Cut edges, welds and the station record

A flame-cut edge is not just a surface. The steel beneath it has been heated to burning temperature and cooled again in seconds, and that thermal history can make it harder than the rest of the plate. A weld made with an unstable flame carries the instability into its metal. Both effects were examined in 1921 with the metallurgical tools of the day, and both point to the same conclusion as the rest of this book: what happens at the torch should be controlled, measured and written down.

This final chapter covers the edge and the weld, then brings the whole book together in a record of the cutting station that someone other than its operator can check.

What cutting does to the edge

The 1921 investigators cut sections through flame-cut blocks of their 1/2 inch, 2 inch and 6 inch test steel and examined them. Etched at natural size, the sections showed clearly how deep the flame had affected the metal. Under the microscope at 100 diameters, the structure at the cut surface could be compared with the unchanged steel at the centre of each block.

The three steels differed in carbon content, and the effect differed with them. The 6 inch billet steel, with about 0.46% to 0.51% carbon, was changed far more at its surface than the lower-carbon plates. The micrographs suggested that carbon might have been added to the surface by the flame. To test that, the investigators annealed a sample of the 6 inch material at about 760 °C, above its critical temperature, taking care to avoid losing carbon during heating, and compared the surface layer with the centre. The carbon content was the same in both.

They concluded that the altered surface was not carburised. It was the result of the cold mass of the block chilling the thin layer of hot metal at the cut face once the flame moved on, a self-quench that hardens higher-carbon steel more than lower-carbon steel. That, they noted, explained the hard surface metal machinists often met when working up flame-cut steel.

The modern description is the same in substance. Thermal cutting leaves a narrow heat-affected zone along the edge, and in hardenable steels that zone can be hard and brittle. READOS’s chapter on heat treatment and surface hardening explains why rapid cooling from high temperature hardens steels with enough carbon. A hard edge can crack when bent, break drills and taps, resist machining and crack when welded.

Judging how hardenable a steel is

A simple first screen for hardenability uses a carbon equivalent, which weights the alloying elements by how strongly they promote hardening. The full IIW formula, discussed in the Weld Repairs That Hold chapter on preheat, hydrogen and cooling, includes chromium, molybdenum, vanadium, nickel and copper as well as carbon and manganese. For plain carbon-manganese steels, the first two terms carry most of the weight.

This is an illustration using only those two terms. The 1921 6 inch steel, at 0.46% to 0.51% carbon and about 0.60% manganese, gives C + Mn ÷ 6 ≈ 0.56 to 0.61. A grade 250 plate with, say, 0.16% carbon and 0.70% manganese gives about 0.28. The base plates’ edges are far less likely to harden than the 1921 billet’s. The figures are not a substitute for the mill certificate, which gives the actual composition, or for the full formula where other elements are present.

Figure 1Flame-cut edges and carbon content

	Likely edge condition	Usual response
Low carbon, e.g. CE about 0.28	little hardening	clean off slag; grind where welded or fitted
Medium carbon, e.g. CE about 0.56 to 0.61	hard, chilled surface layer	preheat before cutting; slow cooling; machine or grind off
Unknown steel	cannot be judged	identify it first; test hardness on a trial cut

The usual responses follow from the mechanism. Preheating the plate before cutting reduces the temperature difference that drives the self-quench, slow cooling after cutting does the same, and removing the hardened layer by machining or grinding takes it away altogether. Where an edge will be welded, standards and procedures for the work commonly require cut edges to be free of slag, notches and hardened or cracked material.

This is an illustration. Ten of the base plates will have two edges welded to a frame. Grinding 1.5 mm from each of those 20 edges, 0.4 m long and 25 mm thick, removes about 300 cm³, or 2.36 kg, of steel. That is a measurable job in time and abrasives, and it belongs in the estimate of Chapter 9.

What the 1921 welding tests showed

The welding half of the 1921 investigation is less directly relevant to a modern cutting workshop, because oxy-acetylene welding of steel has largely given way to arc processes. It remains in use for thin sheet, small pipe, repair, brazing and heritage work. Its findings, though, are a sharp lesson in process stability.

Each torch made test welds in mild steel plate with a single-V butt joint of 90° included angle, using one grade of welding rod, at the specified pressures and at pressures above and below them. The welds were tested in tension and by cold bending. The welders were chosen for skill and impartiality, and the investigator thought the welds were probably made with more care than any comparable set before. The results were still unsatisfactory: the properties varied widely, the bottom of the V was almost always oxidised and the upper part of the weld varied.

The explanation came from the flashback work of Chapter 6. During welding, the tip is constantly moving towards and away from the pool, and the operator’s hand is never perfectly steady. Each movement changes the back pressure, and each change of back pressure shifts the mixture in the torch, with the tendency, in the designs tested, towards an oxidising flame. The gas ratios measured during welding were much higher than those measured with the torch burning freely in air, and different operators doing identical tests produced different ratios. Flow readings taken every few minutes fluctuated even though the pressure at the torch was held within a few hundredths of a pound per square inch. The investigator concluded that no torch tested could maintain a neutral flame during welding, and that a torch that cannot do so cannot be expected to make sound welds consistently.

One further observation is worth keeping. Each 2-foot test weld was made by joining two 1-foot lengths of plate laid on a heavy cast-iron support with a channel beneath the joint. The second plate welded in each test was consistently stronger. The investigator explained this by the flame running along the channel and preheating the second plate, and by the warmed support slowing the cooling and relieving some of the contraction strain. Preheat and controlled cooling, discussed in Weld Repairs That Hold, appear here as an accidental result in a test designed for something else.

