What a Compression Joint Depends On
Why this matters
A compression joint keeps its seal in the permanent shape of a small metal ring that has been driven into the tube. Once it is set, that ring is part of the tube, its geometry cannot be measured without destroying it, and every subsequent decision anybody makes about that joint is really a decision about whether the shape survives. That is why the rules around compression fittings sound fussy and arbitrary until you see what they are all protecting: do not re-set one as if it were new, do not mix components between makers, do not side-load the tube, do not put one on tube harder than the ferrule. Four different rules, one underlying reason. If your shop keeps no record of how a compression joint was set, you have no way to know which of those four you are about to break.
Before opening a compression joint on any pressurised line: an instrument or gauge line holds pressure between the root valve and the instrument even after the root valve is shut, so close the root valve, open the bleed, and confirm zero on a gauge downstream of the valve you closed. 29 CFR 1910.147 requires stored energy relieved, disconnected or restrained before service, and a trapped instrument leg is exactly that. Never tighten a compression fitting that is leaking under pressure. Advancing the nut on a live joint can push the ferrule past its set position and release its grip on the tube, and a tube that lets go under pressure leaves the fitting at speed with the line's contents behind it. Depressurise, then work. Where the fluid is hot, let it cool and wear face and body protection rated for the service, because these joints are common on small-bore lines carrying things that scald or burn.
What the ferrule actually does
The nut advances, driving the ferrule into a tapered surface inside the fitting body. The taper forces the ferrule's leading edge radially inward until it swages into the tube wall. That single motion creates two seals and one grip:
- A seal between the ferrule and the fitting body, on the taper.
- A seal between the ferrule and the tube, at the swaged edge.
- A mechanical grip that resists the tube being pushed out by pressure.
On a two-ferrule design the work is divided: the front ferrule makes the seals and the back ferrule drives it, grips the tube, and acts as a hinge that keeps a small amount of tube flexure from being fed straight into the sealing edge. That hinge is the reason two-ferrule sets are specified where vibration exists, and it is the reason mixing a front ferrule from one series with a back ferrule from another is not a shortcut, it is a different mechanism.
Two things follow immediately. The tube has to be softer than the ferrule, or the edge cannot swage in and the joint holds by friction on a smooth surface, which is not the design. And the set is a plastic deformation of both parts, so the joint's condition after assembly is a permanent record of how it was assembled, readable only by cutting it out.
The three geometries that all have to be right
The tube end. Cut square, deburred inside and out, round, and free of scratches anywhere the ferrule will land or slide. A longitudinal scratch under the sealing edge is a leak path the ferrule cannot bridge, and a tube cut with an abrasive wheel arrives out of round, work hardened at the edge, and full of chips that go downstream into the instrument you were trying to protect. Use a tube cutter or a saw with the tube supported, then deburr, then blow through before it goes into the fitting.
The insertion depth. The tube has to bottom in the fitting before the nut is set. If it does not, the ferrule swages at the wrong axial position and the tube has no shoulder behind it, so the first pressure event moves the tube inside a ferrule that is already at its final shape. Mark the tube at the insertion depth and check the mark after assembly.
The pull-up. Compression tube fittings are set by a specified number of turns past finger tight, not by torque, because torque measures the friction between the nut and the body and the set is about the distance the ferrule travels. The number of turns is published by the fitting manufacturer for that series and size, and it is not transferable between series. After the set, a gap inspection gauge confirms the nut advanced far enough, which is the only non-destructive check that the ferrule actually moved.
The connection record
This is a five-line artifact per joint. It exists because none of the information on it can be recovered later from the joint itself.
| Field | Why it is on the record |
|---|---|
| Tube material, temper, OD and wall | Establishes that the tube is softer than the ferrule and thick enough not to collapse under the swage |
| Fitting series and ferrule count, plus insert if used | Series is the matched set; a soft or plastic tube needs the correct insert or the wall folds |
| Cut method, deburr and roundness confirmed | The three end-condition failures, checked at the moment they are correctable |
| Insertion mark checked after set | Catches a tube that was not bottomed, which is invisible afterwards |
| Turns past finger tight, and gap gauge pass | The set itself, and the only non-destructive proof of it |
| Re-make count | The one field that has to survive the technician who wrote it |
| Support distance to nearest clamp | Records whether the joint is carrying a bending moment it was not designed for |
The last two fields are the ones people leave blank and they are the ones that pay. A re-make count is worthless on a sticker that stays with the tech and valuable on a tag that stays with the joint or a line in the job record keyed to the connection number.
