EVIDENCE / GENERATED FROM THE MODELS
Follow the claim.
Inspect the check.
16 checks, explicit expectations, and the results computed when this page was built. The same models power the interactive experiments.
Build-time evidence · Built
Source commit: 856ab1ddaeae2984027bb06ecd8859057463c5c0
This is a finite set of model checks, not live service monitoring or evidence of a calibrated facility. Each model version is recorded below.
Synthetic zone heat balance
- Expected
- 27.60°C at 120 s; all modeled readiness checks pass at 240 s
- Computed at build time
- 27.60°C; 3 passing checks
Incomplete evidence withholds readiness
- Expected
- Stale telemetry fails freshness; missing current evidence is inconclusive; acknowledgement does not repair equipment
- Computed at build time
- Fail / Inconclusive / Inconclusive
Sensitivity direct-run equivalence
- Expected
- Nine paired cells equal direct runs
- Computed at build time
- 9 paired cells checked
Intervention direct-run equivalence
- Expected
- All four alternatives equal direct runs
- Computed at build time
- 4 alternatives checked
Service interval union
- Expected
- Power loss + thermal exceedance − overlap = combined duration
- Computed at build time
- Both arrangements checked
Thermal equilibrium
- Expected
- Maximum temperature deviation = 0°C
- Computed at build time
- 0.000000°C
Saturated energy balance
- Expected
- 14.000°C rise over 300 s; error < 0.001°C
- Computed at build time
- 14.000000°C rise
Sampling convergence
- Expected
- Peak difference < 0.03°C for 0.5 s vs 0.25 s, both tunings
- Computed at build time
- 0.002366°C maximum difference
Versioned link reproduction
- Expected
- Identical sampled trace after decoding the link
- Computed at build time
- 0.000000°C peak difference
Failure propagation
- Expected
- UPS A exhausts at 90 s; alternate path supplies 360 kW; cooling drops to 250 kW
- Computed at build time
- unavailable; load 360 kW; cooling 250 kW
Shared versus independent cooling
- Expected
- Loop fault: shared cooling = 0 kW; independent cooling = 250 kW at 60 s
- Computed at build time
- Shared 0 kW; independent 250 kW
Alarm flood and rationalization
- Expected
- As found: more than 10 alarms in 10 minutes, root cause not first at 2 min; all three changes: at most 10, root cause first
- Computed at build time
- As found 41 (root cause 7); rationalized 5 (root cause 1)
Alarm design leaves the process unchanged
- Expected
- Identical process traces for every design; delays remove all chattering
- Computed at build time
- Identical traces; chattering 3 → 0
One fault, three redundancies
- Expected
- Pump trip: survived with auto-start (4 of 8 arrangements); supply A: also needs the standby pump on bus B (2); loop A: also needs separate loops (2); only separate/split/auto survives all three
- Computed at build time
- Pump 4, supply A 2, loop A 2; survives all three: separate/split/auto
Control buys seconds, not a heat sink
- Expected
- Without cooling water every tuning trips; faster tuning trips later, by less than a minute in total
- Computed at build time
- Trips at 250, 276, 286 s (conservative, balanced, aggressive)
Resilience comparison reproduction
- Expected
- Both full traces are identical after decoding the versioned link
- Computed at build time
- Identical traces
What makes a useful check?
Equilibrium and a saturated heat balance compare the model with a result derived independently from conservation of energy. Convergence checks whether a finer controller sampling interval materially changes the answer. Link reproduction checks the entire sampled trace. Failure propagation checks both electrical supply and cooling after a dependency is lost.
The wider repository also checks browser interactions, accessibility, input validation and release contracts. Those checks are not represented as live green badges here. The table reports only the computations executed for this build.