Services · Acceptance & MSA

Measurement system analysis and acceptance of optical inspection systems

A vision system is not considered reliable simply because it correctly identifies a handful of sample parts. For use in production, it must be verifiably established that measured values and good/bad decisions are repeatable under defined conditions.

THOP plans and supports the acceptance of industrial machine vision systems. Depending on the inspection task, we carry out measurement system analyses or suitable attributive capability studies. The results are transparently evaluated and documented.

01 · Scope

When does a vision system need a measurement system analysis?

A measurement system analysis makes sense when results from the vision system are used for quality decisions, process releases or the automatic rejection of parts. It shows whether an observed deviation actually originates from the product or is caused by the camera, lighting, part positioning, evaluation algorithm or other influences of the inspection system.

Dimensional & position inspection
Contour & geometry inspection
Completeness & presence
Surface & defect detection
OCR, code & classification
AI anomaly detection
Acceptance of new systems
Re-qualification
02 · Approach

Measuring or attributive inspection?

Measuring vision systems

The system delivers a numerical measured value – for length, diameter, distance or position, for example. We examine how much this value scatters across repetitions, whether a systematic deviation is present, and how robust the measurement remains under real production conditions.

Cg / CgkCapability & systematic deviation
Guide value: ≥ 1.33
%GRRShare of measurement system variation
Guide value: < 10 %
  • Procedure 1: Repeated measurement on a defined reference part
  • Procedure 3: Assessment of the automated overall system with real part feeding
  • Influence of part position, lighting, feeding and mechanical handling
  • Comparison against an independent reference value
  • Assessment against the agreed customer and process requirements

Attributive vision systems

The system delivers no numerical value but a decision – good/bad, OK/not OK or detected/not detected, for example. What matters is whether it reliably and repeatably distinguishes real good parts, defective parts and critical borderline samples.

KappaAgreement with the reference verdict
Guide value: ≥ 0.80
  • Procedure 7: Repeated inspection of defined good, defective and borderline samples
  • Establish the reference verdict for good, defective and borderline samples
  • Deliberately include borderline samples close to the decision boundary
  • Verify the repeatability of the classification
  • Make false-accept and false-reject decisions visible
  • Assess robustness against agreed production variation

Cg/Cgk, %GRR and Kappa are common guide values in practice. The actual acceptance limits are defined together with you, depending on standard, customer and inspection task.

03 · Process

How a measurement system analysis runs

1. Define inspection features & requirements

  • We establish which features are examined and which incorrect decisions are especially critical.

2. Determine reference parts & borderline samples

  • Suitable good parts, defective parts and critical borderline cases are compiled.

3. Define the test plan

  • Number of parts, repetitions, positioning and ambient conditions are defined.

4. Carry out the inspections

  • Parts are inspected repeatedly according to the test plan, with results recorded reproducibly.

5. Evaluate the results

  • Variation, repeatability and deviations are traced back to their possible causes.

6. Accept or optimise the system

  • Results are compared with the agreed criteria and serve as a basis for improvement.
04 · Existing systems

Measurement system analysis for existing systems too

THOP can examine not only newly developed THOP systems. Once the technical prerequisites have been clarified, we also support the assessment of existing vision systems – for example after modifications, product changes, software adjustments or in the case of recurring problems in production. These may be customer-specific solutions or systems from established manufacturers.

05 · FAQ

Frequently asked questions

What is needed for a meaningful measurement system analysis?

Suitable reference parts and clear quality criteria. If the definition of good and bad is not yet unambiguous, we first develop a sound basis for the decision together.

Does a measurement system analysis replace an accredited calibration?

No. We document the performance of your vision system within the agreed application and inspection context. For an accredited calibration or independent certification we involve an appropriately authorised body if required.

Which MSA is suitable for a camera system?

This depends on whether the system delivers numerical measured values or attributive good/bad decisions. THOP determines the appropriate approach based on the inspection task and customer requirements.

Can borderline samples also be tested?

Yes. Parts close to the decision boundary are particularly important for assessing the robustness of the classification under realistic conditions.

Does THOP carry out MSA for existing vision systems?

Yes, provided system access, suitable reference parts and sufficient information about the inspection task are available.

Is an MSA already part of the feasibility study?

The feasibility study includes technical repeat trials, but it is not yet a full acceptance of the later production system. A formal measurement system analysis is planned separately and carried out on the defined system.

Is THOP an accredited testing body?

No. THOP carries out application-related investigations and documented system acceptances. If an accredited calibration or independent certification is required, this must be performed by an appropriately authorised body.

Contact

Reliable acceptance for your vision system

Would you like to demonstrate the repeatability or detection performance of a new or existing vision system? Together we define a suitable test procedure and the necessary reference parts.

Discuss a measurement system analysis