danesabo/Obsidian
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
Obsidian/Writing/ERLM/risks-and-contingencies/assumptions.md
Dane Sabo 59e3923cd2 Auto sync: 2025-10-27 12:34:41 (15 files changed) 
M  .task/backlog.data

M  .task/pending.data

M  .task/undo.data

A  Writing/ERLM/SaboRisksAndContingencies.pdf

M  Writing/ERLM/broader-impacts/v1.tex

M  Writing/ERLM/main.aux

M  Writing/ERLM/main.fdb_latexmk

M  Writing/ERLM/main.fls
2025-10-27 12:34:41 -04:00

2.2 KiB
Raw Blame History

Risk and Contingencies Assumptions Exercise

The outcome I want to achieve is?

  • Turn written reqs into discrete controller
  • Build continuous modes that ensure hybrid stability
  • Implement on industrial controller with HIL simulation

What can't anyone solve this today?

  • Nobody has tried to build system like this with stability in mind from the ground up. NUCE is a specific domain this is useful. Reliance on human operators for safety.

The research approach I am using is?

  • Formal Methods + Control Theory
  • FRET - Reachability
  • Reactive Synthesis

This research approach relies on these fundamental principles?

  • Temporal logic precision
  • automata
  • differential and difference equations
  • procedure writing

The experiment that I will perform is?

  • trying to make an autonomous start up procedure for a SmAHTR reactor

The equipment I will use is?

  1. FRET
  2. STRIX
  3. Simulink
  4. Reachability tools
  5. Ovation

I will analyze the results using?

  1. Prose. How hard was this to do, what MacGuyvering needed done? What TRL?

The expected outcome of this experiment is?

  1. A working autonomous start up controller can take a simulation from cold to critical without needing a human operator to intervene.

What happens if this experiment does not work?

  1. We'll shift to a smaller, simpler problem where we can overcome the limits.

What happens if the hypothesis or prediction is false?

  1. We'll show the gap between current procedure writing and where we need to be to actually do synthesis.

What assumptions do I have that, if proven wrong, would derail this project?

  1. Temporal logic from FRET is easy to synthesize with STRIX
  2. I'm not going to have state-space explosion happen
  3. Writing a start-up procedure for SmAHTR isn't that hard
  4. People give a crap about molten salt reactors
  5. This whole discrete boundary thing is not going to be really hard to implement. The idea is conditions for the transitions between modes to be boolean variables for the temporal lgoic, but that they correspond to some surface in the continuous state space. How am I going to keep track of that?
  6. Computational cost. Center for Research Computing is the answer.