Fair comparisons: what the 1921 blowpipe trials teach

Most workshops inherit their oxy-fuel settings. A nozzle chart taped to the trolley, a pressure someone was taught years ago, a torch brand chosen because the last one was the same. When a supplier offers a new nozzle or a cheaper gas, the comparison is usually a few cuts on scrap and an opinion. The opinion may be right, but nobody can check it, and the next person to ask the question has to start again.

In 1918 the United States Army faced the same problem on a much larger scale. It had bought large numbers of oxy-acetylene welding and cutting sets for field repair and demolition and found that no reliable data existed on which to write a specification. It asked the Bureau of Standards to test the efficiency, safety and workmanship of the makes on the market. The way the Bureau designed that test is the subject of this chapter, because it is still a good model for any comparison a workshop needs to trust.

What the Bureau set out to do

The investigation, reported by Robert S. Johnston in Technologic Paper No. 200, was first meant to be a quick, rough comparison. Once the war ended the pressure eased, and the extra time let the work grow into a planned programme. Before testing started, the Bureau consulted the Army, the Navy and the manufacturers, then fixed a written schedule and sent it to every maker. Makers had to deliver their apparatus before learning the details of the tests, so nobody could tune a torch to the test.

The stipulations covered four things a workshop trial also needs to control.

• The gas. All gas came from banks of cylinders drawn through identical equipment, of the same analysis within commercial limits. Oxygen purity was analysed cylinder by cylinder, and banks were assembled so their mixed purity matched a standard of 98.3%, held within about 0.2 percentage points.
• The pressure. One selected set of regulators controlled every torch of a type, so differences between regulators could not be credited to the torches.
• The operator. One group of experienced welders and cutters did all the work under the Bureau’s direction, and every cutting test was made with the torch on a mechanically guided carriage.
• The material. Test steel was 1/2 inch, 2 inch and 6 inch plate and billet of known composition, with one grade of welding rod.

Manufacturers could send three witnesses, had to confirm in writing that the apparatus was standard stock in working order, could object in writing within 24 hours and could ask for a demonstration with their own operator under identical conditions. A committee of Army, Navy and Bureau representatives settled disputes. Some makers disagreed with the method and withdrew. Their apparatus was bought on the open market and tested in the same way, so withdrawal did not remove a product from the comparison.

Controlling the personal equation

The 1921 report uses an old phrase, the personal equation, for the difference that the person makes. A skilled operator can make a poor torch look acceptable, and a tired or unfamiliar one can make a good torch look poor. The Bureau attacked this in three ways.

First, it took the hand out of cutting. A track-mounted carriage moved the torch at a set speed. Before each test the operator made trial cuts on cold plate to find the fastest speed at which the torch would cut through, and the test was then run at that speed without change. Only the stand-off had to be adjusted by hand, because the plate warped as it was cut.

Second, it measured the effect of the operator where it could not be removed. Two of the welding tests were identical except for the welder, so the difference between them estimated how much the person mattered.

Third, it compared hand and machine cutting directly. On 1/2 inch plate, with the same torch and pressures, an operator with years of experience achieved a speed about 25% below the machine’s. The report expected the gap to widen with thickness. That single comparison tells a modern workshop something useful: a speed taken from a machine trial overstates what a hand cutter will achieve, and an estimate should say which it assumes.

Figure 1What the 1921 trials held constant, and how

	How it was controlled	What a workshop can do now
Gas	banked cylinders, purity analysed and matched	same cylinders or manifold for every setting; record purity grade
Pressure	one selected set of regulators	same regulators; read pressure at the torch end
Operator	one team; machine-guided cutting	one operator; track machine wherever possible
Material	plate of known analysis and thickness	strips from one plate, same edge condition
Order and temperature	trial cuts on cold plate	alternate the order; start each cut on cold plate

Measuring rather than estimating

The heart of the investigation was measurement. The gas banks hung from equal-arm balances of about 3,000 lb and 1,000 lb capacity, for oxygen and acetylene, and consumption was found from the loss of weight. Weighing a bank of heavy cylinders to a small fraction of a pound is not easy, and the method is instructive.

