Resources
Thin-wall aluminium parts: what the drawing must say
Published 2026-08-20
The classic failure on a thin-wall part is not that a dimension cannot be held. It is that every number was right while the part sat in the fixture, and something moved the moment the clamps came off. That is rarely solved by finding a more accurate machine — it comes from material, fixturing and what the drawing did or did not say. Here is where the movement comes from, and the handful of notes that let a shop catch it while quoting rather than during delivery.
Distortion has three sources — know which one you are looking at
The first is residual stress the material brought with it. A profile leaves the press and cools and ages with uneven stress locked inside; T6 goes through solution treatment and a sharper quench than T5, and generally carries that stress more visibly. Machining a layer away breaks the balance the part was holding, so the part finds a new equilibrium — and bows doing it. No amount of machine accuracy prevents this.
The second is clamping force. A thin wall is already bent or pinched the instant the fixture closes, and you cannot see it while it is still clamped. The textbook version is a bore that measures round under load and springs oval when released. The third is cutting force and heat, worst on unsupported walls, where the tool pushes the workpiece away and the deflection lands directly in the dimension.
The split matters, because each has a different owner. Residual stress is answered in material condition and operation sequencing; clamping in the fixture design; cutting deflection in toolpath and support. What the drawing can do is tell the shop, before anyone cuts metal, where the part is thin, where it must not be gripped, and under what condition a given number is true.
If you add only one note, make it the measurement condition
The most valuable sentence on a thin-wall drawing is usually this one: are the critical flatness, profile and hole positions measured in the free state, or may they be measured restrained — that is, in a simulated assembly condition?
This is not wordplay. A long, thin cover panel may sit visibly bowed on a surface plate and pull dead flat once it is bolted to its base. If bolted-down is the only condition the part ever lives in, demanding free-state flatness means paying for a scenario that does not exist — and that payment shows up as extra fixturing, extra operations and scrap, all of which come back in the price.
The call-out does not have to be elaborate. Both the ISO and ASME systems have formal free-state notation, but at a quoting desk a plain sentence is harder to misread: 'flatness of surface A measured in the free state', or 'may be measured restrained: fastened to the mounting plate with M6 screws at assembly torque'. Write the restraint down — otherwise the first article has no single basis on which to pass or fail.
Can the shape hold itself up?
Extruded aluminium walls start at 0.7 mm, but extrudable and stable-after-machining are different questions. Stability comes mostly from shape: a closed cavity behaves better than an open U, a ribbed wall better than a plain one, a section with flanges or end plates better than one that runs uniform end to end. Where a thin wall has to be open, consider leaving a sacrificial tie somewhere non-functional and cutting it off at the end.
The other thing designers routinely forget is to leave something to grip. Drawings that optimise every wall down to its thinnest often end up with no location where the part can be held safely at all. A non-functional clamping pad — even one that gets machined off later — is far cheaper than the argument afterwards about who owns the witness marks and the bow.
If sacrificial tabs are acceptable, say so on the drawing, and say whether they must be removed and what the surface has to look like after removal. A shop that reads that note does not have to come back and ask; a shop that does not see it will usually assume the strict answer and price accordingly.
Use geometric tolerances instead of a row of wall thicknesses
The most common thin-wall drawing error is describing a surface with a dozen linear wall-thickness dimensions. Every one of them measures inside tolerance, and the part still will not seat — because a linear dimension controls the distance between two points, not the shape of the face they belong to.
What you want is flatness, profile and parallelism, referenced to stated datums. ISO 2768 part 2 (classes K and H) can carry the general geometric tolerances so the non-critical features are covered by default; but the one or two surfaces that actually decide assembly deserve their own call-out — together with the measurement condition from the section above.
Do not over-tighten in the other direction either. A purely cosmetic cover face held to the same flatness as the mounting datum turns straight into fixtures, extra operations and inspection minutes. On most thin-wall parts, only one or two surfaces genuinely have to be flat. Deciding which one is worth more than tightening all of them.
Three questions worth pressing your supplier on
Given a thin-wall drawing, a competent supplier does not answer with a price first. They raise process: how much stock is left for finishing, whether roughing and finishing are separated, whether material is removed symmetrically, and whether a stress-relief step belongs between them. Fixturing is the same conversation — vacuum, soft jaws or a dedicated fixture each move price and lead time.
So press on three points. Where is the thinnest section, and how do you intend to hold it? How are roughing and finishing sequenced, and is there stress relief in between? In what condition will the first article be measured, and does that match the note on my drawing? If a quote comes back noticeably low with no mention of the thin walls at all, ask again — the hours may have been estimated as though the part were solid, and that surfaces at volume, not at sampling.
| Check before you send | Why it matters |
|---|---|
| Thinnest wall identified on the drawing | The estimator sees the process risk without hunting through three views |
| Free state or restrained stated for critical features | Decides whether you pay for a condition the part never sees |
| Sacrificial tabs allowed, and removal stated | Gives the shop a legitimate place to hold the part |
| Flatness and profile instead of a row of wall dimensions | Linear dimensions control distance, not the shape of a face |
| Alloy and temper in full (6061-T6, 6063-T5) | Temper drives residual stress level and cutting behaviour |
| Final condition stated: bolted down or free-standing | Tells us whether the distortion has to be controlled at all |
When in doubt, send the drawing and we will settle it together
Thin-wall judgements depend heavily on the actual geometry: a 1 mm wall on a closed cavity and a 1 mm wall on a cantilever are not the same part, and words rarely capture the difference. In our flow, every drawing gets an engineering read before it is priced — where clamping will move it, which geometric call-out needs a dedicated gauge, which face would be better measured restrained — and the reply comes back with those suggestions attached.
So you do not have to lock the tolerancing scheme down before you talk to us. Put the function, the mating parts and the final service condition in a note, send the drawing, and we will decide the call-outs together. Distortion is always cheaper to argue about at the quoting desk than on the line.
Keep reading