Catch web defects and registration drift before the roll ships
Willowark builds inspection and data systems for converters and packagers: line-scan web inspection with defect mapping, print registration monitoring, barcode and label verification, and case-count checks at the pack station. The failure we exist to prevent is the expensive one — a repeating defect that ran an entire roll and was found by your customer's receiving inspector instead of by you.
The engineering starts with the physics of imaging a moving web. Line-scan cameras are synchronized to an encoder so every scan line maps to a known position, lighting is chosen for the defect class — backlight for pinholes and gels, darkfield for scratches and streaks, coaxial for print quality — and results are written to a defect map indexed by footage and cross-web position. That map is what lets slitting and rewind pull the affected section instead of quarantining the whole run.
A first project for a converter is usually one press or one lane, chosen because it runs the tightest customer specification or takes the most chargebacks. We start by imaging your actual defect samples on a bench — the ones your customer sent back — to settle resolution, lighting, and line rate before any hardware goes on the machine. Margins in converting are thin and runs are short, so the system has to earn its place on scrap avoided and disputes won, and it has to be set up by a press operator in the changeover window you already have.
Reviewed
Sound familiar?
If you've said any of these, we should talk.
“A repeating defect ran the whole roll and the customer found it, not us.”
Line-scan inspection across the full web width flags defects as they occur and records position by footage and lane. Repeating defects show up as a pattern at a fixed interval, which points straight at the cylinder or roller causing them instead of leaving it to trial and error.
“Barcodes pass on our verifier and then fail at the distribution center.”
We grade codes in-line to ISO/IEC 15416 and 15415 rather than only checking that they decode, so contrast, modulation, and decodability trends are visible before a grade drops below spec. Fixed geometry and controlled lighting make the reading defensible in a dispute.
“Changeover between SKUs takes an hour because every setting is manual.”
We build recipe management so one SKU selection loads registration targets, camera parameters, inspection tolerances, reject timing, and label templates together. Settings become versioned records, so a bad changeover traces to a specific recipe revision.
“We ship short cases and eat the chargeback every time.”
Vision counts units at the case-pack station and cross-checks against a checkweigher, so both a miscount and a wrong-item substitution get caught. Each case carries a serialized label tied to its verified contents, which turns a chargeback dispute into a lookup.
“Our waste number is a guess at the end of the month.”
We meter waste where it is created: footage in versus footage rewound, splice and setup waste per job, and scrap flagged by the inspection system with a reason. Each job closes with a real material yield tied to substrate lot and operator, so waste reduction targets specific presses and causes instead of a plant average.
How this industry actually runs
The operation as we usually find it.
Converting is fast and continuous. Web moves through print, lamination, coating, slitting, and rewind at speeds where a defect born at one cylinder repeats at its circumference and propagates into every downstream roll. Registration between color stations is held to fractions of a millimeter by tension zones and servo-driven units, and undetected drift becomes a trap or shade problem across thousands of impressions. Downstream packagers face different failure modes: carton erection and glue coverage, code and label placement, case count, checkweighing, shrink wrap, and palletizing, where one missing unit becomes a retailer chargeback. Both ends work under customer specifications, GS1 barcode grade requirements, and changeover windows that shrink as SKU counts grow.
Machine signals to the people who decide
Components:
- PLCs & sensors (counts, states, current)
- Legacy machine (dry contact / clamp)
- Edge gateway (normalize, buffer)
- Production dashboard (downtime, OEE)
- Alerts & reports (who acts, when)
Connections:
- PLCs & sensors to Edge gateway (EtherNet/IP, Modbus)
- Legacy machine to Edge gateway
- Edge gateway to Production dashboard (MQTT)
- Edge gateway to Alerts & reports
What we build
Starting projects that fit Packaging.
- Line-scan web inspection with encoder-synchronized defect maps and per-roll quality reports
- Print registration and color-to-color misregistration monitoring with drift alarms
- Barcode grading, variable-data OCR, and label verification with reject control
- Case-count, carton-integrity, and fill-height verification at pack and palletizing stations
- Recipe management so SKU changeover is one selection instead of thirty manual settings
- Roll and pallet traceability linking press, operator, substrate lot, and shift
- OEE and waste tracking per press with scrap reason codes and substrate cost reporting
- Glue and seal verification on carton erectors and case sealers with vision or thermal imaging
Capabilities we bring
Working in Packaging?
Tell us the line.
What runs by hand, what is not connected, what you are trying to build. An engineer replies within one business day with whether and how we would approach it.
Common questions
What Packaging teams ask first.
How fast can a web run and still be fully inspected?
It comes down to resolution math: detection size, web width, and line rate set the pixel budget, and lighting has to deliver enough energy at very short exposures. High speeds are achievable, but the honest answer comes from working that math against the smallest defect you need to catch.
Should defect classification use deep learning or rule-based algorithms?
Rules win when the defect has a stable signature — a hole, a streak, a registration offset — because they are deterministic and easy to justify to a customer. Deep learning earns its place on defects obvious to a person but hard to describe mathematically, such as subtle gels or mottling.
Do we have to stop the line every time a defect is detected?
Usually not, and stopping is often the wrong response on a continuous process. Most converters mark and map instead, tagging the location so the defect is removed at slitting or rewind. Line stops are reserved for conditions that keep producing scrap, such as a lost registration lock.
We already have a print inspection system on the press. Why would we need anything else?
You may not. Press-mounted inspection is usually good at what it was bought for, typically print defects on that press, and the gap tends to be everything around it: data that stays inside the unit, no defect map reaching the slitter, no link between a roll and the job or substrate lot it came from, and nothing at all downstream of the press. We would rather integrate what you have and fill the gaps than replace it.
Can inspection handle clear film, metallized, and printed substrates on the same line?
Usually, but not with one lighting setup. Clear film wants a backlight for pinholes and gels, metallized and foil substrates need diffuse or coaxial illumination to control glare, and printed stock is a different problem again. We typically design the station with more than one light source and switch by recipe, and confirm each substrate family against real samples before committing to the hardware.
Strategy. Software. Systems.
Engineering for Packaging.
Describe the problem in your own words. An engineer reads it — not a sales script — and tells you plainly what it would take.