Selvage Issue Detection with AI Vision in Textile and Apparel Manufacturing
How NorrStudio by NorrSpect identifies tight selvages, loose selvages, curling edges, draw-in faults, and loom-induced selvage structural failures in woven fabric rolls at production speed preventing selvage geometry defects from causing cutting room yield loss, spreading failures, and garment dimensional problems.
91%
Reduction in selvage-related spreading failures and cutting room rework
2mm
Minimum selvage draw-in deviation detectable inline at production speed
97.3%
Detection accuracy for tight selvage, loose selvage, and curling edge conditions on woven fabrics
Overview
Selvage issues are geometrical defects that originate at the loom but cause their most expensive damage in the cutting room. A tight selvage where the weft thread tension at the selvedge edge is higher than at the fabric centre causes the fabric to draw in at the edges during spreading, producing a curved lay that makes accurate marker placement impossible and forces the cutting room to either accept reduced marker efficiency or re-spread at a slower speed to tension the fabric flat. A loose selvage causes the opposite problem: excess fabric at the edges buckles and waves during spreading, again preventing flat lay and accurate cutting.
NorrStudio, developed by NorrSpect, continuously monitors selvage geometry at both fabric edges at production speed measuring draw-in, curl, waviness, and structural integrity providing the weaving and finishing departments with real-time process correction data and the cutting room with roll-level selvage quality profiles that enable informed spreading decisions before the roll reaches the cutting table
About NorrSpect
NorrSpect is a Swedish AI company headquartered in Umeå, Sweden, specialising in industrial visual inspection for precision manufacturing. Its NorrStudio platform is deployed and validated in automotive and industrial sectors including by manufacturers such as Volvo Cars and is now purpose-built for textile and apparel quality inspection. Selvage geometry detection models are defined and validated during the pilot phase using real production fabric samples and loom parameter data from each client facility.
Industry challenge: how selvage geometry defects cascade through production
Selvage geometry defects are unique in that they do not merely affect a zone of the fabric they affect the entire spreading and cutting process for every roll that carries the fault. A roll with a tight selvage on the right edge cannot be spread flat without either stretching the right selvedge zone which introduces dimensional distortion into the cut panels or accepting a curved lay that reduces cutting accuracy across the full marker width. Neither outcome is acceptable for precision garment production, and neither is visible from the roll label or the incoming inspection report unless selvage geometry has been explicitly measured.
The process causes of selvage defects are rooted in loom tension management. Weft thread tension at the selvedge is controlled by the shuttle, rapier, or air-jet insertion mechanism, and small deviations from the optimum selvage tension setting which varies with weft yarn count, crimp, and elasticity produce tight or loose selvage conditions that compound along the roll length as the warp beam depletes and tension equilibrium shifts. Without inline measurement, the only feedback available to the weaver is the finished roll's spreading behaviour by which point thousands of metres of affected fabric may already be in the dye house or finishing queue.
Tight selvage
Weft threads at the selvage edge interlacing under higher tension than the fabric body — drawing the edge inward and causing the fabric to narrow progressively along the roll, producing a curved lay and reduced usable width at spreading
Loose selvage
Weft threads at the selvage edge interlacing under lower tension than the fabric body — leaving excess yarn at the edge that buckles and waves during spreading, preventing flat lay and causing cutting inaccuracies across the full marker width
Selvage curl
The selvage zone rolling upward or downward relative to the fabric body — caused by unbalanced warp and weft tension or by finishing shrinkage differential between the edge and centre — preventing flat spreading and causing edge panels to cut inaccurately
Draw-in variation
The fabric width varying along the roll length due to fluctuating selvage tension — producing a roll where usable width changes from metre to metre, making it impossible to set a reliable selvedge allowance for marker planning
Leno selvage failure
The leno selvedge weave structure failing over a roll section — caused by a leno thread break or a doup heddle failure — leaving a zone where the selvage has no locking structure and the edge is held only by plain interlacement, causing progressive unravelling during handling
Tucked selvage defect
The tucked selvage — where the weft selvedge end is tucked back into the shed — failing intermittently due to tucker needle wear or timing deviation, producing a zone of loose weft ends at the selvage that create a structurally weakened edge
Solution: NorrStudio AI selvage geometry detection and loom tension diagnostics
NorrStudio uses dedicated full-edge imaging at both selvage positions simultaneously, with edge geometry analysis algorithms tracking selvage position, curl angle, waviness amplitude, and structural integrity at every metre of the roll. Draw-in is measured as the deviation between the actual selvedge position and the specification width — a negative deviation indicating tight selvage draw-in, a positive deviation indicating loose selvage excess. Curl is measured as the out-of-plane angle of the selvage zone relative to the fabric body. Both measurements are expressed as continuous roll profiles that give the weaving and finishing departments the process correction data they need to address the root cause.
