IMPORTANT: YOU NEED TO BRING A PHOTO ID WITH YOU TO GET ON-SITE!
IMPORTANT: YOU NEED TO BRING A PHOTO ID WITH YOU TO GET ON-SITE!
IMPORTANT: YOU NEED TO BRING A PHOTO ID WITH YOU TO GET ON-SITE!

• Please register on the following link:
  https://www.eventbrite.co.uk/e/farscape-february-2024-culham-science-centre-abingdon-oxford-ox14-3db-tickets-753865811937

• Please see at the bottom of this document for the detailed agenda.
• Members of FARSCAPE can log-in (top-right of this page) to either edit the agenda or upload/download presentations.

• John Adams Lecture Theatre (JALT)
  Conference Centre
  Above Costa (E3 Building)
  Culham Campus
• Lunch and Coffee refreshments will be served in the main lobby outside the ZETA Room (1st floor, conference centre, just through the double doors outside the main JALT theatre)

• Bus and rail connections are available to the Culham Campus, please find details here: https://culham.org.uk/contact/
• IMPORTANT: YOU NEED TO BRING A PHOTO ID WITH YOU TO GET ON-SITE!

• All members of FARSCAPE should be able to log-in to the FARSCAPE wiki (top-right of this page) to upload their slides.
• If you do not have a login and require one, please contact stanislas.pamela@ukaea.uk
• Note, you need to log-in to be able to download presentations too, ie. external people can see this page, but they will not be able to download your presentations.
• Please create an upload link of your presentation inside the agenda below. An example of how to do this can be seen with the “13:40 - 14:00, Overview of Meeting, S.Pamela” slides. Contact Stan if you have any difficulty doing this.

Please use the following templates for your presentations, documents and reports:

• FARSCAPE-Powerpoint-template (pptx)
• farscape_report_template.docx
• Draft LaTeX Report Template

• Zoom link for the whole week: https://ukaea.zoom.us/j/85374696824?pwd=d3gyYmlDa2QycUpET3pxZ2x0LzNVQT09

• Please follow Google form link or QR code:
• https://docs.google.com/forms/d/e/1FAIpQLScR-zHas3ZL1NPElYfbCV_dOPP3O4NKo_qydJMS0P9yMhK3HA/viewform?usp=sf_link
• View existing suggestions: https://docs.google.com/forms/d/e/1FAIpQLScR-zHas3ZL1NPElYfbCV_dOPP3O4NKo_qydJMS0P9yMhK3HA/viewanalytics

• Poll for new name: https://docs.google.com/forms/d/e/1FAIpQLSdiy6KGvc3i0Fc4tnZmjSCF_3Y0ct8qICf_M3qYWbMv2ViAhw/viewform?usp=sf_link

• If at any time you would like to give feedback but not sure where to put it, do it here please (anonymous):
• https://docs.google.com/forms/d/e/1FAIpQLSf_3IGLpXegKJPti5L_BeaweJALt1LGg3q8YTGJD8B5dObLdA/viewform?usp=sf_link

• Attendance list
• If you are not registered for dinner and would like to come, please let Lou (lou.pierce@stfc.ac.uk) or Stan know asap (cannot be guaranteed, but we'll try…)

1. Data Survey:

• How much data
• What codes
• What workflows
• What containers
• What HPC resources
• How quickly are you ready

2. Data management plan

3. How do we incentivise people to get their data+workflows integrated:

• Hard way: you can’t publish/conference until data+codes are integrated
• Demonstrate the advantages of having that, eg. Omniverse really high-quality viz
  • ⇒ GS2/C-Gyro workflow
  • ⇒ MOOSE workflow

Task 0: Surfacing experitse with knowledge graphs and LLMs (from Rob Firth and Adri Agnello for Shaun and Rob A)

• Info sharing: digitised JET log corpus, recordings of interviews with retiring engineers, GPU-acelerated compute, expert to generate a shared ontology
• Inter-WS: WS1, WS3
• Risks: data quality, data quantity, data access, staff availability

Current projects

• Construction of KBM models
• Curation/opening of existing datasets – catalogue in SimDB
• Uptake of SimVue.
• Comparisons of methodologies
• Development of quasi-linear models
• Benchmarking of non-linear simulations
• Consideration of parameter space of STEP ramp up
• 3D Hermes3 model building in BOUT++ with StyleGAN

New projects

• Development of plan for other JINTRAC components that need emulation on the path to development of plasma flight simulator
• Question of core-edge integration wrt surrogate models.
• What components need to be emulated and how to move towards Chimera digital twin (/hive).
• Expanded optimisation of components on Chimera using Bayesian Optimisation or other ML based optimisations tools (Adri’s MCMC thingy?).
• Hive → Predict thermal deformation / stress from coil voltage? Synthetic diagnostic?

