Infrared Moisture Sensors on Corrugators

An infrared moisture sensor on a corrugator watches the running web. It does not replace the laboratory. The oven method in board-moisture-testing-corrugator-quality-lab — about 50 g dried near 105°C to constant weight — remains the reference. The sensor’s job is to show a drift while the order is still on the machine, early enough to change wrap, speed, or a reel, and late enough that the number has been tied to that oven. A plant that trusts the first glowing digit on an uncalibrated head will chase adhesive for a reading that was only steam on the window. A number that has not been checked against the oven does not go on the shift release record, and it is not used to compare reels between shifts. The head identity and the grade-curve name go on the same line. If either is missing, that number is not handed to the next shift. After the curve name for a grade changes, the old release record is not carried forward. Run a full oven comparison again the same day and write the gap.

Near-infrared heads read moisture from water’s absorption near 1.94 µm, compared with one or more nearby reference wavelengths that are less sensitive to water. The ratio is turned into percent moisture only after the head is taught on that paper. Clay, coating, white-top, and a heavy ink film change the reference path. A library built on brown kraft will read a coated liner as if the sheet were wetter or drier than the oven says. That is not a small offset to ignore. It is a different grade, and it needs its own curve before the number is allowed onto the quality card. If the grade has changed and the old curve is still in use, withdraw that head’s release.

Place heads where the web is flat, visible, and out of the steam plume. A useful set is one view of the liner before the single facer, one of the medium, and one of the combined board after the double backer and before the cutoff. Do not bolt a head in the cloud that leaves a preheater or a hot plate. Condensate and vapour scatter the beam, and the display then moves when the steam valve moves, not when the paper changes. If the only free bracket is inside that cloud, move the bracket. A convenient mount that cannot see the sheet is not an installation. After the bracket is moved, measure the standoff again. Do not keep the old mark from inside the steam cloud.

Keep the standoff the head was built for. Many process heads specify a gap on the order of 150–250 mm; use the figure on that head, and do not copy the gap from the next machine. Web flutter eats the reading when the sheet leaves that band. After a splice or a loose wrap, watch whether the trace spikes in time with the flutter rather than with a real moisture change. If it does, fix the sheet path and the roll alignment before you touch the calibration constants. A sensor that is too close also collects starch dust and condensate on the window, which is a cleaning failure and a measurement failure at the same time. If the window is already fogged, that shift’s infrared readings are void. Wipe it and check it against the oven before you use it again.

One head in the centre of the web cannot see a dry edge. The cross-web rule from the lab still applies: a left-to-right spread larger than about 1.5 percentage points on one sheet is a process flag, even when the centre number sits inside the plant window. That means at least a left, centre, and right view, or a traversing scanner that actually reaches both edges. Name operator side and drive side on the screen. “Left” flips when someone stands at the other end. A scanner that stops 100 mm short of the trim will miss the dry edge that becomes warp in the stack. Incoming liner and medium are still judged near the plant window, commonly about 6–9%, and a liner-to-medium gap above about 2 percentage points remains a warp flag. The online trace should be able to show both facts, not only a single average. A screen that shows only the centre value cannot decide whether a dry edge is already making warp.

Calibrate against the oven, not against a handheld pin meter. Cut a sample from the spot the head just read, within about 15 minutes, and dry it by the written oven method. If, after the grade curve is set, the infrared number and the oven number on that spot differ by more than about 0.5–1.0 percentage point, do not release the infrared reading for that grade. Pin meters and pinless handhelds read a different zone and must not be the master. Check the correlation at least once a week, and again after a lamp or window change, a crash stop that shook the bracket, or a move to coated or white-top liner. Write both numbers. A pass stamp without the oven result is not a calibration. When the check is outside the allowed gap, leave the infrared column blank for that grade and keep only the oven result.