The record filled in: an 18-connection survey
A shop took over maintenance on a small pumping installation with 18 small-bore instrument and drain connections, all two-ferrule compression, none with any assembly record. They surveyed all 18 by the same method on the same day. Four were leaking, which is about 22 percent of the connections surveyed, not of the joints in the plant.
They cut out all four and reconstructed a record for each.
| Connection | Tube and fitting | Insertion mark | Turns and gap gauge | Re-make count found | Support distance |
|---|---|---|---|---|---|
| A | Correct series, correct temper | Correct | Gauge pass | 4 | 26 inches |
| B | Correct series, correct temper | Correct | Gauge pass | 3 | 30 inches |
| C | Correct series, correct temper | Short by a visible amount | Gauge pass | 3 | Clamped close |
| D | Tube harder than the ferrule | Correct | Gauge pass | 1 | Clamped close |
Three of the four had been re-made more times than the manufacturer's instruction for that series permits, and connection C had additionally never been bottomed. The fourth, D, had been set once, correctly, on the wrong tube: a harder grade that the ferrule could not bite. All four passed a gap gauge, which is the useful and slightly deflating finding. The gauge proves the nut travelled. It does not prove the ferrule found anything to bite into, and it cannot see how many times the ferrule has already been driven.
Two of the four, A and B, also ran more than two feet to their nearest clamp. On both, the cut-out ferrule showed a polished, fretted band at the bite mark rather than a clean swage line. That is the signature of a tube being worked back and forth in a ferrule that is no longer gripping it, and it explains why the re-made joints were the ones that failed rather than simply the ones that had been touched most: each re-make leaves a slightly less positive grip, and an unsupported span keeps testing it.
The rebuild replaced all four with new tube and new ferrule sets, corrected the tube grade at D, added clamps at A and B, and started the record. The same survey by the same method six weeks later found one leaking connection out of the same 18, about 6 percent. Both figures are counts of leaking connections out of the same 18 by the same test, so the comparison is like for like, and the remaining one was a joint that had not been rebuilt.
What the record caught that inspection could not. Every one of these four joints looked correct. Nothing about a re-make count, an unbottomed tube, or a hardness mismatch is visible with the nut on, and all four passed the only non-destructive check available. The record is not a paperwork ritual; it is the only instrument that reaches the failure mode.
What the record still cannot tell you. It cannot tell you the condition of the ferrule on a joint someone else assembled without one, which is why the first survey had to destroy four joints to learn anything. Inheriting a system with no records means the first opportunity to know anything about a connection is the first time you open it, and that is the moment to start the record rather than the moment to re-tighten and move on.
Where a compression joint is the wrong choice in the first place
The record assumes the joint belongs there. Four cases where it does not, and the reasoning is the same each time: the stored ferrule geometry cannot survive the service.
- Tube that must be broken regularly for calibration, cleaning, or seasonal removal. Every break spends a re-make. A joint designed to be opened repeatedly belongs at that location instead.
- High or resonant vibration with no way to shorten the support span. The hinge in a two-ferrule set absorbs a limited amount of flexure and the bite mark is a stress riser in the tube wall, so a tube that will be cycled hard belongs on a flexible connection with the compression joint outside the vibrating zone.
- Soft or plastic tube without the manufacturer's insert. Without a support insert the wall collapses under the swage, and the joint may pass a test on the day and creep for months afterwards.
- Anywhere the joint will be buried, encased, or otherwise unreachable. Compression joints are serviceable connections and their whole value proposition is access. Putting one where nobody can get to it takes the cost of the joint without the benefit, and a joint with no access is a joint with no re-make and no record.
Where any of those apply, the choice is a tolerance and access comparison across joint types rather than a better compression fitting, and this library covers that comparison separately.
References
- 29 CFR 1910.147, control of hazardous energy, for isolating and bleeding a trapped instrument or drain leg before opening a compression joint
- Fitting manufacturer documentation for the matched ferrule series, published turns past finger tight by size, gap inspection gauge use, permitted re-make count, tube hardness and wall range, and insert requirements for soft tube
- Tube manufacturer documentation for temper and hardness, which is what determines whether the ferrule can swage into it at all
- See related: What a Flare Joint Depends On; The Mechanical Pipe Joints and What Each One Tolerates; Why a Joint Fails at the Interface and Not in the Metal