See a filled instrument run before you try your own.
Every live instrument takes real inputs and runs real logic — nothing here is a guessed or invented output. Each of the 17 live instruments below carries one worked wind-sector example: the inputs filled in, the instrument's actual computed result, and what to notice before you run your own numbers.
A condition-monitoring team manually reviews raw vibration and oil-debris traces for every alarm before deciding whether to schedule a bearing inspection. A proposed triage tool would pre-rank alarms and cite the exact trace segment behind each ranking.
category: appraisal · production cost 0.3 · verification cost 0.8 · value 15
Puck B — Weld defect classification
category: prevention · production cost 0.6 · verification cost 0.7 · value 10
Puck C — Torque log auto-fill
category: appraisal · production cost 0.2 · verification cost 0.2 · value 5
ACTUAL COMPUTED RESULT
Puck A quadrant
automate-first
Appraisal share of portfolio value
67%
Result: Automate-first · 67% of value sits in appraisal
WORKED EXAMPLE 03 · THIRTY-SECOND LAB
The same finding, two presentations, two verification times
Tower-flange material certificate finding
The instrument runs one fixed, built-in comparison: the same underlying finding about a mismatched material certificate on a welded structural package, shown once as a vague one-line statement and once with the rule, exact location and evidence lines.
B · traceable — rule, exact location (Certificate 4 · page 2 · line 11), evidence lines and coverage statement shown
Reviewer action
Started the stopwatch, read the evidence, pressed a decision button once genuinely sure
Time on the clock at decision
14 seconds
ACTUAL COMPUTED RESULT
Decision
Verified
Elapsed time
14s
Result: Verified at 14s on Version B
WORKED EXAMPLE 04 · RISK CLASSIFIER
One high-consequence dimension is enough to reach Level C
Blade-root bolt preload rework advisory
A tool ranks blade-root bolt assemblies by predicted preload-loss risk and pushes the ranked list into the torque-rework work queue for an engineer to prioritise, never re-torquing anything itself.
high — a missed prediction could leave a real preload loss uncorrected
Influence
medium — informs prioritisation of an existing queue, not a pass/fail call
Criticality
high — bolt preload on the blade root is a safety-critical characteristic
Sensitivity
low — internal torque and inspection data only
Traceability
medium — each ranking links to its source torque log
Propagation
medium — a systematic misrank could delay several assemblies before caught
Volatility
low — the ranking model is refit quarterly
Dependency
low — runs on internally held data
Reversible
Yes — output only reorders a queue a human still works
Output destination
Engineering work queue
ACTUAL COMPUTED RESULT
Level
C
Driver
Could a wrong result materially affect conformity, safety, customer commitment or release?
Result: Level C
WORKED EXAMPLE 05 · DATA & EVIDENCE READINESS
A 3.2 average score still isn't ready if one requirement is unevaluated
Offshore wind-farm document set across three site languages
Before building a cross-document retrieval tool for an offshore wind farm spanning English, German and Danish technical documentation, the evidence foundation is scored on five dimensions.
welded structural packages, named suppliers, evaluated language, production approval support only
Result: Ceiling 78% · Approved with restrictions
WORKED EXAMPLE 07 · CORRELATED ERROR SIMULATOR
A clean sample is common even with a real, live failing class
Blade-bonding void inspection sampling
A 2% per-item defect rate is quietly present in a bonding line's void-inspection population. The simulator runs 500 trials of a 40-item sample to see how often that failing class hides inside a clean-looking sample.
Result: 46.4% of samples come back clean despite the failing class
WORKED EXAMPLE 08 · SEEDED-CASE PROTOCOL
A healthy weekly reading, seen for what it actually is
Rotor-hub casting defect queue
Known planted defect cases are injected into the live casting-inspection queue at a controlled rate so drift can be caught between formal revalidations.
All three conditions affirmed (reviewers told seeding happens; never used for individual performance management; quarantine designed, tested and audited)
ACTUAL COMPUTED RESULT
Seeds per week
6
Diagnosis
Working; reduce the rate
Export status
Ethics gate passed
Result: 6 seeds/week · Working; reduce the rate
WORKED EXAMPLE 09 · GOLDEN SET BUILDER
Enough no-finding cases to stop a trigger-happy classifier from scoring well
Blade transport-damage classification reference set
A reference set is being assembled to test a tool that distinguishes transport handling damage from a manufacturing defect on receiving inspection.
A small evidence graph is uploaded as CSV, tracing a cable bend-radius requirement through its characteristic, a known failure mode, a control, an inspection record and a resulting non-conformity.
Half the artefact chain still needs a cross-document check
Tower-section supplier package readiness
A tower-section flange flatness tolerance is being traced across the six artefacts that should all carry it, at the Industrialize & supplier readiness stage.
Requirement, Drawing / characteristic, Control plan
ACTUAL COMPUTED RESULT
Verification quadrant for this stage
Automate-first / appraisal
Artefacts still needing evidence or review
3 of 6 (Risk analysis, Work instruction, Inspection record)
Result: 3 of 6 connections still need evidence
WORKED EXAMPLE 13 · CONFIGURATION ATTESTATION & GATE PACK
A complete, durable record built to answer a question five years from now
Cable partial-discharge screening go-live record
The partial-discharge screening tool from Instrument 02 goes live on a cable lot. The exact configuration and the human decision behind it are recorded at the moment of use.
"Golden set GS-CAPA-01, 10-case recall test, qualification file QF-07"
Rung 04 — Shared knowledge layer
(left blank — not yet built)
Rung 05 — Embedded in APQP workflow
(left blank — not yet built)
ACTUAL COMPUTED RESULT
Highest evidenced rung
3 — First validated tool
Result: Rung 3 — First validated tool
WORKED EXAMPLE 16 · ROADMAP GENERATOR
Two of six horizons done, four to go, one named owner throughout
Offshore division 90-day AI-quality plan
The offshore division's implementation plan tracks progress against the guideline's six staged horizons, from 30-day legitimacy work through to the three-year durable capability.
WORKED EXAMPLE 17 · PROHIBITED LIST & SIGNATURE TEST
A described task matches a decision reserved for a named human
Blade root laminate disposition
A team considers letting an automated tool propose the disposition of a batch of blade-root laminate flagged during lay-up inspection, and checks that task against the standing prohibited list.