Skip to content
SMART Measuring Instruments

For NMIs · Volume overview

Overview

This section is for you if you are a national metrology institute or a legal-metrology authority, OIML’s own members. It explains the program in legal-metrology language: traceability, uncertainty, the GUM-adjacent discipline, and why this tooling is yours.

1. What OIML SMART is to an NMI

OIML SMART is the OIML’s own executable-standards program: the Recommendations are modeled once (in Primmel, the executable modelling language) and the whole certification workflow runs from those models , applications, dispatch, testing, evaluation, certificates, the register, and the live surveillance of certified digital twins. For an NMI this is not vendor software: it is your own Recommendation, executing under your own discipline, with every judgment traceable to a clause.

The three things an authority asks of any such system, answered:

  1. Is the judgment traceable? Every requirement carries its clause URN (source: { doc: urn:oiml:pub:r:60-1:2021, clause: "5.2.4" }); every verdict records WHICH definition judged it (the INV-8 definition-version pin) and the resolved inputs it judged from (the re-judgment basis, INV-5).
  2. Is the uncertainty honest? The reference channel is gated, never assumed (§2); a stale value degrades to indeterminate, never a silent pass (§3).
  3. Can we reproduce it? The whole stack stands up from a checkout and reproduces a type evaluation end to end, gated by the same commands you will run (the authority quickstart).

2. Traceability, mechanized

The platform’s traceability chain is not a report, it is the runtime:

  • Reference materials are first-class. A reference reading arrives through a declared channel (reference_instrument | observer_attestation | sim_ground_truth, the C99 probe-channel provenance facet), and an observer channel carries its declared traceability limitation (“twin ≡ display, not twin ≡ mass”, never a comment).
  • The U:MPE ≤ 1:3 discipline is a process in the model. The reference-traceability gate (the oiml-integrated-ref package’s reference_traceability_gate process) computes reference_uncertainty / judged_allowance and voids the setup when the ratio exceeds one third, INVALID (the setup was wrong), never an instrument failure, never a silent admission. This is the same discipline the Recommendations’ reference materials enforce (the CGM U:MPE ≤ 1:3 constraint).
  • Pairing windows are declared, never invented. A served value and its reference pair within the declared pairing window; a pair outside is inadmissible evidence, the run is void, never a fail.
  • Clause URNs everywhere. Requirements, test procedures, and form fields carry their published-print addresses; the reconstruction congruence gate proves coverage, order, and text identity against the published source (the named gaps have reasons and only burn down).

3. The uncertainty + freshness treatment

The platform’s answer to “what is this value worth?”:

  • Freshness is first-class. Every served register declares fresh_within; a value older than its window degrades verdicts to indeterminate (silence is not evidence), never a fail on physics, never a pass on faith.
  • Skew is recorded, not hidden. The served–reference pair’s skew is computed from the two evidence times and judged against the declared window, the same pairing discipline as a laboratory’s simultaneous readings.
  • ISO 14253-1 zones, structurally. Verdict outcomes are pass | fail | invalid | indeterminate: conformance, nonconformance, void (the measurement was wrong, invalid), and undecidable (missing or stale evidence, indeterminate). No verdict ever collapses the last two into a fail.
  • Uncertainty-denominated limits. Acceptance quantities are declared once as typed quantities with units (never bare numbers, INV-1), and verification-interval-denominated limits (the load cell’s v) are a distinct quantity kind, a v-denominated observable against a mass limit is a defect by construction.

4. The member-state narrative

OIML is the authority of legal metrology and its members are the NMIs. This program is the members’ own tooling: the Recommendations are your documents, the models are clause-traceable to them, and the program’s governance is the OIML-CS’s own (B 18:2025, its participant registry, its documents corpus, its operations runtime, all modeled). An NMI running OIML SMART on its own evaluations is exercising its own mandate with better instruments, the standard is the source of truth, the platform is a pure engine, and every claim in the tree has a command that proves it.

Next: the authority quickstart, stand up the stack and reproduce a type evaluation, gated.