QM BenchQuality engineering tools

Acceptance Sampling Plan Designer: c = 0, Attribute and Variables Plans

Design a sampling plan from the two qualities that matter: the good quality you want to accept almost always and the bad quality you want to reject almost always. Compare a c = 0 plan, the smallest attribute plan and variables plans on their OC curves.How to use ↓
  • Free, no sign-up
  • Data stays in your browser
  • PDF report included

Part ofQuality inspection

Risk points

the quality you want to accept almost always, and the quality you want to reject almost always

Smallest attribute plan: n = 132, c = 3; variables plan: n = 54

A lot with 5 % defectives is accepted at most 10 % of the time by every plan. The c = 0 plan needs 45 parts but rejects 36.4 % of good lots at 1 %, so it is strict on the supplier. Variables plans need far fewer parts if the characteristic is measured and roughly normal.

c = 0 (n = 45)Attribute n = 132, c = 3Variables, σ unknown (n = 54)Variables, σ known (n = 19)
What to do next
c = 0 sample
45
Attribute plan n
132
c = 3
Variables n (σ unknown)
54
k = 1.943
Variables n (σ known)
19

Operating characteristic (OC) curves

probability of accepting a lot by its true defect rate
0204060801000.0 %0.3 %0.5 %0.8 %1.0 %1.3 %1.5 %1.7 %2.0 %2.3 %2.5 %2.8 %3.0 %3.3 %3.5 %3.8 %4.0 %4.3 %4.5 %4.8 %5.0 %5.3 %5.5 %5.8 %6.0 %6.3 %6.5 %6.8 %7.0 %7.2 %7.5 %7.8 %8.0 %8.3 %8.5 %8.8 %9.0 %9.3 %9.5 %9.8 %10.0 %Lot defect rate
c = 0 (n = 45)Attribute n = 132, c = 3Variables, σ unknown (n = 54)Variables, σ known (n = 19)

Plans compared

PlanDecision ruleP(accept) at AQLP(accept) at LTPD
c = 0 (n = 45)accept with 0 defectives63.6 %9.9 %
Attribute n = 132, c = 3accept with ≤ 3 defectives95.6 %9.9 %
Variables, σ unknown (n = 54)accept if (limit − x̄) / s ≥ 1.94395.1 %9.8 %
Variables, σ known (n = 19)accept if (limit − x̄) / σ ≥ 1.94395.3 %9.7 %

Acceptance sampling plan designer

Design a sampling plan from two risk points: the AQL with the producer’s risk α and the LTPD with the consumer’s risk β. The tool gives the c = 0 sample size, the smallest attribute plan that meets both points and variables plans with known and unknown standard deviation.

All plans are compared on their operating characteristic (OC) curves and by their probability of accepting lots at the AQL and at the LTPD, so you see the price of each plan in parts and in risk.

Worked example: Sampling Plan Designer

AQL 0.65 %, LTPD 4 %, α 5 %, β 10 %.
  1. 01Risk pointsAQL / LTPDα 5 %, β 10 %
  2. 02c = 0n = 57strict on good lots
  3. 03Attributesmallest n, cmeets both points
  4. 04Variablesfar fewer partsmeasured characteristic

01Method

Attribute plans use the binomial distribution: the c = 0 sample is the smallest n with (1 − LTPD)ⁿ ≤ β; the attribute plan is the smallest n for the smallest c that meets both risk points. Variables plans use the Wallis approximation for a one-sided limit: k = (z₁·z_β + z₂·z_α)/(z_α + z_β), n = ((z_α + z_β)/(z₁ − z₂))², multiplied by (1 + k²/2) when σ is estimated from the sample.

The plans are calculated, not taken from ISO 2859-1 or ISO 3951-1 tables. Variables plans assume a roughly normal characteristic.

02When to use it

Use it to agree incoming inspection plans with suppliers, to justify a c = 0 plan or a switch to variables sampling, and to show the real risk of a sampling plan to management.

PDF report in one click

Export a PDF with the risk points, the plans, their acceptance probabilities and the OC curves.

Private: runs in your browser

All calculations run in your browser.

How to design a plan

  1. Set the AQL: the defect rate you can live with, and α, the risk of rejecting such a lot.
  2. Set the LTPD: the defect rate you must not accept, and β, the risk of accepting it.
  3. Compare the plans: sample size, decision rule and probability of acceptance at both points.
  4. Read the OC curves to see how each plan treats lots between the two points.
  5. Choose: variables plans for measured, normal characteristics; attribute plans otherwise.

For contract inspection with switching rules use the ISO 2859-1 AQL calculator; to feel the risks, play the AQL simulator. Background: OC curve and LTPD. All tools: quality inspection.

Frequently asked questions

›What is a c = 0 sampling plan?

A plan that accepts a lot only if the sample contains no defective part. It is popular in automotive because it sends a clear message, but for the same protection against bad lots it rejects good lots more often than a plan with an acceptance number above zero.

›How is the sample size calculated?

For c = 0, n is the smallest sample with (1 − LTPD)ⁿ ≤ β, for example 45 parts for LTPD 5 % at 90 % confidence. For other attribute plans the tool searches the smallest n and c that meet both risk points with the binomial distribution.

›Why do variables plans need fewer parts?

A measured value carries much more information than a pass / fail result. If the characteristic is roughly normal, a sample mean and standard deviation compared with the limit by a factor k give the same protection with a fraction of the parts.

›Are these the ISO 2859-1 or ISO 3951 tables?

No. The plans are calculated from the risk points you enter. Where a contract requires ISO 2859-1 with its switching rules, use the AQL calculator; ISO 3951 tables must be taken from the standard.