Why Joint Preparation Decides More Than Technique

Why this matters

By the time an arc is struck, the four numbers that decide how much metal must go into the joint, whether the welder can reach the sidewalls, and how much shrinkage the assembly will have to absorb are already cut into the steel. A welder cannot out-run a groove. When the same defect appears at the same place on assembly after assembly, the shop is usually looking for a person when it should be measuring a bevel, and the measurement takes about a minute per joint.

Preparing the joint is its own set of hazards, and they are not the welding ones. Flame cutting and gouging use fuel gas and oxygen under 29 CFR 1910.253, with cylinders secured upright and caps fitted when not in use. Air carbon arc gouging throws molten metal several feet and produces both a heavy fume plume and noise levels that put it inside a hearing conservation program under 29 CFR 1910.95, so it needs local exhaust or respiratory protection under a written program meeting 29 CFR 1910.134, hearing protection, and a screened area. Grinding a bevel on coated steel releases what the coating carries, commonly zinc from galvanizing or a chromate primer under 29 CFR 1910.1026, which is an inhalation route needing tool-mounted capture or respiratory protection rather than eye protection alone. If the part being prepared is a used tank, drum or pipe that held a flammable or its residue, it is not cut, gouged or welded until it has been cleaned or inerted under 29 CFR 1910.252(a)(3). The fire prevention side of any of this routes to the shop's hot work permit and fire watch, which the related hot work cards own.

The four numbers, and the convention they are stated in

Single-V groove butt weld, plates seen end-on, not
to scale. The included angle is the total angle
between both prepared faces; each plate is beveled
half of it.

  \                                        /
   \                                      /
    \                                    /
_____\                                  /______
     |                                  |
     |  root face             root face |
_____|                                  |______
     |<----------- opening ------------>|

• Included angle. The total angle between the two prepared faces. Say "included" every time, because the other convention in live use is the bevel angle of one member measured from the plate's perpendicular, and it is half the number. A 60 degree included angle is two 30 degree bevels.
• Root face, also called the land: the unbeveled square edge at the bottom of the bevel.
• Root opening, also called the root gap: the space between the two root faces.
• Alignment, meaning high-low, the offset between the two plate surfaces at the joint.

Every one of these carries a tolerance on the WPS, and a tolerance is part of the number. "1/8 in root opening" without its plus and minus is not a specification, it is a wish.

What each one buys and what it costs

Included angle buys electrode access to the sidewalls and costs volume. At constant plate thickness, the cross-sectional area of a single-V groove goes with the tangent of half the included angle. Going from 60 to 75 degrees takes tan 30 to tan 37.5, or 0.577 to 0.767, which is 33 percent more metal to deposit at the same thickness. Going the other way, down to 45 degrees, takes it to tan 22.5, or 0.414, which is 28 percent less metal and correspondingly less room for the electrode to see the sidewall. Both directions are real trades, and neither is free.

Root face buys resistance to burn-through and costs penetration. Too large and the root faces do not fuse; too small and the root melts away.

Root opening buys access to fuse both root faces from one side and costs volume and burn-through margin. It also interacts with the root face: a bigger opening tolerates a bigger land, which is why the two are specified as a pair rather than independently.

Alignment buys nothing. High-low is pure loss: it puts one root face out of the arc's reach and it creates an eccentricity in the load path that persists after the weld is perfect.

A double-V changes the arithmetic. For the same thickness and the same included angle, preparing from both sides makes each side half as deep, and since area goes with the square of depth, the two halves together come to half the metal of a single-V. Half the deposited volume is half the shrinkage, and putting it on both sides of the mid-thickness lets the two angular movements oppose each other. The distortion card owns that mechanism; the point here is that it is bought at the prep, not at the arc.

Dilution is the last consequence. Dilution is the fraction of the finished weld metal that came from melted parent metal rather than filler. A tight groove with a high-penetration process gives high dilution, which matters whenever parent and filler chemistries differ enough for the mixture to behave like neither.

The case: three assemblies, same defect, same place

A shop fabricating 3/4 in plate butt welds got sidewall lack of fusion on the same side of the joint, roughly a third of the way up, on three of four assemblies. The welder had been on the job two years.