The report noted that critics of oxy-acetylene welding had called for welders to be specially trained, tested and licensed, and for better welding rods. It called those good and useful measures, but argued that they could not succeed if the operator was given a tool that could not hold a neutral flame. In modern terms, the welder, the procedure and the equipment all have to be qualified and controlled. Welding of structural steel in Australia is specified to standards such as AS/NZS 1554.1, which deal with procedures, welder qualification and inspection; the requirements that apply are those of the standard and specification the work is designed to.

A failure: the bracket that cracked

This is an illustration. Alongside the base plates, the workshop flame-cut a batch of small brackets from an offcut in the rack. Several cracked along the cut edge when they were bent to shape, and a drill broke on another. A file skated over the cut edges without biting, a simple sign of hard metal, and a hardness test on a trial cut confirmed it. The offcut had no identification: it was not the grade 250 plate it had been assumed to be, but a higher-carbon steel left over from a previous job.

The failure had two causes, one metallurgical and one of record keeping. The metallurgical cause was the 1921 self-quench in a hardenable steel. The record-keeping cause was that nobody could say what the steel was. The workshop scrapped the batch, recut the brackets from identified plate, and made the plate’s identity, linked to its mill certificate, an entry on the cutting record for every job, so that any offcut could be traced before it was used.

The station record

The chapters of this book each add entries to one record: the evidence that a cutting station is set up, measured and protected as its settings say. A workshop does not need the 1,900 log sheets of 1921. It needs a short, current record that answers the questions an inspector, a new supervisor or an estimator would ask.

Figure 2Cutting station record, base-plate track machine

• Avoid: Gas supplyoxygen manifold, acetylene cylinders; contents by weight; withdrawal limit
• Avoid: Regulatorsidentities; creep test; set while flowing
• Avoid: Hoses and fittingslength, bore, condition; torch-end pressure checked
• Avoid: Flashback arrestorsboth gases, both ends; tested within 12 months
• Avoid: Flow metersgas and pressure scaled for; last check against weighing
• Avoid: Settingssetting B: flows, torch-end pressure, speed band, maximum drag
• Avoid: Nozzleidentity; checked each shift
• Avoid: Materialplate grade and mill certificate for each job
• Do: ? Trainingrefresher on stand-off and backfire signs, due next month

Each entry should say who made it and when, and the record as a whole should have an owner, usually the supervisor responsible for the station. The 1921 investigation gave the manufacturers a copy of every test log for their apparatus and let them witness and object in writing; the equivalent today is a record that operators, supervisors, safety officers and estimators can all read and question. An entry that only its author understands is not evidence for anyone else.

The record is only useful if it is kept current and reviewed. A change to the hose, the nozzle, the gas supplier, the plate grade or the setting is a reason to check the entries it touches. A backfire, a flashback, a failed trial or a quality problem is a reason to review the whole record, because each of them is evidence that something in the chain has changed. The work health and safety duties that apply to the workplace, the regulator’s code of practice on welding processes, AS 4839 and the equipment and gas suppliers’ instructions set the minimum; the record is how a workshop shows that it meets them and knows why.

A rhythm for keeping the record true

Different entries go stale at different rates, and the record works best when each is checked at the interval its failure mode demands. A nozzle can be damaged in one pierce, so it is checked every shift. A plate grade changes with every job. A flow meter drifts slowly, so a monthly check against weighing is usually enough. Flashback arrestors are inspected and tested by a competent person every 12 months, as WorkSafe ACT’s guidance describes, and immediately after any flashback.

Figure 3When each part of the station record is checked

1. Each shiftNozzle and flamenozzle bore inspected; flame checked against recorded flows
2. Each jobMaterial and settingsplate identity and certificate; settings sheet confirmed
3. Each monthMeasurementflow meters checked against a weighed run
4. Every 12 monthsSafety devicesarrestors inspected and tested by a competent person
5. After any eventWhole recordbackfire, flashback, failed trial or quality problem

The record also changes how improvements are made. When a supplier proposes a new nozzle, a different fuel gas or a cheaper oxygen contract, the record already holds the baseline: the flows, torch-end pressure, speed, drag and cost per metre of the current setting. The proposal can then be tested by the method of Chapter 1, with the same plate, operator, equipment and weighing, and costed by the method of Chapter 9, including handling and any finishing the new setting needs. A change that survives that comparison goes into the record with its evidence. One that does not is set aside with the reason written down, so the question does not have to be asked again from nothing.

That is the most durable lesson of the 1921 investigation. Its specific findings about torches of the day were overtaken within a few decades. Its method, controlling what can be controlled, measuring what matters two ways, writing everything down and drawing conclusions only as far as the data support, is as useful at a workshop cutting station today as it was on the Bureau’s test floor.

Exercise: screening an unknown steel

A supplier’s certificate for a replacement plate gives 0.20% carbon and 1.20% manganese, with no other significant alloying elements. Calculate the two-term carbon equivalent, compare it with the base plate and the 1921 billet, and say what the station record should require before the plate is cut.

The two-term carbon equivalent is 0.20 + 1.20 ÷ 6 = 0.40. That is well above the grade 250 base plate’s 0.28 but below the 1921 billet’s 0.56 to 0.61. Edges in such a steel may harden noticeably, especially in thick sections that chill the cut surface quickly. The record should require the full carbon equivalent from the certificate, a hardness check on a trial cut, and a decision, by someone competent to make it, on whether to preheat before cutting or remove the edge layer where the edges will be welded, bent or machined. The 1921 investigators found the cause of hard flame-cut edges with an annealing furnace and a microscope; a modern workshop can usually avoid the problem with a certificate, a trial cut and a line in its record.