Slitter Edge Fray and Width Miss Detection with AI Vision

By David Cook on October 29, 2026

slitter-edge-fray-width-miss-detection

On a converting line, the slitter can be cutting one lane slightly narrow or fraying one edge for hundreds of metres before anyone looks. Manual width checks happen at doff or on a timer, and a frayed edge often shows up first as a complaint from the customer's sewing or laminating line. By then the roll has been wound, labelled and packed. Line-scan vision measures every slit edge as the web runs, grades each roll and holds bad ones at the exit. See it on your own fabric in a slitter vision demo.

CONVERTING LINE AI · SLIT EDGE AND WIDTH

Grade Every Slit Roll for Edge Fray and Width Before It Leaves the Rewinder

iFactory measures each lane's width and edge condition with line-scan cameras at the exit, then grades every roll against its spec. A roll that misses width or carries fray beyond your limit is held at the rewinder, with the metre positions on record.

Slitter 2 · Rewind set R-0609 to R-0614Parent web 320.0 mm
6 lanes × 50.0 mm spec · 20.0 mm edge trim · line-scan at exit
10.2T
50.1L1
49.9L2
50.0L3
49.6L4
50.2L5
50.1L6
9.9T
Lane 4 · 49.6 mm · spec 50.0 ±0.3Below the 49.7 mm minimum from metre 1,140
Lane 2 · edge fray markerLoose fibres past the right cut line, under watch
Roll R-0612 held at exit
SLITTER QUALITY CONTROL

Slitting Methods and the Edge Each One Leaves

Fray starts with the cutting method. Razor slitting is simple and quick to set up, but its blades need frequent changes to keep a good edge. Rotary shear knives stay sharp longer but take more time to position. Crush cutting presses a knife against an anvil and suits nonwovens and foams, though its edge is the roughest of the three. Hot knife and ultrasonic slitting seal the edge as they cut, but only where the fibre is thermoplastic, so cotton and other natural fibres will not seal. Check your method against your fabric with our converting support team.

Method
How it cuts
Edge on textiles and nonwovens
What to watch
Razor
Thin blade through the web, in air or over a grooved roll
Clean while the blade is fresh, fuzzier as it dulls
Blade change interval
Rotary shear
Paired upper and lower circular knives, like scissors
Clean, holds edge quality over longer runs
Knife setting and wear
Crush or score
Rounded knife pressed against a hard anvil roll
Roughest edge, more loose fibre on wovens
Anvil wear and knife load
Hot knife
Heated blade pressed against an anvil
Sealed edge on synthetics, no seal on natural fibre
Blade temperature, speed
Ultrasonic
Vibrating blade or wheel melts and cuts the fibre
Sealed edge on thermoplastic content only
Fibre content, energy, speed
SLITTER EDGE FRAY

Six Line Conditions That Turn a Clean Edge Ragged

A frayed edge is usually a symptom, not a mystery. The same few conditions show up again and again, and most of them build slowly, which is why edge quality drifts during a run instead of failing at once. Logging fray against blade hours and settings is what turns a complaint into a cause. Review your fray history with us in a converting quality session.

01
Blade wearA dulled razor or shear knife tears fibres instead of cutting them, so loose ends grow along the edge.
02
Knife settingOn shear slitters, poor overlap or side load between upper and lower knives leaves a chewed edge.
03
Tension and wanderLow or uneven tension lets the web move at the knife, giving wavy edges and lane width swings.
04
Method and fibre mismatchHot or ultrasonic slitting on a blend with too little synthetic content cannot form a proper seal.
05
Anvil or groove wearA worn anvil on a crush cutter, or a worn groove under a razor, stops the cut going cleanly through.
06
Seal energy too lowLine speed raised without raising heat or ultrasonic energy leaves edges partly sealed and open in places.
WIDTH MISS DETECTION

Width Tolerance, Lane by Lane

Each lane has its own width spec, and the lanes plus edge trim must add up to the parent web. That arithmetic is a useful check. In the set below, the six lanes measure 299.9 mm and the trims 20.1 mm, which makes 320.0 mm. When one lane runs narrow and its neighbour runs wide, the knife between them has usually moved. When every lane drifts together, look at web wander or tension. Map your own lane specs with our process support team.