Task 1: Divertor detachment 1.1 RL control in 1D, 1.2 fold the active learning into the dataset generation, 1.3 move to 2D

• Info sharing: WS1 (info comes from Moulton in the form of emails, but we need real data)
• Inter-WS dependence: WS1, WS2 for obvious reasons
• Risks: staffing, computing resources (post-EurAtom…)

Task 2: Building a realistic MAST-U into FreeGSNKE 2.1 Vacuum calibration shots, 2.2 shot reconstruction with a plasma in it

• Info sharing: useable data, both on the machine model (what Lucy Harris was doing before she moved), on the probes, and the P6 control
• Inter-WS dependence: WS1 for data
• Risks: same as every year…

Task 3: Control Design and Validation on an “ideal ST” (FreeGSNKE simulations of dynamics with some classical control policies on the coils)

• Info sharing: targeted to being useful for STEP, we need someone who can work on a simple PID control and have some useful numbers
• Inter-WS dependence: WS1
• Risks: same as every year…

Task 4: Explore Coupling free-boundary GS with thermal evolution (starting with 0D for pressure from W_th)

• Info sharing: would benefit from in-house expertise, and order-of-magnitude numbers relevant to MAST-U and thereabouts
• Inter-WS dependence: WS1, WS2
• Risks: staffing

Task 5: RL for (nonlinear) magnetic shape control Including porting FreeGSNKE to GPU to enable scaling to exascale (for experience gathering and RL training). Continuing work begun in 2023 with Matthijs Mars to train RL agents on coil-sweeps, with memory and uncertainties (statistical and systematic).

• Info sharing: Jax-GS for the GPU portability, work more closely with Ubaid, all available info from “good” Ohmic MAST-U shots
• Inter-WS dependence: Ubaid might just bill some of his time to WS3 for this
• Risks: same as every year… plus staffing and computing resources (post-EurAtom)

Task 6: Unsupervised Disruption Prediction This is on hold because the necessary, curated dataset has not been provided yet. If it can be done, this can go forward.

• Info sharing: useable data, meaningful data
• Inter-WS dependence: WS1
• Risks: staffing, data quality, absence of data with useful info (e.g. actuator control data)

• Complex efficient mesh generation
  • Apply current developments to a real problem/more representative problem
  • Tet vs hex – which is actually better for “our” problems?
  • High-order geometries / IsoGeometric Analysis
  • Conformal meshing of immersed geometry
  • Complex geometries and boundaries conditions for structured meshes

• High order FE
  • Implement new elements as per Roadmap
  • Identify key components from PETSc to leverage into MOOSE, and then implementation [MOOSE roadmap required] – what is being done at UKAEA outside of FARSCAPE?
  • Custom preconditioners for MOOSE and high-order finite elements.
  • Explore MOOSE hybrid CG-DG capabilites

• Use cases – need to valdiate and expand current implementation
  • CHIMERA with EM
  • BeeHIVE
  • Blankets (STEP/ITER/DEMO)
  • Software environment coupling at scale
  • Include UQ
  • Include optimisation (topological optimisation using GANNs?)

• Fluid simulations at the exascale
  • NekRS / NekO / Nektar++ / Firedrake / ExaDG (plus other GPU codes?) comparison
  • Identify best code to use as basis for exascale level MHD simulation and implement required MHD extensions
  • Move OpenFOAM towards the exascale through enabling hybrid and/or GPU solution

• amrPX – real applicaton demo phase
  • Implement built-in AMReX MLMG solver into amrPX - risk: feasibility/technical difficulty unknown
  • Improving scalability amrPX
  • Test blob 2D / Hasegawa-Wakatani 2D (Neptune)

• Neutronics
  • Examine performance portability of OpenMC-GPU across multiple architectures
  • Survey MC landscape for status of potential alternatives
  • GPU neutronic analysis direct on CAD (rather than CSG based input)

• Data generation/orchestration, documentation of processes
• Integration of UK exascale strategy

1. Design of a breeding blanket.

• a. Objective: create a simulation framework that can be used to design a breeding blanket while quantifying uncertainties associated with the neutronics.
• b. Dependencies: Parametric CAD and material generator, OpenMC, uncertainty propagation from neutron energy data to TBR.
• c. Outputs: a design that is demonstrated to be good over a range of conditions and not just one, and that is robust enough to take into account uncertain input data.

