When a part does not fit, the argument usually starts before anyone has separated the two things that went wrong. Size error is a feature being too big or too small. Form error is a feature being the right size in the wrong place, or the wrong shape, or not flat. A hole can be exactly the right diameter and still be in the wrong position by enough to stop a bolt going in.
The distinction matters because the two errors are fixed in different ways. Size error on a machined feature is corrected by the tool. Form and position error on a formed part is corrected by the tooling, the sequence or the fixture, and sometimes it cannot be corrected at all. Knowing which one you are chasing tells you whether to call the machinist or the toolmaker.
Forming is very good at some things. A bend angle repeats closely once the tooling is set, because the tooling sets it. Wall thickness in a pressed or hydroformed part is controlled well enough for the application. The overall shape of a bent tube is repeatable, and the relationship between two bends made in the same operation is stable.
Forming is also good at producing a surface that is smooth and free of tool marks, which is worth something on a visible exhaust component. That is a real advantage of the process rather than a consolation prize, and it is the reason formed parts are not simply replaced by machined ones wherever the budget allows.
Forming cannot hold the flatness of a face that has to seal. The face is produced by a tool that is not flat, from material that springs back unevenly, and then it is welded to something that pulls it. Gasket sealing needs flatness, and flatness is a machining property.
Forming cannot hold the position of a hole relative to a face on the other side of the part with any confidence, because the two features are produced in different operations and the material between them moves. Nor can it hold concentricity between an inside diameter and an outside diameter after a bend, because the bend ovalises both and does not ovalise them identically.
What forming can do is get close, and closeness is not the same as control. A formed part that is close enough to assemble will still produce a percentage of parts that will not, and that percentage is a property of the process rather than of the operator's attention.
A tube with several bends accumulates error. Each bend contributes its own angular variation, and the position of the far end is the sum of all of them. The far end can be a long way from where the model says it is, even when every individual bend is within its own tolerance, because tolerances add rather than cancel.
This is why a formed part that measures well at each bend can still fail at the flange. The flange is dimensioned from the far end, and the far end has moved. The machinist then has to decide whether to machine the flange where the drawing says, which produces a part that will not fit, or where the tube actually is, which produces a part that fits and does not match the drawing.
The honest resolution is to dimension the critical face from the features that locate the part on the car, and to machine that face last. Everything measured from a floating end is a guess dressed up as a tolerance.
The standard engineering answer is to give the formed part something to machine. A raised pad, a boss with stock on it, or a locally thickened section gives the machinist a surface that can be faced flat and drilled in position, without trying to hold a thin formed wall. The pad is designed into the part, and it is cheap at that stage.
The pad also solves the datum problem. It is a defined surface in a defined place, so the machine can be told to trust it, and the features cut from it inherit its position. What the pad cannot do is rescue a part whose overall shape is wrong, because facing a pad on a part that is bent the wrong way just produces a flat face in the wrong place.
A formed part is usually proved by assembly: bolt it to a fixture that represents the car and see whether it goes on. That is a functional check, and it is a good one, but it does not tell you where the error is when the part does not fit. A machined part is proved by measurement against a drawing, which tells you exactly what is wrong and by how much.
The two checks answer different questions, and a mature process uses both. Assembly proves the part will work. Measurement proves the process is under control. A shop that only does the first is always reacting; a shop that only does the second can pass every part and still ship a batch that will not bolt up.
A drawing that puts a tight tolerance on a formed surface does not make the surface accurate. It makes the part expensive, because the shop has to either machine a surface that was never going to be machined or argue about a deviation that does not affect function. Either way the buyer pays and the part does not improve.
The useful discipline is to decide, feature by feature, whether the requirement is a sealing face, a locating feature, a fastener pattern or a cosmetic surface. Only the first two usually need machining. The third needs position rather than size. The fourth needs nothing at all, and putting a tolerance on it is a way of transferring cost from the drawing to the invoice.
The split between formed and machined features is decided on the drawing, so it is best settled before the drawing is released rather than after the first parts arrive.
You can see how titanium parts of this kind are finished on our titanium welded parts pages, and how the machined and formed sides compare on the machined and fabricated parts pages, the second group of small components, the brackets and adapters, our titanium fabrication components and the aluminium welding work that sits beside them.
