What Thermal Movement Does to a Joint

Why this matters

"Thermal expansion" is the most over-used explanation in the trades and one of the least understood. It gets written on work orders for leaks it did not cause and left off work orders for leaks it did. The reason is a single missing idea: a joint does not feel expansion. It feels restraint. A run that grows a quarter of an inch and has a quarter of an inch of somewhere to grow puts essentially nothing into its joints, no matter how many times it cycles. The same growth with nowhere to go puts a load into whichever element in that load path is softest, and that element is very often not the joint the customer is looking at.

Most of what follows is about what thermal movement does not do, because the wasted repairs in this subject all come from misattribution rather than from ignorance.

Before opening a joint on a system that runs hot

Isolate, relieve pressure and drain before breaking any joint on a hot or pressurized run, and let it cool: hot water above its atmospheric boiling point flashes at the moment a joint cracks open, and a partly loosened union is not a way to bleed pressure. Lock and tag the isolation under 29 CFR 1910.147, which is the right standard for mechanical and stored thermal energy.

If the repair involves soldering or brazing, that is hot work: permit, combustibles cleared, and a fire watch during and after under 29 CFR 1910.252(a), with NFPA 51B applying in the edition your authority having jurisdiction or your insurer has adopted. Flux fumes and heated solder are an inhalation route, so ventilate and keep your head out of the plume rather than relying on gloves. If any brazing filler on the truck contains cadmium, its fume is acutely dangerous and is regulated under 29 CFR 1910.1027; the practical answer for field work is to carry cadmium-free filler and know which spools are which.

What thermal movement does not do

It does not load a joint that is free to move. This is the whole subject in one line. Free thermal movement is stress-free. A run that cycles 200 F twice a day for twenty years, with anchors, guides and an absorber doing their jobs, subjects its joints to essentially no thermal load at all. If a joint is failing from thermal movement, some part of the run is restrained, and finding the restraint is the repair.

It does not act only at the hot end. Expansion pushes; contraction pulls. Most joints in the trades seal by compression, whether that is a gasket squeezed between two flanges, a ferrule biting a tube, or an O-ring in a groove, and compression is the direction they tolerate. Pulling is the direction they do not. That is why a restrained run's joint typically leaks cold and typically on the first cold start after a shutdown, not at design load, which is the opposite of what most people expect and the reason the leak is often reported by someone who says "it only does it when it's off".

It does not fail on day one. This is a cycle-count failure. A run can be badly restrained from the day it was installed and behave for years, then start leaking within a season of somebody adding a night setback or a weekend shutdown, because the change added cycles rather than temperature. When a joint starts failing on a system nobody touched, ask what changed about the schedule.

It is not fixed by torque. Torque manages preload inside the fastener. It does nothing about the run's need to move, and a joint that was pulled open by restraint will be pulled open again at a higher preload. What preload does and does not control has its own article; the relevant point here is that adding torque changes the load path's stiffness, which usually moves the failure to the next-softest element rather than removing it.

What gets blamed on it and is not it

• Pressure thrust. An unanchored bellows or slip-type expansion joint transmits a force equal to system pressure times the joint's effective area, and that effective area is larger than the pipe bore and is published by the joint manufacturer. At ordinary system pressures this force runs into thousands of pounds, and it exists at constant temperature. It is a pressure phenomenon that happens to live at the device installed for thermal reasons, which is exactly why it gets logged as expansion damage. An expansion joint that pulls a run apart was almost always missing its main anchor.
• Water hammer. A transient of very short duration and very high magnitude, driven by valve or pump events. It cracks solder joints and rocks hangers, and it has nothing to do with temperature.
• Vibration loosening. Fasteners backing off under repeated small relative motion is a different mechanism with its own article, and it produces a slow leak at a threaded joint that looks a great deal like a thermal one.
• Differential expansion inside the joint itself. Two materials with different coefficients in the same clamp stack is a real and separate problem, owned by its own article in this library. That article is about what happens inside the joint. This one is about what the run outside the joint delivers to it. Diagnose which you have before you buy parts.

What restraint actually does: it finds the softest element

A restrained section develops a stress equal to the material's elastic modulus times its expansion coefficient times the temperature change, assuming a uniform temperature change on a straight member behaving elastically with no relief from bending. That stress is the same regardless of section length, so the force delivered into the load path is the stress times the metal cross-sectional area, and it too is independent of length.

