Cameras in Paper Mills: Web Breaks, Reel Handling and Wet-End Losses

In a paper mill the camera's job is replay, not surveillance. A web break destroys its own evidence within seconds, so the only usable clue is video of the moments before it parted. On a 100 TPD Indian mill, each lost machine-hour costs roughly ₹33,000–47,000 in contribution — enough that classifying breaks by cause pays for the system.

You run a 60–200 TPD mill — kraft in Muzaffarnagar, duplex board in Vapi or Morbi, writing-printing in Punjab or coastal Andhra. Two hundred to a thousand people across four shifts. Your paper machine is the whole business: when it runs you make money, when the web parts you make broke. This page is about what a camera can and cannot do about that.

Key points

• India has a lot of small mills, and they run hard. Roughly 900 paper mills, of which about 553 operate at some 25.6 million tonnes of operating capacity, with Gujarat (~150 mills), Maharashtra (~100) and Uttar Pradesh (~70) leading. Mid-size mills like yours are the modal Indian mill, not the exception.
• Price a machine-hour before you price a camera. Industry turnover of about ₹80,000 crore (2024) across roughly 25 million tonnes implies a blended realisation near ₹32,000/tonne (same source). At 100 TPD that is ~₹1.33 lakh of gross revenue per running hour, and at a 25–35% contribution margin, ₹33,000–47,000 per hour of downtime. "Lakhs per hour" describes a large machine, not a 100 TPD one.
• The evidence is gone before anyone arrives. The web runs from a headbox at under 0.4% consistency to about 25% at the couch, above 40% leaving the press section and ~6% moisture after calendering. A break at the wet end is a wet edge crack that has vanished into the pulper by the time the shift in-charge looks up.
• Dedicated web-break camera systems are a mature category. Procemex claims 1,100+ systems and 22,000+ smart cameras delivered, including a customer success story headlined "80% Improvement in Wet-End Web Break Analysis". Treat vendor numbers as vendor numbers — but the technology is not experimental.
• Cameras will not survive the dryer hood. IP66 means protection against powerful water jets — 100 litres/minute at 100 kPa from 3 m — which is a washdown rating, not a saturated-steam-at-temperature rating. General-purpose cameras are commonly specified to around 60°C ambient. The hood needs purged or cooled housings, or you keep out.
• Bale yard and gate are separate money. Fibre is the biggest cost line in a recycled mill; weighbridge and bale-yard video is a different project from web-break replay, and it is usually the one that pays first. See /efficiency/gate-weighbridge-material-pilferage-cameras-india/.
• Stop quoting the Factories Act, 1948. It was repealed when the Occupational Safety, Health and Working Conditions Code, 2020 was brought into force on 21 November 2025, with the Central Rules notified on 8 May 2026.

Why a web break is a video problem

Every other production loss leaves a trace you can inspect afterwards. A web break does not. The sheet parts, the machine dumps to broke, the crew re-threads, and the physical evidence is a slurry in the pulper within a minute or two.

So the root cause becomes folklore. "Wet edge." "Doctor blade." "That felt is due." "Night shift ran it too fast." Your machine superintendent has an opinion, your production manager has a different one, and neither can be tested.

A camera trained on the open draws — couch to press, press to first dryer, dryer to size press, size press to reel — with a rolling buffer changes the argument into an observation. When the break alarm fires, you keep the preceding thirty seconds. Over a month you have thirty or forty clips sorted by position. Patterns appear that no logbook would ever show: eight of eleven breaks at the same draw, always within twenty minutes of a grade change.

That is the whole value proposition. Not prevention by AI — classification by replay.