2. Prototype application for UQ for chimera

• a. Objective: Demonstrate the value of UQ for a digital twin of Chimera
• b. Dependencies: coupled simulation model within Moose that captures heat transfer, structural mechanics and/or magnetic loading, WS4 output
• c. Outputs: Probability distributions of material properties, stresses so that users can make informed decisions for conditions that they don’t test in.

3. Uncertainty aware control algorithm for plasma-shape control

• a. Objective: plasma shape control that quantifies distributions in the plasma equilibria
• b. Outputs: Distribution of changes to coil currents so that you can make better decisions of the change to be made. Compute optimal policies instead of having to train a black-box controller.
• c. Dependencies: A really good fast model of the plasma equilibria and how changes in current affect it, and maybe a gradient associated with it.

4. Uncertainty-aware FE simulations

• a. Objective: Build in uncertainty into FE model in MOOSE.
• b. Dependencies: Knowledge/resources/expertise in stochastic FEM, understanding of a stochastic FEM example problem.
• c. Output: Proxy app showing stochastic FEM in MOOSE

5. First principal based uncertainty of material properties

• a. Objective: Be able to produce

• Topic: FARSCAPE

• Time: Sep 25, 2023 09:00 AM London
  • Every day, until Sep 29, 2023, 5 occurrence(s)
    Sep 25, 2023 09:00 AM
    Sep 26, 2023 09:00 AM
    Sep 27, 2023 09:00 AM
    Sep 28, 2023 09:00 AM
    Sep 29, 2023 09:00 AM

• Please download and import the following iCalendar (.ics) files to your calendar system. Daily:
  https://ukri.zoom.us/meeting/tJwuf-mrrzssH9BTiRZNvqtEhNQivP7OXeFB/ics?icsToken=98tyKuCgrTMqGdWVthyDRow-A4qgc-nzmGJdgvp_qBbXOQ5FcxfTPsxzO6RPR_fo

• Join Zoom Meeting:
  https://ukri.zoom.us/j/98380310654

• You can test your connection prior to the meeting here:
  https://ukri.zoom.us/test

• One tap mobile * <color blue>+442080806591,,98380310654# United Kingdom
  +442080806592,,98380310654# United Kingdom

• Dial by your location * <color blue>+44 208 080 6591 United Kingdom
  +44 208 080 6592 United Kingdom
  +44 330 088 5830 United Kingdom
  +44 131 460 1196 United Kingdom
  +44 203 481 5237 United Kingdom
  +44 203 481 5240 United Kingdom
  +44 203 901 7895 United Kingdom
  0 800 358 2817 United Kingdom Toll-free
  0 800 456 1369 United Kingdom Toll-free
  0 800 031 5717 United Kingdom Toll-free
  0 800 260 5801 United Kingdom Toll-free

• Meeting ID: 983 8031 0654
• Find your local number: https://ukri.zoom.us/u/ab1F5Vte4v

• Join by SIP
  98380310654@zoomcrc.com

• Join by H.323 * <color blue>162.255.37.11 (US West)
  162.255.36.11 (US East)
  115.114.131.7 (India Mumbai)
  115.114.115.7 (India Hyderabad)
  213.19.144.110 (Amsterdam Netherlands)
  213.244.140.110 (Germany)
  103.122.166.55 (Australia Sydney)
  103.122.167.55 (Australia Melbourne)
  69.174.57.160 (Canada Toronto)
  65.39.152.160 (Canada Vancouver)
  Meeting ID: 983 8031 0654

• Join by Skype for Business
  https://ukri.zoom.us/skype/98380310654

• Or Skype on a SurfaceHub:
  SIP: 98380310654@lync.zoom.us