Skip to content

Learn

Quality control in manufacturing: inspection, sampling, CAPA

How inspection plans, sampling, statistical process control and measurement system analysis fit together, and how nonconformances and corrective actions are handled.

Published · Updated · 7 min read

Control and assurance

Quality control (QC) is the checking of products and processes against requirements. Quality assurance (QA) is the wider system that makes good quality likely: procedures, training, change control and records. They work together: QC finds problems, QA reduces the chance that they occur.

Where inspection happens

Inspection is applied at different points depending on risk and cost. A control plan states what is checked, how, how often and what happens when a result is out of limits.

  • Incoming inspection: checking material and components from suppliers
  • First-article inspection: verifying the first part after a setup or change
  • In-process inspection: measuring during production to control the process
  • Final inspection: confirming the finished product before release

Sampling and full inspection

Checking every unit gives the most information but costs time, and manual inspection is not perfect. Sampling checks part of a lot and decides from it. Sampling plans such as those in ISO 2859 define how many units to check and when to accept a lot.

Suppose a single-sample plan specifies 50 units and an acceptance number of one. With two nonconforming units in the random sample, hold the lot for the disposition defined in the plan. Accepting a lot under a sampling plan leaves a risk of undetected defects. Choose the actual plan for the requirement and risk of the characteristic.

Statistical process control

SPC watches the process rather than the product. A control chart shows whether variation is normal or whether something has changed, so you can act before parts go out of tolerance. Capability indices such as Cpk show how much room the process has within the specification.

Trust the measurement

Every measurement includes error. Measurement system analysis, for example a gauge repeatability and reproducibility study, estimates how much of the observed variation comes from the gauge and the operator. Calibration keeps gauges traceable to a reference. Do this before drawing conclusions from the data.

Nonconformance, containment and disposition

When a result is out of limits, the first steps are to stop the spread (containment), identify what is affected and record a nonconformance report (NCR). The disposition then decides what happens to the affected units:

  • Accept under concession with the appropriate authorisation and customer approval where required
  • Rework to meet the requirement
  • Repair to an accepted state
  • Scrap

Corrective and preventive action (CAPA)

An NCR deals with the unit; CAPA deals with the cause. Containment depends on traceability: you can only limit the damage to what you can identify.

A corrective action removes the cause of an existing problem, and a preventive action removes the cause of a potential one. Good CAPA follows a structured analysis such as DMAIC or the five whys, assigns an owner and a date, and verifies afterwards that the action worked.

Standards and further reading

Book a call about your factory

Tell us what you make and which machines you run, and we will walk through how Qynetic fits. If you would rather get news and early access, join the waitlist.

Learn