The Elastane Problem: Why Fabric Inspection Still Fails at Core-Spun Yarn

Every defect in circular knitting has a story. Holes are loud — yarn breaks, needles fail, fabric tears, and everyone on the floor knows within minutes. Needle lines are repetitive — the same damaged needle prints the same vertical streak until someone replaces it. Oil stains are obvious, barre emerges after dyeing, contamination ruins the export roll.

Then there's Lycra miss. Silent on the machine. Invisible on the grey roll. Undetectable by most inspectors, even experienced ones. Unambiguous only after the fabric is stretched, worn, or dyed — at which point it has already moved three stages downstream and cost an order of magnitude more to fix.

Elastane defects are the hardest class of defects in circular knitting to detect, and they are also among the most expensive when they escape. This post is about why — a technical look at what makes core-spun yarn defeat ordinary inspection, and what it takes to actually catch it at the machine.

What core-spun yarn actually looks like

To understand why elastane is hard, you have to understand how it shows up in the fabric in the first place.

Elastane — also called spandex or Lycra, depending on region and brand — is a highly elastic synthetic fiber usually introduced into fabric at a small percentage (typically 2–10% of total weight). It rarely appears as a pure yarn in knitted fabrics. Instead, it's delivered as core-spun yarn, in which a fine elastane filament is wrapped by staple fibers of cotton, polyester, or another primary fiber.

The elastane sits inside. The outer wrap is what the camera sees.

This structural fact is the root of the inspection problem. The elastane core is mechanically critical to fabric performance — it provides the stretch, the recovery, the shape retention that makes the fabric worth producing — but it contributes almost nothing to the visual surface of the grey fabric. When the elastane is absent, the fabric surface often looks similar to fabric where elastane is present, because the cotton wrap is doing the visible work in both cases.

This is why a human inspector can hold a length of Lycra-miss fabric under a lamp and see almost nothing. And it's why a standard RGB vision system trained on color and texture alone will miss a meaningful fraction of elastane defects.

The three classes of elastane defect

Elastane defects are usually discussed as a single bucket — "Lycra miss" — but in practice they fall into three structurally different classes, each with its own cause and its own detection signature.

1. Complete miss

The elastane feeder fails entirely. No elastane is delivered into one or more courses. The feeder may be snapped, jammed, out of package, or disconnected. The affected courses have zero elastane, and once the fabric is stretched or dyed, the missing courses show up as a sharp horizontal band of reduced recovery and altered shade.

Complete miss is the cleanest detection case. The structural difference between a course with elastane and a course without is real and consistent, and the fabric geometry (loop shape, loop density) shifts enough that a well-trained vision system can pick it up with high reliability. The challenge isn't detecting it in principle; it's detecting it fast enough and reliably enough to halt the machine before dozens of meters are produced.

2. Intermittent drop

The elastane feeder is unstable but not entirely broken. Elastane appears and disappears across consecutive courses as the feeder loses and regains tension. This happens when an elastane package is low, when the feeder's tension device is partially clogged, or when the ambient humidity shifts enough to change how the elastane draws off the cone.

Intermittent drop is harder. The signal is partial — some courses have elastane, some don't — and the pattern is irregular. On a grey roll, you may see nothing at all. After dyeing, you see faint streaking that looks like barre to a buyer but is actually periodic elastane absence. Intermittent drop is the most common cause of "mystery barre" complaints from buyers who return rolls that the mill's own inspectors couldn't replicate on the grading table.

3. Slippage and tension drift

The elastane is present, but at the wrong tension. Either it's too loose (and the fabric has inconsistent stretch) or too tight (and the fabric shrinks unevenly after finishing). Tension drift is usually gradual, accumulating over hours as a feeder slowly deviates from its set point.

This is the hardest class. There is no absence to detect — every course has elastane — just a subtle change in how the yarn is being delivered. The fabric surface reveals this change only in geometry: slightly different loop heights, slightly different course density, slightly different widthwise dimension. These differences are small enough that they vanish under normal front-lighting and normal resolution.

Elastane defects are not one problem. They are three problems, each with its own physics, each defeating a different kind of inspection method. Solving one does not automatically solve the others.

Why standard vision misses so many elastane defects

If you hand a general-purpose computer vision system a set of fabric images and ask it to find defects, it will do reasonably well on holes, needle lines, and contamination — defects that produce strong contrast signals in ordinary lighting. It will do poorly on elastane defects. Three reasons.

The wrap hides the core

The visible surface of core-spun yarn is dominated by the outer staple fiber wrap. The elastane underneath is a weak visual contributor on grey fabric. A model trained to recognize visible defect signatures has little to work with, because the visible difference between elastane-present and elastane-absent fabric is often subtle.

Flat lighting flattens the signal

The structural differences caused by missing elastane — changes in loop geometry, course density, surface topology — produce small but real shadow signatures. These are exactly the signatures that flat front-lighting eliminates. A camera shooting straight-on with diffuse white LEDs sees the fabric surface as a texture; an oblique light source turns the same surface into a landscape of ridges and valleys that reveal elastane behavior.

Most off-the-shelf fabric inspection systems use flat lighting because it works well for holes, stains, and contamination. For elastane, it's the wrong tool.

Training data is thin

Generic models are trained on publicly available fabric image datasets, most of which are heavy on dramatic visible defects and thin on elastane-specific examples. A model that has never seen ten thousand examples of Lycra miss, Lycra drop, and tension drift — across different fiber blends, different yarn counts, different machine gauges, different lighting conditions — will not generalize to the quiet signal that elastane defects actually produce.

