The National Nuclear Security Administration (NNSA), an agency the Department of Energy (DOE), responsible for enhancing national security through nuclear science, recently awarded ǿմý a $14.2 million grant for a new Center of Excellence (COE) named “Materials Data Science for Stockpile Stewardship (MDS3).”
The new center, MDS3, is directed by Case School of Engineering faculty members Roger French, Kyocera Professor of Materials Science and Engineering; Laura Bruckman, associate professor of materials science and engineering; and Yinghui Wu, Theodore L. and Dana J. Schroeder Associate Professor of Computer and Data Sciences. The mission of the new COE is to develop, demonstrate, and deploy novel data science tools, frameworks, codes, and computing infrastructure to advance understanding of materials degradation and the failure of materials, parts, and subsystems using novel computer science and data science. In addition, it aims to deliver a pipeline of diverse data-enabled workforce for the future.
ǿմý leads the effort in partnership with University of Central Florida and in collaboration with Lawrence Livermore National Laboratory, Los Alamos National Laboratory, Sandia National Laboratory and Kansas City National Security Campus. The tools and frameworks developed through MDS3 will be broadly applicable across NNSA programs and the computational and experimental research components of the MDS3 COE will demonstrate the full lifecycle of data-science-enabled stockpile stewardship to address challenging NNSA topics and missions. This unified cycle of learning will integrate stockpile surveillance of materials aging and lifetime, leading to actionable insights for advanced manufacturing design and production agency teams.
Understanding and transforming historical and newly generated data to extract scientific knowledge and insights is a non-trivial task, given both the complexity and size of the data. Advanced analytical tools and infrastructure are necessary to interpret and generate value from these datasets. In the MDS3 COE, researchers develop the physical, computational, and intellectual resources to understand, process, and build predictive models to predict the lifetime performance of materials and systems.
The MDS3 COE has a matrixed structure consisting of two computer science thrusts (Computing Infrastructure and Knowledge Management & Learning) that cross-cut with two materials science thrusts (Field-lab Aging & Reliability and Next-Generation Design Optimization & Production). The intersections of these thrusts contain defined data science and materials science research projects and provide an avenue to explore and establish a suite of capabilities and tools. These projects can evolve over the course of the Center of Excellence to fill collaborators’ needs.
The Department of Energy offers phenomenal instruments such as the synchrotron beamlines that provide terabytes of data, but analyzing such data to its fullest cannot be done on a personal computer at scale. An organized materials data science pipeline with distributed and high-performance computing is necessary. The evident need of such pipelines is highlighted by a project in the MDS3 COE on beamline powder X-ray diffraction at the Advanced Photon Source (APS), in collaboration with LANL, looking at metal alloys in situ using deep neural nets. A major challenge is to generalize and scale these models into general-purpose computational frameworks to support large-scale materials analytical pipelines.