Technical resource
The shielding gas you pay for and never use
Walk into most workshops I go into and the flowmeters are cranked to 20 to 25 litres a minute. Watch someone purge a pipe for a stainless root and the disregard for how much gas is going out the end of it is total. I have heard the line more than once. I am not paying for it. Someone is.
That setting is only the loss you can see. Underneath it sit four more: a flowmeter that does not read what it says it does, the hose volume dumped on the floor every time a trigger is pulled, leaks that run whether anyone is welding or not, and the gas handed back inside part full cylinders that your supply contract gives you no credit for. Nobody counts any of it, because gas is a small share of the cost of a finished weld and a very large share of what you spend on consumables.
So if you own the place or you manage it, go and do this. Check twenty per cent of your machines while they are actually being welded on, not sitting idle in the corner. Write the flow rates down, average them, and put that number in below. You will be surprised. And the saving is not just a saving: it is the deposit on new equipment, or it is money you can push into the parts of the business that actually lift productivity.
Set your shop up below. The infographic and the numbers move with it.
How to use this
For lecturers and industry trainers
This one is aimed at the person who signs the gas invoice, so run it as a costing exercise rather than a physics lesson. Put it on the screen and work the controls:
- Start with the operating factor. Set it to 100 per cent and watch the number triple. That is the mistake in nearly every gas saving claim you will read: they cost the whole shift instead of the arc-on time. Bring it back to 30 and explain why.
- Then move the set flow from 18 down to 12. That is not a theoretical number. It is what BAE Systems Govan actually implemented, with preset controllers and the welders’ ability to adjust taken away, after which no defects were found.
- Switch the cylinder size from G to E. Same gas, same waste in cubic metres, and the bill jumps. The small shop is punished twice, once on the rate and again on having no volume to spread it over.
- Push the leak hole from 0.2 mm to 1.0 mm. It goes up twenty five times, because leakage scales with the square of the diameter. Then turn on out of hours pressurisation and watch it move again. Use that to explain why a single leak figure should never be published, only a range and a method.
- Wind the arc starts up to 400 for tack heavy work. Surge grows, but not as fast as people expect, because the hose needs about three to four seconds to refill between pulls. Point out that the tool caps it for that reason, and that any figure which does not is inflated.
The point to land: the largest single loss is a number on a flowmeter, and correcting it costs nothing but a document revision. Under ASME IX a change in shielding gas flow rate range is a nonessential variable at QW-408.3, so it is a WPS revision, not a requalification. No coupon, no test plate, no NDT. Some editions of AWS D1.1 are stricter, so name the code you are working to before you rely on it, and check the Australian position against your own copy of the standard rather than against this page.
Walk your shop with this list
Nine checks, no instruments needed for the first six. A supervisor can do the lot in an hour.
- Read every flowmeter and write it down before anyone knows why you are asking. Compare against the WPS. You are looking for the gap between what is specified and what is set.
- Check what gas each flowmeter is scaled for against what is actually connected to it. An argon scaled meter on an Ar/CO2 mix is only about one per cent out, so that panic is overdone. On a helium tri-mix it reads over fifty per cent low, and that one matters.
- Look for a flowmeter mounted off vertical or with the float fouled. Both read wrong and neither announces it.
- Find the regulator delivery pressure. If it is set at 550 kPa (80 psi) where 170 kPa (25 psi) would do, every trigger pull is dumping three times the hose volume it needs to.
- Measure the hose run from the gas solenoid to the torch. Long leads on 1/4 inch hose store a lot of gas. A shorter or narrower delivery hose cuts the surge on volume alone.
- Soap test every joint from the cylinder to the torch, cold, with the machine off and the gas on. Hose barbs, solenoid valves, torch connections, regulator seats, quick connects, perished hose.
- Shut the cylinder valves at knock off and check in the morning whether the pressure has moved. If it has, you have quantified your own leak rate, which is more than the published literature has managed.
- Weigh or gauge cylinders on return. Anything going back part full is gas you bought and surrendered, and most supply terms give no credit for it.
- Count the cylinders on site against the ones in use. Rental accrues whether or not the gas is consumed, so an idle cylinder in a corner is a standing charge with no output.
Why this matters
Ask a room of welders whether more shielding gas gives better protection and nearly every hand goes up. The belief survives because the failure mode is invisible and delayed. Subsurface porosity does not announce itself at the arc, it turns up at radiography or in service. Nobody ever gets the feedback that would correct the instinct, so it compounds: a weld fails, the flow goes up, the next weld passes for unrelated reasons, and the lesson is learned backwards.
The controlled work says otherwise. Campbell, Galloway and McPherson produced acceptable welds at 6 L/min in a draught free bay against a 15 L/min baseline, and recorded that there was no scientific evidence supporting the higher figure. Be careful with that 6. In the same programme the 6 L/min case failed once a 0.7 mph cross draught was introduced, so the number to work to in a real workshop is 9 L/min, which they found good even in a light draught. The figure actually put into production at BAE Systems Govan was 12. Against a shipyard’s own measured 24 to 26, twelve still halves the bill.