What to point cameras at, in order

1. The open draws. Highest value, hardest environment. This is where dedicated web-monitoring kit with stainless housings belongs, not a ceiling dome.
2. The pope reel and reel change. Reel-change sequence timing, wrinkles, edge condition at the turn-up. Also a common place for a crew to lose a few minutes they never report.
3. The winder and rewinder. Set changes, splices, core loading, roll rejects. Winder throughput is frequently the second bottleneck after the machine itself.
4. Finishing and roll despatch. Roll number, wrapping, weight ticket, loading. This is where customer claims are won or lost.
5. The pulper and bale conveyor. Feed rate, contraries, plastics and metal going in with the bales.
6. The bale yard and weighbridge. Fibre in, moisture in, and what leaves after dark.

Placement logic for a wet, hot, tall building is genuinely different from a machine shop — see /cameras/where-to-place-cameras-shop-floor/.

The rupee model on a 100 TPD mill

Build it yourself with your own numbers. This is the template, with every assumption exposed:

Four honest caveats. The ₹32,000/tonne is a blended national average derived from public turnover and capacity figures, and turnover includes value that never reaches a small mill's realisation — kraft liner sells well below it and coated writing-printing well above, so substitute your own number. The 4% break loss is an illustration, not an industry statistic; pull your true figure from your DCS downtime log. The 15% improvement is what a disciplined root-cause programme with evidence can plausibly do, not a guarantee — the cameras find the cause, your maintenance team fixes it, and if nobody acts on the clips the saving is zero. And the capex line covers general floor cameras; hardened wet-end web-monitoring systems cost considerably more than a standard PoE install. For the general costing frame see /cost/factory-camera-ai-system-cost-india/ and /cost/does-camera-ai-pay-for-itself-india/.

Even on those assumptions, recovering 48 machine-hours a year would be worth roughly three times a full year of the general system. Check the arithmetic against your own log before you believe it.

Steam, heat and fibre dust: where cameras physically fail

This is the section vendors skip, and it is the one that decides whether your project works.

The dryer section is not a camera environment. Saturated steam, condensate and high ambient temperature near the hood will kill a general-purpose camera regardless of its IP rating. An IP66 body survives a 100 kPa water jet; it does not survive continuous hot wet vapour finding the gasket. If you must see inside the hood, you need a purged or air-cooled housing designed for it — the kind of stainless enclosure serious web-monitoring suppliers ship as standard.

Fibre dust and mist coat lenses fast. Anywhere near the pulper, the broke chute or the winder, you are on a cleaning schedule. Budget it, or your saving quietly becomes a wall of grey rectangles by month three.

Wash-down means the wet end is a corrosion problem. Aluminium housings and mild-steel brackets will not last. Stainless mounts, sealed glands, drip loops.

Chemical and recovery areas may be classified. In mills with a recovery boiler, chlorine dioxide plant or LPG/solvent storage, electrical equipment used in a hazardous area under the Petroleum Rules must be approved for that zone — PESO permits flameproof, intrinsically safe and encapsulated equipment in Zone 1, and non-sparking or increased-safety types in Zone 2 (PESO FAQs). An ordinary PoE camera in such an area is an insurance and safety exposure, not a technicality.

What cameras will NOT do in a paper mill

They will not predict a break. There is no credible ceiling-camera model that says "this web will part in forty seconds." Wet-end break physics depends on consistency, retention, vacuum, felt condition and draw — variables your DCS sees and a camera does not. Anyone selling break prediction off general video is overselling.

They will not replace your QCS. Basis weight, moisture profile, ash, caliper — that is your scanner, and the camera has no opinion about any of it.

They will not measure roll hardness or telescoping reliably. A visibly telescoped roll is obvious on video; a roll with a bad hardness profile that will fail at the customer's press is not.

They will not fix a mill that does not log downtime honestly. If "web break" is the catch-all reason code for every stoppage including a boiler trip and a tea break, video adds pictures to a lie. Fix the reason codes first.

They will not deputise for guarding. Nips, winders and rewinders are a well-known entanglement and crush risk. A camera records the injury; a guard and an interlock prevent it. See /safety/machine-guarding-danger-zone-detection/.

