Ask a duct line supplier what his length tolerance is and he will say plus or minus 0.5 mm. Ask him where that number comes from and most of them cannot tell you. I had to work it out the hard way on a job in Saudi Arabia where the customer measured every blank with a steel tape and rejected three deliveries in a row.
The truth is that the tolerance is not set by the servo. A modern 750 W servo with a 2,500 pulse encoder on an 80 mm feed roll has a theoretical resolution of about 0.01 mm, which is far finer than anything the duct cares about. The error comes from everything around the servo: the coil, the straightener, the roll pressure, the shear, and the way the operator loads the machine at seven in the morning.
This is the breakdown I give to buyers who ask why their blanks are 2 mm short at the end of a coil, and what to do about each part of it.
A finished blank's length is the sum of five errors. If you want plus or minus 0.5 mm in production, each one has to be under about 0.2 mm, because they do not cancel, they add.
The feed rolls grip the sheet by friction alone. Spring pressure or pneumatic pressure determines how much grip you have, and lubrication determines how much you need. On dry galvanized at 0.8 mm, 0.45 MPa on a 100 mm pneumatic cylinder gives roughly 3,500 N of clamping force, which is enough for a 60 m/min feed with no slip. Drop to 0.3 MPa and you will see short blanks of 0.3 to 0.8 mm, always in the same direction, and always worse on the last two metres of a coil where the sheet is less flat.
On the other hand, over-pressure marks the coating. Above about 0.7 MPa on 0.6 mm sheet you will see roller marks on the finished duct, and on prepainted material those marks are rejects. Find the pressure that holds tolerance and stop there. Mark the gauge with a paint pen so nobody "helpfully" turns it up.
Every coil has some memory of the coiler. A 5 tonne coil of 1.25 m wide galvanized sheet typically shows 2 to 4 mm of crossbow across the width and 1 to 3 mm of camber over 3 m of length. The straightener removes most of the crossbow but not all of the camber.
Camber is the one that ruins your diagonal. If the strip curves 2 mm over 3 m, a 1.5 m blank will be a parallelogram by about 1 mm on the diagonal, and when you join four of them into a duct, the flanges will not line up. The fitter blames the folding machine. The folding machine is fine.
The fix is a straightener with at least five working rolls and independent adjustment on the entry and exit pairs, plus an edge guide that actually touches the strip. Set the entry gap to about 85 percent of sheet thickness and the exit gap to 95 percent, then look at a 2 m cut strip on the floor. If it curves, adjust the exit pair in 0.02 mm steps. It takes twenty minutes and it is the best twenty minutes you will spend on the machine.
The factory ships a default acceleration profile that is gentle, because a gentle profile works on every material and never alarms out. It is also slow, and on thin sheet a slow feed with a long deceleration is where a lot of length error hides.
What I set on a Line 5 running 0.8 to 1.2 mm galvanized:
The settle time matters more than people think. The shear takes 60 to 90 ms to actually close after the signal. If the strip is still creeping during that window, you lose whatever distance it travels. At 20 m/min that is 0.03 mm, which is nothing. At full speed it would be 0.09 mm, which is why the two-stage feed is worth having.
Two mechanical items cause repeatable, one-directional error that no amount of servo tuning will fix. First, backlash in the feed roll drive coupling. Grab the roll and rock it by hand with the power off; anything more than 0.05 mm of rotation is worth fixing, and a worn coupling will give you 0.2 to 0.4 mm. Second, shear blade clearance. Excessive clearance lets the blade drag the strip down before it cuts, and the blank comes out long. On 1.0 mm sheet, blade clearance should be 0.06 to 0.08 mm per side. I have measured 0.25 mm on machines that were never set up after shipping, and those machines were producing blanks that were consistently 1.5 mm long.
| Symptom | Most likely cause | Check | Fix |
|---|---|---|---|
| Blanks consistently short by 0.5-1.5 mm | Feed roll slip | Roll pressure gauge, coating condition | Raise to 0.45-0.55 MPa, clean rolls with solvent |
| Blanks consistently long by 1-2 mm | Shear blade clearance too large | Feeler gauge between blades | Reset to 0.06-0.08 mm per side |
| Length drifts over a shift | Incremental positioning or encoder coupling slip | Compare first and last blank of the shift | Switch to absolute mode, tighten encoder coupling |
| Random 1-3 mm variation | Strip not seated against the edge guide | Watch the strip through the straightener | Adjust guide to 0.5 mm clearance, check camber |
| Square blank, wrong diagonal | Coil camber | Cut a 2 m strip, lay it on the floor | Adjust straightener exit rolls, re-thread |
| First blank after coil change is off | Strip not purged through the straightener | Operator procedure | Run and discard 1.5 m after threading |
That last row is worth a paragraph. On a coil change, the new strip is threaded by hand and the first 1.5 m is usually still bowed from the coil wrap. If you cut a production blank from it, that blank will be 2 to 4 mm off and nobody will know why. Write it into the work instruction: after threading, run 1.5 m to the scrap bin, then start. It costs eight seconds and saves a rejected duct.
