Nonwoven Defect Attribution: A Technical Deep Dive into Roller Geometry, Thermal Lag and Process Noise

Date Issued:2026-11-11

Most troubleshooting advice on nonwoven lines stops at "check the roll". This article goes one level deeper: what the roll actually does to the web, how that signature can be measured with confidence, and where the measurement stops being reliable. The aim is to let an engineer defend a conclusion with numbers rather than with seniority.

The imprint mechanism, stated precisely

Consider a web moving at speed v through a nip formed by a roll of diameter D rotating at n revolutions per minute. Surface speed at the roll face is πDn*/60. When the web and roll are synchronous, the roll surface advances exactly one circumference per revolution relative to the web. Any surface or geometric feature at a fixed angular position therefore contacts the web at machine-direction intervals of:

L = πD (in millimetres, when D is in millimetres)

A 300 mm roll imprints every 942 mm. A 450 mm roll imprints every 1,414 mm. A 600 mm roll imprints every 1,885 mm. This is why the first diagnostic act should always be a roll list sorted by circumference: the list converts an unbounded search into a lookup table.

Two refinements matter in practice. First, slip. If the web is not perfectly synchronous with the roll — because of tension differences across the width or a driven roll running slightly fast — the observed spacing is scaled by the slip ratio. A 2% slip on a 942 mm circumference shifts the interval to 923 mm or 961 mm, which can cause a genuine match to be dismissed. Measure across twenty repeats rather than seven to reduce this error.

Second, non-circular error. An oval shell contacts the web twice per revolution at the two high points, producing spacing of πD/2. A three-lobe error produces πD/3. Engraved rolls with two-start patterns behave the same way. When the observed spacing is a simple fraction of a circumference, the diagnosis shifts from "damaged surface" to "geometry error", and the repair path changes from polishing to regrinding or replacement.

Metrology: what to measure, and with what

Runout. Total indicator reading is the difference between maximum and minimum dial reading as the roll turns one revolution. The critical detail is temperature. A 400 mm diameter steel shell with an asymmetric internal channel layout can exhibit 0.012 mm TIR at 20 °C and 0.04 mm at 200 °C, because the hot side grows more than the cold side. Acceptance and rejection decisions must therefore be made at operating temperature, with the roll at steady state — which for a large shell means at least 45 minutes after reaching setpoint.

Measure at four axial stations: drive end, quarter, centre, far end. A taper shows as a monotonic trend across stations; a local defect shows as a peak at one station and one angular position. Record the angular position with a chalk mark so the same spot can be found after regrinding.

Crown. Crown is a deliberate diameter increase at the centre of the face, expressed in micrometres or as a radius difference. It exists to compensate for deflection under nip load, so the crown value is only meaningful in combination with the design load. Verify it with impression film or lead foil: insert the film, apply production nip load, and read the contact width across the face. Uniform contact width indicates correct crown and parallelism for that load. Narrow contact at the edges indicates insufficient crown or excessive deflection; narrow contact at the centre indicates over-crown or excessive load.

A useful sanity check: a 0.03 mm crown error over a 2,500 mm face is enough to drop local nip pressure below the bonding threshold for many spunbond products, even though 0.03 mm sounds negligible.

Surface topography. Roughness average (Ra) alone is insufficient for release behaviour. Two surfaces can share Ra and behave differently because the peak density and the presence of plateaus differ. For bonding rolls, specify Ra plus a peak count or a bearing ratio curve if release problems are recurrent.

Thermal uniformity. An infrared camera at low rotational speed (2–5 m/min surface speed) gives a full-face map in under a minute. The key metrics are maximum cross-face deviation, axial gradient, and the size of any cold or hot sector. Thermocouples in the control loop cannot substitute: they read one point, often inside the shell wall rather than at the surface.

Thermal lag and why it destroys causal reasoning

Heated shells are first-order thermal systems with time constants that scale with mass and heat transfer path. For a 400 mm diameter, 25 mm wall shell, a step change in heating medium temperature typically produces 63% of the final surface change in 12–20 minutes and 95% in 35–60 minutes.