What makes this worth an owner’s attention is the shape of the problem, not just the size. Shielding gas is a small share of the total cost of a finished weld, which is exactly why it never gets looked at, and a large share of what you spend on consumables, which is why the saving is real when you do look. And the biggest single component of it, over-setting the flow, has no capital cost attached to fixing it and no requalification burden. It is a document revision and a locked flow limiter.
The other four need hardware or discipline, and they are worth knowing in proportion. On a busy shop the flowmeter setting dwarfs the rest. On a quiet one with cylinders sitting idle, rental and residual gas climb up the list. Leaks are the wild card: there is no published leak rate for shielding gas systems anywhere in the literature, so the honest treatment is a range and a method, which is what the slider above gives you.
What is verified, what is derived, what is a modelling choice
Currency date 31 July 2026. verified peer reviewed primary. invoice taken off a real Australian gas invoice. published a published Australian plan or retail price. derived calculated here, working stated. trade press non refereed, cited through the literature. illustrative a modelling choice with no published basis.
| Figure | Status | Basis and what a reviewer will say |
|---|---|---|
| Flow the weld needs: 9 L/min | verified | Campbell, Galloway & McPherson, Int J Adv Manuf Technol 63 (2012) 975 to 985, DOI 10.1007/s00170-012-3961-2. Good quality welds down to 9 L/min in the presence of a light draught, Ar/CO2 80/20 on DH36. The better known 6 L/min is a draught free result and failed at 0.7 mph in the same programme, which is why 6 is offered here as a laboratory figure and not the default. 12 L/min is the figure implemented at BAE Systems Govan with preset controllers and welder adjustment removed, after which no defects were found. |
| Flow actually in use: 24 to 26 L/min against a norm of 18 | verified | Beyer et al. 2013, Heriot-Watt and Strathclyde, on BAE Systems Naval Ships practice with a 16 mm nozzle. The same paper records that 15 to 20 L/min is typically specified but “is often adjusted to as high as 36 L min-1 by welders in practice”. Used here in place of the frequently quoted “surveys show approximately 25 L/min”, which appears in a 2012 conference paper with no citation for the surveys. |
| Metering error, 10 per cent floor | derived | An error budget, not a measurement: instrument accuracy of 3 to 5 per cent of full scale, float reading error, gas mismatch, and back pressure on a non compensated meter. No welding manufacturer publishes an accuracy figure for any flowmeter, flowgauge regulator or flowmeter regulator. A single common installation fault gets you 20 to 30 per cent. Presented as a budget with the components named, which is defensible; presented as a measured figure, which it is not, it would be demolished. |
| Surge volume per trigger pull | derived | Hose volume times [(regulator pressure less the pressure at the solenoid while welding) divided by atmospheric] times 1.13 for hose expansion. 1/4 inch ID hose holds 0.0317 L per metre. Ten metres at 345 kPa (50 psi) gives 1.10 L per start. This corrects the vendor’s own arithmetic, which ratios to a hose at zero gauge pressure and overstates by about a quarter. The result matches ESAB’s independently published 1.835 L per start to three significant figures once you account for their unstated 550 kPa (80 psi). The tool caps the loss where the hose has not had the three to four seconds it needs to refill, corroborated by Air Products reporting surge falling to 0.2 L at 50 per cent duty cycle. |
| The 95 L/min surge peak | excluded | Deliberately absent. It traces to Uttrachi’s own US patent 7,015,412, a year before the trade article usually cited, he owns the competing product, no test conditions are stated anywhere, and the same author has published 100, 150, 200 and 225 CFH in different documents. There is no independent measurement of it. |
| Leak rate as a percentage of gas used | excluded | No published leak rate percentage exists for welding shielding gas distribution systems from any primary source. Not from TWI, EWF, AWS, any gas major, or any peer reviewed study. Every figure circulating in the trade press traces to a vendor or to a single trade article from 2000. This tool asks for an equivalent hole size instead, which is honest, and shows how violently the answer moves with it. |
| Leak flow from an equivalent hole | derived | Argon is comfortably choked at shielding gas line pressures, so flow is linear in absolute upstream pressure. Q in L/min equals 0.0604 times area in mm2 times absolute pressure in kPa, at a discharge coefficient of 0.61 and 20 degrees C. The method reproduces a US Department of Energy published leak table to within 3.4 per cent. Read the axis as an equivalent sharp edged orifice, as DOE does, because real leaks are not neat holes. One 1 mm hole at 275 kPa (40 psi) leaks 18 L/min, more than a whole station’s set flow. |
| Compressed air leak analogue, 20 to 30 per cent | verified | US Department of Energy. It transfers as methodology, ISO 11011, and not as a leak rate for welding gas. Do not read the Australian government’s “half of this energy is lost in leakages” as a 45 per cent leak rate; it conflates conversion efficiency with leakage. |