The decoiler is treated as a dumb drum. It is not. How the coil pays off determines how much the straightener has to work, and a decoiler fighting the line will produce tension spikes that show up as length error at the shear.
You want the strip to pay off with slight tension, not slack and not tight. On a 5 tonne coil with a 508 mm bore, a pneumatic brake at 0.15 to 0.25 MPa is usually right. Too little and the coil overspins, feeding a loop that tangles when the line accelerates. Too much and the strip is stretched elastically, so the length you measure is the length under tension, and it springs back 0.2 to 0.4 mm when the shear releases it.
The test is simple: stop the line mid-feed and look at the strip between the decoiler and the straightener. It should be straight with a very slight sag, not banjo-tight and not touching the floor.
As the coil pays down from 1,500 mm to 600 mm outside diameter, the payoff speed at a given rpm drops by more than half. If your decoiler is driven at constant rpm rather than constant tension, the line will be fine at the start of the coil and starving at the end. That is exactly when the operator reports "short blanks at the end of the coil". Either use a dancer arm with a potentiometer controlling decoiler speed, or accept a lower line speed and set the decoiler by hand every 200 mm of coil diameter. The dancer is worth the money.
After all of that, here is what I am willing to put in writing for a well-set-up line on 0.8 to 1.2 mm galvanized: plus or minus 0.5 mm on length for 95 percent of blanks, plus or minus 0.8 mm worst case, and diagonal difference within 1.5 mm on a 1.5 m by 1.2 m blank. On 0.6 mm sheet the numbers get worse, around plus or minus 0.8 mm, because the sheet is more prone to buckling between the feed roll and the shear.
What you should not promise is the same tolerance on stainless or aluminium. Stainless work-hardens and springs back, so the same settings give you a blank that is 0.3 to 0.6 mm short. Aluminium at 1.0 mm is worse still, because it marks easily and you cannot use enough roll pressure to eliminate slip. Plan on plus or minus 1 mm for those materials, and set the shear to compensate if the run is long enough to matter.
If your downstream process is a Pittsburgh lock, remember that the lock consumes 8 to 10 mm of edge on each side. That allowance is where a lot of apparent length error actually comes from. Measure the flat blank before forming, not the formed part, or you will chase a tolerance problem that does not exist. The auto duct line overview shows where the shear sits relative to the lock former on each model, and the electric shearing machine page covers the standalone shear for shops that cut offline.
None of the above holds for long without a weekly routine. Mine is short enough that a shop will actually do it.
The written log is the part people skip and the part that pays. When a customer complains about a shipment, the log tells you whether the machine was drifting that week or whether the problem started at the folding station. Without it you are guessing, and guessing on a container that has already sailed is expensive.
Yes, and it is one of the better upgrades for a line older than about ten years. You need the servo, a planetary reducer, a new feed roll assembly, and a controller that can talk to the existing shear signal. Expect two days of installation and a genuine improvement from around plus or minus 1.5 mm to plus or minus 0.5 mm.
Almost always decoiler tension. As the coil diameter shrinks, a constant-rpm decoiler cannot keep up and the strip goes tight, so the feed roll slips against a stretched strip. Check for a dancer arm or reduce line speed in the last 20 percent of the coil.
Yes, but not the way people expect. Thicker sheet holds tolerance better because it resists buckling, but it also needs more shear force and more blade clearance. The worst case is 0.5 to 0.6 mm, where the sheet buckles between the roll and the blade and you get 1 mm variation from a machine that is perfectly set up.
Every coil change, with a tape measure on the first good blank, and a full ten-blank check weekly. Full servo calibration with a gauge block is a twice-a-year job, and only if the weekly checks show drift. Do not let anyone "calibrate" by adjusting the nominal length to match a tape measure, because that hides the slip instead of fixing it.