Guide
Will a Steel Bearing Stay Tight in an Aluminum Housing Across the Operating Temperature Range?
An aluminum housing is a bearing seat that grows and shrinks with temperature at roughly twice the rate of the steel bearing ring pressed into it, so a fit that is correct on the inspection bench can loosen when the machine runs hot and squeeze the ring when it sits cold.
By Saola Manufacturing engineering team · Published · Updated
Short answer: not by default. Bearing makers say an aluminum or other light-alloy housing needs more interference than a steel one, and the mismatch can still open the fit at the hot end of the range. Decide the temperature range first, estimate the loss with the arithmetic below, then choose between a tighter fit, a steel sleeve or a tolerance ring, and write the choice on the drawing.
On this page
Why does a steel bearing loosen in an aluminum housing when it warms up?
Aluminum has a much higher coefficient of thermal expansion than bearing steel, so the housing bore opens faster than the outer ring grows. The interference you set at room temperature shrinks as the assembly warms.
Bearing-design data from Dynaroll lists 52100 chrome bearing steel at 12 ppm/°C and aluminum at 23 ppm/°C, which is close to a factor of two. The page does not name an aluminum alloy, so treat the aluminum figure as typical: the coefficient differs a little between wrought and cast alloys and rises with temperature. Take the number for your own alloy from its datasheet before you rely on it.
SMB Bearings states the practical consequence in one line: an aluminum housing expands more than a steel housing and therefore requires a greater interference fit. The same page says shaft and housing fits should always accommodate differences in expansion coefficients. JTEKT says the same thing in its fit-selection notes: when the housing is aluminum or another light alloy, relatively greater than normal interference is needed, and for thin-section housings greater than normal interference is needed as well.
The failure mode is not subtle. A manufacturer of tolerance rings notes in a white paper that a typical press fit or slip fit at room temperature can loosen at higher temperature in aluminum, magnesium or plastic housings, that the loss grows with bearing diameter, and that the result is outer-race slippage and eventually destruction of the bearing and the housing surfaces. The paper's example is a 6205 bearing, about 2.00 in outside diameter, in an aluminum housing: with a 0.0012 in interference fit the bearing became loose at about 170 °F.
How much interference do you lose across the temperature range?
Multiply the seat diameter by the difference in expansion coefficients and by the temperature change. For a 52 mm seat the free mismatch works out to roughly half a micrometer per degree, so a typical 60 °C swing is the same order as a normal interference fit.
The table is a first-order estimate. It assumes the whole assembly sits at one uniform temperature, that the ring and the housing expand freely, and the 11 ppm/°C difference between the two Dynaroll coefficients. In a real assembly the stiff steel ring restrains the aluminum, so the contact pressure changes by less than the free mismatch suggests, and how much less depends on wall thickness and the housing shape. Treat the table as a way to see whether the problem is a few micrometers or a few tens of micrometers, not as a design calculation. JTEKT notes that its own stress equations apply to steel only and that it should be consulted for other housing materials.
| Temperature change from 20 °C | 30 mm seat (µm) | 52 mm seat (µm) | 80 mm seat (µm) |
|---|---|---|---|
| +40 °C (hot, loses interference) | 13 | 23 | 35 |
| +60 °C (hot, loses interference) | 20 | 34 | 53 |
| +80 °C (hot, loses interference) | 26 | 46 | 70 |
| −40 °C (cold, gains interference) | 13 | 23 | 35 |
Read it with a fit class in mind. In typical cases, if the room-temperature interference you intend on a 52 mm seat is smaller than the loss in the +60 °C row, the ring can go loose at that temperature even before you count manufacturing tolerance on the bore. If it is much larger, the same arithmetic tells you what the cold end adds. Fit classes themselves come from the ISO 286 system, which is a paid standard, so this guide does not reproduce its values; your bearing supplier's catalog lists the recommended housing classes for each load case.
What happens in the cold, and can the fit be too tight?
Yes. Cooling closes the aluminum bore around the steel ring and adds interference, and too much interference can fracture rings or remove the running clearance inside the bearing.
SMB Bearings warns that excessive interference fits can cause high stress that may fracture rings. SMB Bearings also notes that an interference fit can reduce a bearing's radial play by up to 80% of the interference amount, so every micrometer of extra squeeze comes out of the internal clearance. A bearing that runs smoothly at room temperature can therefore turn rough, hot or noisy at the cold end of the range, especially if you have already picked a heavy interference to cover the hot end.
That squeeze between a loose hot fit and a tight cold fit is why the selection has to look at both ends of the range at once. A bearing used outdoors, in a freezer or in a vehicle may start at a temperature far below room temperature and warm up under load, so the interference swings through its whole range on every cycle.
JTEKT adds one more twist: a difference in temperature, as well as a difference in expansion coefficient, can change the interference at the ring seats, and it may sometimes increase the interference between outer ring and housing. Estimate both seats, not only the one in the aluminum.
What are the options: more interference, a steel sleeve or a tolerance ring?
There are three common ways to keep the ring seated: choose a tighter housing fit class, move the bearing seat into a steel sleeve, or let a tolerance ring take up the changing gap. Each one changes what has to be machined, so pick it before the drawing is released.
