Digitally Twinning a Fusion Powerplant - Everything Everywhere All at Once…

Digitally Twinning a Fusion Powerplant - Everything Everywhere All at Once…

Tuesday, May 23, 2023 1:20 PM to 1:40 PM · 20 min. (Europe/Berlin)
Hall Z - 3rd Floor
Focus Session
AI ApplicationsDigital TwinsEmerging HPC Processors and AcceleratorsExascale SystemsIndustrial Use Cases of HPC, ML and QC

Information

The UK is embarking upon a grand challenge – to deliver fusion power to grid in the early 2040’s through an ambitious project named STEP. To realise this bold mission, radical developments are urgently required across the engineering process. The timeline for delivering Net Zero is incredibly challenging due in part to existing engineering processes and established design tools being inefficient and insufficiently scalable. The engineering process itself must therefore be reimagined: we must develop new, accessible, state-of-the-art, and “actionable” simulation tools that will underpin a new capability for engineering, built upon a synthesis of very large but sparse data, intelligently combined with uncertain models. This will require new methods, new algorithms, and the world’s largest supercomputers, such that we can take a “probabilistic” approach to the design of complex, strongly coupled systems built upon the confluence of Artificial Intelligence and HPC at the exascale. We must simulate plasma, materials, and engineering (everything) in high fidelity, across the entire plant (everywhere), simultaneously (all at once) to deliver new predictive capability that exposes unforeseen “emergent behaviour” which will likely arise due to the strongly coupled nature of fusion plant (initially as plasma “flight” and “plant lifetime” simulators, as simulations for modelling the Plasma Wall Interface (PWI) and eventually as full system digital twins) to help optimise, accelerate and de-risk design. Capability must then be democratised across the engineering world through the “Industrial Metaverse” where products will be designed “in-silico” rather than via extensive, expensive, and time-consuming test-based engineering design. In this presentation I will outline some early proof of concept studies that are now starting to exploit the power of convergent HPC, Big Data and AI to simulate our long established exascale use case – a digital twin of the world’s first fusion powerplants.
Format
On-siteOn Demand
Beginner Level
70%
Intermediate Level
20%
Advanced Level
10%

Log in