Compressed Air for Pulp and Paper Manufacturing

A 500,000 ton per year paper mill runs about 3,000 kW of compressor capacity. Around 10% of the electricity bill. The recovery boiler and steam system have had dedicated engineering teams for decades. Compressed air has a utilities supervisor who is also responsible for four other systems.

Compressed air problems in paper mills get misdiagnosed because the symptoms appear on the paper machine, not in the compressor room. Felt life drops and the wet end team investigates felts. Coating profiles go bad and the coating team changes formulations. Seal air carries oil mist at 0.08 mg/m³ onto the felt surface, degrades felt hydrophobicity, pushes up dryer steam consumption over weeks, and the investigation into rising dryer energy use looks at the dryer section and stays there.

System Overview

Paper Mill Compressed Air Infrastructure

This topic takes up a disproportionate amount of space here because the financial exposure is high and the root cause mechanism is widely misunderstood.

Put a data logger on the downstream side of the threading air reducing valve. Set the recording interval to 100 milliseconds. Leave it there through a few threading attempts. Look at the trace.

On most machines the threading air supply is a take-off from the main 7 bar header through a spring-loaded reducing valve. Maybe a small receiver behind it. Maybe not. If the receiver exists, it was sized during construction based on judgment, not calculation. The reducing valve is a general-purpose globe valve, whatever the valve supplier had in the correct bore size at the time.

So what does the data logger show? Most of the time, a stable trace. Then a pressure screen fires a backwash somewhere on the network, or someone opens a blowdown valve on the machine floor, or a compressor unloads in the compressor room a quarter kilometer away. Main header pressure drops half a bar for a fraction of a second. The transient arrives at the reducing valve. Spring-loaded diaphragm, 6 cm² area. Designed for slow drift in supply pressure. A transient that arrives and leaves in 200 milliseconds is not slow drift. The valve tries to respond, overshoots, and the downstream pressure oscillates through maybe 0.1 to 0.15 bar for a fraction of a second. If this coincides with a threading attempt, the web is gone.

The hardware for fixing this is simple. Dome-loaded precision regulator fed from a dedicated receiver, with a check valve between the receiver and the main header. The check valve prevents network transients from propagating backwards into the receiver. Without the check valve, the receiver is coupled to the network and provides no isolation, which is a mistake that has been observed on installations where someone understood the need for a receiver.

Receiver sizing. The threading nozzle flow rate comes from the nozzle manufacturer. Different machines have different nozzle counts and different nozzle types, so there is no universal number, but the data is available for any specific installation by looking up the nozzle part number and checking the flow table. Threading duration depends on machine configuration. Reel-up threading is different from Pope reel threading is different from center-wind threading, anywhere from 30 to 90 seconds. Multiply flow rate by duration, convert to volume at receiver pressure, and size the receiver so the pressure decay during the event stays within ±0.02 bar. For most machines running above 1,000 m/min, the answer falls between 500 and 2,000 liters. The spread in that range is driven mainly by nozzle count. A machine with twenty threading nozzles pulls much more air during a threading event than one with eight.

At most mills, it has not been done because of the capital approval process. Even a small project needs a written business case. The business case requires quantifying the cost of threading failures. Quantifying the cost requires extracting threading failure frequency and downtime duration from the downtime log, and at many mills, threading failures are not tracked as a separate downtime category. They go into "unplanned stops" or "paper machine breaks." The information is in the database but it has not been queried for this purpose, and the person who would need to write the business case, probably the utilities engineer, is not the person who owns the downtime database, and may not have access to it, and certainly does not know the per-minute machine cost off the top of their head, which is something the production planning group tracks. So writing the business case requires cooperation across three groups (utilities, production, planning) for a project that each group individually considers minor.

