The event is free of charge
May 18, 2022 - May 19, 2022
In future manufacturing, the design of materials to exactly match application purposes and the design of processes to produce workpieces with desired material microstructures and properties plays a crucial role. Sophisticated materials and process designs not only enable the production of highly performant products but also reduces production costs and makes production more sustainable for example by avoiding defect products. To tackle this challenging materials science problem, the application of data-driven methods from the highly dynamic field of machine learning are promising.
In the ongoing DFG research project Taylored Material Properties via Microstructure Optimization: Machine Learning for Modeling and Inversion of Structure-Property-Relationships and the Application to Sheet Metals, we committed ourselves to push forward the development of data-driven methods for materials and process design by bringing together the communities of materials science, of data-driven modeling and machine learning, specifically the communities working on data-driven materials design and process design.
Materials design, incorporating the modeling and optimization of microstructure-property relations, as well as the inverse modeling of these relations using data-driven methods.
Process design, incorporating any kind of tools to design processes and process chains, such as applications of reinforcement learning and surrogate modeling of process simulation in order to reach targeted microstructures or properties.
Process control, incorporating process control approaches that link and optimize processing-structure-property relations, as well as issues like interoperability of data and process interfaces.
Opportunities for data-driven methods in materials sciences, incorporating the presentation of recently developed machine learning methods that are promising for future developments in materials and process design (addressing for example the modelling of uncertainty, generative models, optimal experimental design, the integration of domain knowledge).
For each session, invited keynote speakers will share their thoughts and give impulses. Besides, we encourage interested researchers to present their work and ideas on the above-mentioned topics (presentation, tutorials, live demos, …).
Dr. Dirk Helm (Fraunhofer IWM)
|9:45||Machine Learning for Process and Alloy Design
Prof. Dierk Raabe (Max-Planck-Institut für Eisenforschung)
|10:30||Taylored material properties via microstructure optimization: A general introduction with focus on data generation
Lukas Morand (Fraunhofer IWM)
|11:15||Progress and Challenges in Physics-Aware Machine Learning for Material Modeling
Prof. Felix Fritzen (University of Stuttgart, Institute of Applied Mechanics)
|12:00||Structure-Guided Processing Path Optimization with Deep Reinforcement Learning
Dr. Johannes Dornheim (Karlsruhe Institute for Technology, Institute for Applied Materials)
|13:30||Smart forming processes: real-time state estimation with hybrid models
Jos Havinga (University of Twente)
|14:00||A multi-task learning-based optimization approach for finding diverse sets of material microstructures with desired properties
Tarek Iraki (Karlsruhe University of Applied Sciences)
|14:45||Service-oriented streaming platform for machine learning in production
Dr. Christian Kühnert (Fraunhofer IOSB)
|15:15||Virtual get-together with demonstrations|
|16:30||End of day 1|
Dr. Dirk Helm (Fraunhofer IWM)
|10:45||Addressing Annotated Data Scarcity and Materials Diversity with Advanced Deep Learning Architectures
Ali Riza Durmaz (Fraunhofer IWM)
|11:15||Physics-based machine learning via correcting analytical model predictions towards high-fidelity simulation solutions – a hybrid modelling approach
Frederic Bock (Helmholtz-Zentrum Geesthacht, Institute of Materials Mechanics)
|12:00||Accelerating Material Modeling with Machine Learning
Dr. Jaber Mianroodi (Max-Planck-Institut für Eisenforschung)
|12:30||Convolutional neural networks for efficient computational homogenization
Dr. Fadi Aldakheel (Leibniz University Hannover & Swansea University)
|14:00||Deep learning driven multiscale simulation in the presence of uncertainties
Alexander Henkes (Technische Universität Braunschweig)
|14:30||Machine learning advances in mechanics: addressing the elephant in the room
Miguel Bessa (Delft University of Technology, Department of Materials Science and Engineering)
|15:15||Artificial Intelligence and High-Performance Data Mining for Accelerating Materials Discovery and Design
Prof. Ankit Agrawal (Northwestern University, Department for Electrical and Computer Engineering)
|16:00||Beyond forward ICME models: A perspective on materials design as inverse problems
Dr. Anh Tran (Sandia National Laboratories, Department of Optimization and Uncertainty Quantification)
|16:30||Virtual get-together with demonstrations|
|17:30||End of workshop|