Birds of a Feather Archive

Digital Twins in HPC: From Shared Definitions to Shared Standards


Authors: Barton Fiske (NVIDIA Corporation), Andrea Townsend-Nicholson (University College London), Dan Isaacs (Digital Twin Consortium)

Abstract: The HPC community now has a working definition of the digital twin, anchored in the 2024 NASEM report. What it lacks are standards to operationalize that definition across the full lifecycle. This BoF brings together practitioners from novice to expert, spanning scientific computing, biomedicine, infrastructure, and data centers, to surface the most urgent gaps: data homogenization across heterogeneous sources, verification, validation, and uncertainty quantification (VVUQ) at every lifecycle stage, and the emerging role of agentic AI as both enabler and new source of complexity. We will map the landscape, identify active efforts, and determine what coordinated community action looks like.

Long Description: Motivation and timeliness

Digital twins are proliferating across HPC-adjacent domains — climate modeling, computational biomedicine, fusion energy, data center optimization, urban infrastructure, and materials discovery. The 2024 NASEM report, Foundational Research Gaps and Future Directions for Digital Twins, provided a consensus definition and named VVUQ as the single most urgent class of unsolved problems, explicitly calling on federal agencies to make it integral to all new digital twin programs. Two years on, the community has the vocabulary but not the standards infrastructure to act at scale.

The three gaps this BoF addresses

At the ISC 2026 Digital Twins for Science & Industry workshop, practitioners from biomedicine, fusion energy, data centers, climate modeling, and urban infrastructure independently converged on three blocking problems. First, data homogenization: twins ingest from sensors, simulations, historical records, and real-time feeds with incompatible schemas, sampling rates, and provenance; no agreed interoperability layer exists for scientific HPC contexts. Second, VVUQ across the lifecycle: NASEM identified a gap between traditional modeling and simulation VVUQ and the continuous, adaptive VVUQ that a living twin requires as its physical counterpart and model stack evolve. Third, agentic AI integration: as twins evolve from reactive dashboards toward autonomous systems — a shift from reactive to agentic — they increasingly drive closed-loop scientific workflows and autonomous scientific discovery. This introduces new standards requirements around agent governance, trust, and uncertainty propagation through AI-generated model updates.

The existing landscape and where it falls short

Several bodies are active but fragmented, with no single body spanning the scientific HPC domain. The Digital Twin Consortium (DTC), now part of the EDM Association, maintains the Capabilities Periodic Table and a new AI Agent Capabilities Periodic Table (June 2025), with 12+ active testbeds spanning HPC, healthcare, and climate. ISO/IEC JTC 1/SC 41 has produced 49 standards with 20 more in development, including a maturity model and a digital thread standard published in 2026. The A2SD community roadmap (June 2026) calls explicitly for standards around agent-based orchestration and digital twins in closed-loop HPC workflows. This BoF asks directly: is DTC, HPSF, ISO/IEC SC 41, or a new liaison structure the right vehicle for the standards infrastructure this community needs?

Three proposed collaboration areas

This Bof seeks to ask if a VVUQ working group anchored in HPC use cases and liaising with DTC's Digital Engineering Working Group is needed; whether data homogenization profiles for sensor-to-HPC pipelines are needed; and also ask whether agentic AI governance frameworks for autonomous scientific twins are “ready for science”. Each of these is aimed at addressing trust, uncertainty propagation, and human-in-the-loop requirements. This BoF continues a Digital Twins series at SC and ISC since SC22, with workshops and BoFs at SC22, SC23, SC24, SC25, ISC26, and SCAsia. The series has built a consistent and growing practitioner community across academia, national labs, and industry. Now is the right moment to shift from what is a digital twin to how do we build them reliably, with integrity, and autonomously at scale.

Website: https://sites.google.com/view/sc26-digital-twin-bof/home



Back to Birds of a Feather Archive Listing