• Substitution. As gas was drawn off, water was run into bottles on the platform to keep the balance near its zero, so the reading was always taken in the same part of the scale.
• Drift correction. The zero was found before and after each test from several swings of the beam, and the drift between them was corrected to the start and finish times.
• Sensitivity check. The balance’s sensitivity was measured every morning with a standard weight, because it changed from day to day.
• Leak correction. Small leaks at cylinder valves were too small to find easily but large enough to matter. The lines were pressurised and the drift measured for about half an hour to find the leak rate, which was then subtracted. If the loss exceeded about 0.01 to 0.02 lb an hour, testing stopped until the leak was found.

With those corrections the report put the gas consumption figures within about 0.005 lb. The Bureau also built orifice flow meters, read alongside the balances, to give a second and independent measure of flow. When the two disagreed, the disagreement led to a discovery about the meters that Chapter 5 describes. The principle is that a measurement you cannot cross-check is a measurement you cannot defend.

When the specified settings did not work

The schedule required each torch to be run at the maker’s specified tip size and pressures. A large proportion would not run at them. The report suggests why: operators were commonly told to set the regulator a few pounds per square inch above the specified torch pressure and throttle the gas at the handle valves, which gave a margin against regulator fluctuation and helped avoid flashback. Run with the handle valves fully open, many tips delivered gas so fast that the flame would not stay on the tip.

The Bureau’s response is a rule worth keeping. At least one handle valve on each gas had to be fully open during a test, because only then is the torch’s consumption defined by its design and the delivered pressure rather than by where an operator happened to leave a valve. Where the specified pressure would not hold a stable flame, the maker’s representative was asked for a modified pressure, and that was recorded.

The welding tests then deliberately ran the same tip at 50% above and 25% below the working pressures, because the investigator expected a well-designed torch to tolerate the pressure errors common in real shops. The cutting tests did the same with oxygen pressure alone. Testing a setting either side of its nominal value is still good practice: it shows whether a result sits on a plateau or at the edge of a cliff.

Designing the workshop’s trial

This is an illustration. The base-plate workshop has a nozzle chart for 25 mm plate and a supplier suggesting two changes: trim the preheat flame, and raise the cutting oxygen pressure for speed. That gives three settings to compare.

• Setting A, the chart: cutting oxygen 4.6 m³/h, acetylene 0.70 m³/h with preheat oxygen 1.1 times the acetylene, track speed 380 mm/min.
• Setting B, trimmed preheat: the same cutting oxygen, acetylene 0.50 m³/h, 400 mm/min.
• Setting C, high oxygen: cutting oxygen 5.8 m³/h, acetylene 0.50 m³/h, 420 mm/min.

The speeds are the fastest at which each setting cut cleanly through in trial cuts on cold plate. The flows are what the trial measured. They are invented for teaching and are not data for any real nozzle.

Figure 2A fair trial of three cutting settings

1. 1Preparestrips from one plate; same edge condition; cold start
2. 2Fix the equipmentsame regulators, hoses, torch and track machine
3. 3Weighoxygen and acetylene cylinders before each run
4. 4Find speedtrial cuts on cold plate at each setting
5. 5Cut6.0 m per setting, in the order A, B, C, then C, B, A
6. 6Weigh againconsumption by mass, converted to volume
7. 7Inspecttop edge, face, drag lines, dross, squareness

Each run cut 6.0 m. The cylinders feeding the torch were weighed on a platform scale reading to 0.05 kg before and after each run. Oxygen weighs about 1.353 kg per cubic metre and acetylene about 1.101 kg per cubic metre at 15 °C and atmospheric pressure, so the mass lost converts directly to volume. At setting B, for example, the run took 15.0 minutes and used 1.742 kg of oxygen, which is 1.287 m³. The scale’s 0.05 kg resolution is about 2.9% of that mass, good enough to separate the settings, but not good enough to resolve the acetylene alone on a short run, which used only 0.138 kg. The acetylene result was therefore cross-checked against the reading of a flow meter, as Chapter 5 explains.

Figure 3Trial results, 6.0 m of 25 mm plate per setting

	Time	Oxygen used	Acetylene used
A, chart	15.79 min	1.912 kg = 1.413 m³	0.203 kg = 0.184 m³
B, trimmed preheat	15.00 min	1.742 kg = 1.287 m³	0.138 kg = 0.125 m³
C, high oxygen	14.29 min	2.046 kg = 1.512 m³	0.131 kg = 0.119 m³

How large a difference can the trial see?

