Cap torque is usually checked by hand, on a few bottles, every so often. The other thousands that pass the capper in between are trusted on faith. Vision cannot read torque directly, but it can see the signs a poorly tightened cap leaves behind: a tilt, a raised crown, a band out of place. Food plants weighing that approach can ask iFactory AI how closure checks are connected to capper data.
Best Vision Torque Verification Software for Food Plants: A 2026 Guide
Learn what vision can and cannot tell you about cap torque, how to calibrate it against a torque tester and what to ask software suppliers. iFactory AI links closure vision to capper health for food lines.
The Problem With Checking a Few Bottles
A torque tester gives a trustworthy number on one bottle at a time. That makes it ideal for setting up a capper and poor for watching a fast line. Between checks, anything can change.
The bars show an illustrative line running 600 bottles a minute, with five bottles tested every 30 minutes. In that half hour 18,000 bottles are capped and 5 are checked, about 0.03 percent. A head that starts to slip right after a check can run for nearly 30 minutes unseen.
Accurate but sparse
A direct torque reading, in a laboratory-like setting, on a handful of bottles.
Dense but indirect
Every bottle is imaged, but the camera sees geometry, not newton-metres.
Calibrated proxy
Vision features are tied to torque readings, then used on every bottle.
What Vision Can See and What It Cannot
Be clear on the limits before buying anything. A camera measures what it can see. A cap's torque is a force, so vision can only infer it. A packaging press item describing one inspection system says it checks for missing, incorrect, skewed and under- or over-torqued caps. That is a supplier claim, so ask any vendor how the torque judgement is made and proven.
| Feature | Seen Directly? | What It Suggests About Torque |
|---|---|---|
| Cap present | Yes | Nothing on torque, but catches the worst closure fault |
| Cap skew or tilt | Yes | Cross-threading or poor seating, often with abnormal torque |
| Cap height above the finish | Yes, with the right lighting and side view | A high cap may be under-tightened or not fully seated |
| Tamper band position and integrity | Yes | A stretched or split band can mean over-torque or damage |
| Liner present | Sometimes, before capping | Missing liners cause poor sealing |
| Thread engagement | Only by inference from height and angle | Short engagement can mean a loose cap |
| Torque in newton-metres | No | Must be calibrated against a torque tester |
The honest claim for vision is closure verification: it finds caps that look wrong, on every bottle, and gives an early signal when the capper drifts. It does not replace the torque tester, and it should never be described to auditors as a torque meter.
Application Torque and Removal Torque
Two torque numbers matter. Application torque is what the capper applies and cannot be measured directly after the fact. Removal torque is what it takes to open the cap, and it is the number a tester can read. O.Berk explains that removal torque is used as an indication of application torque.
As an illustration, a 28 mm closure suggests about 14 inch-pounds of application torque on that rule. It is a starting point only, since liner, thread style and product change the answer. The Frain paper adds that once the relationship is established, a given application torque should consistently give the same removal torque.
Check Every Cap, Not Just Samples
Book a 30-minute session and iFactory AI will show how closure images, capper data and torque tests combine into one view for a food line.
Why Torque Wanders: Causes the Camera Can Help Spot
The Frain Group paper on common capper problems lists many causes of torque variation. Some show up in the image of the finished cap, and some do not. Knowing which is which tells you where vision adds value.
| Cause in the Source | Effect | Vision Signal |
|---|---|---|
| Clutch contamination from washdown water or product | Lower friction, or lock-up with sugar-based product, giving low or high torque | Head-to-head scatter in cap height and skew |
| Chuck slippage from low downforce | Uncontrolled slip sets the torque; scuffed caps | Scuff marks and cap rotation offsets |
| Missing cap liners | Poor sealing, feed jams | Liner check before capping |
| Bottle and cap variation | Neck outside diameter and cap inside diameter change the fit | Height and angle shifts by lot |
| Bottle neck ageing | Necks shrink or expand, especially light bottles stored in uncontrolled warehouses | Slow drift after a new bottle lot |
| Bottle slippage or deformation | Bottles turn or collapse while capping | Tilted caps and bottle shape changes |
Many of these have a visible footprint on the cap. That is the case for vision: not that it reads torque, but that most of the causes of bad torque show on the outside.
Building the Proxy: Calibrating Vision Against a Tester
A trustworthy proxy comes from data, not from a vendor's claim. The method is plain: collect paired vision and torque readings, look for the relationship and set limits you can defend.
| Cap Height Above Nominal | Skew | Removal Torque Seen | Vision Call |
|---|---|---|---|
| 0.0 to 0.3 mm | Under 1 degree | Within range | Pass |
| 0.3 to 0.6 mm | 1 to 2 degrees | Low end of range | Watch |
| Above 0.6 mm | Above 2 degrees | Below range | Reject |
| Below nominal, band stretched | Any | Above range | Reject |
All the figures in this grid are invented to show the method and are not limits to copy. Your bottle, cap and liner set their own. Re-check the link whenever a new bottle or cap lot arrives, since both change torque.