Measures selvage draw-in continuously at both edges simultaneously detecting width deviation of 2mm or greater from the specification width at production speed
Quantifies selvage curl angle at both edges flagging curl exceeding the spreading flat-lay tolerance as a process correction alert to the finishing department
Detects loose selvage waviness by measuring selvage boundary irregularity amplitude distinguishing genuine loose selvage buckling from normal selvedge weave texture variation
Identifies draw-in variation along the roll length providing the cutting room with a metre-by-metre usable width profile that enables precise selvedge allowance setting per roll section
Detects leno selvage failures and tucked selvage defects by monitoring selvage structural integrity flagging zones where the selvage locking mechanism has failed before the edge unravels during subsequent handling
Correlates tight and loose selvage patterns to specific loom tension settings and warp beam depletion stages providing the weaving department with actionable tension correction data rather than generic selvage quality complaints
Generates roll-level selvage geometry reports for the cutting room confirmed usable width profile, curl severity map, and spreading risk classification per roll
Solution
NorrStudio AI Inspection Selvage Issue Detection Module
Inspection scope
Woven fabric rolls at loom exit batching frame and stenter exit across all weave constructions
Hardware
Dual selvage line-scan cameras, oblique edge illumination, out-of-plane curl sensor, motion-sync encoder
Output
Real-time selvage alerts, draw-in profiles, curl maps, usable width reports, loom tension signals, PDF QA archive
Integration
Loom tension control systems, stenter pin rail spread control, ERP / WMS, cutting room CAD and marker planning
Deployment time
Pilot phase calibrated to client fabric construction, selvage type, and buyer width specification before full deployment
Use case: bottom-weight woven mill tight selvage elimination for tailored trouser buyers
The problem: A bottom-weight woven fabric mill producing wool-polyester blend trouser fabric for European tailoring brands was experiencing a persistent tight selvage problem on its rapier looms approximately 12–16% of rolls per production run were being returned or downgraded by the cutting-and-sew factory due to tight selvage draw-in that prevented flat spreading and caused curved marker lays, reducing cutting accuracy on the trouser leg and waistband panels that required precise grain alignment. The mill's outgoing inspection had no objective selvage geometry measurement QA sign-off relied on the inspector pulling the roll end flat by hand and assessing visually, a method that consistently underestimated draw-in on rolls where the tight selvage was gradual rather than sudden.
The NorrStudio solution: NorrStudio was installed at the loom exit batching frame across twelve rapier looms. Draw-in measurement was calibrated to the buyer's ±4mm usable width tolerance. The system immediately revealed that tight selvage draw-in was occurring progressively as the warp beam depleted selvage tension increasing as beam diameter decreased and the let-off mechanism failed to compensate proportionally. A let-off tension compensation curve was programmed into the loom control system for each fabric weight, correcting the progressive draw-in. Roll-level usable width profiles were transmitted to the cutting factory with each roll, replacing manual spreading assessment with confirmed width data.
Results:
Metric | Before NorrStudio | After NorrStudio |
|---|---|---|
Roll return rate from tight selvage at cutting factory | 12–16% per production run | <1% per production run |
Tight selvage draw-in detection method | Visual pull assessment — subjective | Continuous 2mm-resolution inline measurement |
Let-off tension compensation fault identified | Unknown — generic selvage complaint | Progressive beam depletion draw-in — identified in first week |
Cutting factory spreading failures per month | 8–12 per month | 0–1 per month |
Usable width data provided to cutting factory | Nominal specification only | Confirmed per-metre usable width profile with each roll |
Roll-level selvage QA documentation | None | Full draw-in and curl profile per roll, archived and shareable |
What is the difference between selvage issues and edge fraying and does NorrStudio detect both?
Edge fraying is a structural defect weft threads physically pulling free from the selvage weave, reducing the fabric edge to loose threads that must be trimmed away. Selvage issues are geometric defects the selvage structure is intact but its tension, position, or geometry deviates from specification, causing spreading and cutting problems without any visible thread loss. NorrStudio detects both categories: the edge fraying module monitors for thread loss and structural failure, while the selvage issue module monitors for draw-in, curl, waviness, and leno or tucked selvage mechanism failures. The two modules operate simultaneously from the same edge imaging hardware.
How does NorrStudio measure selvage curl a three-dimensional out-of-plane deformation with a flat camera system?
NorrStudio uses an oblique edge illumination configuration where the light source is positioned at a low angle relative to the fabric plane. A selvage zone that is curling upward or downward relative to the fabric body casts a characteristic shadow profile on the fabric surface adjacent to the selvage the shadow width and shape are directly proportional to the curl angle. This shadow geometry measurement provides a reliable proxy for curl angle without requiring a separate out-of-plane sensor, enabling curl detection at full production speed using the same camera hardware as the draw-in measurement system.
Can NorrStudio identify whether a tight selvage is caused by weft tension or by stenter over-stretching during finishing?
Yes. Loom-induced tight selvage produces a draw-in pattern that is present on the greige fabric at the loom exit detectable at the batching frame before any finishing. Finishing-induced tight selvage from inadequate stenter pin spread or differential shrinkage during heat-setting only appears after the stenter exit. Deploying NorrStudio at both points allows the facility to attribute each selvage condition to its process origin: if draw-in is present at the loom exit, it is a weaving tension issue; if it only appears after the stenter, it is a finishing process issue.
Can the draw-in measurement data be fed back to the loom's let-off tension control in real time?
Yes. NorrStudio's continuous draw-in measurement can be integrated with the loom's electronic let-off tension control system to provide closed-loop selvage width correction automatically adjusting the let-off tension as the warp beam depletes to maintain consistent selvage tension and usable width throughout the beam's full length. This closed-loop integration is particularly valuable for fabrics where the optimal selvage tension changes significantly between full beam and depleted beam conditions.
How does the cutting factory use the roll-level usable width profile delivered by NorrStudio?
The cutting factory receives the confirmed usable width profile expressed as minimum usable width at each metre of the roll before the roll reaches the spreading table. The marker planner uses this data to set precise selvedge allowances per roll section, maximising marker efficiency on conforming sections while allocating appropriate allowances on sections with measurable draw-in. For rolls where draw-in variation is significant, the profile enables the roll to be split at the worst draw-in sections, treating each conforming section as a separate spreading unit rather than applying a conservative blanket allowance across the full roll
Similar Topic
