FIELDMIND LABCOMPUTATIONAL R&DRU

METHOD / REPRODUCIBILITY / CLAIM CONTROL

Claim-admission method

The objective is not a favourable plot but an auditable chain: the question can fail, the model assumptions are described, controls precede interpretation, and the conclusion does not exceed the evidence.

01 / EVIDENCE LADDER

Claim levels are not interchangeable

Each transition requires a new evidence class; wording cannot turn a numerical PoC into a laboratory result.

  1. 01

    Hypothesis

    A proposed mechanism

    No result yet
  2. 02

    Numerical result

    An observable from one calculation

    Requires robustness checks
  3. 03

    Simulation study

    A documented series under one model

    Bounded by equations and parameters
  4. 04

    Numerical PoC

    A linked chain of controls and stress tests

    Supports a mechanism at model level
  5. 05

    Laboratory validation

    An independent physical experiment

    Tests correspondence with reality
  6. 06

    Experimentally confirmed result

    A reproducible physical measurement

    Not automatically industrial readiness

CASE / DB1

From observation to a defensible conclusion

In DB1, the localized interface response survives controls. This supports interface organisation within the models considered. A strengthening claim requires separate mechanical tests; three such branches did not support an improvement. The positive interface result and negative mechanical result are therefore retained together.

See the DB1 checks and split verdict ↗

02 / WORKFLOW

Six required transitions

Specific tests change with the physics; the verification logic remains stable.

  1. 01

    Define the falsifiable question

    Translate the idea into state variables, observables, assumptions, and an outcome that can fail.

  2. 02

    Build the smallest defensible model

    Choose equations, discretisation, boundaries, initial conditions, and diagnostics appropriate to the question.

  3. 03

    Establish controls and balance checks

    Add passive, null, negative, ablation, conservation, and numerical-validity controls before interpreting a favourable branch.

  4. 04

    Run a structured experiment campaign

    Sweep parameters, seeds, perturbations, and resolutions; retain negative and failure results.

  5. 05

    Audit robustness and claim boundaries

    Separate model-level evidence from assumptions, artefacts, untested extrapolation, and laboratory claims.

  6. 06

    Package evidence and the next test

    Deliver the agreed methods, diagnostics, limitations, and a clear laboratory or engineering follow-up, under the project access terms.

03 / CONTROL FAMILIES

Controls grouped by function

Not every project requires every test, but omitting a family must be justified.

A

Conservation and numerical validity

  • Mass, energy, and flux balance
  • Grid and timestep convergence
  • Numerical error and invariant checks
B

Counterfactual branches

  • Passive and null baseline
  • Negative control
  • Ablation of the claimed mechanism
C

Result robustness

  • Seeds and initial states
  • Noise, defects, and perturbations
  • Parameter and metric sensitivity
D

Transfer boundary

  • Untested extrapolation
  • Required physical measurements
  • Confirmation and falsification criteria

04 / CLAIM ADMISSION GATE

A conclusion passes only with its boundary

Admit → reproducible · robust · bounded

Retain → negative · ambiguous · open

Do not transfer → model effect into physical validation

A numerical result establishes model behaviour under stated assumptions and does not replace independent laboratory validation.
FIELDSTABLE