Tolerance & Fit Calculator
ISO 286 hole and shaft limits for the preferred fits — plus the thermal expansion term that quietly destroys interference fits and is in no fit table.
Two things people get wrong about fits
A tolerance grade is a rule, not a number. H7 is not "plus fifteen microns" — it scales with diameter. An H7 hole is +0/+15 µm at 10 mm, +0/+25 at 50 mm and +0/+35 at 100 mm, because the grade holds relative precision roughly constant rather than absolute. Quoting a fit without a size says almost nothing about how much slack you have.
Differential thermal expansion eats interference fits, and this is the one that bites. Aluminium expands at about 23 µm/m/K and steel at about 12, so an aluminium hub on a steel shaft has a hole growing at twice the rate of the shaft it is meant to be gripping. At 50 mm an 80 K rise opens the joint by about 46 µm — and a standard H7/p6 press fit at that size has a maximum interference of 42 µm. The fit is not merely reduced by running hot. It is entirely gone, at a temperature a gearbox or motor housing reaches with nothing wrong. That term appears nowhere in the fit designation, and it is routinely larger than the whole tolerance band it is acting on.
- The extremes pair up crosswise. The tightest condition is the biggest shaft in the smallest hole; the loosest is the smallest shaft in the biggest hole. Pairing them the obvious way gives a fit that looks fine on paper and interferes in a third of assemblies.
- A transition fit is a choice, not a defect. H7/k6 comes out with slight clearance on some assemblies and slight interference on others, by design. What you are buying is accurate centring, not grip — if the joint has to carry torque, a key or a pin is doing the real work.
- Hole basis is standard because holes are expensive. You get a bore from a reamer or a standard tool in whatever sizes exist, then grind the shaft to suit. Shaft-basis fits exist and are the right answer for drawn bar stock, but almost everything else is quoted H-something for that reason.
- A thin hub stretches instead of gripping. Contact pressure depends on the hub outside diameter as much as on the interference, so doubling the interference on a thin-walled part buys far less grip than it looks — while moving you closer to splitting the hub.
- Heat the hub, chill the shaft, or both. The same arithmetic that destroys a hot fit is how you assemble it: open the interference into clearance, drop the parts together, let it cool on. Doing both halves the temperature each part has to reach, which matters when heating would temper a hardened part.
- Limits are not a measurement. Both parts still have to be round and straight. A bore with lobes or 20 µm of taper does not behave like its diameter suggests, and measuring a part warm off the machine is its own error.
How to use
- Enter the nominal size and pick a fit from the preferred list.
- Set the hub and shaft materials — the pairing matters more than the fit does.
- Enter the temperature rise the joint actually sees in service.
- Check whether the fit survives the heat, then read the press force and assembly temperature.
Frequently asked questions
Why is H7 not just one number?
Because a tolerance grade is a rule rather than a value. An H7 hole is plus nought to plus fifteen microns at 10 mm, plus nought to plus twenty-five at 50 mm, and plus nought to plus thirty-five at 100 mm. The grade holds relative precision roughly constant, not absolute precision, because holding half a thou across a 400 mm bore is a completely different manufacturing problem from holding it across a 10 mm one. Quoting a fit without a size therefore says very little about how much slack you actually have.
Can a press fit come loose just from getting hot?
Yes, and it is the commonest way interference fits fail in service. Aluminium expands at about 23 microns per metre per kelvin and steel at about 12, so an aluminium hub on a steel shaft has a hole growing at twice the rate of the shaft it is gripping. At 50 mm diameter an 80 kelvin rise opens the joint by about 46 microns, and a standard H7/p6 press fit at that size has a maximum interference of 42. The fit is not reduced. It is entirely gone, at a temperature a gearbox or motor housing reaches with nothing wrong.
How do I design around differential expansion?
Design against the hot condition rather than the cold one, because the cold figure is the temperature you will never be running at. In practice that means either choosing a tighter fit so the loosest assembly still holds when hot, matching the materials so the mismatch disappears, or accepting the fit is only a locating feature and adding a key, a pin or a shoulder to carry the load. Matching materials is the cleanest answer whenever the design allows it, since it removes the term completely rather than budgeting for it.
What actually is a transition fit for?
Accurate centring. A transition fit such as H7/k6 comes out with a little clearance on some assemblies and a little interference on others, depending on where in their tolerance bands the two parts happen to land, and that is deliberate rather than a defect. What you are buying is that the part sits concentric with very little play either way. If the joint has to transmit torque, a transition fit is the wrong family and whatever key or pin is in there is doing the real work.
Why are fits nearly always quoted as H-something?
That is hole basis, and it is standard because holes are the expensive half. A bore comes from a reamer or a boring head in whatever sizes exist, so it is far cheaper to fix the hole at H and grind the shaft to whatever suits than to make a non-standard hole for every fit. Shaft basis exists and is the right choice when you are working from drawn bar stock that is already accurate, which is why it turns up in linear rail and hydraulic rod work, but hole basis covers almost everything else.
How much force does a press fit need?
It depends on the hub outside diameter about as much as it depends on the interference, which surprises people. A thin hub simply stretches rather than gripping, so contact pressure and press force both fall away sharply as the wall gets thinner. That also means doubling the interference on a thin-walled part buys far less grip than the arithmetic suggests while moving you noticeably closer to splitting the hub, so the bore hoop stress is worth checking against the hub material yield strength rather than assuming it is fine.
Should I heat the hub or chill the shaft?
Either works and both is better. The arithmetic that destroys a hot fit is exactly the arithmetic that assembles it: open the interference into clearance, drop the parts together, and let the joint close as temperatures equalise. Heating the hub is usually simpler, but chilling the shaft is preferred where heating would temper a hardened part, damage a seal or a bearing, or discolour a finish. Doing both halves the temperature change each part has to reach, which often turns an awkward job into an easy one.
Why is my measured fit different from the calculated one?
Because these are limits, not a measurement. Both parts still have to be round and straight: a bore with lobes from a three-jaw chuck, or 20 microns of taper down its length, does not behave the way its nominal diameter suggests, and neither number will match what a bore gauge reads at one position. Measuring a part warm off the machine is its own source of error, and at these tolerances a few degrees is worth real microns on anything above about 50 mm.
🔒 This tool runs entirely in your browser. Nothing you enter is uploaded, logged, or stored.