Aramid (Kevlar) spinning control with AI — the micro-break is invisible, the ballistic test is not.

In aramid for bulletproof vests, a tension spike or a broken filament during spinning goes unnoticed — it goes into the bobbin, moves on to weaving and shows up in the V50 ballistic test, when the panel is already made. iLEAN Edge watches tension per end, speed and filament in line, holds the suspect bobbin and alerts the spinner. The person decides.

← See all technical textile solutions

Aramid spinning room with godets, a tension sensor and an Edge camera over the end — micro-break control with AI
The problem

A fiber that does not forgive — and a test that arrives too late.

Aramid (para-aramid of the Kevlar/Twaron type, meta-aramid of the Nomex type) is one of the few fibers where process control directly decides whether the final product passes a life-or-death test.

  1. High tenacity, low elongation — the yarn does not "give". Any tension spike during drawing turns into an internal micro-break; the end keeps running, the bobbin looks healthy, but the ballistic panel is not.
  2. The defect is invisible to the eye — subtle loose fuzz, a disordered halo, diameter irregularity. The operator, with six ends to watch and the light available, cannot see it.
  3. The final test is destructive and late — the panel is made, the vest is assembled, and the V50 is run at the end. If it fails, the whole batch is reworked or scrapped, and the traceability needed to go back to the guilty roll has almost always been lost along the way.

The quality manager knows this, but cannot put a technician in front of every end on every shift. The classic system (bench tensiometers, sampling, an expert eye) works 99% of the time — and that 1% is the vest that fails the final test, and the claim from the public agency that had bought it.

How it fits the IRIS system

iLEAN does not change your spinning machine — it puts eyes and memory where they do not reach today.

The problem with aramid spinning is not a lack of information: it is information living on islands (the set point in the machine control, the tensiometer reading, the veteran's eye passing by once a shift) that at the critical moment (the tension spike at 3:47 on the night shift) does not reach anyone in time. iLEAN acts as the putty that fills those gaps, without asking you to replace your machine or your PLC.

The terminal watches the end and the tension at the same time. The agent cross-references it with the speed and the spindle history. The suspect bobbin is held at second zero, before it leaves the line. The person signs.

The three iLEAN pieces applied to aramid spinning control:

  • Edge — a terminal with machine vision (CNN) over the end at the drawing exit, connected to a tension sensor per end. It marks the suspect stretch, fires an actuator (stack light, soft stop, paint mark on the bobbin) and records the event. It works with no network. If the plant loses WiFi, Edge keeps watching.
  • Connect — captures the machine set points (godet, drawing, temperature), the spindle history, the polymer batches. And it also captures what arrives from outside (the military or law enforcement customer's specification, the test from the purchasing public agency's laboratory) at second zero, without anyone having to forward anything.
  • Agents — cross-reference the visual signature of the end, the tension, the godet speed and the spindle history. If the combination exceeds the threshold, they hold the bobbin and alert the spinner and the quality manager through whichever channel they use. The person decides; the machine does not stop on its own if the operator judges it to be a controlled transient anomaly.

See the full IRIS architecture →

Before and after

Manual spinning vs. spinning assisted by iLEAN Edge

AspectTensiometer + sampling + the eyeWith iLEAN Edge + Connect + Agents
Tension per endSpot measurement, on a sampleContinuous, in line, on every end
Vision of the endThe veteran's eye, one pass per shiftA CNN on the end, the whole shift, never blinking
Transient spike at nightOften not detectedFlagged and traced to the affected bobbin
Holding a suspect bobbinMoves on to weaving, detected at the endHeld before it leaves the line
Operation without a networkn/aEdge keeps running on cabinet power
Traceability for a failed V50Manual reconstruction, sometimes impossibleAutomatic per-bobbin dossier, with video of the event
Impact estimate

Impact estimate for your plant — to be validated with your numbers.

The block below is an estimate to be validated with the specific data of your spinning room. We put it forward so the committee has an order of magnitude; we refine it during the diagnostic.

