MIE Special Seminar: High-performance Computational Science and Engineering: A Path Towards Virtual Materials Testing
The Department of Mechanical and Industrial Engineering
Professor Karel Matous
Department of Aerospace & Mechanical Engineering
University of Notre Dame
367 Fitzpatrick Hall of Engineering
Notre Dame, IN 46556
Topic: High-performance Computational Science and Engineering: A Path Towards
Virtual Materials Testing
Date: Friday, February 28, 2014
Time: 11:30am to 1:00pm (Refreshments from 10:30am to 11:00am)
Location: 333 Curry Student Center
Abstract: With concentrated efforts from the material science community over the past several decades to
Develop new multi-functional materials, that inherently span several length scales due to the presence of disparate phases, the need for modeling tools that accurately describe the physical phenomena at each scale has only further been emphasized. For example, knowledge of the size of a failure zone is crucial in numerical modeling of damage since it introduces a length-scale that must be numerically resolved for the physics of the problem to be captured. Opposite to the characteristic length of damage localization is the macroscopic size of the structure, O(m), and the size of a material domain (i.e., Representative Unit Cell (RUC)) that characterizes the material morphology with sufficient detail, O(mm). Numerically resolving such a wide range of scales, O (103) from macro-to-micro and O (103) from micro-to-damage zone, is a challenge.
In this lecture, I will present a multiscale theoretical and computational framework for modeling the macroscopic/mesoscopic behavior of heterogeneous materials and heterogeneous material layers in particular. I will show that this framework is capable of capturing the large range of spatial scales. The novel computational multiscale cohesive model is based on the variational energy equivalence, Hill's Lemma. Thus, we efficiently couple physical processes at the mesoscale to the macroscopic response with a point-wise attached heterogeneous meso-continuum in order to derive a homogenized cohesive law. Simulations involving this wide range of scales are inherently expensive, requiring the use of high- performance computing. Therefore, I will discuss a high-performance computational solver that executes on thousands of processing cores. I will present highly nonlinear simulations consisting of over 1 billion computational cells that lead to more than 574 million nonlinear algebraic equations.
Any serious attempt to model a heterogeneous system must also include a strategy for constructing a complex computational domain. This work follows the concept of data-driven (image-based) modeling. I will delineate a procedure based on topology optimization to construct a RUC with the same statistics (n-point probability functions) to that of the original material, which is tomographically characterized. Our current micro-computed-tomography (micro-CT) instrument provides tomographic imaging of a material structure with ~3 µm resolution, resulting in a direct link between experimental observations and the material system. We show that high-performance direct numerical simulations of these statistically meaningful mesoscopic domains are possible. Understanding simulation data consisting of billions of computational cells and millions of highly nonlinear equations requires data-mining, and I will present a strategy related to damage patterns. Therefore, well resolved mesostructure-statistics-property relationships can be obtained.
Finally, micro-CT can become a much more powerful tool by combining it with quantitative abilities for complete comparison with the corresponding simulations. In particular, the resulting stress- strain measurements and displacement data, although limited to moderate levels of quasi-static thermal and mechanical loads, are providing unprecedented detail of a fully 3D deformation field, including in
situ damage propagation.
Bio-Sketch: Dr. Matouš is an Associate Professor in the Aerospace and Mechanical Engineering Department at the University of Notre Dame. He received his M.S. and Ph.D. in the Theoretical & Applied Mechanics from the Czech Technical University in Prague. Dr. Matouš’ interests are in the area of predictive computational science and engineering at multiple spatial and temporal scales including multi-physics interactions, the development of advanced numerical methods and high-performance parallel computing. His research focuses on the interplay between applied mathematics, computer/computational science and physics/materials science. Moreover, he developed a research program in micro-tomography (micro-CT) based computational and experimental modeling of heterogeneous materials focusing on co-designed simulations and experiments based on statistically representative analysis. Dr. Matouš leads the Computational Physics Team and is the Software Architect for the Center for Shock Wave-processing of Advanced Reactive Materials (C-SWARM, $8 Millions over 5 years) that has been established as one of six National Nuclear Security Administration's center of Excellence. The center is devoted to the development of a parallel multiscale and multi-physics computational framework for predictive science. In particular, C-SWARM’s goal is to predict the behavior of heterogeneous materials, specifically the dynamics of their shock-induced chemo-thermo-mechanical transformations and resulting material properties. He has authored or co-authored more than thirty refereed journal papers and sixty conference proceedings articles and abstracts. As the PI or co-PI, he has directed research projects of a combined value over $10 millions. Dr. Matouš received the Rector's Award for the best Ph.D. students from the Czech Technical University in Prague. Two articles from Dr. Matouš’ group have been featured on Science Direct Top 25 Hottest Articles in their respective engineering areas. He is a member of ASME, SES, USACM, and IACM. Dr. Matouš is a Fellow of ASME.
Friday, February 28, 2014 at 11:30am to 1:00pm
Curry Student Center, 333
346 Huntington Avenue, Boston, MA, Boston