Detecting elastane defects reliably is, in practice, a specialization. It requires an inspection system designed specifically around the physics of core-spun yarn in circular knitting, not a generalized vision platform that happens to be deployed on a knitting machine.

What actually works

After several years of deployment across mills running everything from 2% Lycra interlock to 8% Lycra single jersey, the detection approach that consistently holds up has three components.

Structured and oblique illumination

The single biggest change from generic inspection is the lighting. Oblique illumination — light arriving at the fabric surface at a steep angle rather than head-on — creates shadows that reveal the geometric differences caused by missing or mis-tensioned elastane. Structured lighting (patterns projected onto the fabric) goes further, exposing dimensional variation that flat light simply hides.

This is hardware, not software. You cannot post-process your way to the signal if the signal isn't in the captured image. Getting the illumination right is the first and most important step in elastane detection.

Models trained on elastane-specific data

Once the image has the signal in it, the model has to know what to look for. That means training on large volumes of labeled examples covering all three elastane defect classes — complete miss, intermittent drop, tension drift — across the full range of fiber blends, yarn counts, and machine configurations in production use.

This is where fleet scale matters. A system deployed on a handful of machines sees a handful of elastane configurations. A system deployed across hundreds of mills sees thousands of configurations, and the model trained on that data generalizes to new mills in a way a small-sample model cannot.

Course-level analysis, not frame-level

Elastane defects are inherently a course-level phenomenon — they happen in specific knitted courses, propagating horizontally across the fabric. A detection pipeline that reasons about fabric frame-by-frame without aligning to course structure will miss patterns that only become visible when you compare course to course.

Aligning inspection to the actual knitting geometry — knowing which pixels belong to which course, which course belongs to which feeder — turns elastane detection from "find a visual anomaly" into "watch the tension signature of each feeder over time." It's a different problem, and a more tractable one.

Why this matters for export mills

For mills producing 100% cotton or 100% polyester, elastane detection is irrelevant. For everyone else — which is most of the circular knitting industry — elastane defects are disproportionately expensive for three reasons.

Lycra-containing fabric is higher value. Mills run elastane because buyers pay more for stretch. A Lycra miss on a premium roll is not the same economic event as a structural defect on a basic roll.

Lycra defects escape most often. Because the defect is quiet on the grey roll, it is far more likely to reach dyeing, finishing, cutting, and shipping undetected. Each stage downstream multiplies the cost of the eventual fix.

Lycra defects damage buyer relationships. A buyer who receives Lycra-miss rolls — which don't feel or wear correctly — rarely asks for a discount. They ask for a new supplier. The economic event is not a rebate; it is a lost account.

For export-oriented mills in Tiruppur, Dhaka, Colombo, and the rest of the global knitting cluster, elastane inspection is quietly one of the most consequential investments in the quality stack. It is the defect class most likely to be missed, the defect class most likely to cascade, and the defect class most likely to end a customer relationship.

How Knit-I handles the elastane problem

Knit-I is designed specifically for the circular knitting environment, and elastane detection is one of the capabilities we've iterated hardest on. Three choices matter.

First, the system uses purpose-designed illumination that combines structured and oblique light to expose the geometric signatures of elastane behavior. This is not incidental — it's the part of the stack that makes every downstream detection step possible.

Second, the models are trained on elastane-specific data drawn from mills across seven countries and a wide range of fiber blends and gauges. A new mill coming online does not need to generate its own training set; the generalized model transfers well because the underlying elastane physics are the same everywhere.

Third, the system performs course-aligned analysis and tracks each elastane feeder's behavior as a time series. This is how intermittent drops and tension drifts get caught — not as isolated frames, but as anomalies in a per-feeder signature that evolves over minutes and hours. When an elastane defect is detected, the system communicates directly with the machine's control system to halt production, limiting the scrap to the seconds between detection and stop.

Frequently asked questions

What is elastane defect detection?

Elastane defect detection is the process of identifying errors in the feeding, placement, or tension of elastane (spandex/Lycra) yarn during knitting. It includes Lycra miss, Lycra drop, and slippage or tension-driven variations that affect fabric stretch and recovery.

Why is elastane so hard to detect visually?

In core-spun yarn, the elastane core is wrapped by cotton or polyester staple fibers, so the elastane itself is rarely visible on the fabric surface. Lycra miss often produces no strong color or texture signal on grey fabric; it becomes obvious only after stretching, dyeing, or in finished-garment wear. Standard RGB vision and manual visual inspection both miss many elastane defects for this reason.

What are the main types of elastane defects in circular knitting?

Three main classes: complete Lycra miss (the elastane feeder fails and one or more courses have no elastane), intermittent Lycra drop (the feed is unstable and elastane appears and disappears across courses), and Lycra slippage or tension drift (elastane is present but at the wrong tension, producing visible banding or uneven stretch after dyeing).

Can AI detect Lycra miss in real time?

Yes. Modern AI inspection systems like Knit-I use high-resolution cameras with controlled illumination and deep learning models trained specifically on elastane defects. The system can detect a Lycra miss within milliseconds of formation and halt the knitting machine before more defective fabric is produced.

What illumination works best for elastane detection?

Elastane defects respond much better to structured and oblique lighting than to flat front-lighting. Oblique illumination creates subtle shadow signatures from the slight structural differences caused by missing or mis-tensioned elastane, which standard flat lighting flattens out. This is why custom-designed illumination is a key part of reliable elastane inspection systems.

For a broader view of circular knitting defects beyond elastane, see our guide to circular knitting defect detection. For an industry-wide look at what our deployment data reveals, read what 4,500 cameras across 7 countries taught us about fabric defects.