MBSE Colloquium: John G. Michopoulos, "Inverse Multiscale Design of Materials"
Monday, December 5, 2011
1146 A.V. Williams Building
301 405 6579
Model-Based Systems Engineering Colloquium
On A Framework for Data and Specification-Driven Inverse Mulitscale Design of Materials
John G. Michopoulos
Naval Research Laboratory
| video |
We are presenting an overview of a framework for treating material behavior and material properties as design models and variables respectively, in the context of designing materials that need to perform such as they satisfy specific data- and specification-driven performance requirements. After an overview of the general principles and strategies two examples will be presented at the opposite ends of the length scale spectrum. In the atomistic scale level we demonstrate the determination of the Lenard-Jones (L-J) potential parameters that govern the dynamics of a problem of fracture such that a specific load history of the associated medium is satisfied. The problem is approached from an inverse problem perspective. A global Monte Carlo optimizer along with a legacy molecular dynamics code are implemented on Graphical Process Unit computational infrastructure, to compute the design variables of the problem that in this case are the L-J constants. In the macro scale we demonstrate the determination of a general set of material parameters that define the elastic and inelastic (with damage) constitutive response of an anisotropic medium made from fiber-epoxy composite material laminates. The specification requirements to be satisfied are those of a set of load-displacement histories in the full 6+6 dimensional kinematic space. The problem is also approached from an inverse problem perspective. Instead of arbitrarily specifying the load-displacement histories of a material system as the target performance of the system, we used our custom-developed recursive 6-DoF robotic loader to collect the actual kinematic responses of an existing material system. The desired material characterization problem is homomorphic to the problem of determining the material parameters when the behavior is known. The determination of the design (material) constants is achieved by minimizing objective functions such as the difference between experimentally measured and analytically computed system responses as described by strain fields and surface strain energy densities. Small and large strain formulations based on strain energy density decompositions are developed and utilized for determining the constitutive behavior of composite materials. Finally, examples based on both synthetic and actual data demonstrate the successful application of design optimization for constitutive characterization.
As a Research Scientist/Engineer and director of Computational Multiphysics Systems Lab (CMSL) of the Center of Computational Materials Sciences at the Naval Research Laboratory (NRL), Dr. Michopoulos oversees multi-physics and information technology (IT) research and development, operations and initiatives at CMSL. Current major initiatives include research and development of linking performance to material trough data and specification driven methodologies, electromagnetic launcher dissipative mechanism modeling and simulation, heterogeneous integrated computational, sensing and communication grids via data-driven multidisciplinary and holistic approaches and environments, engineering sciences research, development and management in areas of computational, theoretical and experimental multiphysics, platform/structure simulation based design, mechatronic/robotic data-driven characterization of continua, automation of research, distributed supercomputing, and multiphysics design optimization. Dr. Michopoulos also currently serves as the Chair of the Computers and Information in Engineering Division of the American Society of Mechanical Engineers. He is a member of the editorial board if the Journal of Computers and Information Science in Engineering and the Journal of Computational Sciences. He is a founding member of the International Science and Technology Outreach Society and prior to joining NRL he has been a senior research scientist for Geo-Centers Inc and prior to that director of the Image Processing Laboratory of the Institute of Fracture and Solid Mechanics at Lehigh University. He has also served as founding member and chief technology officer of three companies. He has authored and co-authored more than 230 publications and books and has been honored with 56 performance and leadership awards. Dr. Michopoulos holds an AS degree in Electronics and Telecommunications from the Capitol Radio Engineering Institute, a M.Sc. in Civil Engineering and a Ph.D in Applied Mathematics from the National Technical University of Athens, Greece, and has pursued post-doctoral studies at Lehigh University Bethlehem PA, on computational multi-field modeling of continua and Fracture Mechanics.