Machining Welded Structures: Why the Sequence Decides
Knowledge · Large parts & steelwork
Why weld first and mill second?
Because welding heat destroys any previously created precision. The seam shrinks on cooling, pulls at the part and leaves residual stress; details in welding distortion. A mounting face milled before welding would afterwards be neither flat nor in position. The proven chain: cutting and bending, welding, stress-relief annealing at 550 to 620 °C for precision parts, then machining of all functional faces in as few set-ups as possible.
How much machining allowance does a welded structure need?
As a rule of thumb 3 to 5 mm on all faces to be machined, more for very large or distortion-prone structures. The allowance must cover both the welding distortion and the coarse tolerances of the welded assembly itself (ISO 13920). Too little allowance is expensive: if the face does not clean up after distortion, the assembly is scrap or needs weld repair. Too much allowance only costs cutting time. In doubt: one millimetre more.
Which tolerances can welded assemblies reach?
| Feature | As welded only | With machining |
|---|---|---|
| Flatness of a mounting face | millimetre range (ISO 13920) | tenths to hundredths, size-dependent |
| Hole-pattern positions | ±1 to ±3 mm typical | tenths range and tighter |
| Fits (bores, seats) | not achievable | H7 and finer possible |
| Parallelism of functional faces | millimetre range | tenths and tighter, best in one set-up |
Actual values depend on size, annealing state and clamping concept; on multi-metre parts the effects from large-part machining add on top, above all temperature and dead weight.
What matters for datums and clamping?
Consistency between drawing, production and installation. The drawing datum system should use faces that are actually machined first and on which the part later rests. The assembly is clamped the way it will be installed, so dead weight acts as in service. And wherever possible, related functional faces are machined in one set-up, because every re-clamp adds alignment error.
Why is single-source supply particularly valuable here?
Because the interface between weld shop and machine shop is exactly where accuracy is created. If the assembly travels between two suppliers, two parties argue in case of error about whether allowance, distortion or clamping was at fault, plus transport and schedule chains. From one source, work preparation plans welding sequence, annealing, allowance and machining as one process, and responsibility for the final dimension sits in one place. A practical tip: mark all faces to be machined unambiguously in the assembly drawing and dimension finished sizes from the datum system; the most common real-world defect is a drawing where nobody can tell which dimensions apply to the welded and which to the machined state.
Frequently asked questions
Why can functional faces not be milled before welding?
Because welding distortion makes them useless: the seam shrinks on cooling and distorts the part by a multiple of the later tolerance. Machining always follows welding, with sufficient allowance.
When is stress-relief annealing necessary before machining?
Whenever tight tolerances must hold permanently or a lot of material is removed. Without annealing, locked-in stresses can release during milling and distort the part afterwards, even days later.
How much allowance is usual on welded structures?
3 to 5 mm on functional faces as a rule of thumb, more on large or distortion-prone structures. The allowance must safely cover distortion and the coarse tolerances of the welded state.
What happens if the allowance is insufficient?
The face does not clean up: unmachined patches remain after milling. Depending on the case, straightening, weld repair with re-machining, or scrap. That is why allowance is set deliberately in work preparation.