported by the U.S. Department of Energy National Nuclear Security Administration (NNSA) Office of Defense Nuclear Nonproliferation Research and Development (NA-22), as well as by the Next Generation Safeguards Initiative under the Office of Nuclear Safeguards and Security supported by NNSA’s Office of Nonproliferation and International Security.
Research Team: The PNNL team was led by Principal Investigator Ken Jarman from the Computational Sciences & Mathematics Division's Computational Mathematics group and includes Andy Gilbert, Ben McDonald, and Tim White from the National Security Directorate’s (NSD) Simulation & Analysis group and Sean Robinson from NSD’s Human-Centered Analytics group. Co-author Mark Deinert is from the University of Texas at Austin.
Reference: Gilbert AJ, BS McDonald, SM Robinson, KD Jarman, TA White, and MR Deinert. 2014. “Non-invasive material discrimination using spectral x-ray radiography.” Journal of Applied Physics 115(15):154901. DOI: 10.1063/1.4870043.