Woodruff School of Mechanical Engineering

Faculty Candidate Seminar

Title:

Uncertainty Quantification and Stochastic Multiscale Methods in Mechanics of Materials

Speaker:

Dr. Johann Guilleminot

Affiliation:

University Paris-Est Marne-la-Vallee, France

When:

Monday, March 28, 2016 at 11:00:00 AM

Where:

MRDC Building, Room 4211

Host:

Surya Kalidindi
surya.kalidindi@me.gatech.edu
404-385-0555

Abstract

With the unceasing development of computational capabilities and advanced experimental facilities, researchers and engineers are now concurrently facing the identification, the representation and the simulation of material behavior over unprecedented levels of resolution – ranging from the nanoscale to the structural scale. In this context, accounting for parametric and model uncertainties is a highly topical issue of major importance, since the credibility of the multiscale predictions (and therefore that of all related tasks, such as design optimization) may be strongly impacted by information exchange and uncertainty mitigation across scales. By nature, such a modeling effort requires an interdisciplinary approach at the interface of applied mathematics, computational mechanics and materials science. Indeed, and in addition to the construction of robust computational solvers, the construction, calibration and validation of probabilistic representations are nowadays recognized as key ingredients for performing accurate simulations and multiscale data assimilation. The aim of this talk is twofold. First, we provide a self-contained treatment of information-theoretic models for tensor-valued random fields. Such models are typically used as random coefficients in stochastic differential operators, and can be calibrated by solving underdetermined statistical inverse problems with experimental, potentially multiscale data. In particular, they allow for the modeling of fluctuating physical properties, such as elasticity or diffusion fields, at mesoscale. Algorithmic issues related to sampling schemes are subsequently reviewed in a high-dimensional setting. In particular, a new strategy involving a family of stochastic differential equations is presented for random fields with values in arbitrary n-dimensional sets. Throughout this first part, the multiscale modeling of interfaces at nanoscale is discussed in detail, on the basis of Molecular Dynamics simulations. This class of problems is one of the most important in multiscale analysis, due to its critical contribution in failure mechanisms and surface effects, and is used as an illustrative, multidisciplinary benchmark problem. Second, we discuss several applications of interest in computational science and engineering, and more precisely in the modeling of composite manufacturing processes, with the modeling of anisotropic diffusion fields; in computational homogenization of heterogeneous microstructures with nonseparated scales; and in the modeling of hyperelastic materials. The latter notably paves the way for the simulation, upscaling and design of materials with stochastic material nonlinearities, such as polymers and soft biological tissues.


Biography

Dr. Johann Guilleminot is currently an Assistant Professor in the Multiscale Modeling and Simulation Laboratory (UMR 8208 CNRS) at the Universit´e Paris-Est Marne-la-Vall´ee, France. His research activities are mainly devoted to Uncertainty Quantification (UQ) in Computational Mechanics and Materials Science, including the multiscale analysis of heterogeneous materials and complex systems from the nanoscale to the macroscale, stochastic homogenization and statistical inverse problems for data-driven model calibration and validation. His research has benefited a wide range of multiscale and multiphysics problems, such as the modeling of composite manufacturing processes, atomistic-to-continuum coupling and the investigation of the (non)linear behavior of porous materials, polycrystalline microstructures and living tissues for example. He serves as co-chair for the UQ-related activities and member of the steering committee at the Laboratory of Excellence in Research (LabEx) on Multi-Scale Modeling and Experimentation of Materials for Sustainable Construction, and as chair for the graduate program in Mechanics of Materials and Structures at its current institution. He is the recipient of a multiyear Young Investigator Grant on Nonlinear Multiscale Mechanics, awarded by the French National Research Agency, and selected honors include prizes for best M.S. and Ph.D. theses, and the French Award for Excellence in Research. He obtained his Habilitation1, with majors in Applied Mathematics and Mechanics, from the Universit´e Paris-Est Marne-la-Vall´ee in 2014, and his Ph.D. and M.S. in Theoretical Mechanics and Mechanical Engineering from the Lille 1 University - Science and Technology in 2008 and 2005, respectively. He joined his current institution in 2009, and spent nine months as a visiting scholar in the Department of Aerospace and Mechanical Engineering at the University of Southern California in 2010.

Notes

Refreshments will be served.