What a Flare Joint Depends On

Why this matters

A flare is the only common joint where you manufacture the sealing surface on site, out of the tube itself, using a hand tool. That surface is also the thinnest and most heavily worked metal in the whole assembly, because forming it stretched the wall out and hardened it. So a flare is a seal and a stress riser occupying the same few millimetres of material, and the failure that keeps a shop coming back is not usually a leak past the face. It is a crack in the cone, made during forming or during the life of a badly supported tube, that a wrench can only ever make worse.

Before opening a flare on a charged or pressurised system: recover the refrigerant to the level the equipment and the regulation require and confirm zero at a gauge on the section itself. Venting most refrigerants during service, maintenance, repair or disposal is prohibited under 40 CFR 82.154(a), and the technician doing the work has to hold the certification required by 40 CFR 82.161. This is not only a paperwork requirement: a flare nut backed off under pressure releases liquid that flashes at the joint, which causes frostbite and eye injury on contact, and heavier-than-air vapour that collects in pits, crawlspaces and mechanical rooms and displaces oxygen, so ventilate the space rather than relying on gloves and a face shield to cover an asphyxiation route. On fuel gas tubing, shut off and lock the supply and confirm zero at the appliance before the nut moves, and do not use an ignition source to check anything.

What the joint is actually made of

Four parts, and only one of them comes out of a box.

• The cone you formed on the tube. This is the sealing surface. Its finish, its roundness, its wall thickness and its freedom from cracks were all decided before the fitting was touched.
• The fitting seat. A machined cone at a defined angle. Its job is to be flat, clean, and the same angle as yours.
• The nut. It pulls the two cones together. It does not seal anything.
• The tube behind the flare. This is the part everybody forgets, and it is where the fatigue lives.

The sealing action is metal-to-metal contact over a band, not a line and not a point. When the nut pulls up, the softer of the two cones yields slightly and conforms to the harder one, and that conformed band is the seal. Everything that makes a flare joint work or fail is about whether that band is continuous.

Angle is a gate, not a preference

The two common flare angles are not interchangeable and they do not announce themselves. A 45 degree flare is the general tube and refrigeration convention under SAE J513, and a 37 degree flare is the higher-pressure hydraulic and instrument convention under SAE J514. Put a 45 degree tube cone against a 37 degree seat and the contact band collapses to a narrow line at one diameter instead of spreading across the face. It will often seal on a pressure test and then leak in service, because a line contact has almost no tolerance for the small movements a band absorbs.

Check the angle on the fitting before you set the tool, every time you are working with fittings you did not buy yourself, and treat a mixed box of loose fittings on a truck as unknown until measured. This costs seconds and it is the single failure on this list that no amount of good technique downstream can recover.

Why forming is the whole game

Forming a flare takes a straight tube wall and stretches it radially outward into a cone. Three consequences follow, and all three are physical, not procedural:

• The wall gets thinner. The cone material has been spread over a larger circumference than it started on. How much thinner depends on tube diameter, starting wall, and temper, so take the acceptable range from the tube and tool manufacturer rather than from a remembered figure.
• The material work hardens. Cold deformation raises hardness and drops ductility in exactly the region that now has to yield against the seat and then flex for the life of the system.
• Any surface defect gets opened. A scribe from a dull cutter wheel, an inside burr, a die mark from the tool's cone, a scratch from a dropped tube end. Stretching the wall pulls those open into a leak path across the seat, or into a crack starter at the cone root.

That is why the tool matters more than the wrench. An eccentric, rolling flaring cone works the surface progressively and burnishes it, and it produces a visibly smoother, more uniform cone than a straight screw-down cone at the same size. It is also why temper matters: soft, annealed tube flares cleanly, and hard-drawn tube resists the deformation and splits unless it has been annealed first.

The narrative: two visits, the same joint, a worse leak each time

The signal. A split system was called in low on charge. The tech found bubbles at the liquid-line flare at the outdoor unit, snugged the nut about a quarter turn until the bubbles stopped, topped the charge, and closed the call. Six weeks later the same system was low again at the same joint. The previous interval between top-ups on that system, from the service history, had been about five months, so the leak rate after the quarter turn was roughly four times what it had been before it.

That ratio is the first real piece of evidence and it eliminates the obvious explanation. A joint that was simply under-tightened gets better when you tighten it. This one got worse, and by a factor you can point at.

