Deep material network for multiscale nonlinear problems and virtual product design

Prof Chuin-Shan (David) Chen

National Taiwan University

Abstract: Deep Material Network (DMN) is a micromechanics-informed deep-learning architecture for multiscale materials modeling that uses a hierarchical network of mechanistic building blocks with analytical homogenization solutions to ensure physical consistency. Notably, DMN has demonstrated the ability to extrapolate from linear-elastic training data to complex nonlinear behaviors such as plasticity and damage with remarkable accuracy. Recent advances extend DMN’s generality via graph neural network (GNN) integration, which extracts microstructural features to parameterize the network and thus enables accurate predictions across multiple disparate microstructures. In parallel, new foundation model approaches leverage large pre-trained DMNs to enable efficient transfer learning, dramatically reducing the domain-specific data required for high-fidelity material predictions. DMN has transitioned from academia to industry through integration into virtual product design tools (e.g., ANSYS LS-DYNA), powering real-world simulations from automotive crash analyses and electronics drop tests to giga-casting property prediction. This development exemplifies how academic–industry collaboration can accelerate progress in concurrent multiscale simulation, product design, and ICME (Integrated Computational Materials Engineering). In this semi-plenary talk, I will overview the DMN framework and its latest enhancements, highlighting their generality and impact on multiscale simulation and virtual product design.

Bio: Prof. Chuin-Shan (David) Chen is a Distinguished Professor at National Taiwan University (NTU), with joint appointments in the Department of Civil Engineering and the Department of Materials Science and Engineering. He currently serves as President of the Association of Computational Mechanics Taiwan (ACMT) and as an Executive Council Member of both the International Association for Computational Mechanics (IACM) and the Asian-Pacific Association for Computational Mechanics (APACM). Prof. Chen has been recognized with numerous honors, including the IACM Fellow Award (2012) and the APACM Computational Mechanics Award (2019). His current research focuses on integrating physics-based modeling with emerging artificial intelligence techniques to advance next-generation industrial product design, multiscale materials modeling, and manufacturing simulation.