There is a secondary issue that shows up at mills that have installed a dedicated threading receiver. After a threading event depletes the receiver, it has to refill through the supply line before the next attempt. If the machine breaks during threading, clears, and re-attempts, and then breaks again, a receiver that takes four minutes to recharge through a small-bore supply line will not be ready for the third attempt. This catches people off guard because the installation seemed to work perfectly on the first threading attempt after it was commissioned, and then a few weeks later during a bad run with multiple breaks, the third threading attempt fails because the receiver is depleted and has not had time to refill. The supply line bore has to be sized for an acceptable recharge time given the threading frequency that the machine actually experiences, which is different from the threading frequency under normal operation. Under normal operation, the machine threads twice a day. During a bad run with repeated breaks, it might thread five or six times in an hour.

And the check valve between the receiver and the network has to be correctly oriented. Flow direction arrow stamped into the body of a wafer check valve, installed between flanges in an overhead pipe rack, fitter reaching above head height, arrow facing the wall or the ceiling, turnaround schedule running. A backwards check valve couples the receiver to the network and the installation provides no isolation. Air still flows. The receiver still fills. The regulator still regulates. Everything appears to work.

The air knife on a blade or jet coater profiles coating weight by blowing excess coating off the web surface. Pressure fluctuation in the supply gets amplified by the slot geometry. About 0.5% pressure ripple maps to about 0.5% coating weight variation on a roughly linear basis for small perturbations. On a 10 g/m² target, 0.05 g/m². Shows up on the printing press as uneven ink laydown even when the on-machine scanner does not catch it, because the spatial frequency of the variation can fall within the scanner noise floor.

Dedicated blowers are the correct supply for air knives on modern coating lines. The problem exists on older retrofit installations where the knife is fed from the 7 bar plant network through a reducing valve. Rent a portable blower, supply the knife in isolation for 48 hours, compare CD profiles. If the CD improves, convert the supply permanently. If not, cross the air supply off the list and continue investigating other variables. The test is cheap and fast and eliminates a variable either way.

Compression

Oil-Free Rotors

Filtration

Post-Treatment

Quality

Monitoring Systems

Main headers are sized at construction. After that, everything that happens to the piping makes it worse. This is a long section because it covers a lot of ground and because accumulated network degradation is usually the single largest contributor to compressed air waste at any mill that has not done a systematic pipe walkdown.

During a machine rebuild, a new branch is needed. Engineering specifies 80 mm. At the tie-in point, the routing through the existing pipe rack is tight. The only way to fit the branch is a 3-meter section of 50 mm. The correct fittings would take two weeks to fabricate. The machine is down and the turnaround schedule is running. The 50 mm section goes in. Air flows. Machine starts. The as-built drawing gets updated from the engineering specification document, which says 80 mm, because the draftsman was working from the engineering output and was not informed about the field change. Five years later, the people who made the decision to install 50 mm have moved to other positions or other mills. The institutional memory of the deviation is gone. The 50 mm section is inside a pipe rack, insulated, not visible from the floor. It might be discovered during a future pressure drop investigation. Or it might not, because during that investigation, two other restrictions are found and remediated and the investigator declares the problem solved and moves on.

Gate valves partially closed. The cause is usually unknowable by the time someone notices. Throttled during commissioning to balance branch flows, with or without documentation? Closed during maintenance and incompletely reopened? Deliberately throttled by an operator trying to solve a local pressure problem? Gate valve handles give no clear visual indication of percent-open from the operating floor six meters below. Could be 100% open. Could be 60%. Looks the same from down there.

Flanged joints with gaskets that shifted during bolt-up. The gasket ring protrudes into the bore by 10-15% of the cross section. Completely invisible without disassembly. Discovered only if someone disassembles the joint for another reason and notices the gasket position, or if someone does an ultrasonic flow measurement that shows higher-than-expected velocity at a joint location.

Check valves backwards. The flow direction arrow stamped into the body of a wafer check valve is not always visible after installation, especially in overhead pipe racks where the body is between flanges and the visible surface faces the ceiling or the wall. Installed backwards, a check valve does not block flow. The disc opens against its spring. Air pushes through. Pressure drop is higher than it would be with correct orientation, but the line works, and the elevated pressure drop gets attributed to general network aging. Could persist for the remaining life of the valve.