That force then travels along the load path until it reaches whichever element yields first. Not the element that looks weakest. The one with the lowest actual strength in the direction the load is applied.

And here is the part that catches experienced people: on a soldered or brazed copper system, the softest element is frequently not the fitting and not the tube in general, it is the band of tube immediately adjacent to the joint, because the heat used to make the joint annealed it. Hard-drawn copper tube has roughly three times the yield strength of annealed copper. Making the joint created a soft ring around it, and the run then loads that ring every cycle.

The case: three fittings, one cause

A 1-inch type L copper heating branch in a rooftop mechanical room, 24 feet between two points where it was hard-clamped to structure at both ends. Idle in winter at about 55 F, service temperature about 140 F, so an 85 F swing. A fitting near one end had been replaced three times in two heating seasons, always found weeping, always weeping when the system was cold.

Free movement. Copper runs about 9.2 parts per million per degree F near room temperature. Over 288 inches at 85 F that is 9.2e-6 times 288 times 85, which is 0.225 inch. A quarter of an inch, on a run that had nowhere to put it.

Restrained stress. Copper's elastic modulus is around 17 million psi, so 17e6 times 9.2e-6 is about 156 psi per degree F, and 85 F gives about 13,300 psi.

Which element yields. Hard-drawn copper's yield sits around 30,000 psi and annealed copper's around 10,000 psi, both approximate and both varying with temper and temperature. The restrained stress of 13,300 psi is comfortably below the hard-drawn number and above the annealed one. So the straight tube between the clamps was never going to yield. The annealed ring beside the fitting was always going to.

The force involved. One-inch type L copper has about 0.169 square inches of metal. At 13,300 psi that would be about 2,250 pounds, except that it never gets there: once the annealed ring reaches its roughly 10,000 psi yield it flows, which caps the force at about 1,690 pounds. The soft ring is acting as a fuse, protecting the clamps and the structure, and paying for it with a little plastic deformation every cycle.

Why cold. Each hot cycle pushed the annealed ring into compression and it shortened slightly. Each cool-down left the run a little shorter than the joint spacing wanted, so the ring went into tension, and tension is the direction a soldered socket does not like. The weep appeared cold, cleared when the system came up, and made the customer sound unreliable.

Why three fittings did not fix it. Every replacement cut out the fitting and soldered in a new one, which annealed a fresh ring in the same place and reset the cycle count to zero. The repair reliably bought most of a season. The load path was unchanged, and the element that failed was never the part being replaced.

What the fix was and how they proved it

One of the two rigid clamps was converted to a guide: the same clamp body, a wear pad, and the fasteners set so the pipe is captured laterally and free axially. There was already a change of direction near that end, so 0.225 inch had somewhere to go the moment the run was allowed to reach it. Nothing was added to the pipe and no fitting was replaced.

They proved it with a scribe mark rather than by waiting. A mark on the tube against the guide bracket, made cold with the temperature recorded, then read at operating temperature: the run moved close to the computed figure. Had it moved a fraction of that, another restraint would still have been in the path, and the next place to look would have been the wall sleeve and the branch clamp on the drop.

Reading a joint that has been through this

Before you cut anything out, look at the tube rather than the joint.

• A slight barrel or wrinkle in the tube just outside the socket is plastic deformation from repeated compression, and it says the run is restrained.
• A circumferential crack right at the edge of the socket on a soldered joint, with the solder itself intact, is a fatigue crack in annealed base metal, not a joint failure. Replacing the fitting will reproduce it.
• Clamps at both ends of a straight section with no offset between them is the finding, and it is visible from the floor with a flashlight before any tools come out.
• Bright wear at an intermediate hanger means part of the run is already sliding, which tells you the restraint is elsewhere and narrows the search.

Photograph the failed section in place before you cut it. Once the tube is out of the run, the evidence about which direction it was being pulled goes with it.

References

• 29 CFR 1910.147 for isolating and controlling stored thermal and mechanical energy before breaking a joint on a hot or pressurized run.
• 29 CFR 1910.252(a) for welding, cutting and brazing fire prevention including the fire watch, with NFPA 51B applying in the edition your authority having jurisdiction or insurer has adopted; 29 CFR 1910.1027 for cadmium, where a brazing filler contains it.
• Tube, pipe and expansion-joint manufacturer data for temper-dependent strengths, metal areas and published effective areas for pressure thrust.
• See related: Thermal Expansion Mismatch Inside a Joint; How to Support a Run That Has to Move; Why Things Vibrate Loose.