Attribute Acceptance Sampling (Create)

On this page
  1. Purpose
  2. Key Capabilities
  3. Statistical Methods
  4. Input Fields & Options
  5. Output Descriptions
  6. Worked Example

Purpose

Attribute Acceptance Sampling (Create) designs a single sampling plan for inspection by attributes, where each unit is judged good or defective. You state the quality level a lot should be accepted at (AQL), the quality level it should be rejected at (RQL), and the two risks you can live with, and the tool returns the sample size n and the acceptance number c: inspect n units at random from the lot, accept the lot if c or fewer are defective, otherwise reject it. Use it to set up receiving or final inspection for a supplier or a product when the check is pass or fail (visual defects, go/no-go gauges, functional tests). To compare plans you already have, use Attribute Acceptance Sampling (Compare).

Key Capabilities

  • Finds the smallest single sampling plan (n, c) that meets both the producer's risk at the AQL and the consumer's risk at the RQL.
  • Reports the actual probability of accepting a lot at the AQL and at the RQL for the plan found.
  • Operating characteristic (OC) curve: probability of acceptance against lot quality.
  • With a lot size: average outgoing quality (AOQ) and average total inspection (ATI) curves for rectifying inspection.
  • Quality levels in defectives per million (ppm).
  • PDF export via Download.

Statistical Methods

The probability of accepting a lot with fraction defective p is the binomial probability of finding c or fewer defectives in n units: Pa(p) = Σ (k = 0 to c) C(n, k)·pᵏ·(1 − p)ⁿ⁻ᵏ. The plan must give Pa(AQL) ≥ 1 − α, so that good lots are rarely rejected (producer's risk α), and Pa(RQL) ≤ β, so that bad lots are rarely accepted (consumer's risk β). AQL and RQL are entered in ppm and divided by 1,000,000 to give proportions.

For each acceptance number c from 0 to 60 the tool finds the smallest n with Pa(RQL) ≤ β and checks that this n still gives Pa(AQL) ≥ 1 − α. Of the acceptance numbers that work, it keeps the one with the smallest sample size. If none works, it asks for a wider gap between AQL and RQL or looser risks. The binomial model assumes the lot is large compared with the sample; with a lot size entered, the sample size must be smaller than the lot.

With a lot size N, the AOQ and ATI curves assume rectifying inspection: rejected lots are inspected 100% and defective units are replaced. Then AOQ(p) = p·Pa(p)·(N − n) / N, the average fraction defective that leaves inspection, and ATI(p) = n + (1 − Pa(p))·(N − n), the average number of units inspected per lot. The highest point of the AOQ curve is the average outgoing quality limit (AOQL): the worst average quality that can reach the customer, whatever the incoming quality. All curves are drawn for lot quality from 0 to three times the RQL.

Input Fields & Options

Field

Description

Units for quality level

Defectives Per Million (the only choice).

Acceptable Quality Level (AQL)

Required, in ppm. The quality level at which lots should usually be accepted, e.g. 10000 for 1% defective.

Rejectable Quality Level (RQL / LTPD)

Required, in ppm, larger than the AQL. The quality level at which lots should usually be rejected (lot tolerance percent defective).

Producer's risk (Alpha)

Default 0.05. The highest acceptable chance of rejecting a lot at the AQL.

Consumer's risk (Beta)

Default 0.1. The highest acceptable chance of accepting a lot at the RQL.

Lot size

Optional, at least 2. Needed for the AOQ and ATI curves (rectifying inspection).

Operating characteristic (OC) curve / Average outgoing quality (AOQ) curve / Average total inspection (ATI) curve

All ticked by default. The AOQ and ATI curves are drawn only when a lot size is entered.

Title

Default Acceptance Sampling by Attributes; shown as the heading of the output.

create-attribute-acceptance-sampling-inputs


Figure 1 — Options for the example: AQL 10,000 ppm (1%), RQL 50,000 ppm (5%), α = 0.05, β = 0.1 and lots of 2,000 units.

Output Descriptions

  • Plan table — Sample size (n), Acceptance number (c), AQL (ppm), P(accept | AQL), RQL (ppm) and P(accept | RQL); with a lot size, also the AOQL (ppm) and the incoming quality (ppm) at which it occurs.
  • Operating Characteristic (OC) Curve — probability of acceptance against lot defectives per million.
  • Average Outgoing Quality (AOQ) Curve — when a lot size is given: average outgoing quality in defectives per million against incoming lot defectives per million; the peak is the AOQL.
  • Average Total Inspection (ATI) Curve — when a lot size is given: average number of units inspected per lot against incoming lot defectives per million.
  • PDF export via Download.

create-attribute-acceptance-sampling-output


Figure 2 — The plan for the example: n = 132, c = 3, accepting 95.6% of lots at the AQL and 9.9% at the RQL, with its OC and AOQ curves.

Worked Example

A supplier quality engineer sets up receiving inspection for moulded connectors, delivered in lots of 2,000. Each connector is checked with a go/no-go gauge. Lots that are 1% defective are acceptable to the plant, and lots that are 5% defective must be stopped. The engineer enters AQL = 10000 ppm, RQL = 50000 ppm, keeps Producer's risk = 0.05 and Consumer's risk = 0.1, and enters Lot size = 2000.

The plan is n = 132, c = 3: take 132 connectors at random from each lot, accept the lot if 3 or fewer are defective, reject it if 4 or more are. A lot that is 1% defective is accepted with probability 0.9557, just above the required 0.95, and a lot that is 5% defective with probability 0.0992, just below 0.10. The OC curve shows the full picture: the chance of acceptance falls steeply between about 10,000 and 50,000 ppm. With lots of 2,000 units and rejected lots inspected 100%, the AOQL is 13,747 ppm, reached when incoming lots are about 22,197 ppm defective: whatever the incoming quality, on average no more than about 1.4% defective connectors reach the plant.

If rejected lots are sorted 100% and the defectives replaced, the AOQ curve peaks at about 0.0137, when incoming lots are about 22,000 ppm defective: on average no more than about 1.37% defective connectors reach the line, whatever the supplier sends. The ATI curve shows the cost: about 215 connectors inspected per lot when lots are at the AQL, rising to about 1,815 when they are at the RQL, because most lots are then rejected and sorted. Asking for more protection costs more inspection: β = 0.05 needs n = 181, c = 4, and an RQL of 30,000 ppm needs n = 390, c = 7.