Knowledge-informed Learning 1: Natural and Biological Systems
The focus of this session is on identifying challenges associated with learning and building robust, accurate knowledge-informed digital twins of natural and biological systems. Potential applications where such capabilities may be relevant include personalized health, gaining insights into large-scale natural systems, and improving predictive capability and decision making in environmental systems, including climate systems. Topics of interest include but are not limited to the creation of the digital twin, data and validation needs, assurance requirements, and technology and infrastructure needs specific to natural and biological systems.
On-site Chair: Jeph Wang, Los Alamos National Laboratory
Virtual Chair: Amanda Howard. Pacific Northwest National Laboratory
Knowledge-informed Learning 2: Complex Engineered Systems
The focus of this session is on identifying challenges associated with learning and building robust, accurate knowledge-informed digital twins of engineered systems. Potential applications include design optimization or automating control in complex engineered systems. Topics of interest include but are not limited to the creation of the digital twin, data and validation needs, assurance requirements, and technology and infrastructure needs specific to complex engineered systems.
On-site Chairs: Nat Trask and Carianne Martinez, Sandia National Laboratories
Virtual Chair: Malachi Schram, Jefferson Lab
Assurance 1: Uncertainty Quantification
The goal of this session is to assess the landscape of uncertainty quantification and identify data needs, challenges in developing UQ methods to improve digital twin assurance and demonstration needs for UQ. The session will also attempt to identify aspects of AI and UQ relative to digital twins that may be missing from the existing scientific dialogue.
On-site Chair: Guannan Zhang, Oak Ridge National Laboratory
Virtual Chair: Daniel Ratner, SLAC National Accelerator Laboratory
Assurance 2: Trusted AI
This session will focus on the needs and challenges associated with trustworthy AI as part of digital twins. Topics for discussion include but are not limited to the state of the art within trustworthy AI, priority components necessary for achieving trustworthy digital twins, reliability, robustness, causal analysis, and explainability.
On-site Chair: David Stracuzzi, Sandia National Laboratories
Virtual Chair: Abhinav Saxena, GE Global Research
Edge Deployment and Co-design 1: AI and Models for Speed and Power Efficiency
This session will explore deployment issues for AI and other models supporting digital twins, from the point of view of speed and power efficiency. Relevant architectures for edge deployment that meet or could address the deployment speed and power efficiencies, and integration approaches for hardware, algorithms, and data are among the topics of interest.
On-site Chair: Sudip Seal, Oak Ridge National Laboratory
Virtual Chair: Michael Churchill, Princeton Plasma Physics Laboratory
Edge Deployment and Co-design 2: Workflow and Integrated HPC with Edge Deployment
This session will explore the workflow and scalability/integration aspects of edge deployment and co-design of AI for digital twins. Included in these are needs for research facilities and testbeds for validation and assurance, challenges associated with integrating HPC with edge deployment, and infrastructure requirements.
On-site Chair: Jibonananda Sanyal, National Renewable Energy Laboratory
Virtual Chair: Draguna Vrabie, Pacific Northwest National Laboratory