Head-by-Head View of a Rotary Capper
On a rotary capper each head is its own small machine. A single worn chuck can produce most of the rejects while the line average looks fine. Software that tags each bottle with its head number turns scattered rejects into a clear list.
In this illustration, head 7 rejects 3.6 percent of its bottles while others sit under 1 percent. The line average across 12 heads would read well under 1 percent and hide the problem. A maintenance job to check the chuck and clutch of head 7 is the obvious next step.
Keep the Torque Tester Honest Too
The tester is the reference for the whole scheme, so its results must be sound. The Frain paper makes a point many teams overlook: variability may come from improper testing rather than inconsistent capping.
Record who tested, which tester and when it was last calibrated, next to each reading. Without that, a bad reading cannot be traced to its cause, and the vision calibration built on it is weaker.
What to Do When a Cap Fails the Check
A rejected cap is only half the story. The line needs a clear response so the cause is fixed and not just the bottle removed.
Reject bottles should leave the line automatically and be counted by head and by reason. The count is the data that shows whether a fix worked.
How to Judge the Best Software: A Buyer Scorecard
Closure inspection tools vary from fixed camera units to flexible AI platforms. Score each against the same criteria. The weights are a starting point.
| Criterion | What to Ask | Weight |
|---|---|---|
| Honest claims | Does it say plainly what it infers and how that was validated against torque readings? | 20 |
| Imaging quality | Can it see height, skew, band and thread area at your line speed and cap colours? | 20 |
| Head tracking | Does it tag each bottle with the capper head and report by head? | 15 |
| Changeover | How fast can a new bottle or cap be taught, and how is it re-validated? | 15 |
| Evidence and audit | Are images, limits and decisions stored and traceable? | 15 |
| Capper link | Does it feed maintenance and adjustment, not just reject bottles? | 10 |
| Cost and support | Is pricing clear and is local support available? | 5 |
| Total | 100 |
Insist on a trial with your own bottles, your own caps and your worst lots. A demo with clean parts under perfect light says little about a real line.
What Closure Faults Cost
A loose cap can mean a leak, a spoiled product, a returned pallet or a recall in the worst case. A tight one can mean a broken band or a customer who cannot open the pack. These figures are illustrative.
The direct cost looks small, but a fault left for a shift is eight times that, and the real exposure is a complaint or a retailer rejection, which cost far more per event. With vision, the same head is flagged within minutes, the bad bottles are rejected and the exposure is a handful of bottles.
Also count the quiet costs: time spent sorting suspect pallets, extra torque tests after a complaint, and the product held while the team decides what to do. A closure record that shows which bottles came from which head, and when, shortens every one of those investigations from days to hours.
Camera and Lighting: Getting a Clean View of the Cap
Vision results are only as good as the picture. Closure inspection is harder than it looks, because caps are shiny, curved and often coloured, and the line moves fast. A good set-up is worth more than a clever model.
Side and top
A top view shows presence and band colour, while a side view shows height and tilt. Most torque proxies need the side view.
Control the glare
Diffuse or angled light and shields reduce reflections from glossy caps and wet bottles.
Trigger consistently
A fixed trigger point and stable bottle position keep measurements comparable from one bottle to the next.
Plan for the awkward cases at the start: dark caps, clear caps, sports caps with pull tops, wet bottles after a pasteuriser and condensation on cold fills. Test the camera on each, because a set-up that works for white caps may fail on black ones.
Keep the optics clean. A smudged lens or a loose bracket moves the measurements slowly, and the drift looks like a capper fault. Add the camera to the cleaning and maintenance plan, and use a reference part to check it at each start-up.
Common Closure Verification Mistakes
Each of these shows up often in plants that adopt vision. Each has a plain fix.
A 90-Day Path to Closure Verification
iFactory AI is built as a turnkey server that racks in, plugs in and goes live in roughly 6 to 12 weeks, so a one-line closure pilot fits inside a quarter.
What iFactory AI Adds
iFactory AI is smart manufacturing and industrial software. For closures it joins images, capper data and maintenance in one record for food lines.
Plain-Language Glossary
Shared words keep quality, packaging and maintenance aligned.
Frequently Asked Questions
Can a camera measure cap torque?
Not directly. It sees position, tilt and band condition, which relate to torque. Calibrate those features against a torque tester before using them as a proxy.
Does vision replace manual torque testing?
No. Manual testing stays as the reference and for periodic recalibration. Vision adds checks on every bottle between those tests. Ask support how the two are combined in one workflow.
What target torque should we use?
Use your closure supplier's guidance and your own correlation work. A common starting rule is about half the closure millimetres in inch-pounds, but each package differs.
How many bottles do we need for calibration?
Enough to cover every capper head and each bottle and cap lot, often a few hundred pairs. Your quality team should set the plan.
How fast can we change over to a new product?
That depends on the tool. Ask for a changeover demo with your own parts and confirm how limits are re-validated for the new closure.
Verify Every Closure on Your Food Line
iFactory AI links closure images, torque tests and capper health so your team can catch loose and tight caps before they ship. Book a walkthrough on your own line data.