A tighter fit is the simplest and costs nothing in parts, but it narrows the window between loose when hot and too tight when cold. A sleeve moves the dissimilar-metal joint to the outside of the sleeve and gives the bearing a steel-to-steel seat; the sleeve then needs its own retention, such as a shoulder, a knurl or a cast-in design, and a second tolerance. A tolerance ring is a thin corrugated band whose waves act as springs, described in the white paper as retaining a bearing across a wide service temperature range; the paper's example reports the ring still holding axial retention at 270 °F where the plain interference fit had already let go.
| Option | Best when | Main risk | What changes on the part |
|---|---|---|---|
| Tighter housing fit class | Modest temperature range, small bearing, no thin walls | Too much squeeze when cold; bore tolerance is hard to hold on a thin wall | Tighter bore size and roundness on the housing |
| Steel sleeve or insert | Wide range, large bearing, repeated thermal cycles | Sleeve can creep in the aluminum if its own retention is weak | Sleeve ID, OD and the aluminum bore it sits in; an extra part |
| Tolerance ring | Wide range where a plain fit cannot cover both ends | Axial and torque capacity depend on the ring and the bore it is paired with | Bore size and finish per the ring maker; a groove or lead-in if needed |
| Axial retention only (snap ring, cover plate) | As a backup to any of the above | Holds position along the axis but does not stop a loose ring from turning | Groove or shoulder; not a substitute for a correct radial fit |
What goes on the drawing and in the RFQ?
State the bearing, the fit class or bore size and the temperature range on the drawing, and tell the supplier which seat is the datum for the bore. A bare "press fit" note leaves the supplier guessing about temperature and material.
Give the bearing part number, the housing alloy and the temperature range the machine sees in service, from cold start to the hottest steady state. Give the bore as a size with a tolerance class, and add cylindricity or roundness where the bearing maker asks for it. Add the temperature at which the bore is inspected: the 2016 revision of ISO 1 confirmed 20 °C as the standard reference temperature for dimensional specification, as the NIST summary of that revision states, so say the bore is measured at that temperature unless you need otherwise.
If the housing is a casting, say whether the bearing bore is cast to size or finished by machining. A cast bore usually needs finishing to reach a bearing-seat tolerance, so the usual route is to leave stock and finish the bore. The casting is made through partner plants we manage; the bore is then finished by CNC milling or CNC turning, depending on the shape. Ask the supplier to measure the bore after any later coating step, because those steps can move it.
For the sleeve option, draw the sleeve as its own part with its own fit to the housing, and show how it is retained. For a tolerance ring, copy the bore size and finish from the ring maker's drawing rather than from a general fit table.
How do you verify the fit before volume production?
Measure the finished bore on the first articles, then test one or two assemblies through the full temperature range with the bearing in place. The arithmetic only tells you whether the design is plausible.
On the first articles, check bore diameter and roundness at the bearing seat and record the housing temperature while you measure. Then run the soak that matches the machine's duty: hold the assembly at the hot end, then at the cold end, and check for ring rotation in the bore (witness marks, fretting dust or a bright polished band on the housing). Loose-ring failures can show up as fretting, so keep a sample of the first lot for inspection.
On the incoming side, make sure the supplier's measurement report names the instrument, the reference temperature and the number of parts measured. At production volume the question is not whether one bore passes, it is whether the process drifts, so ask for a bore measurement on a defined sample from each lot and agree the response when a reading falls near the limit.
Frequently asked
Does an aluminum housing always need a tighter fit than a steel one?
Both SMB Bearings and JTEKT say it needs relatively more interference, because the bore opens faster when warm. That is a starting point, not a rule: a very wide cold range or a thin wall can make the extra interference harmful, which is when a sleeve or a tolerance ring becomes the better choice.
Is the fit on the inner ring affected as well?
Yes. JTEKT notes that when the bearing runs hotter than the shaft, greater interference is required at the inner ring. Estimate each seat on its own materials and its own temperature.
Can I use the same housing alloy for the casting and the machined version?
The expansion coefficient differs a little between alloys and drifts with temperature, so check the datasheet for the alloy you use.
At what temperature should the bore be inspected?
State it on the drawing. Per the NIST summary, ISO 1 keeps 20 °C as the standard reference temperature for dimensional and geometrical specification, and a bore measured at a different temperature carries a thermal error that is easy to avoid by stating the reference.
When should I ask the bearing maker?
Whenever the housing is not steel or cast iron, the wall is thin, or the temperature range is wide. JTEKT states that its stress equations apply to steel and that it should be consulted for other materials, and the same advice applies to any bearing supplier.
Sources
- Bearing shaft and housing fits, SMB Bearings technical page, accessed 2026
- Bearing knowledge: fits (housing fit selection and Table 9-1), JTEKT Corporation, accessed 2026
- Tolerance Rings can save your design when mounting ball bearings (white paper), USA Tolerance Rings, 2016
- Thermal expansions of bearing mating materials affecting radial play, Dynaroll, accessed 2026
- The 2016 revision of ISO 1: standard reference temperature for the specification of geometrical and dimensional properties, NIST, 2016