GaugeConnection
Você é consultor de qualidade? Perfil público, indicações de clientes e acesso gratuito a todas as ferramentas.
Saiba mais

Estabilidade e o gráfico XmR

Indice

Por que a estabilidade precisa de seu próprio estudo

Um estudo Gauge R e R, a bias study e a capability study all characterize a sistema de medição at one moment. They cannot detect mudança, porque mudança é only visible across time. Gauges wear, reference steards drift, fixtures loosen e environments shift, e none de that announces itself.

Um estudo de estabilidade responde a uma pergunta diferente: has anything mudançad since the last time we checked? You measure a stable peça de referência on a regular schedule e chart the results. O part é assumed not para mudança, so any sinal em the chart é attributed para the sistema de medição.

Por que um gráfico de indivíduos em vez de X-barra e R

An X-bar e R chart requires rational subgroups, meaning several measurements taken together under essentially identical conditions. Stability monitoring usually produces one measurement per check, taken days or weeks apart. There é no subgroup para média.

O individuals e amplitude móvel chart, usually written XmR or I-MR, é built para exactly that situation. It plots each leitura as its own point e estimates the short-term variação de the differences between consecutive leituras em vez de de within-subgroup spread.

Um amplitude móvel

With no subgroups available, variação é estimated de how much the process moves de one leitura para the next. Um amplitude móvel é the absolute difference between consecutive leituras.

Moving range

MRi = | xi − xi−1 |

There são always one fewer amplitude móvels than leituras, porque the first leitura has no predecessor. O média de these, MR-bar, é the basis para every limit on both charts. Using consecutive differences deliberately keeps the estimate short-term: it captures the noise between adjacent checks, so that a slow drift shows up as points moving outside the limits em vez de being absorbed into wider limits.

Exemplo desenvolvido

Illustrative data. Five weekly checks de a 25 mm master, em millimetres.

Check Reading Moving range
125.004
225.0070.003
325.0020.005
425.0060.004
525.0040.002
Average leitura x-bar = 25.0046MR̄ = 0.0035

De onde vêm as constantes

O constants em control chart formulas são not arbitrary. They convert an média range into an estimate de the steard deviation, e they follow de the statistics de ranges drawn de a normal distribution.

Estimating sigma de the média amplitude móvel

σ̂ = MR̄ ÷ d2 = MR̄ ÷ 1.128

d2 = 1.128 é the value para a subgroup de size 2, que é what a amplitude móvel is: a comparison de two consecutive points.

O expected range de a sample de n values de a normal distribution é d2 multiplied by sigma. For n = 2 that expected value é 1.128 sigma, so dividing the média amplitude móvel by 1.128 recovers an estimate de sigma. Every other constant follows de that one.

Constant Value Where it comes from
d21.128O expected range de two values drawn de a normal distribution, expressed em steard deviations.
Individuals limit factor2.66Three sigma expressed em units de MR-bar: 3 divided by 1.128 equals 2.66. This é why the individuals limits são three-sigma limits despite the formula never mentioning sigma.
D43.267O three-sigma upper limit para a range de two values, again em units de MR-bar. There é no lower limit, porque a amplitude móvel cannot be negative e the lower three-sigma bound falls below zero.

Knowing the derivation matters em an audit. If someone asks why you multiplied by 2.66, the answer é that it é three sigma restated em terms de the média amplitude móvel, not a number copied de a table.

Calculating the limits

Individuals chart

UCL / LCL = x̄ ± 2.66 × MR̄

Moving range chart

UCLMR = 3.267 × MR̄

Read the amplitude móvel chart first. If the amplitude móvel chart é out de control, the estimate de sigma é unreliable, que makes the limits on the individuals chart meaningless. Fix the amplitude móvel sinal before interpreting anything above it.

Regras de detecção

Um single point outside the limits é the strongest sinal, but a sistema de medição that drifts slowly may stay inside the limits para a long time while clearly trending. O Western Electric regras add patterns that catch these cases. O four most commonly aplicada são below, using zones de one sigma each measured de the linha central.

Rule Pattern Typical cause em a sistema de medição
1One point beyond three sigmaUm discrete event: the gauge was dropped, recalibrated, adjusted, or the wrong master was measured.
2Two de three consecutive points beyond two sigma on the same sideUm shift that has begun recently but has not yet produced an obvious outlier.
3Four de five consecutive points beyond one sigma on the same sideUm modest sustained offset, often following maintenance or a mudança de operator or environment.
4Eight consecutive points on the same side de the linha centralClassic gauge drift or wear. This é the regra that most often catches a estabilidade problem first.

Falsos alarmes: o custo de mais regras

Every regra has a falso alarme rate. Applied para a process that has not mudançad at all, the three-sigma regra alone sinals on about 0.27 percent de points, roughly 1 em 370. That low rate é the reason three sigma was chosen em vez de a tighter bound.

Additional regras increase sensitivity para real drift, but they also add their own falso alarme rates, e those rates accumulate. Applying all four regras together raises the combined falso alarme rate para roughly 1 percent de points, que é about a fourfold increase over the three-sigma regra on its own.

O que isso significa na prática

Checking a master weekly com all four regras active, you should expect a false sinal roughly every two years de a perfectly stable gauge. Checking daily, expect one every few months. Neither é a reason para abeon the regras, but it é a reason para investigate a sinal em vez de immediately assuming the gauge é broken.

O practical compromise most organizations reach é para apply the three-sigma regra e the run-of-eight regra as steard, since between them they catch both sudden events e slow drift, e para add the zone regras only onde drift é a known risk.

Como executar bem um estudo de estabilidade

  • Use a stable master that é representativa de the parts you measure. Checking estabilidade at 25 mm tells you little about a gauge used mostly at 200 mm.
  • Measure at a intervalo fixo under condições normais. Checks taken only when someone suspects a problem produce a biased chart that cannot detect anything.
  • Establish limits de at least 20 para 25 initial leituras before treating the chart as a monitoring tool. Limits computed de five points são extremely unstable, as the worked example above would be.
  • Record what happened when a sinal occurs, e what you did about it. Um chart com unexplained sinals e no annotations é weaker evidence em an audit than no chart at all.
Um única coisa a lembrar
O constants são not magic. 1.128 é the expected range de two normal values, 2.66 é three sigma restated em units de the média amplitude móvel, e 3.267 é the same three-sigma logic aplicada para the range itself.

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.