Hypothesis: the nut was under-torqued. Killed on inspection. With the nut off, the seat carried a complete, continuous witness ring around the full circumference, so the cones had made full contact. An under-torqued flare shows a partial or faint witness ring, or a ring interrupted on one arc. This one had been fully pulled up on the first assembly, and the quarter turn had gone past that into extruding the cone material outward.

Hypothesis: mismatched angle. Killed. Both the fitting and the tool were the 45 degree convention, verified against a gauge rather than by eye, and the witness ring's width was consistent with a matched pair.

Hypothesis: debris on the seat. Killed. The seat was clean and the witness ring showed no interruption or crater where a particle would have held the cones apart.

What was true. With the cone cleaned and inspected under magnification, there was a hairline crack running from the root of the cone, where the flare meets the straight tube, part way around the circumference. Three things had stacked to put it there. The tube was hard-drawn and had not been annealed before flaring. It had been cut with a dull wheel that left a rolled edge, and the burr had been reamed out hard enough to thin the lip further. And the line was clamped 30 inches away from the joint, so every bit of compressor vibration between the clamp and the fitting was reacted at the one place in the assembly with thinned, work-hardened metal and a geometric change of section.

The crack was the leak. Tightening the nut did not close it, it opened it further, because pulling the cone harder against the seat puts the root of the cone in tension. That is the mechanism behind the four-times number: the correction applied on the first visit loaded the actual defect in the direction that grows it.

Confirmation. Two independent checks. The removed cone was dye checked and the crack showed as a continuous line, not a scratch. And the repair held: the joint was re-made with the tube cut with a sharp wheel, deburred without thinning, annealed, formed with a rolling flaring cone, torqued to the fitting manufacturer's published value with a torque wrench rather than by feel, and clamped within a short distance of the fitting with a vibration loop between the clamp and the compressor. It ran a full cooling season with no measurable loss.

What would have changed the conclusion

• A witness ring interrupted on one arc would have pointed at seat contamination or a non-round cone, not at fatigue, and the repair would have been a re-flare rather than a re-flare plus a support change.
• A leak rate that dropped rather than rose after tightening would have pointed at simple under-assembly, and the finding would have been a torque discipline problem rather than a forming and support problem.
• A rigid, non-vibrating source changes the third factor entirely. The same thinned, hardened cone on a static water or instrument line may sit there for years, because nothing is cycling it. Vibration is what converts a marginal flare into a cracked one, which is why the identical technique produces failures on one system type and not on another and why crews get away with it until they do not.
• A tube that must run hard-drawn for support reasons does not get flared. That is a case for a different joint, and the tolerance comparison for the alternatives is a separate article.

How to verify you got this right

1. Gauge the seat angle on the fitting and set the tool to match, before the first flare of the day on unfamiliar fittings.
2. Cut with a sharp wheel and deburr without thinning the lip. A dull wheel work hardens and rolls the edge inward, and the reaming needed to clean that up removes wall you need.
3. Slip the nut on before forming. This gets its own line because forgetting it means cutting off a good flare, and a re-cut shortens the tube, which pulls the run tighter and moves you toward the support problem above.
4. Inspect the cone before assembly, off the fitting. Look for a uniform surface, a complete circular lip, no radial scoring, no split at the lip, no visible crack at the root. Reject and re-make; there is no acceptable repair to a cracked cone.
5. Torque to the fitting manufacturer's value with a wrench. The manufacturer owns this number, not the code and not the shop's habit, and the value is specific to size, material and whether a sealing washer is used. Feel is not a method here because the correct value is close to the value that extrudes the cone.
6. Support the tube within a short distance of the joint, with the manufacturer's or the equipment maker's spacing where one is published, and put the flexibility between the support and the vibration source rather than between the support and the flare.
7. Leak check to a method appropriate to the fluid, and record the interval. The interval between top-ups is data. A shortening interval after a repair means the repair loaded the defect, and the next step is removal and inspection, not another quarter turn.

References

• 40 CFR 82.154(a), prohibition on venting refrigerants during service, maintenance, repair or disposal, and 40 CFR 82.161, technician certification, for any flare opened on a refrigerant-bearing system
• SAE J513 (45 degree flare tube fittings) and SAE J514 (37 degree flare tube fittings), which reach you through the equipment manufacturer's specification or your project documents rather than as regulation
• Tube, fitting and flaring tool manufacturer documentation for temper requirements, acceptable cone dimensions, assembly torque, and support spacing
• See related: What a Compression Joint Depends On; The Mechanical Pipe Joints and What Each One Tolerates; Cross-Threading Prevention