What Happens When a Weld Seam Reaches a Servo Feeder?
The weld arrives without warning and takes about 1.5 seconds to pass through the feed rolls at 200 SPM.
In those 1.5 seconds a 0.12 mm thickness step can turn a 45 mm feed into a 44.4 mm feed, and the die finds out before the operator does.
A servo feeder handles a coil join well enough if the setup anticipates it. The failures we see are almost never the weld itself; they are the machine running the weld on settings written for continuous strip.
What a Weld Seam Does to Feed Length
A strip weld is not flush. Even a good join carries a step of 0.05 to 0.15 mm on one or both faces, and a hand-made repair can carry 0.3 mm.
The feed rolls have to climb over that step twice, once entering and once leaving.
On the way in, the step lifts the roll and releases grip for a fraction of a rotation. On the way out, the roll drops and the material behind the weld is briefly under less pressure than the controller assumes.
The error is not symmetrical. Entering the weld the roll climbs the step and loses grip; leaving it the roll drops and briefly over-feeds.
Two errors of opposite sign on consecutive strokes are harder to spot than one.
On 0.6 mm strip the same step is proportionally twice as damaging, because the change in roll opening is a larger fraction of the material thickness.
At 200 SPM with a 45 mm feed, one stroke lasts 300 ms. A 0.35 mm loss on the stroke that crosses the weld is a 0.8 percent error, which is sixteen times the ±0.05 mm the line is expected to hold.
The servo sees it as a torque disturbance, not a position error. By the time the encoder reports the shortfall the strip has already been released, so the controller cannot correct the stroke that was lost.
Hardness matters as much as thickness. A weld is typically harder than the parent strip, so the same roll pressure produces less bite exactly where more is needed, and the roll surface takes the wear.
How to Detect the Weld Before the Die Does
Mount the detector 300 to 500 mm before the feed rolls so the controller gets at least one stroke of warning. Closer than that and the signal arrives after the weld is already between the rolls.
Three detection methods are in common use, and they differ in what they can see rather than in accuracy.
A contact weld detector rides the strip surface on a small roller and trips on a step above about 0.03 mm. It is cheap and reliable, and it misses a flush weld entirely.
A thickness gauge downstream of the weld station measures the actual section and will catch a flush weld that a contact detector skips. It costs more and needs calibration, but it sees what the feed rolls will feel.
The third method is the servo itself. A repeatable torque spike of 15 to 20 percent at a known strip position is a usable signal, provided the controller logs torque against feed length and someone reads it.
Which Settings Change for the Seconds It Passes
Write the numbers on the setup sheet rather than leaving them inside a recipe. A weld mode nobody can read is a weld mode applied at the wrong moment.
The aim is not to run the weld perfectly. It is to get the weld through the die without a misfeed and without marking the rolls, then return to production settings immediately.
| Action | What it gains you | What it takes back |
|---|---|---|
| Drop to 60 percent speed for 1.5 m before the weld | More time per stroke for the roll to recover grip | About 12 lost strokes per coil, or roughly 4 seconds of output |
| Open the feed roll gap by 0.1 mm | The step passes without lifting the roll off the strip | Less grip on the strokes either side, so pressure has to rise with it |
| Raise roll pressure one step | Restores bite on the harder weld section | Faster roll and bearing wear, and more marking risk on coated strip |
| Disable pilot release for one stroke | Removes the chance of the pilot pulling a short feed straight | Requires a controller that can do it on position, not on a manual switch |
Opening the gap without raising pressure is the common mistake. The step passes cleanly and the strip slips for the next twenty strokes, which produces a run of short feeds that looks unrelated to the weld.
What the Die Protection Must Do in the Same Second
Tonnage monitoring catches the consequence rather than the cause. A short feed usually shows as a 10 to 15 percent tonnage rise on the station after it, which is late but better than nothing.
Feed length monitoring has to be active on the stroke that crosses the weld, not armed after it. A window of ±0.3 mm will catch the 0.35 mm loss described above, and it will not nuisance-trip on normal variation.
Pilot release timing needs to follow the same logic. If the pilots enter before the feed is confirmed, a short stroke gets pulled into position and the sensor never sees a fault.
The strip is then stretched, and the next station produces a part that measures correctly and cracks a week later.
Set the stop to latch. A weld-induced misfeed that clears itself on the next stroke will otherwise run for a full coil before anyone inspects the die.
Where This Procedure Stops Working
Stainless behaves differently again. The join work-hardens, and the step often comes with a flat spot that no roll pressure recovers, so the strip has to be cut back further than on CR4.
A detector that trips after the weld has already passed the rolls records history. It does not control anything, and it should not be counted as protection in the setup sheet.
It cannot save a weld that is thicker than the die clearance allows. If the join stands 0.4 mm proud on a 1.2 mm strip running 0.08 mm clearance, the punch will take the load no matter how the feeder behaves.
It cannot compensate for a weld placed inside a critical form. Some dies simply have no tolerant station, and the only answer is to cut the weld out before the die or plan the coil change to avoid it.
It also cannot be automated on every machine. Position-based pilot release and torque logging are controller features, and a feeder without them relies on the operator hearing the change, which is not a procedure.
Finally, the 4 seconds of lost output per coil is real but small.
What is not small is the habit of running welds at full speed to protect a rate target, which trades a few seconds against a die repair measured in days.
The Steps to Write Into the Setup Sheet
Six steps, written in the order the operator performs them, are enough to make weld passage repeatable across shifts.
- Record the weld position on the coil tag when the coil is loaded, in metres from the outside end.
- Enter weld mode 1.5 m before the mark, which drops speed and opens the roll gap in one action.
- Confirm the pressure step came up with the gap, because the two settings belong together.
- Watch feed length on the two strokes either side of the weld, not on the weld stroke alone.
- Return to production settings 1.5 m after the mark, and verify the first five parts against the gauge.
- Log the weld, the speed and any misfeed, so the next coil with a join starts from data rather than memory.
Feed accuracy is a budget spread across the whole line rather than a property of the feeder alone, and our coil line knowledge base breaks that budget down contribution by contribution.
The rated feed length, strip width range and roll configuration for each model are listed in the equipment range index.
That is the quickest way to confirm whether a machine can do this on position rather than on a manual switch.
How much feed length does a 0.15 mm weld step cost at 200 SPM?
Typically 0.3 to 0.5 mm on the crossing stroke. That is six to ten times a ±0.05 mm tolerance, which is why the weld stroke has to be monitored rather than averaged.
At what weld step does a servo feeder start to slip, 0.05 or 0.1 mm?
Around 0.1 mm on smooth CR4 at normal roll pressure. Coated or oily strip can slip at 0.05 mm, and textured material tolerates more.
How far before the weld should speed drop, 0.5 m or 1.5 m?
1.5 m is a practical figure for 1.2 mm strip at 200 SPM. It gives the roll about 30 strokes to settle and still costs only 4 seconds of output.
Can the feeder run a weld at full speed if the die has good protection?
Yes, if the protection window is tight enough to catch a 0.3 mm short feed and the stop latches. Without a latched stop, a single clear stroke hides the fault.
Where to go next
This article belongs to the feed accuracy cluster, where a single local disturbance has to be traced through the whole error budget.
The linked guide shows how each contribution is measured and what it costs to remove.
our coil line knowledge base on feed accuracy