Build a small grade library instead of one slope for the whole mill. Separate at least virgin liner, recycled medium, coated or white-top liner, and a heavy-print sheet if you run one. Recycled and high-filler sheets swing more, so they need a fresh oven pair at the start of the reel and after a splice, not only at the first setup. When two reels of the same nominal grade differ by more than about 2 percentage points at the head, do not average them into one curve. Split the order or reset preheat, and record why. base-paper-moisture-management-corrugator covers how those reels got wet or dry before they reached the sensor. The head only reports what arrived. A grade that is not in the library does not borrow the slope of a neighbouring grade and go on the machine.

Clean the window on a schedule, not when the trace looks strange. Steam and starch dust fog the glass. On a line that runs glue and heat all shift, wipe the window at least once per shift with the method the sensor maker allows, and again after a steam leak or a wrap that threw wet spots at the head. Do not use a shop rag soaked in an unknown solvent. A scratched or filmed window adds noise that operators then treat as moisture. If the trace jumps by more than about 1 percentage point in under a minute on a steady grade and a steady speed, treat it first as a window, a splice, or a flutter event. Open the adhesive recipe only after those three are ruled out. Write the cloth and the solvent on the method sheet. The shift does not change them on the spot.

Do not let the sensor rewrite the glue formula. A wet medium is not fixed by opening the glue gap or raising viscosity. Extra adhesive on a wet sheet makes a wetter bond line and a slower dry. A dry liner is not fixed by adding water in the kitchen to “match the sensor.” Hold the reel, change wrap, or slow the order, and keep the moisture card current before you discuss /products/corrugated-board-glue. Concentrate powders and water-only corrugator powders have their own mix ratios. None of them absorbs a false infrared number. If the head and the oven disagree, the head is not yet a reason to change starch, caustic, or water. Until the moisture card is current, the glue kitchen does not add or remove water because of the head.

Installation is a mechanical job with an electrical ending. Rigid bracket, no shared vibration with an unguarded drive, cable routed off the hot plate, and an air purge if the head calls for one so the window stays clear. Confirm the purge pressure is the maker’s figure, not “whatever the mill air header has today.” After the bracket is tight, mark the standoff and the cross-web position with a punch or a painted witness so the next roll change does not leave the head 30 mm closer and 40 mm more toward the centre. Power-up should show a live temperature and a live moisture, both named. A display that only says “OK” is not commissioned. If the witness marks are painted over, the next shift remakes them before start-up.

Set alarms on the facts you already use in the lab, not on a tighter band the sensor salesman prefers. A practical start is to alarm when the centre reading leaves the plant window, when left-to-right spread exceeds about 1.5 points, or when liner and medium diverge by more than about 2 points. Alarm the rate as well: more than about 1 point in a minute on a steady grade. Every alarm needs a named response — check window, check reel, check wrap — and a rule that adhesive changes are not the first response. If operators learn to silence the alarm because it fires on every splice, the sensor has become decoration. Widen the splice exception in the software, and keep the steady-state alarm. An alarm silenced with no reason written down counts as not handled.

Keep a correlation log that another shift can audit. Each row needs date, order, grade, head identity, standoff checked, window cleaned, infrared percent, oven percent, the difference, and the name of the tester. Once a week is the minimum for a grade you run every day. A grade you have not run for a month gets a new pair before it is released. When a customer asks why a stack warped, the log should show the liner, the medium, the cross-web spread, and the oven check, not a recollection that the screen was green. Store the log with the laboratory oven sheet for that week so the two systems cannot drift apart in two different files. A row missing the head identity or the oven value is unfinished at review.

Xuegong New Materials Group can help a plant connect a trusted moisture trace to adhesive choice after the oven correlation is real. The sensor maker owns the wavelength, the standoff, and the purge specification. The plant owns the grade library and the decision to hold a reel. Use board-moisture-testing-corrugator-quality-lab for the oven method and base-paper-moisture-management-corrugator for reel storage. Use /products/corrugated-board-glue only when the moisture card and the infrared log agree. An uncalibrated head is not a reason to change the glue, and a clean window is part of the measurement, not a housekeeping extra. If the maker has not given a purge pressure, do not substitute the shop air-header pressure. When the purge air stops, stop infrared release, wipe the window, and compare with the oven again before you use it. A shift without that comparison record cannot claim infrared moisture is under control.