Abandoned piping is a separate problem from restrictions, though both contribute to waste.

Decades of modifications leave behind branches to decommissioned equipment. Construction-phase temporary air lines for pneumatic tools. Reserved connections for expansions that were planned, budgeted, approved, and then canceled when the market shifted. Pipe stubs with brass caps, pipe plugs, blind flanges, or in some cases nothing at all, just an open tee or branch fitting pressurized to 7 bar and venting to atmosphere inside a pipe rack above a drop ceiling.

Thirty percent is typical. Some mills run forty. Mills with a sustained leak program, meaning dedicated labor hours on a recurring schedule, not a campaign that happens once and then lapses, get down to ten or fifteen percent.

Ultrasonic leak detection works in areas away from the paper machine. Near the machine, the broadband ultrasonic energy from press hydraulics, dryer steam, vacuum systems, and mechanical vibration creates a background level high enough to mask everything except the largest leaks.

Network

Pipe Connections

Detection

Leak Assessment

Blow guns. Air wrenches. Hoists. Pneumatic tools of every description, used every shift. Quick-connect couplings on the tools and on the air drops get mated and separated over and over. The internal seal, a small nitrile or polyurethane O-ring, wears with each cycle. Not suddenly. Gradually. A coupling with a half-worn seal still clicks in. Still holds the tool. Still delivers air. A millwright using a blow gun cannot tell the difference between a new coupling and a worn one based on how the tool performs.

A single worn coupling leaks at a rate equivalent to a 2 to 3 mm orifice at 7 bar. Fifteen to thirty liters per minute, continuously.

The couplings are $3 to $8 each in bulk. A blanket replacement, every coupling on the machine swapped during a scheduled shutdown, costs $2,000 to $4,000 in parts and half a day of labor for two people. The energy cost of the air being lost through those worn couplings, calculated from the compressor specific power and the electricity rate and 8,760 hours per year, is a much larger number. The calculation takes ten minutes with a spreadsheet. It has not been done because quick connects live in the mechanical maintenance consumables budget alongside O-rings and grease fittings, and the people responsible for energy analysis do not look at mechanical consumables, and the maintenance planner thinking about quick connects is thinking about whether the tool stays attached, not about whether the coupling leaks while it is attached.

Compression fittings. Constant vibration from the paper machine loosens ferrules over months and years. Individual leak rate per fitting is tiny, maybe 0.1 to 0.3 L/min. Thousands of fittings on a single machine. If 20% have developed micro-leaks, total is a few hundred liters per minute across the machine. Not as dramatic as the quick-connect number. Still significant.

Ultrasonic detection does not work at these leak rates. Pressure decay testing is more systematic: close the isolation valve on an instrument air manifold, pressurize the downstream section, watch the gauge for decay. If the gauge drops, there are leaks in that section, and then soapy water on each fitting in that section finds them.

Higher pressure means more flow through every unregulated opening. Square root of the pressure ratio. Going from 6.5 to 7.5 bar adds about 8% to the flow through every leak and unregulated nozzle.

After leak repairs, the discharge setpoint has to come down proportionally. If the setpoint stays where it was, the capacity freed by the repairs gets consumed by increased flow through the remaining unregulated openings. The energy savings from the leak repairs vanish from the electric bill. This is widely understood in compressed air consulting. It is not widely understood at mills. The leak repair gets done, the energy bill does not improve, and the conclusion drawn is that leak repair was not worth the effort. Wrong conclusion. The setpoint was not adjusted.

"Oil-free" means no oil injected into the compression chamber. The intake air in a paper mill contains oil vapor from lube systems, diesel particulate from forklifts and mobile equipment, volatile organics from the general mill environment. All of that gets compressed and concentrated. Discharge oil content from an oil-free machine without post-treatment is around 0.03 mg/m³. ISO 8573-1 Class 1 requires 0.01 mg/m³. The downstream activated carbon filter is still necessary.

