GaugeConnection
¿Es usted consultor de calidad? Perfil público, referidos de clientes y acceso gratuito a todas las herramientas.
Más información

Designing a R&R del sistema de medición estudio

Tabla de contenidos

Selecting the piezas

The piezas must span the rango your proceso actually produces. %GRR is a ratio of medición variación to pieza variación, so the piezas you choose set the denominator. Choose them badly and the resultado is meaningless regardless of how good the gauge is.

The instinctive approach - grab ten consecutive piezas off the line - is usually wrong. Consecutive piezas are the most similar piezas the proceso makes. Part variación comes out artificially small, %GRR is inflated, and a perfectly good gauge fails.

The correct approach is to sample across time so that normal proceso variación is represented - piezas from different shifts, different days, different material lots. The aim is a set that looks like the real output of the proceso, not a set chosen for a convenient resultado.

Parts do not need to be in specification
Including a few piezas near or slightly outside the specification limits is good practice, because that is where medición error causes the costly mistakes. The estudio is assessing the sistema de medición, not the piezas, so a nonconforming pieza is a legitimate and useful sample.

Por qué 10 piezas, 3 operadores, 3 trials

The patrón design is a compromise between statistical confidence and shop-floor practicality: 90 mediciones, generally completable in a single session. The grados de libertad explain why this piezaicular shape works.

Fuente df Qué it supports
Operator 2 Minimal, but enough to detect a substantial operador effect. This is the weakest pieza of the patrón design.
Part 9 A sound estimate of pieza variación, which is the denominator of %GRR.
Operator x Part 18 Enough to test whether operadores disagree on piezaicular piezas.
Repeatability 60 A precise estimate of equipment variación - the best-estimated component in the estudio.

Notice the imbalance. Repeatability is estimated on 60 grados de libertad and is therefore reliable, while the operador effect rests on 2. Reproducibility is the least precisely estimated component in a patrón estudio, so treat a borderline AV resultado with more caution than a borderline EV resultado.

Adding piezas is generally more valuable than adding trials. More piezas sharpen the estimate of pieza variación and improve ndc; more trials mainly refine a repetibilidad estimate that is already the most precise number in the estudio.

Choosing the operadores

Use the people who actually perform the medición in production. A estudio run by three engineers proves that three engineers can use the gauge consistently, which is not the pregunta. If three shifts run the job, represent all three.

Dónde a single operador performs all inspection, a full R&R del sistema de medición is not the right estudio - there is no reproducibilidad to assess. Run a repetibilidad-only estudio, or reconsider whether the single-operador arrangoment is itself a risk when that person is absent.

Running the estudio

Execution details matter as much as the design. Each of these protects a specific pieza of the análisis.

  • Randomise the medición order within each operador's pass. Measuring piezas 1 to 10 in sequence every time lets any drift over the session masquerade as pieza variación.
  • Keep operadores blind to the pieza identity and to previous readings. Someone who remembers medir 25.004 last time will tend to reproduce it, which understates repetibilidad.
  • Make each trial a genuine re-medición - release the pieza, re-seat it, measure again. Reading the display repeatedly measures the display, not the sistema de medición.
  • Do not let operadores watch each other measure. Reproducibility is precisely the difference between their independent techniques, and observing each other suppresses it.
  • Record every reading at the full resolution of the gauge. Rounding before análisis discards the repetibilidad signal the estudio exists to measure.

Cuándo you cannot run the patrón design

Destructive tests, very long cycle times, and low-volume production all make 10 x 3 x 3 impossible. Smaller estudios are legitimate provided the reduced confidence is acknowledged on the report rather than presented as equivalent.

For destructive testing the crossed design fails outright, because no pieza can be measured twice. This requires a nested estudio using batches of piezas assumed homogeneous, which separates repetibilidad from pieza variación differently and rests on an assumption that must itself be justified.

If you reduce the design, reduce piezas last. Dropping to 2 operadores or 2 trials weakens components that are already estimated with reasonable precisión; dropping piezas directly undermines the pieza variación estimate that %GRR and ndc both depend on.

Rejoining the server...

Rejoin failed... trying again in seconds.

Failed to rejoin.
Please retry or reload the page.

The session has been paused by the server.

Failed to resume the session.
Please retry or reload the page.