Why Extruded Profiles Distort After Machining

A machined extrusion bends because the metal was already loaded before the tool touched it. Extrusion, quenching and stretching leave a stress field locked into the section; machining removes material from one side of that field and unbalances it. The part does not deform because of the cut — it deforms because the cut let the existing stress express itself.

That distinction matters because it changes the fix. If distortion were caused by cutting, slowing the feed or taking lighter passes would solve it. It usually does not. What solves it is either removing the stress before machining, keeping the material removal symmetrical, or accepting the movement and cutting to a datum that will survive it.

Where does the stress in an extruded profile come from?

Three places, in order of contribution.

Differential cooling at the die exit. The outer surface of the profile leaves the die and hits air or water mist while the core is still hot. The skin contracts first and the core contracts later, and because they are metallurgically bonded the result is a locked-in gradient: the skin in tension or compression depending on section, the core in the opposite sense. Thin-walled hollow sections with uneven wall distribution are the worst, because the thin walls cool in seconds while a thick boss behind them holds heat for minutes.

Quenching. 6xxx alloys are solution-treated at roughly 500–530°C and quenched to room temperature. On a press quench the cooling rate has to be fast enough to hold the alloying elements in solution, which means the surface cools far faster than the core. For a 3 mm wall the difference is small. For a 10 mm solid boss next to a 2 mm fin, it is large, and the section warps toward the mass.

Stretching. After quenching, profiles are stretched 1–3% to straighten them. Stretching does relieve a good part of the longitudinal residual stress, and this is why a stretched-and-straightened profile machines much more predictably than an as-quenched one. What stretching does not fix is the through-thickness gradient and the transverse stress, which is what a milling operation exposes.

Why does removing material cause a bend?

Think of the section as two layers in equilibrium: one in tension, one in compression, held in balance by the fact that they are attached. Remove a large fraction of one layer and the remaining material has to re-balance at a new curvature.

The magnitude scales with how asymmetric the removal is, not with how much metal is removed in total. A part that has 30% of its section taken off evenly from all faces will move far less than a part that has 15% taken off one face only.

A representative case: a 6061-T6 hollow section, 500 mm long, 40 × 40 mm, 3 mm wall, with a 12 mm deep pocket milled from one face. Release the clamps and the part typically bows 0.3–1.2 mm over the 500 mm. The same pocket split between two opposite faces bows an order of magnitude less, even though the same volume is removed.

Is the distortion from stress or from cutting heat?

Both happen, and they look different, so it is worth separating them before choosing a remedy.

Observation	Likely cause	What to check
Part springs when unclamped, before any temperature change	Residual stress release	Measure with clamps on and off; the difference is the stress component
Part grows during the cut and shrinks back overnight	Cutting heat	Measure immediately after machining and again after 4 hours at rest
Bowing along the length, consistent direction	Asymmetric material removal	Map where the most metal came off relative to the neutral axis
Twist rather than bow	Uneven wall cooling in the extrusion	Check wall-thickness spread on the incoming profile, not the machining
Dimension off but shape fine	Thermal expansion during gauging	Let the part stabilise before measuring; 10°C on 500 mm is 0.12 mm

That last row is worth pausing on. Aluminium expands at 23.4 × 10^-6 per °C. A 500 mm part measured warm by 10°C reads 0.117 mm long. A shop measuring parts straight off the machine and a customer measuring them in an air-conditioned office are not measuring the same thing, and this is a frequent source of "your parts are out of tolerance" conversations that have nothing to do with the machining.

How do you stop it?

Four approaches, in order of preference.

Start from stress-relieved stock. Specify T651 or T6510 rather than T6. The difference is a controlled stretch after quenching, which removes most of the longitudinal stress. It costs more per kilogram and is not available in every section, but it eliminates the majority of the problem before the machine is switched on. If a part has tight flatness over a long length, this is the first thing to ask for.

Keep removal symmetrical. Design pockets on opposite faces, or if the function requires a one-sided pocket, accept the bow and specify it. A drawing that calls for ±0.05 mm flatness on a 500 mm part with one-sided 12 mm pocket in T6 is asking for something the process cannot deliver without a straightening operation afterwards.