PTFE and ceramic rotor coatings delaminate at the microscopic level after several years of operation. Sub-micron fragments, hard and sharp-edged, enter the airflow. Standard oil monitors do not detect them. Standard particulate counters may or may not, depending on the particle size threshold of the instrument.

Uhle Box seal air carrying oil deposits on the felt changes the felt's water interaction. Dewatering degrades. Dryer steam consumption rises. Happens slowly, over weeks. By the time the dryer section energy trend is noticed, there are so many other possible explanations that the investigation focuses elsewhere. Seal air quality is not monitored at most mills. If it were, 0.05 mg/m³ would be a reasonable alarm threshold.

Oil below 0.1 mg/m³ in air contacting the wet web causes fish eye defects and ink adhesion problems. Activated carbon is the last barrier. Calendar-based cartridge replacement, typically annual, does not account for seasonal variation. Summer air carries more volatile organics. The cartridge loads faster and saturates before the annual replacement date. When saturated, activated carbon does not just stop adsorbing. It releases previously captured contaminants. The contamination spike from a saturated cartridge can be worse than the steady-state concentration that would pass through with no cartridge installed at all. There is no external indication of this happening. No pressure differential change. The evidence shows up in the paper quality lab days later.

Online oil monitoring downstream of the filter, triggering cartridge replacement on measured breakthrough, eliminates this failure mode. The instruments exist and work. Installation cost is not large. The obstacle is that the air quality monitoring program, if one exists, is owned by utilities, and the paper quality program is owned by the quality department, and the activated carbon replacement schedule was set at commissioning and has been followed without question ever since.

Air Quality

Filtration and Treatment Systems

Desiccant dryer tower switching releases a humidity pulse downstream, a few seconds to fifteen seconds per switch event. For most compressed air applications this is completely irrelevant and does not need to be discussed.

Process troubleshooters looking at this defect will try to match the period against known machine-side periodicities. Roll rotation periods, felt loops, pump stroke frequencies, oscillating shower traverse rates. The period will not match any of them. It corresponds to the dryer tower switching timer, which is set in the compressor room, in a different building, managed by a different group. Heatless dryers with 60-120 second cycle times produce a more pronounced pulse than heated regeneration dryers with longer cycles.

The connection between the paper defect and the dryer switching timer requires someone who understands both the paper machine process and the compressed air system well enough to hypothesize the connection and then verify it by either changing the switching interval and watching the streak spacing change, or by measuring the humidity pulse downstream of the dryer and correlating the timing. This combination of knowledge is rare. Paper machine process engineers know the machine. Compressed air specialists know the compressor room. The defect lives at the intersection and can persist for a long time.

Steam systems in paper mills have flow meters everywhere. Compressed air, very often, has no flow metering at all. Production is estimated by counting running compressors and assuming nameplate output, which becomes increasingly wrong as machines age and internal clearances open up and output declines.

Metering compressed air in paper mill conditions is technically frustrating. Thermal mass meters lag on transients from pulsed loads. Vortex meters lose accuracy at low velocities. DP meters have limited turndown. Installation conditions are usually poor: not enough straight pipe upstream and downstream, elbows and diameter changes too close to the sensing element. Under these field conditions, 20-30% measurement error is realistic, and the display shows decimal places and looks precise.

A meter at the compressor outlet measures total production. Useful air delivered to process is total production minus leaks, minus drain losses, minus dryer regeneration purge. A 30% leak rate and a 30% demand increase produce the same reading at the outlet meter. Without branch metering downstream, there is no way to separate them.

Same compressor, 0°C intake versus 38°C intake: about 11% difference in air output. Intake air density. Summer air also carries six to seven times more moisture, so dryer load and condensate production both increase substantially in summer.

Mills sized for worst-case summer have surplus capacity in winter. The adjustments that would capture this, lowering the discharge setpoint, changing compressor priority sequence, reducing drying capacity, are all operationally simple. None of them cost capital. They save 3-5% of annual compressed air energy in climates with distinct seasons. Operating procedures for seasonal changeover have not been written at most mills and the adjustments do not happen.