This matters because human operators adjust parameters on a two-minute cycle. When bonding looks weak, the response is usually: raise temperature, raise nip pressure, slow the line slightly, then inspect twenty minutes later. By then, all three changes are active and the thermal response is still in progress. Any improvement is credited to whichever change the observer believed in beforehand.

The remedy is procedural, not technological: change one variable, wait for the thermal time constant, then evaluate. Write the sequence down with timestamps. An engineer who does this consistently will out-diagnose a better-instrumented plant that does not.

Induction-heated shells behave somewhat differently because energy is generated in the shell wall rather than conducted through it, which shortens the effective response at the surface but introduces axial zoning effects. The relationship between coil zoning, gap, and axial uniformity is covered in the technical article on Electromagnetic Induction Heating Rolls: Principle and Performance, and it explains why a zone imbalance can mimic a blocked internal channel.

Time domain, frequency domain, and what each reveals

Time-domain analysis is what most plants already do: look at the fabric, measure spacing, compare with the roll list. It is fast, requires no instrumentation, and resolves perhaps 70% of cases.

Frequency-domain analysis becomes valuable when the defect is subtle, when multiple faults overlap, or when the defect is dynamic. Two accessible techniques:

Vibration spectrum on the roll bearing housing. A bearing defect produces characteristic frequencies (ball pass frequency, cage frequency) that are not harmonically related to rotation. If the fabric defect frequency matches a bearing frequency rather than 1× rotation, the roll surface is innocent and the bearing is not. A simple accelerometer and a phone-based spectrum app is enough to make this call.

Vision system periodicity histogram. Modern inspection systems record each defect with machine-direction and cross-direction coordinates. Sorting the machine-direction intervals into a histogram produces a clear peak at the true period and a flat background from random defects. This is the single most useful analytical output of inline inspection, and it is nearly free once the system exists.

Note the relationship: a period L in space corresponds to a frequency f = v/L in time. At 300 m/min and L = 1,414 mm, f = 3.5 Hz. If the defect vanishes when line speed drops to 150 m/min but the spacing stays constant in metres, the cause is geometric. If the spacing in metres changes with speed while the frequency in hertz stays constant, the cause is time-based — a pump cycle, a control loop, or a reciprocating element.

That single distinction — metres versus hertz — resolves a surprising number of arguments.

Designed experiments on a running line

Formal design of experiments is usually dismissed as impractical in production, but a stripped-down version works. The goal is not a full factorial; it is to establish whether the defect amplitude responds to process variables at all.

Practical protocol: hold the recipe, choose one process variable with a wide safe range (nip load is usually best, because it can be moved ±10% without quality risk), and run three levels for ten minutes each with a five-minute stabilisation before sampling. Measure defect amplitude or defect count per 100 m at each level. If amplitude changes by more than the measurement noise, the cause is at least partly process-side. If amplitude is flat while the defect persists, the cause is geometric.

The same logic applies to temperature, but the thermal time constant forces longer dwell times: twenty minutes per level minimum on a large shell. This is why temperature experiments are frequently abandoned halfway and produce misleading conclusions.

Industry Watch: Recent Developments

Three developments are relevant to how attribution is practised.

Higher line speeds are compressing the tolerance window. ITMA Asia + CITME 2026 runs in Shanghai from 20 to 24 November with 180,000 m² of exhibits and more than 1,900 machinery exhibitors, including a dedicated nonwoven zone and new digitalisation sections. The equipment being launched there is aimed at higher output, which shortens nip dwell time and makes geometric error less forgiving. Source

Inline inspection is becoming standard rather than optional. IndexBox's spunlace bonded nonwoven forecast to 2035 projects about 6.8% baseline CAGR with electronic-grade grades at 7–9%, and identifies "higher-speed automated lines with integrated online quality inspection" as a structural shift in the competitive landscape. For engineers, that means periodicity analysis will increasingly be a data query rather than a manual measurement. Source