Set R-0609 to R-0614 · spec 50.0 ±0.3 mmScale 49.5 to 50.5 mm · shaded band in spec
Lane 1 · R-060950.1
Lane 2 · R-061049.9
Lane 3 · R-061150.0
Lane 4 · R-061249.6
Lane 5 · R-061350.2
Lane 6 · R-061450.1
Lane 4 is 0.1 mm under the minimum. Lane 5 sits 0.2 mm wide, so the knife between them is the first thing to check.
CONVERTING VISION SYSTEM LIMITS

What Line-Scan Vision Measures, and What It Can't Judge

A line-scan camera builds an image one line at a time as the web passes, which suits continuous material. Dedicated edge cameras of this kind report edge position in hundredths of a millimetre. So width and visible fray are well within reach. Seal strength, fibre content and anything inside a wound roll are not. See where the line is drawn in a live measurement demo.

Line-scan can measure
  • Every slit edge position, so each lane width, along the whole run
  • Loose fibres and fringe visible beyond the cut line
  • Wavy edges and lane wander from scan to scan
  • The metre position of each event, from the line encoder
It cannot judge
  • Whether a sealed edge will hold under load or washing
  • Fibre content, so whether a blend can seal at all
  • Relaxed width of knits and stretch fabrics after tension is released
  • Damage inside the roll once the layer is wound over
How the rules fill the gap

Width is graded against the spec at line tension, and stretch fabrics carry a correction set from relaxed lab samples. Seal strength and fibre content stay with the lab and the material spec. The camera grade only releases what it can actually see.

Ask Which Roll, Which Lane, Which Metre

The line head does not need another trend screen. iFactory answers in plain words, names the roll and the metre range, and points to the knife or setting most likely behind it.

Slitter 2 · Line head and iFactory
Line headWhy is R-0612 held?
iFactoryLane 4 measured 49.6 mm from metre 1,140 to the end of the set. The minimum is 49.7 mm. Lane 5 ran 0.2 mm wide over the same length.
Line headAnd the fray on lane 2?
iFactoryWithin limit so far, but rising since the last blade change. R-0610 is released with a note. Check knife 4/5 position before the next set.
SLIT ROLL AUDIT

Roll-Level Grading From Metre-Level Data

The camera sees thousands of edge readings per roll, but the decision is per roll. iFactory rolls the readings up into one grade for each lane roll, using the width spec and fray limits you set for that product and customer. The grades below show the logic. The limits are yours, not ours. Our quality support team can help you write the grading rules.

A
ReleaseWidth in spec over the full length, no fray beyond your limit.
B
Release with noteWidth in spec, short fray events under the limit, logged on the roll record.
C
Downgrade or re-slitFray over the limit on one edge, or width out briefly within an allowed length.
H
Hold at exitWidth out of spec beyond the allowed length, or fray over the limit on both edges.
CONVERTING LINE AI

How a Held Roll Is Stopped at the Rewinder

Holding a roll only works if it cannot slip through packing. iFactory maps each event to its lane, roll and metre, grades at doff and sets a hold that the line or MES respects. Whether that hold is a PLC signal, a label that will not print or an MES status depends on how your line is wired. Plan the hold path in a line integration call.

1
MeasureLine-scan tracks every slit edge across all lanes at the exit, scan line by scan line.
2
Map to rollThe encoder ties each event to a lane, a roll ID and a metre position.
3
Grade at doffWhen the set doffs, each roll is graded against its product spec.
4
HoldA held roll gets a hold status and no release label, so it cannot be packed.
5
DecideQA reviews images at the flagged metres and chooses re-slit, downgrade or scrap, with the cause logged.
SLITTER ANALYTICS

A Shift Report the Line Head Can Act On

Roll grades answer whether to ship. The shift view answers what to fix. Linking fray and width to blade hours, knife positions and line speed shows which setting is drifting before it costs a roll. The cards below are an illustration of one shift, not measured results. Get a sample report from our analytics team.

Lane rolls held: 1 of 36R-0612, width under minimum from metre 1,140
Fray trend: lane 2 risingEvents per 1,000 m climbing since the last blade change
Knife to check: 4/5Narrow and wide neighbours point at one position
Blade hours loggedEach knife's run time shown next to its edge quality
Speed versus sealFray events compared with speed steps on sealed products
Width by parent rollLane drift compared across parent rolls and suppliers
SLITTER QUALITY COST WORKSHEET

Put Your Own Numbers on Fray and Width Misses

We don't publish defect rates or savings figures, because every line and fabric is different. Every input below is assumed, and the worksheet only multiplies them. Put your own complaint log and check intervals in their place, or work through it with us in a cost review call.

