Neoritma · Editorial team · 16 October 2026
Choose one inspection question
Start with a question rather than a technology list: which defect should be detected, on which product, and where? A missing-label check and print readability need different acceptance criteria. Write down the boundary between accepted and defective products with production and quality teams.
Separate a representative sample set
Collecting only ideal products makes a pilot look easier but does not represent the line. Include shifts, print quality, orientation and packaging variants. Label known defects separately and retain an independent acceptance set instead of testing only on the samples used during development.
Record imaging conditions
The camera model alone is not the solution. Test distance, field of view, movement, lighting and trigger position together. Check whether image capture is repeatable first. Assuming software can always correct reflections or motion blur may lead to the wrong equipment choice.
Measure error types separately
One accuracy percentage can hide the type of failure. Count missed defects, false rejects and no-decisions separately. Record sample size and product mix. The customer team should set acceptance limits against process risk rather than applying one universal percentage.
Connect decisions to line actions
An OK/NOK result does not define the line response. Test a failed reject mechanism, network loss and operator reinspection. The technical scope should identify PLC outputs, recorded events and the conditions that require the line to stop.
Test changeovers in the pilot
If production changes products, recipe selection belongs in the test plan. Labels, camera references and production records must identify the same product. Define operator warnings for a wrong recipe or a previous work order left open.
Document records and access
If images or product records are retained, agree access, location and deletion rules. An operator screen and remote-support access need different permissions. Keep unnecessary personal or confidential information out of pilot datasets and use approved company channels for file sharing.
Turn the result into a decision record
The report should show samples, conditions, error types and unresolved issues. The next decision may be installation, another trial or a different approach. Discuss maintenance, training, spares and acceptance testing together with price. A successful demonstration is not a guarantee for every production condition.
Related solutions
Questions before a project
Will it work with our existing system?
The team reviews machines, PLCs, software and interfaces. Compatibility, connection responsibility and additional hardware are defined in the technical offer; universal compatibility is not assumed.
Can we start with a pilot or trial?
Start with a limited task and acceptance criteria. Samples or site data define the trial, measures and next decision. Pilot availability and terms are agreed for the specific project.
Does installation stop production?
Connections and safety requirements may need a planned stop or test window. Agree the schedule with production and maintenance teams; zero downtime is not guaranteed.
How are capacity and price determined?
Product or load samples, target flow, conditions and integration scope are reviewed. The final configuration and commercial offer are based on these inputs.
How are data and software interfaces planned?
Define data sources, interfaces, permissions, retention and recovery responsibilities. On-site or cloud deployment is assessed against the customer environment.
What is the service and maintenance scope?
Discuss maintenance access, training, spares and remote support in the offer. Coordination with STU Servis is available through the group; warranty and response commitments require a written scope.