Validating an Automated Counting System: Requirements, Test Plans and Ongoing Verification

Written by Sciotex Engineering
In this article

Validating a counting system means proving, with documented evidence, that it counts within your required accuracy under production conditions, and then keeping that proof current. For operations with a formal quality system, such as medical device, aerospace and automotive suppliers, the work follows a familiar pattern: define requirements, verify installation, test operation across the range of conditions, demonstrate performance over a production run and set a routine to keep verifying. The sample size follows from the accuracy you are trying to prove.

This article extends our count accuracy article for quality-controlled environments and supports the automated counting guide.

Step 1: Write the requirements

  • Accuracy: the maximum acceptable count error, for example no more than 0.05% of parts.
  • Part types and ranges: sizes, materials, finishes and the variation within each.
  • Throughput: parts per minute or lots per hour.
  • Count range: minimum and maximum lot size.
  • Environment: lighting, vibration, dust, temperature.
  • Records: what must be saved for each lot (count, time, operator, part number, image).
  • Failure behavior: what happens on a jam, a power loss or an uncertain count.

Step 2: Installation checks

Confirm the equipment is installed as specified: model and software versions, power and utilities, safety devices, network and data connections, and calibration or setup records. Record serial numbers and settings.

Step 3: Operational tests

Run known-count lots to show the system behaves correctly across the operating range:

  • Minimum and maximum lot sizes and throughput.
  • Each part type and each recipe.
  • Worst-case parts: oily, bent, burred, sticky or densely packed.
  • Foreign or mixed parts, to see whether they are flagged.
  • Interruptions: power cycles, jams, lot changes and restarts.
  • Operator variation, if operators load or handle the parts differently.

Each test needs a defined pass criterion and a reference count verified independently, for example by two manual counts.

Step 4: Performance demonstration and sample size

Run enough production-like lots to support the claimed accuracy. With zero errors, the rule of three says the error rate is below 3 divided by the number of parts counted, at roughly 95% confidence.

Required error rateParts to count with zero errors
1%300
0.1%3,000
0.05%6,000

Count that many parts, split over many lots, different shifts and operators, and each part type. If errors occur, the statistics change and more parts are needed, so investigate every error and its cause before continuing.

Step 5: Ongoing verification

Validation is not permanent. Define a routine:

  • A reference-lot check at the start of each shift or each part change.
  • A fuller check after any change to parts, lighting, location, software or recipes.
  • Records of each check, reviewed periodically for drift.
  • A procedure when a check fails: stop, correct and quarantine or re-count everything since the last good check.
  • Change control for software, recipes and settings, with access limits.
  • Periodic re-validation and review of the requirements.

Common weaknesses in validation

  • Using easy parts instead of worst-case parts.
  • Too few parts to support the accuracy claimed.
  • Reference counts that were not independently verified.
  • No plan for new part numbers.
  • No documented response to a failed verification.

For the method details, see count accuracy and counting technologies compared.

Need a counting system that will pass your quality audit?

Sciotex builds automated counting machines and supports validation with free sample testing and documentation. Tell us your accuracy requirements.

Explore automated counting machines

Frequently asked questions

What does validating a counting machine involve?

Defining requirements, verifying installation, testing operation across lot sizes, parts and conditions, demonstrating accuracy over enough production, and setting up ongoing verification.

How many parts do I need to count to validate accuracy?

With zero errors, the error rate is below 3 divided by the parts counted at about 95% confidence. Proving 0.1% needs about 3,000 error-free parts, and 0.05% needs about 6,000.

How often should a counting machine be verified?

At least a quick reference-lot check at the start of each shift or part change, plus a full check after any change to parts, lighting, location, software or recipes.

What should I do if a verification check fails?

Stop counting, find and correct the cause, and quarantine or re-count the lots counted since the last passing check.

Do new part numbers require revalidation?

New parts should at least be verified with reference lots and, depending on your quality system, validated through the same change-control process as the original.

Related articles

Written by

Sciotex Engineering

The Sciotex engineering team has designed and built vision, automation and test systems since 1997.

Have an inspection problem to solve?

Talk it through with a Sciotex engineer.

In this article
Have a counting problem to solve?

Talk it through with a Sciotex engineer.

Next step

See your parts counted