Rolls rejected by customers per month6, assumed
Cost per rejected roll: freight, replacement, handling$310, assumed
Customer rejection cost per month6 × $310 = $1,860
Width drift events per month4, assumed
Lane metres run before a manual check finds each one1,000 m, assumed
Fabric value per lane metre$0.45, assumed
Off-spec material per month4 × 1,000 × $0.45 = $1,800
Total per month, then per year$1,860 + $1,800 = $3,660 · × 12 = $43,920

This is the cost exposure on the assumed inputs, not a promised saving. How much of it continuous measurement removes depends on how fast your team acts on a hold.

TURNKEY AI DEPLOYMENT

Line-Scan Cameras, Edge Server and Roll Holds in One Delivery

iFactory delivers slitter vision as a single package. Line-scan cameras and lighting are mounted at the slitter exit, and a pre-configured NVIDIA AI server ships racked and ready with the edge and width models loaded, so measurement runs on site at line speed. You rack it, plug in power and Ethernet, and the AI is live. Our deployment team can survey your slitters.

What's Included
  • Pre-configured NVIDIA AI edge server, racked and ready
  • Line-scan cameras and lighting sized to your web width
  • Edge, width and fray models tuned to your fabrics
  • Encoder link for lane, roll and metre mapping
  • Roll grading rules per product and customer
  • Hold signal to the rewinder, labeller or MES
  • 24×7 remote monitoring of models and hardware
Live in 6–12 weeks
Wk 1–4
Mount and Calibrate

Cameras fitted at the exit, width calibrated against gauge measurements, encoder linked.

Wk 5–8
Grade Without Holding

Rolls graded silently and compared with manual checks while fray limits are agreed.

Wk 9–12
Holds Go Live

Exit holds switched on, operators trained, shift report started for the line head.

HOW iFACTORY FITS

Made for Slitting and Rewinding Lines

iFactory works alongside your slitter controls, not in place of them. It adds a measurement and a decision at the point where rolls leave the line. Start with a pilot scoping call for one slitter.

1
Measures every lane. Width and visible fray are tracked edge by edge over the full run.
2
Grades every roll. Metre-level readings become one grade per lane roll, against your spec.
3
Holds bad rolls at exit. A failed grade stops release at the rewinder before packing.
4
Points at the cause. Fray and width tie back to knives, blade hours and speed.
1000+
Industrial clients running iFactory
99.9%
Platform uptime on round-the-clock slitting
6-12 wks
From racked server to graded, held rolls
FREQUENTLY ASKED QUESTIONS

What Converting Line Heads Ask About Slitter Vision

How is lane width measured on a moving web?
Line-scan cameras at the slitter exit find each slit edge on every scan line. The distance between neighbouring edges is the lane width, and the line encoder gives the metre position of each reading. Calibration against gauge measurements is done at install and checked on a schedule.
Can it see fray on sealed synthetic edges?
It can see loose fibres and open sections that stand out from the cut line. It cannot test whether a hot knife or ultrasonic seal will hold under load, which stays a lab check. You can see fray detection on your own material if you book a 30-minute demo.
What about knits and stretch fabrics?
These are measured under line tension, and their relaxed width can differ. A correction is set from relaxed lab samples for each product, and the grade uses it. If a product's relaxed width varies too much, it stays on a manual check alongside the camera.
How does a roll actually get held at the exit?
Through whatever your line already uses: a PLC signal to the rewinder, a release label that will not print, or a hold status in your MES or ERP. The path is mapped during the survey with our integration support team.
Who sets the fray and width limits?
You do, per product and per customer. iFactory applies those limits to every roll and records the result, and the shadow weeks are used to agree fray limits from real images. To see how limits are set up, walk through the grading screen with us.
Where should a converting plant start?
Start with the slitter that draws the most edge or width complaints. Grade silently for a few weeks and compare with manual checks, then switch on exit holds. To plan that first slitter, see it running on a live line.

Let No Off-Spec Roll Reach Packing

Measure every lane, grade every roll and hold the ones that miss width or fray limits, with metre-level evidence. It ships as one turnkey package and runs on your slitter within 6 to 12 weeks.

Example set summary · Slitter 2
Lanes measured6, edge by edge, full length
Width spec50.0 ±0.3 mm per lane
Graded5 released, 1 held
Watch itemLane 2 fray, knife 4/5 position
Every hold traced to lane, roll and metre

Share This Story, Choose Your Platform!