Thermal Expansion Mismatch Inside a Joint

Why this matters

Everybody learns that a hot pipe run gets longer. Far fewer people work out what happens when two different materials are clamped together and heated as a unit, which is the situation inside almost every joint you build. The aluminum housing grows faster than the steel bolt holding it. The plastic body grows faster than the metal fitting threaded into it. The gasket sits between two members that are pulling apart at different rates. None of that movement is visible, none of it shows on a torque wrench, and it is the reason a joint you assembled correctly can arrive back at your bench loose, cracked, or weeping with nothing obviously wrong with any single part.

The rule that predicts the direction

Two members clamped together and taken through the same temperature change do not move the same amount. The one with the higher coefficient of thermal expansion tries to grow more. Because they are clamped, it cannot, so the difference turns into force.

• Heating a high-expansion member held by a low-expansion fastener raises clamp load. The housing swells against the bolt, the bolt stretches further, tension goes up.
• Cooling the same joint drops clamp load below where you set it, if anything in the stack yielded or took a set while hot.
• A low-expansion member held inside a high-expansion one goes into compression when hot, then loosens when cold. This is the ceramic-in-metal and glass-in-metal case, and the reason those parts crack on the cold side rather than the hot side.

Say the direction out loud before you calculate anything. Half of the wrong conclusions in the field come from getting the sign backwards and then hunting for a leak on the hot cycle when the joint actually opens on the cold one.

Coefficients you should recognize

Coefficient of thermal expansion is how much a material grows per unit of length per degree of temperature. The field unit is millionths of an inch per inch per degree F. You do not need to memorize them, you need to recognize which class is roughly double which.

Treat these as class-level figures for reasoning. Specific alloys and specific compounds vary, and the manufacturer's data sheet is the number to use when the answer is close.

Where the differential has to go

Once you know the differential exists, ask the only question that matters: what absorbs it? There are four possible sinks, and every joint uses one of them whether you chose it or not.

1. Elastic stretch in the fastener. The best outcome. The bolt takes the movement as extra tension and gives it back on cooling. This is what a long-grip bolt and a spring washer are for.
2. Compression in a resilient member. A gasket or a sealing element squashes further. Fine while the gasket is still springy, useless once it has taken a permanent set.
3. Slip. The members move relative to one another. Fine if the joint was designed to slide, destructive if it was not, because it wears the faces and fretting begins.
4. Permanent deformation. Something yields. A soft boss crushes, a thread pulls, a brittle member cracks. This is the sink that consumes the joint.

The whole design question is moving the differential from sink four into sink one.

The joint movement worksheet

This is the artifact. Fill in every field before you commit to hardware. It takes about five minutes and it is the only thing that catches a mismatch before it is buried in a wall.

• Field 1 - Members. Every material in the clamped stack, in order, from bolt head to nut.
• Field 2 - Coefficients. The published coefficient for each member.
• Field 3 - Temperature swing. The highest and lowest temperature the joint itself reaches, measured or reasoned from the process, not the room.
• Field 4 - Constrained dimension. The grip length of the fastener, or the length of the run between anchors. Movement scales with this, so a long grip is a friend.
• Field 5 - Free movement per member. Coefficient times dimension times temperature swing, one line per member.
• Field 6 - Differential. The largest free movement minus the smallest.
• Field 7 - Sink. Which of the four sinks above is taking it, and whether you chose that or inherited it.
• Field 8 - Accommodation. What you are changing so the answer is sink one.

The worksheet filled in

An aluminum pump housing bolted to a steel adapter plate, running warm.

• Field 1: aluminum housing, steel adapter, steel cap screws, no gasket in the load path.
• Field 2: aluminum about 12.8 millionths, steel about 6.5 millionths per inch per degree F.
• Field 3: shop assembly at 70 degrees F, measured 250 degrees F at the housing on a sustained run. Swing of 180 degrees F.
• Field 4: grip length 2.0 inches, 3/8-16 cap screws.
• Field 5: aluminum free growth is 12.8 millionths x 2.0 inches x 180 degrees F, which is 0.0046 inch. Steel bolt free growth over the same 2.0 inches and 180 degrees F is 0.0023 inch.
• Field 6: differential is 0.0046 minus 0.0023, which is 0.0023 inch. Two and three tenths thousandths, on a joint you cannot see move.