A 250 kW screw compressor at 35°C ambient and 80% relative humidity produces over 50 liters of condensate per hour. Condensate that stays in the pipe gets picked up by the airflow and carried downstream. Water hammer in pneumatic actuators. Erratic instrument signals. Wet spots on the paper surface.

VFD compressors are the standard recommendation for compressed air energy efficiency. They save energy. The recommendation is valid.

Pressure/flow controllers between the compressor room and the main network are the other standard recommendation. They hold downstream pressure stable and allow compressors to operate at a lower average discharge pressure. Also valid.

Screw compressors reject the large majority of their input power as heat. At a 3,000 kW installation, this represents a substantial heat source. Boiler feedwater preheating is the most common recovery application. Process water heating and building heating in winter are others. Many mills reject this heat through cooling towers.

Running the entire plant at the highest required pressure tier, which is typically 8-10 bar for the threading system or for specific pneumatic actuators, means that every reducing valve on every 6 bar instrument air user and every 4 bar purge user is converting the pressure differential into heat. Two pressure tiers, with the majority of the plant at 6-7 bar and a separate smaller system or local boosters for the few high-pressure users, reduces this thermodynamic waste. Whether the economics justify the modification depends on the ratio of high-pressure consumption to low-pressure consumption and the cost of the additional piping, which is different at every mill.

Manufacturer audits recommend buying compressors. That is the business model. An audit conducted by a party whose revenue comes from equipment sales produces recommendations that favor equipment sales. Network improvements, leak remediation, pressure optimization, and control tuning generate no revenue for the auditing party and receive proportionally less emphasis in the report. Independent audits, from firms with no equipment sales revenue, weight the recommendations differently.

Master controllers installed during a project, commissioned by the vendor's field engineer, and never subsequently re-tuned, are running on parameters that were set for a load profile that no longer exists. Product mix changes, capacity additions and removals, equipment aging, new pulsed loads, all of these shift the load profile over the years following commissioning. The control parameters need to be updated to follow. At most mills they have not been.

Controls

System Integration

Automation

DCS Connection

Trending

Long-Term Data

DCS integration opens up feedforward control. A grade change is scheduled in thirty minutes. The DCS knows this because the production planning system sends the grade change command. If the compressed air master controller is connected to the DCS, it can pre-load a standby compressor before the demand increase from the grade change arrives. The pressure dip that would occur during the changeover is absorbed before it propagates to the paper machine. This is substantially faster and more stable than waiting for the pressure to drop and then responding with feedback control.

Implementing this connection is a controls integration project that requires cooperation between the process automation group and the utilities group, agreement on a communication protocol, and middleware if the compressor PLCs and the process DCS are from different vendors, which they usually are. The project spans two departments and belongs fully to neither.

Tracking specific power (kW per m³/min) over years, using the flow metering installation described above, reveals compressor degradation gradually. A machine whose specific power is trending upward over a period of months is losing internal clearance and heading toward the point where output falls below useful levels or where the efficiency loss justifies an overhaul. Three years of site-specific trend data is more informative for scheduling overhauls than the manufacturer's maintenance interval in running hours, which was determined under test conditions that do not match any specific mill's intake temperature, humidity, dust loading, or load profile.

Kraft mills are corrosive environments. Sulfide compounds from pulping, chloride compounds from bleaching, persistent high humidity everywhere. This is harder on compressed air piping and equipment than most general industrial environments.

Aluminum piping resists corrosion and has the additional characteristic that its internal surface does not generate iron oxide scale. Carbon steel pipe networks accumulate internal rust over their service life. The rust flakes off, enters the airstream as particulate, and has to be captured by filters downstream. Over a twenty or thirty year pipe life, the filtration burden and the filter replacement cost on a carbon steel system is higher than on an aluminum system. Aluminum piping costs more to install. Whether total cost of ownership favors aluminum depends on the expected remaining life of the facility and what is already installed.