Applications

Computing for Licensing Evidence

Producing the traceable, reproducible analyses a regulator needs to evaluate a plant's safety case.

What a licensing case needs

A regulator does not accept conclusions on trust; it evaluates evidence. A licensing case is a structured argument backed by analyses that are documented, reproducible, and traceable from inputs and assumptions to results. Computing produces most of that evidence, so the way computations are managed determines whether the case is defensible.

Properties the evidence must have

Kronos motion — safety factor

Traceability: every number links to its inputs, code version, and assumptions
Reproducibility: an independent party can re-run and get the same result
Verification: the code solves the equations correctly (checked against benchmarks)
Validation: the equations describe reality (checked against experiment)
Uncertainty: results carry honest error bounds, not point estimates

Verification versus validation

Verification asks whether the software correctly solves its intended mathematics; validation asks whether that mathematics matches the physical world. Both are required. A code can be perfectly verified and still validated poorly, giving precise answers to the wrong equations. Licensing evidence documents both, with the benchmarks used for each.

Why reproducibility is central

If a reviewer cannot regenerate a result, they cannot check it, and an unchecked claim carries little weight. Kronos treats the reproducible-science pipeline (see Reproducible Science Pipeline) as the backbone of its licensing evidence: analyses ship with their inputs, code version, environment, and seeds so any result can be rebuilt from scratch.

python

manifest = {
  'result': 'shielding_dose',
  'code_commit': 'a1b2c3d4',
  'inputs_hash': 'sha256:...',
  'nuclear_data': 'ENDF/B-VIII.0',
  'seed': 12345,
}   # travels with every reported number

Honest framing

This evidence supports pre-application engagement and the eventual license for the Hyperion breeder, whose construction begins in the second quarter of 2027. Projected results are labeled as projections; no hardware net-gain is claimed before first-of-a-kind first tritium around 2030.