Every dimensional check after welding depends on where you measure from, and the choice of that origin is made long before anyone picks up a gauge. On a fabricated exhaust or a welded tank, the datum is usually a machined face, a flange register or a pair of hole centres that survived the process without moving. Pick the wrong reference and the part will read out of tolerance when it is not.
The reason is that welding moves metal. Shrinkage pulls a joint toward the weld, and the amount depends on how much metal was deposited, how it was sequenced and how the part was restrained. A dimension taken from a surface that itself moved tells you about movement, not about the shape you were trying to verify.
A part measured after welding is not the part that was drawn. It is the drawing plus everything the process did to it. Between two points either side of a seam, the distance shortens. Across a closed shape, the form pulls in. Along a tube run with several joints, the errors accumulate in one direction and cancel in another.
Knowing that, an inspector reads the numbers against an expectation of movement rather than against a perfect nominal. A flange that sits slightly closer to its neighbour than the drawing says may be exactly what the welder predicted, and the question becomes whether the assembly still fits the vehicle or the mating part. Function decides, and the drawing is the tool for judging it.
A flange face is the most common place for a post-weld problem to be discovered, because it is the surface that has to seal. Weld shrinkage on one side of a flange pulls it out of flat, and the error is invisible until the part is bolted to something rigid. Checking flatness on the bench with a straight edge and a feeler gauge finds it in seconds.
Distortion of this kind is directional. The flange bows away from the side where the weld was laid first and pulled hardest. That is why the check is carried out in more than one direction, and why the reading is kept rather than simply passed or failed. A record of which way the face bowed tells the next job how to pre-set or sequence.
Tube assemblies are checked against a centreline more often than against individual coordinates, because the centreline is what carries the flow and what has to line up with the mating pipe. A run that is dimensionally correct end to end but kinked at a joint will pass a tape measure and fail a straightness check, and the kink is where the stress will concentrate.
The practical method is to support the part in free state, sight along it, and then confirm with a straight edge or a laser where one is available. Concentricity at a joint is checked with a dedicated gauge or by rotating the part against a fixed pointer. Both methods show the same thing: whether the upstream and downstream bores share an axis.
A steel tape reads to whatever the eye can resolve, and it stretches. A vernier caliper reads a local dimension well and says nothing about form. A dedicated go and no-go gauge answers one question quickly and cannot answer another. Choosing an instrument is choosing which question you are willing to leave unanswered.
For fabricated work the useful discipline is to match the instrument to the risk. Critical interfaces get the slow, direct measurement. The rest gets a gauge that confirms the part will assemble. What is not acceptable is using a fast check on a critical interface and writing the result up as though it were a full measurement.
A part measured in its fixture tells you it fits the fixture. That is worth knowing, but it is not the same as knowing it fits the vehicle. A restrained part can be pulled into shape by clamps and spring back the moment it is released, which is why free-state measurement matters on anything that has to bolt to something else.
The two checks answer different questions and both belong in a competent routine. In the fixture, you confirm that the geometry the welder was aiming at was achieved while the part was held. Out of the fixture, you confirm that the geometry survived release. Where the two disagree, residual stress is doing the work, and the part will move again in service.
A number on its own is not a record. The same figure can be good or bad depending on the datum, the instrument, the temperature of the part and whether it was restrained at the time. A useful record carries all of that, in a form short enough that someone will actually fill it in.
The habit worth building is to record the ones that moved. If a dimension is stable across a run, one confirmation is enough. If it drifts, every part gets a number, and the sequence of numbers is the most valuable thing the shop owns, because it shows the drift before the part goes out of tolerance.
Not every out-of-tolerance dimension is a reject. A part can be outside the drawing and still function, and a shop that rejects on the number alone wastes work that was never at risk. The judgement needs to be made by someone who understands what the dimension does, not by whoever is holding the caliper.
Our custom work is specified with these interfaces agreed up front, because "out of tolerance" is meaningless until someone says which tolerance protects which function.
Dimensional inspection only pays for itself when the numbers return to the people who can act on them. A finding that a flange bows consistently by the same amount is not a quality problem, it is a process instruction waiting to be written. Pre-set the flange the other way and the problem disappears.
That feedback loop is what separates inspection from sorting. Sorting removes bad parts; feedback removes the cause. Shops that only sort spend their time reworking, and the same finding reappears on the next batch, which is how a tolerance that nobody can meet becomes normal.
The same reasoning applies whether the part is a fabricated titanium assembly, a section of general metalwork or a batch of welded titanium parts. Geometry behaves the same way in every material; only the amount of movement changes.
Buyers can read a lot from how a supplier talks about measurement. A shop that asks which interface matters is thinking about function. A shop that quotes a tolerance without asking is reciting a drawing. Our additional parts and company profile pages describe the range of work where this measuring routine is applied.
