The proposed work further extends existing research to (a) improve the usability of NIST hedgehog on an HPC cluster and its applicability to accelerate a physics-based simulation of an MPM model of…

The University of Utah's Office of Sponsored Projects received a $210,000 cooperative agreement from the National Institute of Standards and Technology (NIST) under the Measurement and Engineering Research and Standards program (CFDA 11.609) for the period January 1, 2026 through December 31, 2027. The award supports research extending NIST's computational capabilities and artificial intelligence applications to advanced manufacturing processes. The primary deliverables include an enhanced source code repository for a Material Point Method (MPM)-based simulation of laser sintering in additive manufacturing with improved performance capabilities, and peer-reviewed publications documenting the derivation of surrogate models from physics-based simulations and generalizations applicable across digital twin applications.

The research activities focus on developing AI-derived surrogate models from physics-based computational models to accelerate simulation predictions in laser sintering processes. Key work includes deriving and tuning a heat conduction surrogate model for both reliability and performance, quantifying associated uncertainties, and integrating the surrogate model into a digital twin framework for evaluation and steering of physics-based simulations. The project leverages the DOE Kokkos library and NVIDIA CUDA computing architecture to extend NIST Hedgehog functionality across high-performance computing clusters, with intended benefits to both the additive manufacturing research community and the broader digital twin modeling field.

Project description

This project extends research into using artificial intelligence, NIST hedgehog, novel numerical techniques (material point methods MPM), and digital twins with a demonstration application to laser sintering in additive manufacturing (am); an application of interest to NIST. IT seeks to (a) further extend NIST hedgehog into computing clusters, (b) use the DOE kokkos library above the nvidia cuda language, and (c) derive surrogate models from physics-based models using AI techniques that can be integrated into a digital twin of a physical process, namely laser sintering in am.

Activities

The following activities will be performed: 1. Derive a heat conduction surrogate model. 2. Tune the heat conduction surrogate model for reliability. 3. Tune the surrogate model for performance. 4. Quantify uncertainties associated with the surrogate model. 5. Integrate the surrogate model into a digital twin. 6. Evaluate the surrogate model in the digital twin. 7. Steer the physics-based simulation with the surrogate model.

Expected outcomes

The expected outcomes of this research are:source code repository for an MPM-based simulation of laser sintering with an improved performance. A potential publication documenting (a) the derivation of a surrogate model from a physics-based simulation model, and (b) a generalization of such derivations.

Intended beneficiaries

The intended beneficiaries are primarily two research communities: (a) the first one is centered around additive manufacturing and, more specifically the modeling of additive manufacturing processes; (b) the second one is centered around the integration of digital twins and artificial intelligence to develop digital twins of physical processes. Software & systems will use this research to continue development of key AI resources and to guide the further development of AI-HPC-digital twins technologies.