Improving and accelerating materials development is an important goal in science and industry, as innovative, tailored, and optimized materials ranging from metals and polymers to composites and architected materials are key to a sustainable future. With recent advances in machine learning and manufacturing technologies, tailoring the chemical composition and the microstructure of classical materials as well as the internal structure of architected materials for a targeted property, i.e., inverse design, has become feasible. The inverse design process requires techniques to characterize and reconstruct numerical models of the local material structure. Computational homogenization of these microstructures facilitates prediction and understanding of the interplay between effective properties and microstructural features of complex materials.
To describe the nonlinear, inelastic effective behavior of materials with high precision while explicitly considering design variables, material models are being enhanced by machine learning that adhere to fundamental physical constraints. This enables improved extrapolation and the use of sparse training data, while maintaining a high degree of modeling flexibility. Furthermore, model discovery techniques makes these models interpretable and explainable, increasing their trust in critical applications.
Topics of interest covered within this mini-symposium include but are not limited to:
• techniques for exploration and inversion of process-structure-property linkages
• inverse design approaches for metals, polymers, composites, and architected materials
• design approaches that account for crucial manufacturing constraints
• microstructure characterization and reconstruction, e.g., 2D and 3D image-based methods definition of descriptors
• numerical and experimental analysis of designer materials across scales
• constitutive modeling with NNs (elasticity and inelasticity)
• sparse regression, bayesian learning, symbolic regression
• interpretability and fulfillment of physical constraints
• acceleration of multiscale simulations with NNs
• identification and parametrization of constitutive models from experiments
• use of full-field data for data generation/model training
• data-driven identification for the extraction of stress-strain pairs from experiments
Organized by: K. Meyer , M. Kästner , K. Kalina , D. Kochmann , J. Fuhg , S. Kumar , D. Seidl and N. Vlassis
Keywords: Architected Materials, Constitutive models, Data-driven approaches, Inverse Design, Machine Learning, Microstructure, Neural Networks, PANN, PENN, Process-Structure-Property Linkages