A few years ago a consulting engineer in Riyadh rejected a container of spiral duct because three lengths out of the hundred he pried open had welds he did not like. We argued that the lock seam carried the joint and the welds were a reinforcement option he had specified by mistake. We lost that argument, paid for the reinspection, and learned something worth learning: when a weld appears on a drawing, whoever inspects it gets to define what "good" means unless you arrive with a sampling routine of your own. Since then every welded spiral order from us leaves with a line-side sampling plan, and the rejection rate on third-party inspection has gone to zero. This is the routine, written for the plant floor rather than the quality manual.
On a spiral tubeformer there are two seams and people routinely conflate them.
The spiral lock seam — the folded and locked joint that runs helically along the whole tube — is formed mechanically by the lockFormer rolls. It carries the pressure and the stiffness on standard commercial duct. Inspection here is about lock quality: full closure of the fold, no gapping, no scoring through the zinc. A lock seam that opens under gentle thumb pressure is a forming setup problem — lock roll gap, strip tension or guide alignment — not a welding problem.
Where the spec calls for a welded seam — typically on heavier gauges, on pressure applications, or where the engineer does not trust lock seams at all — a longitudinal seam welder joins the strip edges before they enter the forming head, or a TIG/Plasma torch runs the helical seam continuously. This is the seam that draws inspector attention, and the one the rest of this piece covers.
Quality plans fail on the floor when they demand more than the shift can give. Ours takes about four minutes per sample.
We sample at these moments rather than every Nth piece: first-off after any strip coil change, first-off after any diameter or gauge change, first-off after any welder parameter adjustment, then one piece every 50 joints on stable running. On a line making 90 pipes an hour that is roughly one sample every 35 to 40 minutes on a stable run — manageable, and it catches the two things that actually drift: coil changes and parameter nudges.
Cut a 100 mm ring off the pipe end. Visual pass on the weld bead at 10x: bead width consistent, no visible porosity, no craters at the stop, undercut under 0.2 mm or 10% of gauge, whichever is less. Then a 90-degree hand bend of the ring at the weld — a sound weld on 0.9 to 1.2 mm galvanized bends with the parent metal and does not crack along the line. On the first-off only, cut one ring in half and check penetration from the ID side with a light and a mirror, or a boroscope if you have one: full fusion at the root, no heavy drop-through that would obstruct airflow. Log the result on the board. Two minutes of writing beats two weeks of email archaeology when an inspector asks.
Sampling finds problems; a parameter sheet prevents most of them from recurring. The sheet that lives laminated beside the welder should carry, for each gauge and coil coating weight you run, the four numbers that matter: wire feed or current, travel speed, torch stand-off, and gas flow. On 0.9 mm galvanized our baseline runs roughly 140 amps with the torch at 12 mm stand-off and 12 litres per minute of argon-CO2 mix; on 1.2 mm it climbs toward 170 amps and the travel slows a step. These are starting points, not gospel — every coil batch moves them a little, which is exactly why the first-off sample after a coil change is non-negotiable.
The detail worth more than the numbers themselves is the "adjusted by" column. When a setter tweaks the feed to suit a particular coil, they write what they changed and why, and the next setter inherits the reasoning instead of reverse-engineering it from a weld. We went from three parameter variants of the same truth on the same machine to one living document in about a month. The sheet also shortens new-operator training: an apprentice with a good baseline sheet and the sampling routine runs acceptable welds inside their first week, because they are calibrating against a known point instead of guessing from scratch.
Every weld defect has a short list of usual suspects, and they repeat from plant to plant with remarkable consistency.
| Defect | Most likely cause | Correction | Re-check |
|---|---|---|---|
| Porosity (pinholes in bead) | Zinc boiling off galvanized coating; current or heat too low for speed | Slow travel 10–15%, or raise current one step; grind zinc back 20 mm either side on stubborn batches | Next 2 joints |
| Undercut at toe | Travel too fast; torch angle drifting | Reduce speed; re-set torch to 5–10° push, check clamp alignment | Next joint |
| Crack at crater on stop | No crater fill; abrupt arc stop | Enable crater fill / slope-down on the welder; extend run-off tab on test pieces | Next 2 joints |
| Excessive drop-through (ID) | Current high or gap too wide at strip edges | Drop current one step; check slitter edge condition and strip width tolerance ±0.3 mm | First-off + 1 |
| Bead width wandering | Coil camber or edge waviness; guide wear | Check strip edge quality at decoiler; replace or re-dress edge guides | Next joint |
| Weld cracks on hand bend | Fusion line contamination — oil from slitting, mill scale | Wipe or brush strip edges; check decoiler lube carry-over | First-off |
The pattern to internalise: porosity and cracks usually trace to the coating or contamination, geometry defects (undercut, wandering bead) usually trace to speed, angle or guides. When you see both at once, suspect the coil before you suspect the welder.
Two rules keep a small problem small. If two consecutive samples fail on the same defect, stop and correct — the third sample is not evidence, it is scrap. And if any sample fails the hand-bend test, no further pipe from that parameter set ships until one passed sample follows a documented correction. I have watched crews "run it out and rework later" through a whole shift; 400 reworked joints cost more than the 25 minutes of stopped production would have.
One more stop-the-line trigger that took us embarrassingly long to formalise: any change of operator on the welder gets a first-off sample regardless of the counter. Parameter sheets drift hand to hand. The best welder I have employed ran 8% slower feed than the sheet said, produced the cleanest welds in the plant, and had never once updated the parameter board.
Inspectors rarely argue with data. What they reject is a quality claim with nothing behind it. The pack we ship with welded spiral orders now contains: the parameter sheet signed by the setter, first-off samples per change photographed with the ring bend result, the periodic sample log with time stamps, and a short statement of the acceptance criteria — undercut limit, porosity limit, bend test definition. That last page matters more than people expect. Half of inspection disputes are the two parties holding different unstated criteria. Write yours down, in the factory video and the submittal both, and most arguments never start.
Keep the log on the line, on paper, photographed at shift end. A quality system that lives only in a shared folder on someone's laptop dies the week that laptop's owner takes leave.
If you buy welded spiral duct, ask the supplier three things. What is the sampling plan and how often does it actually sample on a running line. What are the acceptance criteria in writing, including the bend test definition. And can they show a sample log from a recent order, not a certificate template. Suppliers with a real routine answer inside a day and their logs look like logs — ink, timestamps, occasional crossed-out lines. Suppliers improvising an answer are the ones whose containers get opened in Riyadh.
How often should we sample spiral duct welds on a stable run?
First-off after every coil, diameter, gauge or parameter change, then once every 50 joints. That catches the drift sources that actually matter while costing the shift about four minutes per sample. Fixed hourly intervals miss coil-change events, which cause most of the failures we see.
What is the quickest line-side weld quality check?
Cut a 100 mm ring, inspect the bead at 10x for porosity and undercut, then bend the ring 90 degrees by hand. A sound weld on sheet gauge bends with the parent metal without cracking. The whole check takes about four minutes and needs no more than a loupe and a band saw.
Why does our weld porosity spike after a coil change?
Different coils mean different zinc weights and surface condition. Heavier zinc boils harder under the arc and releases gas into the pool. Slow the travel slightly or step the current up for that coil, and grind the zinc back 20 mm either side of the seam if it persists.
Do we need to weld spiral duct at all?
Not usually. The lock seam carries standard commercial pressure classes and most specs accept it. Welded seams earn their cost on heavier gauges, higher pressure classes, or when the project engineer specifies them. Confirm which case you are in before paying for welds nobody requires.