#5. Data-driven and physics-informed multiscale materials modelling

Organizers

  • Tilmann Hickel, MPIE, Germany (hickel@mpie.de)
  • Dennis M. Dimiduk, The Ohio State Universit; BlueQuartz Software, US
  • Sean Donegan, Air Force Research Laboratory, USA
  • Michael Ortiz, California Institute of Technology, USA
  • Katrin Schulz, Karlsruhe Institute of Technology, Germany
  • Laurent Stainier, École Centrale Nantes, France
  • Daniel Urban, Fraunhofer IWM, Germany
  • Isao Tanaka, Kyoto University, Japan

Description

This symposium covers innovative high-throughput and materials-informatics approaches for multiscale mechanics of materials and the discovery and design of novel materials with targeted mechanical and functional properties. The paradigm of a theory-guided data-driven materials research, which is based on the innovative methodologies of the modern information society, is currently extending the traditional means of material science, which are based on fundamental principles and empirical wisdom from experiment, theory, or simulation. The challenge is to use recent developments in the fields of data mining, machine learning, and artificial intelligence for the identification of structure-composition-property relationships in the highly diverse, but often sparse materials data space.

Just recently, experimental methods such as high-resolution transmission electron microscopy and various tomography techniques have experienced significant technological improvements and data are widely available. Simultaneously, numerical modeling has progressed towards the consideration of extended materials defects and microstructures from the atomistic up to a continuum level. This growing applicability of data-driven and physics-informed strategies that account for the multiscale character of materials shall be further encouraged by this symposium.

The symposium, therefore, intends to gather scientists who employ numerical simulations or combinatorial experiments for the high-throughput screening of materials, make use of machine learning approaches, model order reduction or surrogate modeling techniques for analyzing large amounts of materials data. We welcome contributions to any developments of computational or experimental high-throughput techniques for accumulating, analyzing, interpreting, storing, and sharing fundamental knowledge about materials in efficient ways. Contributions may range, and preferably bridge, from physics-based materials understanding to data-driven and application-oriented development.