| Trade account rate, preset at $10 per m3 | invoice | The evidence is a real Australian trade account invoice from a major supplier, South Australia, first half of 2026. The $10 preset is that evidence rounded up, which understates the saving rather than overstating it. Use the input field for your own figure. Method: cylinder price ex GST plus the per cylinder freight surcharge, divided by the supplier’s own published G size gas content. Across four different products on the same account it came out at $6.93, $6.99, $7.03 and $7.69 per m3, on an Ar/25% CO2 mix, an Ar/16% CO2 with oxygen, an Ar/18% CO2 and a pure welding argon. A tight band across four products is a good sign the content figures are sound. Supplier and product identities are withheld because this is contract pricing, not a published rate. Note also that not every supplier states a reference temperature for its cylinder contents, so the cubic metre basis is theirs rather than a normalised one. |
| Published plan and retail rates, presets at $20 and $35 per m3 | published | Derived from publicly advertised prices divided by published cylinder contents, normalised ex GST. A national flat rate plan works out at $21.47 per m3 on a G size refill, rounded here to the $20 preset. A hardware chain swap cylinder works out at $34.14 per m3 on an E size. The smallest retail bottles sold through tool retailers reach $106 per m3 on the same gas. No Australian major publishes a shielding gas price list, so every figure here is a plan or retail rate rather than a contract rate. The full spread on identical gas is roughly fifteen to one, trade account to small retail bottle, and the cylinder size gradient alone accounts for a factor of four to six of it. A small shop is punished twice, once on the rate and again on having no volume to spread it over. |
| Variable freight surcharge, $1.50 per cylinder | invoice | A separate line on the same invoices, charged per cylinder delivered. Small on its own, but it means every unnecessary cylinder you cycle carries a delivery cost on top of the gas, which is a further argument against running more bottles than the work needs. Check your own invoices for it, because it is easy to read past. |
| The high CO2 mix holds more gas | published | On one supplier’s published contents, an Ar/25% CO2 mix holds 12.8 m3 in a G size against 10.6 m3 for pure argon in the same cylinder at the same pressure, roughly 21 per cent more gas in the same bottle. That is part of why the high CO2 mix works out cheapest per cubic metre, and it is worth knowing before you compare cylinder prices without dividing by content. |
| Cylinder contents: E 4.9, G 10.2 to 11.0, pack 217.6 m3 | published | From suppliers’ own published cylinder data sheets and product pages. Watch the units trap: a cylinder letter is a designation, not a gas volume, and pack contents do not scale linearly from single cylinders because packs are filled to a higher pressure. |
| Cylinder rental, $144 per year | illustrative | Still the weakest number on this page. Cylinder rental for welding gas is not published by any Australian supplier, at any size, in any channel. One major publishes a page promising to explain its fees and charges that contains no dollar amount at all. The $144 comes from forum reports of about $12 a month on an E size. Two real delivery invoices were checked against it and they do not settle it either: their rental lines carry quantities but zero dollars, because those lines record cylinders coming back rather than a rental charge, and the rental itself is billed on a separate cycle. One of them does carry a note advising a rental price increase, which confirms rental is charged and rising without saying by how much. Getting a real figure needs a rental or statement invoice, not a delivery invoice. For context, amortising an owner purchased cylinder over its ten year test interval gives $26 to $30 a year, with three independent suppliers agreeing within 12 per cent, so rental plausibly runs at four to six times the cost of ownership. Note also that at least one major’s published standard terms allow rental to be varied with immediate effect on cylinders already in your possession, with silence for fourteen days deemed consent. Read yours. |
| Residual gas surrendered on return | verified | At least one major Australian supplier’s published standard terms give no allowance for residual gas returned, and continue rental after termination for as long as gas remains in the cylinder. A cylinder handed back half full has lost both the gas and the rental days. This is a waste mechanism that sits in the contract, not in the equipment. Check the clause in your own supply agreement. The five per cent assumed here is a modelling choice. |
| Operating factor, arc starts per day, equivalent hole size | illustrative | Modelling choices, all three, and all three are levers on this page so you can put your own in. The operating factor is the one to check first in anybody’s gas saving claim: a model built on attended hours rather than arc-on hours overstates the answer by roughly three times. |
| Gas as a share of cost | verified | Small as a share of the total cost of a finished weld, large as a share of consumable spend. This is why it is never audited and why the saving is real when it is. |
The Strathclyde and Heriot-Watt research programme behind most of the flow rate evidence on this page was funded in part by BAE Systems, whose shipyard practice it reports, and by a manufacturer of gas control equipment of the kind it evaluates. The work is peer reviewed and its data is the best available on the subject. The funding is stated here so you can weigh it yourself.
Graham Fry · grahamfry.com.au · 1300 00 WELD