A straight bar in a vice is a stable object. A bent tube is a spring. It has been formed past its yield point, it has sprung back to where it now sits, and it carries the memory of that forming in every wall section along the bend. Clamp one end and the other end moves. Clamp the middle and both ends move. That is the central difficulty of machining anything that has already been bent, and it does not go away with a heavier fixture.
The practical consequence is that on a bent part the fixture decides the result more than the machine does. A shop with an ordinary mill and a well-made fixture will beat a shop with an expensive machine and a pair of generic clamps, every time, on this class of work.
Every bend is formed slightly past its target and released, and the release is not perfectly predictable. Material strength varies along a length of tube, wall thickness varies within its own tolerance, and the tooling itself wears. The result is that the position of the far end of a bend varies from part to part, in a distribution rather than at a fixed value.
This matters because the drawing usually dimensions from the cut end of the tube. If the end has moved, everything measured from it has moved too. A shop that trusts the cut end as a datum is trusting the least repeatable feature on the part. That is the mistake behind a lot of fixtures that work on the sample and fail on the run.
A clamp does not hold a tube still by magic. It holds it by pushing, and the push travels through the wall. On thin-wall tube, and on aluminium in particular, enough clamp pressure to stop chatter is also enough to make the section oval while it is clamped. The cut is then round in the fixture and out of round when the pressure comes off.
The second effect is preload. A part that has to be forced into position to be clamped is a part that is being bent while it is cut. The tool produces a feature that is correct in the clamped state and wrong in the free state, and the error is invisible until the part is released and measured on a surface plate.
The fix is usually not more clamping but better support: more contact area, softer jaws, a clamp positioned over a feature that can take the load rather than over a thin run of tube.
The most repeatable feature on a bent tube is the tangent point where the bend begins. That point is set by the tooling, and it repeats far better than the end of a saw cut. A fixture that locates on the bend and machines the ends will produce consistent parts even when the tube arrives slightly long or slightly short, because the length error ends up in the stock that gets cut away.
The same logic applies on the other axis. If the part has a flange at one end and a sensor boss near the other, the boss should be positioned from the bend and the flange machined last, not the reverse. Sequence in the fixture is as much a design decision as the fixture itself.
An unsupported overhang on a bent tube will chatter. The tube is stiff along its axis and floppy across it, and the cutting force pushes exactly where it is weakest. Chatter leaves a patterned surface, and the pattern is a record of the tube moving rather than the tool cutting.
Shops solve this in unglamorous ways: an adjustable jack under the free end, a tailstock with a plug in the bore, a sacrificial ring clamped near the cut. What matters is that the support is set to touch, not to force, and that it is set before the final clamping, so the support does not itself push the part out of position.
A split clamp bored to the tube's actual outside diameter is the workhorse. It gives full circumferential contact, it resists ovalisation, and it can be made in an afternoon. Its weakness is that it hugs the tube wherever the tube happens to be, so it locates on the outside surface rather than on the bend.
A vee block with soft jaws is faster to set up and easier to adapt, but it contacts on two lines and will mark soft material. An expanding mandrel from the inside is excellent for thin wall and for parts where the outside must not be touched, but it depends on the bore being round to begin with.
For awkward shapes, low-melt potting is the last resort: the part is set in a castable medium that supports it everywhere, machined, then released by melting the medium away. It is slow and it needs a tank, but it will hold a shape nothing else will. It is also the only one of these methods that supports a part with no good external surface at all.
If the feature sits on a straight run and the bend is far enough away, machining first is usually the better order. The part is rigid, the fixture is simple, and the machine sees a straight tube. The risk is what the bender does afterwards: a bend close to a machined bore will ovalise it, a bend near a slot will close the slot slightly, and a bend after a threaded section can pull the thread off axis.
Where the feature is close to the bend and the accuracy matters, the sequence that works is rough, bend, finish. The feature is opened up before bending with stock left on, the bend is made, and the finishing cut removes the distortion along with the stock. It costs an extra setup and it is the only reliable way to get both a tight feature and a tight bend in the same part.
Bent-tube machining is a fixture problem before it is a cutting problem, so the questions at the quotation stage are mostly about holding.
You can see how these parts look when they leave the bench on our pages for small machined components, the wider machined and fabricated parts range, the brackets and adapters that carry them, our titanium fabrication components, the aluminium welding output and the titanium welded assemblies built to the same drawings.
