DDPS | ML-driven Models for Material Microstructure and Mechanical Behavior by Lori Graham Brady

Description: The mechanics describing material behavior leading to failure is often associated with microstructural features of the material. The random variability of these microstructures leads to uncertain material response. This talk describes approaches to address two challenges: 1) digitally generating ensembles of microstructures that represent key features of the experimentally obtained microstructures; and 2) leveraging limited physics-based analyses of these microstructures to train machine learning (ML) models that capture key localizations. The first challenge of digitally generating microstructures requires that the simulation process captures all the salient features of the microstructure, which may or may not be addressed by simply considering 2- or even n-point correlations. Furthermore, extrapolating from two-dimensional images of experimentally obtained microstructure to digitally generated three-dimensional images is critically important to realistic modeling of the microstructural response. behavior. This talk describes a transfer learning approach to rapid generation of three-dimensional microstructures describing different families of material microstructures. The second challenge of leveraging physics-based analyses makes use of these digitally generated microstructures as a means of creating training data. ML models that connect microstructural images to properties and/or contour plots of local stresses, strains and damage accelerate the analyses by orders of magnitude when compared to the physics-based models. These accelerated representations are particularly important in the context of materials design, in which rapid assessment of samples from a very high-dimensional design space is necessary for any realistic optimization approach. These representations also support real-time decisions in the context of high-throughput experimentation on materials.
Bio: Lori Graham-Brady is Professor and former Chair of the Civil and Systems Engineering Department at Johns Hopkins University, with secondary appointments in Mechanical Engineering and Materials Science & Engineering. She currently serves as Director of the Center on AI for Materials in Extreme Environments and Associate Director of the Hopkins Extreme Materials Institute. Her research interests are in machine-learning-enabled mechanics models, uncertainty quantification, computational stochastic mechanics, multiscale modeling of materials with random microstructure and the mechanics of failure under high-rate loading. She has received a number of awards, including the Presidential Early Career Awards for Scientists and Engineers (PECASE), the Walter L. Huber Civil Engineering Research Prize, and the William H. Huggins Award for Excellence in Teaching. She is a Fellow of the ASCE Engineering Mechanics Institute and the US Association for Computational Mechanics.
DDPS webinar: www.librom.net...
💻 LLNL News: www.llnl.gov/news
📲 Instagram: / livermore_lab
🤳 Facebook: / livermore.lab
🐤 Twitter: / livermore_lab
🔔 Subscribe: / livermorelab
About LLNL: Lawrence Livermore National Laboratory has a mission of strengthening the United States’ security through development and application of world-class science and technology to: 1) enhance the nation’s defense, 2) reduce the global threat from terrorism and weapons of mass destruction, and 3) respond with vision, quality, integrity and technical excellence to scientific issues of national importance. Learn more about LLNL: www.llnl.gov/.
LLNL-VIDEO-848847
#LLNL LivermoreLab #DataDrivenPhysicalSimulations