Rough, stress-relieve, finish. Take the bulk of the material off, let the part move, then finish-machine. On tight parts this is the standard route: rough to within 0.5 mm, leave it overnight or run a low-temperature stabilising cycle, then finish. It costs an extra setup and saves the scrapped parts.

Fixture against the movement. Where the part must be machined in one go, holding it on a vacuum plate or in a low-melt fixture reduces clamping distortion, but it does not remove residual stress — the part will still move when released. This is a mitigation, not a cure.

Can you stress-relieve after machining instead?

Sometimes, but it is the wrong order for a T6 part.

A stress-relief anneal for 6xxx sits in the 150–200°C range for a few hours, which is close enough to the artificial ageing temperature that it will over-age the alloy and drop its strength. A full re-solution-treat and re-age after machining is possible in principle but will introduce fresh quench distortion, so it usually trades one problem for another.

For parts that must be flat and must keep T6 strength, the practical answer stays upstream: buy stress-relieved stock, and design the removal to be balanced. Once the part is machined, the options shrink quickly.

Does the extrusion die design affect how much it moves?

Yes, and this is the part most drawings never consider.

A section with uniform wall thickness cools evenly and carries a low stress gradient. A section with a heavy boss on one side and a thin fin on the other cools unevenly, and no amount of press tuning fully removes the resulting gradient. The die also controls how the metal flows: an unbalanced flow leaves the profile straighter but more highly stressed, which means it looks fine on the rack and moves on the machine.

If a part is going to be machined to a tight flatness, the section design is part of the tolerance chain. Widening a thin wall, adding a rib to balance a boss, or moving a heavy feature toward the centre all reduce the stress before the first chip is cut. That conversation belongs in Profilformentwicklung, not in the machine shop.

Wo dies nicht gilt

This covers machined 6xxx extrusions in T5, T6 and T651 tempers. It does not cover:

Castings and forgings. Their stress fields come from solidification and forging reduction rather than from extrusion and quench, and the remedies are different.

7xxx alloys. 7075-T6 machined from plate has its own well-known distortion behaviour, and stress-relieved plate (7075-T651) is the normal answer. The magnitudes and the annealing windows are not the same as 6xxx.

Very short parts. Below roughly 100 mm in the bending direction, the movement is small enough that it usually falls inside normal tolerance, and the analysis above stops being worth the effort.

Parts where the bow is designed out. If the assembly accommodates movement — a slotted joint, a compliant mount — the flatness callout may be unnecessary in the first place.

As with any dimensional requirement, the value that matters is durch Ihre Zeichnung festgelegt, and what we can hold on a given section and temper is confirmed in writing before the order, not discovered at inspection.

Was Sie uns senden sollen

Send the drawing with the material removal marked and the flatness or straightness intent stated. We will tell you whether the section should be redesigned, whether stress-relieved stock is needed, or whether the flatness callout is tighter than the application requires. See also CNC milling and CNC precision machining for how the operations are sequenced.

E-Mail: linjunbing@truetrustlv.com  |  WhatsApp: +86 136 6644 0600

Sylvia Lin — Shanghai Truetrust Aluminum Industry Technology Co., Ltd.

FAQ

Does T651 really stop the distortion?

It removes most of the longitudinal stress, which is the component that produces bow along the length. It does not eliminate through-thickness stress, so heavy one-sided removal can still move a T651 part. It reduces the problem substantially; it does not make it zero.

Can I just machine it oversize and finish later?

Yes, and this is the most common fix. Rough with 0.5–1 mm of stock left, let the part settle, then finish. The extra setup costs less than the rejected parts, and it works on sections where stress-relieved stock is not available.

My parts were fine in winter and are out of tolerance in summer. Why?

Thermal. A 500 mm aluminium part differs by 0.117 mm over 10°C. If the shop and the inspection room are at different temperatures, or if the part is measured before it has cooled, the numbers will not agree. Check the measurement temperature before changing the process.

Does anodising after machining make it worse?

Not usually. Anodising is a surface conversion and does not relieve or add bulk stress, though the film itself can crack on a part that is flexed afterwards. If the part must be flat and anodised, machine to the finished dimension and account for the 10–25 µm film per surface.