Collaborative Research: Elements: OpenMSG: a cloud-based multiscale modeling toolkit for heterogeneous materials and structures

About This Grant

Advanced materials such as composites, metamaterials, soft materials, and architected materials are inherently heterogeneous and multiscale in nature. Currently, multiscale modeling serves as the most effective approach for analyzing and designing these materials. However, the growing complexity of microstructural features and macroscopic structural configurations presents significant challenges to achieving both computational efficiency and predictive accuracy. While emerging machine learning (ML) models offer a cost-effective alternative, their effectiveness is often limited by the lack of high-quality training data in many real-world engineering applications. Moreover, advanced ML techniques are still not routinely incorporated into traditional mechanics or materials engineering curricula. To address these challenges and support both research and education in multiscale material and structural modeling, this project supports research that develops a cloud-based cyberinfrastructure that integrates new multiscale modeling theory with multi-fidelity ML models. This platform seeks to provide open-access tools, curated datasets, and comprehensive training resources to advance materials science, enable efficient structural analysis and design, and support workforce development in ML-assisted material and structural modeling. The goal of this project is to develop a cloud-based, open-source multiscale modeling software called OpenMSG, providing an ultra-efficient prediction toolkit for mechanical and multiphysics behaviors of highly heterogeneous materials and structures. To achieve the goal, this project first develops new multiscale models based on mechanics of structure genome (MSG), which can discretize analysis domains using efficient beam and shell elements while still considering strong material heterogeneity and anisotropy. The new models will provide an unprecedented combination of computational efficiency and accuracy and generate highly correlated multi-fidelity data. Building on these multiscale models, the project then develops a framework using multi-fidelity neural networks (NNs) for ML-assisted multiscale modeling. This hybrid approach further reduces the computational burden while preserving model accuracy across design spaces. The models and framework are demonstrated using additively manufactured functionally graded materials and composite blade designs, which also showcase OpenMSG’s capabilities in advancing the fundamental understanding of heterogeneous material behavior and facilitating the efficient design of complex engineering structures. OpenMSG will be developed, tested, and maintained on the widely recognized Composites Design and Manufacturing HUB (cdmHUB) with the support of CI experts from HUBzero. Leveraging established user bases and infrastructure on cdmHUB, this project delivers not only a cutting-edge multiscale modeling tool but also fosters a sustainable global user community dedicated to data-driven multiscale materials and structural modeling. This award by the Office of Advanced Cyberinfrastructure is jointly supported by the Division of Civil Mechanical and Manufacturing Innovation within the Directorate for Engineering. This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.