A fabricated part and a machined part are made by opposite logics. Fabrication starts from stock that is already close to the finished shape — tube, plate, a laser-cut flange — and joins the pieces with welds. Machining starts from stock that is deliberately larger than the finished shape and takes the difference away. What differs is where the accuracy comes from, and what kind of error each route tends to leave behind.
The two failure modes are not interchangeable. A welded part that is slightly out of shape will usually still bolt up, and then leak at the flange face once it has been through a few heat cycles. A machined part that is out of shape will not bolt up at all. The first failure is quiet and expensive to find. The second is loud and cheap to find. Buyers who understand the difference ask better questions at the quotation stage.
A weld is a small local casting that shrinks as it solidifies. On thin-wall tube the shrink pulls the joint, and the pull is not symmetric, because one side of the joint usually has more metal than the other. A flange face that was flat on the bench is no longer flat once a tube has been welded to its back. A tube end that was round becomes slightly oval within the heat-affected zone.
None of this is a criticism of the welder. It is geometry. The heat input is local, the restraint is partial, and the metal cools from the outside in. A fabricator can control the sequence, the tacking and the fixture, and can bring the distortion down a long way, but the last part of it is not removable by skill alone. That residue is what a machining operation is for.
Every machining operation is built on a datum: a surface or a set of features that the machine is told to trust. On a casting or a billet part the datum is designed in. On a welded assembly it usually has to be created. That is the step buyers underestimate when they compare a fabrication quote against a machining quote, because the fixture and the datum are most of the difference.
In practice a shop establishes the datum from whatever the assembly does reliably. Sometimes that is the flange face itself, gripped in soft jaws and indicated until it runs true. Sometimes it is a pair of holes that were drilled in the same setup. Sometimes the shop leaves a machining pad on the tube, or a boss with extra stock, precisely so there is something honest to hold on to later. You can see the kind of small hardware that comes out of this work on our machined and fabricated parts pages.
If the drawing gives no datum and the assembly offers no reliable surface, the shop will invent one, and the part will be correct to that invention rather than to your intention. That is the mechanism behind a whole family of arguments about parts that measure right and fit wrong.
Most of an exhaust system does not need a machine tool at all. What needs one is the small set of features where a seal, a fastener or a sensor has to work: the flange face and its gasket groove, the seat of a v-band joint, the bore that receives a slip joint, the boss that carries an oxygen sensor, and any thread in the assembly. Those are the places where a leak starts or a fastener loosens.
That short list is worth keeping in mind when a drawing arrives covered in tolerances. A tube's outside diameter along a bend does not need to be held tightly to work. The face that presses a gasket does. Spending machining time on the first and not the second is the most common way to pay for precision and still get a leak.
Aluminium is soft enough to be forgiving on force and unforgiving on finish. It galls, it sticks to the cutting edge, and a built-up edge will score the surface it is supposed to be finishing. It also moves with temperature, so a part that measures well when warm can measure differently once it has cooled, which matters on a thin section.
Stainless steel is the opposite problem. It work hardens under a dull edge and a dwelling cut, so a timid approach makes the next pass harder than the last. It also smears rather than shears when the edge geometry is wrong, and the smear hides the true surface underneath.
Titanium concentrates heat at the edge because the chip carries so little of it away, and it reacts with the air and with the tool at temperatures the other two materials would tolerate. The same drawing, issued in three materials, is three different processes with three different costs. A quote that treats them as one material is a quote that will be revised later.
On a small part, the time the tool spends in the material is often a minor share of the time the part spends on the machine. The rest is fixture building, program proving, first-article checking and re-clamping. That is why the price of the second identical part is so much lower than the price of the first.
It also explains a pattern that surprises buyers: adding a machined feature to an existing part can be cheap if the part is already set up, and expensive if it forces a second operation, a second fixture and a second datum. The number of setups is a better predictor of cost than the number of features.
The first problem is an interrupted cut. A cutter that crosses from weld bead into parent metal and back again meets a hard-soft-hard sequence several times per revolution. The insert takes a shock at each boundary, and the surface left behind is not uniform. Shops deal with this by taking the weld down in a separate roughing pass, or by machining the joint away entirely and welding after machining.
The second problem is stress release. Removing stock from a welded assembly changes the balance of residual stress that the welds put in. The part can move after the cut, so it measures correctly on the machine and incorrectly on the bench the next morning. Thin flanges are the classic case.
Before a machined feature goes on a fabricated part, four or five points decide whether the result is repeatable or merely correct once.
You can see the range of small machined and fabricated hardware we already produce on our second group of small components, the brackets and adapters we build for custom jobs, our titanium fabrication components, the aluminium welding work and the titanium welded parts that usually sit next to them on the same bench.