Now convert the movement to force, because two thousandths sounds like nothing until you do. The clamped members are far stiffer than the bolt, so treat essentially all of the differential as extra bolt stretch; the real split follows the stiffness ratio and shifts some of it into the members, which makes this a conservative read rather than an exact one. Using published values for a 3/8-16 fastener - a tensile stress area near 0.0775 square inch and a steel modulus near 29 million psi - 0.0023 inch of extra stretch over a 2.0 inch grip works out to roughly 2,600 pounds of additional tension.

Compare that against what the bolt was set to. A grade 5 fastener of that size carries a proof load near 6,600 pounds, and a normal assembly target is around 75 percent of proof, so roughly 4,900 pounds. Add the thermal 2,600 pounds and the bolt is carrying about 7,500 pounds hot, which is about 114 percent of its proof load. The thermal contribution alone is roughly a 50 percent increase over the assembly preload.

• Field 7: the sink is number four. Something yields every hot cycle. On an aluminum housing it is usually the boss under the head rather than the bolt.
• Field 8: accommodation required.

The failure mode: the joint does not leak hot. Hot, it is clamped harder than you set it. It leaks cold, after the yielded material fails to give the movement back and the cold clamp load lands below what the joint needs. A tech called to a cold leak tightens it, which adds preload that gets consumed on the next hot cycle, and the boss loses a little more each time. Three or four visits in, the threads pull.

What changes the answer

Change the bolt material. Swap the steel cap screws for austenitic stainless at about 9.6 millionths. The differential against aluminum drops from 6.3 to 3.2 millionths per inch per degree F, so the free differential over the same 2.0 inch grip and 180 degree F swing falls to about 0.0012 inch and the added tension to roughly 1,300 pounds, half the steel case. That is the fix arithmetic, and it comes with a real cost: common stainless fastener grades have lower strength than a grade 5 steel bolt, so you cannot keep the same assembly preload target. Check the stainless proof load before you treat this as a free win.

Change the grip length. Both the free movement and the bolt's own compliance scale with grip. A longer bolt through a spacer stretches more for the same force, so the same differential produces less tension rise. This is why long through-bolts survive thermal cycling that short cap screws do not.

Add a resilient member. A conical spring washer or a properly sized gasket moves the differential into sink two. It works, and it has a shelf life: once the resilient member takes a permanent set at temperature, the joint is back to sink four with no warning.

Change the temperature swing. Insulating the joint or moving it away from the heat source cuts the swing directly, and everything above scales linearly with it. Halve the swing and you halve the differential.

Reading a joint that has already been through it

You can usually tell which sink took the movement by looking at the parts you removed.

• A bright annular witness mark under a bolt head or washer, wider than the head, means the members were sliding relative to each other. That is sink three, and it points to fretting wear on the mating faces.
• A dished or crushed boss, or a washer pressed into a soft face, is sink four on the soft member.
• A bolt that measures longer than its unused twin has yielded in tension. Compare against a new fastener from the same box rather than against a catalog length.
• A gasket that stays flat and takes a fingernail impression rather than rebounding has taken compression set. It was sink two until it stopped being resilient.
• Cracks radiating from a bolt hole in a brittle member, showing up on the cold side, are the low-expansion-inside-high-expansion case releasing.

Photograph the witness marks before you clean the faces. Once the mating surfaces are wiped, the evidence of which sink took the load is gone and you are back to guessing.

Checking you got this right

• Confirm the temperature you used is the joint's own temperature under sustained load, taken at the joint, not the fluid setpoint and not the ambient.
• Confirm the grip length is the actual clamped stack, including washers and any spacer, rather than the bolt's nominal length.
• Re-check that your differential used the difference between the two coefficients, not one of them alone. Using the higher figure by itself roughly doubles the answer and sends you chasing a problem you do not have.
• Re-torque a hot joint only after a full heat cycle and a return to a known temperature, and record which temperature you checked at. A value read hot and a value read cold on the same joint are two different numbers and neither one is wrong.
• If your accommodation is a resilient member, note it as a wear item on the record so the next visit checks it rather than trusting it.

References

• Published coefficient of thermal expansion data from material data sheets for the specific alloy or compound in the joint
• Fastener proof load and tensile stress area tables published under SAE and ASTM fastener grade standards
• Manufacturer assembly instructions for joints combining dissimilar materials, including any specified re-torque interval
• See related: Expansion and Contraction in the Field; What Temperature Does to a Material's Behaviour; The Torque Sequence That Matters