A dimensional check at the end of a job tells you whether the part is acceptable. It does not tell you why, and it cannot help the next part. A distortion record is different: it is a series of measurements taken at defined points, in a defined state, with the values written down so they can be compared with the previous run.
The distinction is the difference between passing a part and improving a process. Shops that measure at the end produce good parts when the welder is careful. Shops that record distortion produce good parts because the process is set up to make them, which is what a buyer actually needs on a repeat order.
Every measurement needs a reference, and the reference has to exist before welding rather than being invented afterwards. On a welded assembly the natural choice is a machined face, a locating bore or a marked point on a fixture, and it should be the same feature the customer will use.
If the datum is chosen after the part is welded, the measurement measures the part against itself and hides the movement. A flange that is used as the datum and also as the feature being checked will always read zero, which is the most common way a distortion record turns into a useless document.
The datum should also be marked on the part or on the job card, so the next operator measures the same way. Two datums produce two sets of numbers and an argument.
Length is the simplest and the most often skipped. A long welded body gets shorter, and the change is small enough that it is only noticed at assembly. Measuring overall length before and after welding gives a number that can be turned directly into a cutting allowance.
Diagonals cover squareness on frames and rectangular assemblies. Two diagonal readings, taken in the same plane, detect racking that an outline dimension cannot show. This is also the fastest check on a large part, which makes it the one most likely to be done.
Flatness covers flanges and gasket faces. A straight edge and a feeler gauge, or a surface plate and a dial indicator, both work, provided the method is the same every time. The number matters less than the trend.
Roundness covers tube ends, canisters and bolt circles. Two diameter readings at right angles detect ovality, which is the characteristic distortion of a closed shape. The difference between the two is the number worth recording.
Angularity covers brackets, flanges and mounting feet. An angle that changes by a small amount across a long part produces a large positional error at the end, so this is the measurement that predicts fit problems on the vehicle.
Measure cool and measure free. A warm part is still expanded, so its length reads longer than it will be. A clamped part is held in a shape it will not keep. Both errors are in the optimistic direction, which is why they survive in shops: the part passes and the problem appears later.
Measure at stages as well as at the end. The reading after tacking, after the root and after the cap separates the movement into parts and shows which stage is responsible. On a part with many joints this is the only way to know where to intervene.
Allow the part to settle. A welded assembly continues to move for a period after the arc stops, and a measurement taken immediately after the last bead is not the final geometry. Leaving the part to cool undisturbed before the check costs nothing and prevents a false reading.
A record needs the part identity, the datum used, the values before and after, the state at measurement, and the sequence that was welded. Without the sequence the numbers cannot be compared with a different run, because a different order produces a different result.
It also needs a direction, not just a magnitude. A flange that moved up by a small amount and a flange that moved down by the same amount have the same error size and completely different corrections. Sign conventions matter more than decimal places.
Photographs of the measurement setup are worth more than a long description. A picture of where the straight edge sat removes any doubt about what the number represents.
The point of the record is the next job. If a body consistently finishes short, the cut length gets longer by that amount. If a flange consistently lifts, the fixture gets a compensating step. Each change is small, and each one is justified by data rather than by a feeling about the last batch.
Records also reveal when a process has drifted. A sudden change in the numbers usually means a material lot, a fixture wear item or a change of welder technique. Without the record the drift is invisible until a customer complains.
On titanium the record matters even more, because correction after welding is limited. The allowance has to be established on samples and carried into production, and the only way to know it is still valid is to keep measuring.
A shop with a modest method applied the same way every time learns more than a shop with an excellent method applied inconsistently. The comparison that matters is between runs, so the method has to be identical for the comparison to mean anything.
That means the same instrument, the same datum, the same clamping state and the same point in the process. It also means the same person where possible, or a written method clear enough that the person does not matter.
Precision beyond what the process needs is wasted effort. A reading that resolves a fraction of the movement being studied is enough. Chasing more resolution than the welding can hold produces numbers nobody acts on.
If distortion matters for your application, the drawing should say which dimensions are checked and in what state, so the shop can build the record into the job.
How materials behave under welding heat is compared on shop by material, and the route from drawing to approved sample is set out on customization. Welded stainless assemblies sit under stainless steel exhaust systems, tube under exhaust pipe, and titanium fabrication under titanium fabrication components. The plant and its capacity are described on about us.