Component quality is being reframed as an uptime variable. A recent technical note from a hydroentanglement screen supplier argues that when a minor component is bought as a commodity, operators compensate by raising pump pressure, which accelerates wear and creates irregular density lines. The identical compensation loop occurs with tired calender rolls: temperature and load are raised to force bonding, and the defect changes shape instead of disappearing. Source

Process-specific defect taxonomies

Spunbond. Filament-level defects dominate. Coarse filaments from air temperature drift bond poorly at setpoint; shot and drips come from polymer or die issues; width-wise basis weight variation comes from quench air asymmetry. Roll contributions are calender runout, thermal non-uniformity, and crown error. Discriminator: check whether the defect lane moves when quench air is rebalanced.

Meltblown. Collector drum surface condition and die-to-drum distance dominate. Because the web is fragile, drum marks transfer readily, and a scratched drum shows as periodic thin spots at drum circumference.

Spunlace. Perforated drum and Cylinder Roll condition dominate, along with jet strip condition. A partially blocked perforation pattern creates repeating marks that look like bonding faults. Pressure stability of ±2 bar is the process-side benchmark; the drum's hole geometry is the hardware-side benchmark. The Spunlace Drum specification details the perforation tolerances that matter.

Thermal bond (carded). Carding profile is the dominant process variable, Embossing Calender Roll pattern wear and crown are the dominant hardware variables. Pattern depth loss is progressive and produces a gradual softening of hand-feel that is often blamed on fibre lot.

Laminating and coating. Nip parallelism and metering geometry dominate. Embossed Roll pattern fidelity and coating roll concentricity are the usual hardware suspects.

Reconditioning thresholds and their justification

Thresholds should be derived from the process window, not from habit. A defensible approach: determine the maximum TIR that still produces acceptable bonding at the lowest nip load in the product range, then set the maintenance limit at 70% of that value to leave margin for thermal growth. For many spunbond bonding rolls this lands between 0.015 and 0.02 mm.

For crown, the limit derives from contact width uniformity: if impression film shows contact width varying by more than 15% across the face at production load, crown or parallelism is out of specification for that product mix.

For coating, the trigger is usually defect count rather than thickness: when pits or release marks appear more than once per 1,000 m, replating is due. Thickness measurement alone misses localised failure.

Conclusion

Attribution is a measurement problem before it is a technical one. Establish the roll list with circumferences, measure runout hot, map thermal uniformity, read the periodicity histogram, and change one variable at a time with the thermal time constant respected. These steps do not require new capital, and they convert an argument into a record.

For plants specifying new hardware, the difference between a diagnosable roll and an anonymous one is documentation: dimensional records, hot runout, crown method, thermal maps. The specification depth described under Spunbond Nonwoven Rolls and Spunlace Nonwoven Rolls is a reasonable benchmark, and the broader process context sits under Nonwoven Fabric Industry. Method notes and field observations are collected in the Industry Blog.

Related Articles

Related Products

· Induction Calender Roll

· Induction Heat Calender Roll

· Chrome Drafting Roll

· Ceramic Coating Roll

· Spunlace Drum

Related Technical Guides

· Electromagnetic Induction Heating Rolls: Principle and Performance

· PRS Cooling Rolls for BOPP, BOPET and BOPA Lines

· MDO Preheating Rolls for BOPP and BOPET Lines

Related Case Studies

· Chrome-Plated Chill Rolls: South American BOPP Project

· Corrosion-Resistant Chill Rolls: North American BOPP Project

Contact Us

For dimensional records, thermal uniformity data, or engineering review of a recurring defect, our team can work from your line data and roll drawings.

Contact: Jenny Liu / Email: sale05@huataogroup.com / Phone / WhatsApp: +86 17743789775 / Website: https://www.huataomulti-indusrolls.com/

We warmly welcome customers from around the world to contact us and establish mutually beneficial partnerships.

* E-mail :

can not be empty

* Question :

can not be empty

Related News