They will not come free of worker-data duties. Footage of identifiable workers is personal data under the DPDP framework, whose substantive obligations — notice, security safeguards, breach reporting — take effect on 14 May 2027. Put up the notice before the cameras: /compliance/worker-cctv-privacy-notice-dpdp-india/.

The bale yard: the project that usually pays first

In a recycled-fibre mill, fibre is the largest single cost line, and it arrives as bales on open trucks into an open yard, often at night, weighed once. Every classic loss shows up here: short-weight loads, moisture loaded deliberately into the bales, contraries baled in as filler, and bales that leave again.

Do the arithmetic on your own purchase register, not on a vendor's slide. If your landed fibre rate is, say, ₹20,000/tonne (use your real number), a single 400 kg bale is ₹8,000, and two trucks a month running 2% short on a 30-tonne load is about ₹24,000 a month — scale that by your actual truck count and the number stops being small.

The camera work here is unglamorous and effective: number plate at the gate, bed of the truck before and after tipping, weighbridge display in the same frame as the vehicle, and yard coverage that actually holds up at night. The full method is in /efficiency/gate-weighbridge-material-pilferage-cameras-india/.

Compliance and procurement realities in 2026

Two things have changed and both hit paper mills.

First, the legal frame. The Factories Act, 1948 is gone; the OSH Code, 2020 governs, in force since 21 November 2025, with Central Rules notified on 8 May 2026. Nothing in it mandates CCTV as such, but hazardous-process duties, accident reporting and the conditions attached to employing women on night shifts are all easier to evidence with retained footage. Start at /compliance/osh-code-2020-factory-cctv-india/.

Second, procurement. Under MeitY's compulsory registration order for electronics, CCTV cameras sold in India must meet the notified Essential Requirements, and the relaxation for non-compliant stock ends on 1 April 2026. That is a rule for what may be sold, not an order to rip out working cameras — but if you are buying now, buy compliant now. Retrofitting a mill in a wash-down environment twice is painful. See /compliance/bis-stqc-2026-factory-camera-compliance/.

FAQ

How much does a web break actually cost my mill? Take your tonnes per hour, multiply by your realisation per tonne, then by your contribution margin. On a 100 TPD mill at ~₹32,000/tonne and 30% contribution, that is roughly ₹40,000 per hour of lost machine time — plus the steam, power and chemicals already spent on paper now going back through the pulper as broke.

Can AI cameras predict web breaks before they happen? Not from general video. Break physics lives in consistency, vacuum, draws and felt condition — DCS territory. What cameras do reliably is preserve the seconds before the break so you can classify causes and act on the pattern. Any vendor promising prediction from ceiling cameras is selling ahead of the technology.

Will cameras survive near the dryer section? Not general-purpose ones. IP66 is a water-jet rating (100 l/min at 100 kPa), and typical cameras are specified to around 60°C ambient — the hood area challenges both. Use purged or cooled stainless housings designed for paper machines in that zone, and keep ordinary PoE cameras to the drier, cooler parts of the hall.

What should I install first if my budget is limited? The bale yard, weighbridge and gate. Fibre is your biggest cost line, the losses there are simple to see and quantify, and payback is easiest to prove. Wet-end web-break monitoring is the higher-value project but needs hardened kit and a maintenance team ready to act on findings.

Do I need worker consent to record on the shop floor? Consent is not the only lawful basis, but you do owe notice, a stated purpose, a retention limit and security safeguards under the DPDP framework, whose substantive obligations bite on 14 May 2027. Put a clear multilingual notice at every entrance, and brief the union, before the first camera goes live.

Does OSH Code 2020 require CCTV in a paper mill? No provision mandates cameras outright. But it imposes hazardous-process duties, accident reporting and specific conditions for women on night shifts — and footage is a cheap way to evidence all three. Treat cameras as compliance infrastructure you chose, not a statutory box you ticked.