Scram reach via PJ model runs cleanly through all three probe horizons: T=10s: 6919 sets in 118s — n ∈ [0.0347, 0.0355] T=30s: 9900 sets in 156s — n ∈ [0.0153, 0.0156] T=60s: 12340 sets in 198s — n ∈ [0.00682, 0.00698] Factor-of-two power decay per 30s matches the delayed-neutron group structure (lambda_1=0.0124, half-life ~56s). At t=0 the algebraic n drops from 1.0 → 0.15 (prompt jump captured as an instantaneous algebraic adjustment); then tails off on precursor timescales. Scram reach is completely sound across the full 60s horizon — no step-budget truncation, unlike heatup beyond 300s. HOWEVER: X_exit(scram) = n ≤ 1e-4 is not reached in 60s (current n ~ 7e-3). This is a T_max vs plant-decay-rate mismatch, not a control failure. Options documented in journal: redefine X_exit in terms of shutdown margin (industry standard), extend T_max to 600s, or loosen to n ≤ 0.05. Flagged for Dane's review. Scram envelope summaries saved to reach_scram_pj_result.mat. Journal now 33 pages, still compiles clean. Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
pwr-hybrid-3-demo
Preliminary example for the HAHACS thesis — a verified hybrid controller for a small modular PWR startup. Composes three layers into one demonstrable pipeline:
- Discrete layer (
fret-pipeline/): FRET natural-language requirements → LTL → synthesized AIGER controller → state-machine diagram. - Continuous layer (
code/): 10-state point kinetic equation + thermal-hydraulics PWR model with bounded steam-generator heat removal as the disturbance input. Controllers, linearization, LQR, reach-tube propagator, Lyapunov barrier — all Julia. - Verification artifacts (
reachability/): predicate concretizations (single source of truth inpredicates.json) and the standalone reach analysis writeup (WALKTHROUGH.md). - Research context (
thesis/): the HAHACS PhD proposal. - Lab journal (
journal/): chronological invention log in LaTeX. - Predicate explorer app (
app/): Pluto.jl notebook bridging FRET predicates and continuous-state halfspaces.
Layout
pwr-hybrid-3-demo/
CLAUDE.md AI-facing context and architecture map
docs/
architecture.md How the layers compose
figures/ Shared figures for thesis + talks
fret-pipeline/ FRET → ltlsynt → AIGER → state machine
code/ Plant model, controllers, reach (all Julia)
reachability/ predicates.json + WALKTHROUGH.md
app/ Pluto.jl predicate explorer
journal/ LaTeX lab notebook
hardware/ Ovation HIL artifacts (TBD)
claude_memory/ Short AI-context notes
thesis/ [submodule] PhD proposal
presentations/
2026DICE/ [submodule] DICE 2026 abstract
Quickstart
Clone with submodules:
git clone --recurse-submodules <url>
cd pwr-hybrid-3-demo
Run the controller synthesis pipeline:
cd fret-pipeline
python3 scripts/fret_to_synth.py pwr_hybrid_3.json specs/synthesis_config_v3.json
bash scripts/synthesize.sh specs/synthesis_config_v3.json circuits
python3 scripts/trace_aiger.py circuits/PWR_HYBRID_3_DRC.aag diagrams
dot -Tpng diagrams/PWR_HYBRID_3_DRC_states.dot -o diagrams/PWR_HYBRID_3_DRC_states.png
Run the plant model and reach analysis:
cd code
julia --project=. -e 'using Pkg; Pkg.instantiate()' # first time only
julia --project=. scripts/main_mode_sweep.jl # all 5 DRC modes
julia --project=. scripts/reach_operation.jl # operation-mode linear reach
julia --project=. scripts/barrier_lyapunov.jl # Lyapunov barrier
julia --project=. scripts/barrier_compare_OL_CL.jl # OL vs CL barrier
julia --project=. scripts/reach_heatup_nonlinear.jl # nonlinear heatup (10s cap)
Open the predicate explorer:
cd app
julia --project=. -e 'using Pluto; Pluto.run()'
# Browser opens; navigate to predicate_explorer.jl
Soundness note: the current reach tubes are over-approximations
of the LINEAR model, not sound over-approximations of the nonlinear
plant. See reachability/README.md and reachability/WALKTHROUGH.md.
Prerequisites
- Julia 1.10+ (via
juliaup). - Python 3.10+ (FRET pipeline only).
- Spot for
ltlsynt(brew install spot). - Graphviz for
dot(brew install graphviz). - LaTeX (via
latexmk) for the thesis + journal builds.
Further reading
CLAUDE.md— orientation for AI agents working in this repodocs/architecture.md— how the layers composecode/CLAUDE.md— code architecture, conventions, validity rangecode/README.md— usage and dependenciesreachability/README.md— reach scope, soundness statusreachability/WALKTHROUGH.md— standalone analysis writeupjournal/README.md— journal format conventionsjournal/journal.tex— the journal itself, dated entriesthesis/CLAUDE.md— the thesis project structurefret-pipeline/README.md— FRET naming conventions and pipeline details
Description
Languages
Julia
46.5%
TeX
34.5%
Python
17.5%
Shell
1.5%