The mechanism, in numbers
Take a steel bushing with a 50 mm bore, 2 mm wall, 40 mm long, clamped in a three-jaw chuck at 8 kN per jaw. The radial deflection at each jaw is roughly 0.03–0.06 mm, with a corresponding inward movement at 60° between jaws. Machining the outside diameter removes material in the deformed state, so when the jaws release, the bore returns towards round but the outside diameter keeps the three-lobed shape, inverted.
Measured roundness error of 0.02–0.1 mm is normal, and it gets worse as the wall gets thinner: deflection scales roughly with the cube of the wall thickness ratio, so halving the wall multiplies the error by about eight.
Why it does not show up on the machine
On the machine, with the part clamped, the bore is round — because the jaws hold it round. Everything measures well. The error only appears when the part is released, which is why this defect is so often discovered at assembly, by the customer, and not at inspection.
The practical test: measure the part clamped, release, measure again, and record the difference. If the difference is above one third of the bore tolerance, the clamping is consuming your tolerance budget.
What an expanding mandrel changes
A split collet expands evenly along a land that covers 60–100 % of the part length. The load is distributed over the circumference instead of three points, so the deformation is a uniform radial expansion rather than a three-lobed wave. A uniform expansion is still a deformation, but it is axisymmetric, and axisymmetric deformation does not destroy roundness — it only changes the diameter, which is compensated by the machining allowance.
The design details that matter: the number of petals (too few re-introduces lobing, too many will not grip), the land length relative to the wall (a short land on a thin wall creates a bulge at the end), and the clearance between collet and bore at the lower limit (0.01–0.03 mm, then expand).
Alternatives, and why they are worse
Soft jaws bored in place: better than hard jaws, but still three point loads. Reduces the error by about half, does not remove the cause.
Pie jaws / six-point: improves the situation to six lobes, which is finer and often acceptable, but the part is still unsupported between contact points.
Filling the bore with low-melt alloy: works, is slow, and creates a cleaning problem on the next operation.
Grinding after hardening: correct if the part is hardened anyway, useless if it is not, and does not fix the clamping distortion in the preceding turning operation.
| Clamping method | Typical roundness error | Springback after release |
|---|---|---|
| Hard three-jaw chuck | 0.03–0.10 mm | High, three-lobed |
| Soft jaws bored in place | 0.015–0.05 mm | Medium, three-lobed |
| Six-point / pie jaws | 0.008–0.02 mm | Low, six-lobed |
| Expanding mandrel, full land | 0.003–0.008 mm | Minimal, axisymmetric |
Questions
Can I just take a lighter clamping pass?
Does the mandrel need to cover the full part length?
What bore tolerance do you need from me?
Published by Liuzhou Dongben Machinery Equipment Co., Ltd. · More notes