Candidates eliminated on evidence. Technique: the same welder produced sound welds on the fillet work running in parallel that week, radiographed clean. Consumable: fresh H4 rods issued from the holding oven into a quiver and returned within the exposure limit, so this is not the consumable-condition failure the related storage card describes. Machine: output checked against the WPS ranges on the machine's own calibrated meters. Interpass: readings taken with a contact pyrometer before every pass, all inside the WPS window.

That leaves the joint itself, which nobody had measured.

What the measurements found. The bevels came off a hand-guided track torch that had been set up once and never re-checked.

• Included angle, measured with a protractor gauge at 8 stations along the 4 joints, two stations per joint: 44, 45, 45, 46, 45, 44, 46, 45 degrees. Median 45 degrees against the WPS's 60 degrees. Consistent, so this is a machine setting, not a wandering hand.
• Root opening, taken with a taper gauge at 4 points per joint: averaged 1/16 in against the WPS's 1/8 in.
• High-low on the worst assembly: 3/16 in on 3/4 in plate.

Working the geometry. At 45 degrees included, the groove area is 0.414 times the square of the depth. At the specified 60 degrees it is 0.577. So the shop was depositing about 28 percent less metal than the procedure called for, which is why nobody complained about arc time, and it was doing it in a groove whose sidewalls were 7.5 degrees steeper on each side than the electrode had been qualified to reach. Add a 1/16 in opening instead of 1/8 in and the bottom of the groove is narrower still. Add 3/16 in of high-low and one root face is set back behind the other by a quarter of the plate thickness.

Every one of those pushes the same direction: less room, at the sidewall, low in the groove. The defect location, one side and a third of the way up, is where the electrode angle needed to reach past the steepest part of the wall.

The correction, and the thing it costs. Reset the track torch and verify the first bevel of every setup with a protractor gauge before cutting the rest. Re-fit to the specified 1/8 in opening. Set an alignment tolerance and use dogs and wedges rather than accepting what the plates give you. Then the honest consequence: restoring the included angle from 45 to 60 degrees raises the groove area by the ratio 0.577 divided by 0.414, or 1.39, so the assemblies now take about 39 percent more deposited metal. More deposited metal is more transverse shrinkage, so the sequencing plan has to be rebuilt at the same time rather than after the first assembly comes out short. The sequencing card owns that plan.

What the shop had been about to do instead. Retrain the welder, and if that failed, replace him. Three of four assemblies with the defect in the same location on the same side is a process signal. One of four in a random location would have been the other conversation.

What would change the diagnosis. If the defect had been at the root rather than the sidewall, the root face and opening would be the suspects and the included angle largely irrelevant, and the root pass card owns that branch. If the bevels had measured on specification, the next measurement is the electrode angle and the travel technique, and the conversation about the welder becomes the right one.

Check it against the rules the siblings state

• Convention named with every angle: 45 and 60 degrees are stated as included angles, with the bevel-per-member half explicitly given as 22.5 and 30 degrees respectively, so no reader can substitute the other frame.
• Area relationship carries what it holds constant: groove area goes with the tangent of half the included angle at constant plate thickness, and the double-V comparison is stated at the same thickness and the same included angle.
• Arithmetic closes both directions: 0.414 against 0.577 is 28 percent less area, and 0.577 divided by 0.414 is 1.39, a 39 percent increase. Those are the same comparison read from opposite ends and both figures are printed.
• Sample size stated rather than implied: 8 bevels measured across 4 assemblies with a median of 45 degrees, and 4 taper-gauge points per joint. Three of four assemblies carried the defect, and that is written as a count, not as a rate.
• Consumable and interpass rules from the sibling cards were run against this scenario before the geometry was blamed: rods issued from the oven and returned inside the exposure limit, and contact-pyrometer readings before every pass inside the WPS window.

References

• AWS D1.1, structural welding code for steel, in the edition invoked by your project specification or the adopted building code, for prequalified joint details, groove dimensions and fit-up tolerances
• ASME Boiler and Pressure Vessel Code Section IX with the governing piping or vessel code, as adopted by your jurisdiction, for pressure work joint details
• 29 CFR 1910.253 for fuel gas and oxygen used in cutting, 29 CFR 1910.95 for noise exposure from arc gouging, 29 CFR 1910.134 for respiratory protection, 29 CFR 1910.1026 for chromate coatings, and 29 CFR 1910.252(a)(3) for used containers
• See related: What a Root Pass Has to Achieve; What Distortion Is and Why It Is Predictable; What a Welded Joint Changes About the Parent Metal