Before believing a difference, check that it is larger than the measurement can blur. Each weighing on a scale that reads to 0.05 kg can be out by up to half a division, 0.025 kg, and the consumption is the difference of two weighings, so in the worst case it can be out by 0.05 kg.

• Oxygen, A against B. 1.912 − 1.742 = 0.170 kg, more than three times the worst-case error. The difference is real.
• Oxygen, C against B. 2.046 − 1.742 = 0.304 kg, about six times the error. Also real.
• Acetylene, A against B. 0.203 − 0.138 = 0.065 kg, only 1.3 times the error. On its own, the scale cannot prove that A uses more acetylene than B.

There are three ways to sharpen a measurement that is too coarse: cut further so more gas is used, use a finer scale, or add an independent measurement. The workshop added a flow meter reading, which agreed with the scale and resolved the acetylene difference. The 1921 investigators faced the same choice and made both their balances and their test lengths large enough, cutting 50 linear feet of 1/2 inch plate in one test, so that the gas used was large compared with the weighing error.

Setting C was the fastest, by 5% over B, but used 26% more cutting oxygen to get there, and its cut face showed local gouging. Setting A used 40% more acetylene than B and rounded the top edge. Whether B or C is the better choice is a question of cost and quality that Chapters 8 and 9 answer. The trial’s job is to make that later argument rest on numbers both sides accept.

A failure: the confounded trial

The workshop’s first attempt at this trial produced a different answer. Setting A was cut first thing in the morning, on cold plate from the rack. Setting B was cut straight afterwards on strips that had sat beside the first cuts, and C last, after a cylinder change. B appeared about 10% faster than A, and C about 15% faster.

The supervisor noticed two things in the record. The strips for B and C had been noticeably warm to the hand when cutting started, and the oxygen cylinder for C was full while A had run on one near empty. Either could explain part of the difference. A warm plate needs less preheat and cuts faster, which flatters whatever is cut later. The 1921 report had met the same problem: its supplementary cutting tests at varied oxygen pressure showed some inconsistencies that it put down to cutting warm plate.

The cure was the design in the figure above: cold strips for every run, cylinders weighed rather than assumed, and the settings run twice in alternating order. When the order was reversed, part of the apparent advantage of C disappeared, and the repeat runs agreed within the scale’s resolution. A trial that is run once, in one order, cannot tell an effect of the setting from an effect of the order.

What a fair trial record contains

The value of a trial lies in its record. The 1921 investigation produced over 1,900 log sheets, each with the pressures, temperatures, balance readings and flow meter readings for a test, so that every reported number could be traced back. A workshop trial needs far less, but it needs the same kind of entries.

Figure 4Record for a workshop cutting trial

• Avoid: Questionwhich of three settings for 25 mm plate
• Avoid: Equipmentregulator, hose, arrestors, torch, nozzle identified
• Avoid: Materialplate grade, thickness, edge condition, temperature at start
• Avoid: Gascylinder identities, masses before and after, purity grade
• Avoid: Settingspressures at regulator and torch, valve positions
• Avoid: Resultslength, time, gas by mass, cut-quality notes
• Do: ? Repeatreversed order not yet run for setting A
• Avoid: Signaturewho ran it, who checked the arithmetic

Exercise: gas per metre for setting A

From the trial record, setting A used 1.912 kg of oxygen and 0.203 kg of acetylene to cut 6.0 m. Calculate the oxygen and acetylene used per metre of cut, in cubic metres at 15 °C and atmospheric pressure.

Convert mass to volume first. Oxygen: 1.912 ÷ 1.353 ≈ 1.413 m³. Acetylene: 0.203 ÷ 1.101 ≈ 0.184 m³. Dividing by the 6.0 m cut gives about 0.236 m³ of oxygen and 0.031 m³ of acetylene per metre. As a check, the flows recorded for setting A were 5.37 m³/h of oxygen in total (4.6 m³/h cutting plus 0.77 m³/h preheat) and 0.70 m³/h of acetylene, at 380 mm/min, or 22.8 m/h. Dividing flow by speed gives 5.37 ÷ 22.8 ≈ 0.236 m³/m and 0.70 ÷ 22.8 ≈ 0.031 m³/m. The two methods agree, which is exactly the cross-check the 1921 balances and flow meters were designed to provide.