  • Technical spinning room with 2-6 para-aramid or meta-aramid lines, production largely dedicated to ballistic applications and protective aramid.
  • Edge pilot on one line: camera over the end + tension sensor + integration with the existing machine control and with the ERP. First value expected within a few weeks (marking of suspect bobbins and cross-referencing with tension).
  • Indicative payback between 4 and 9 months, depending on the number of bobbins currently scrapped "on suspicion" that the system could have released with data behind it, and on the weight of the cost of failed V50 panel tests.
  • Hard levers: a reduction ≥ 30% in scrap on suspicion, targeted holding of the problem bobbin before weaving, and event traceability so the ballistic lab can go back to the guilty spindle when something fails.

And the quality manager's reasonable doubt

“What if the AI flags as a defect something that is normal spindle variation?” — hallucination is a problem of free generation, not of anchored tasks. Identifying a visual pattern already trained in your own spinning room and cross-referencing it with a tension sensor is an anchored task. In that family of tasks the best models brought error below 1.5% [1]. And even so, what is critical is never decided alone: iLEAN flags and holds, and the spinner or the quality manager signs. The three safety rings exist precisely for this.

[1] OpenAI paper “Why Language Models Hallucinate”, 2025 — on the reliability of AI in anchored tasks.

Frequently asked

What people ask about aramid spinning control

Why is tension critical in aramid (Kevlar) spinning?

Aramid (Kevlar and similar fibers) is a high-modulus, low-elongation fiber: a tension spike during spinning or drawing is not “absorbed” the way it would be in an elastic fiber — it turns into a micro-break of filaments inside the end. That micro-break is not visible to the naked eye, but it locally lowers the tenacity of the yarn and, when the yarn goes into ballistic fabric for a bulletproof vest, the micro-broken zone becomes a weak point in the panel. The V50 ballistic test finds it late and expensively. That is why tension per end, godet speed and yarn guide geometry are parameters that allow no deviation.

How does iLEAN Edge detect a filament micro-break?

Edge installs a camera with a CNN over the end at the drawing exit and a tension sensor per end. The neural network learns the visual pattern of healthy yarn (core + filament halo) and raises the alert when the halo becomes disordered — loose fuzz, broken filaments, diameter irregularity. In parallel, it cross-references that reading with instantaneous tension and godet speed. If all three data points point in the same direction, Edge marks the stretch, holds the bobbin and alerts the spinner through the earpiece. The person decides whether to cut the end, change the yarn guide or carry on.

Does it work with wet spinning or only dry spinning?

It works with both processes. Wet spinning of para-aramid (standard for Kevlar and Twaron) has its own visual and tension signature; dry spinning of meta-aramid (Nomex) has another. iLEAN trains the model with video from your specific line — the aim is not to universalize it from the factory, it is calibrated in the plant. That is the IRIS logic: the AI feeds on the reality of your plant, not on a generic dataset.

What happens if the plant loses its network?

The critical Edge cycle (vision + tension sensor + alert to the spinner + holding the suspect bobbin) keeps working as long as the terminal has power. The connection to the central platform is used to train better models, assemble dossiers and cross-reference with the ERP/MES — all of that can wait. What is critical cannot wait: what is critical, as the book says, cannot depend on there being WiFi.

How much does a pilot cost in an aramid spinning room?

The order of magnitude of an Edge pilot in a technical spinning room is that of any industrial Edge pilot: terminals with camera + tension sensors + integration with the existing machine control, plus an annual license. A reasonable payback is measured in a few months — the hard lever is a single bobbin held back that at any other time would have reached the weaving stage and then the vest, plus the cost of a failed V50 ballistic test on a finished panel. We ask for your spinning room's data and send you the estimated ROI in 48h.

Let's talk

Tell us your case and in 48h we'll send you the estimated ROI of this AI project for your spinning room.

We work on your plant's real data, not ours. Diagnostic with no commitment.

Request estimated ROI in 48h See technical textiles