Grain direction for sheet metal: when rolling direction matters

A louvre blade in 3 mm SS 304 cracked along the outside of a 90° bend at R3. The same drawing, at the same radius, had run for two years in 2 mm CRCA without a single reject. The grain direction for sheet metal decided the outcome, and nothing on the drawing said which way the coil was oriented. The blade had been nested with its bend line along the rolling direction, and the outer fibre was being stretched across the grain at 33% strain.

The fix took one line on the drawing and a change in the nest. It also cost a day of production, because the first 40 blades were already cut and the laser program had to be re-nested before the rest could run.

Grain direction for sheet metal: which way the bend line should run

Run the bend line perpendicular to the rolling direction. That puts the outer surface in tension along the rolling direction, where the coil was rolled, where the inclusions are elongated, and where the mill certificate's elongation figure is measured. A bend line parallel to rolling direction loads the sheet across the grain and will crack at a radius that works in the other orientation.

The mechanism is in the inclusion stringers. Rolling stretches manganese sulphide and oxide inclusions into long bands parallel to the rolling direction. Under tension along those bands, a micro-crack has to cross them, and the material necks and work-hardens ahead of the tip. Under tension across them, the crack opens along the band boundary with far less energy. The same steel, the same thickness and the same radius can therefore behave as two different materials.

Anisotropy also shows in yield strength. The transverse yield is usually a few percent higher than longitudinal, which is why a strip that bends cleanly in one orientation springs back more in the other. That difference is small in absolute terms, about 0.5–1.5° on mild steel, and large enough to fail an angular tolerance on a long flange.

The mill certificate does not settle the question on its own. Elongation is normally measured on a specimen cut parallel to the rolling direction, and the standard test methods require the specimen orientation to be reported. A coil certified at 28% elongation is telling you what the material does along the grain; it says nothing directly about a bend that stretches the sheet across it. That is why the certificate is a raw-material check and the bend trial is the process check.

Coil memory is the second effect. A coil that has been wound and unwound carries residual stress, and a part cut with its long axis along the rolling direction keeps more of the coil curvature than one cut across it. On a 1,500 mm panel that shows up as a 2–4 mm bow after cutting, before any bend, and the bow then biases the bend angle.

What the strain calculation tells you, and what it does not

For a 90° bend, the outer fibre engineering strain is:

ε = t / (2R + t)

where t is thickness and R is inside radius. On 3 mm sheet at R3 the strain is 3 / 9 = 33%. At R6 it falls to 3 / 15 = 20%. At R9 it is 14%.

That number is purely geometric. It does not know which way the grain runs, so it cannot tell you whether 33% is safe. What the direction changes is the strain the material can survive before the outer surface breaks up. The calculation narrows the range you have to test; the direction decides where in that range the limit sits.

Use it as a screening tool. If a design asks for 30% outer fibre strain on a 1.5 mm stainless visible panel, the radius is aggressive whatever the orientation, and the grain note will not save it. If it asks for 12%, orientation still matters for surface finish but the crack risk is low.

Minimum inside radius by material and direction

The values below are for 90° bends, air formed or press formed, on annealed or as-supplied stock, with a sheared or laser-cut edge. They are starting points for a first article, not a substitute for a bend trial on the actual coil.

The right-hand column is where the table meets this plant: a 200 T, 5-axis brake with a 4,450 mm bed, forming to 10 mm. A radius below the table is not a machine limitation, it is a material limitation, and no amount of brake setup removes it.

A worked example: louvre blade in 3 mm SS 304

The blade is 900 mm long, 60 mm wide, with a 90° return along one long edge at R3, and the material is 3 mm 304 with a 2B finish on the visible face.

With the bend line perpendicular to rolling, the table allows 1t = 3 mm, and the design is at the limit. The 33% outer fibre strain is high but workable on annealed stock with a good cut edge. With the bend line parallel, the limit is 2t = 6 mm and the R3 blade is 3 mm under. It cracked, which is what the drawing asked for.

Three options, in order of preference:

Rotate the blank 90° in the nest. The bend line becomes perpendicular to rolling, R3 stays, and the part forms. On a 1,250 mm × 2,500 mm sheet the blade still nests at the same count, so there is no material penalty, but the 2B finish direction now runs across the blade instead of along it. On a visible louvre that is a cosmetic change, and the customer has to accept it.

Raise the radius to 2t. R6 brings the strain to 20%, which 304 handles along the grain. The return flange gets a visible 6 mm radius, the blade loses 3 mm of flat width, and the louvre spacing may need to change.

Change the material. A 430 ferritic grade at 1.5t, or a 304 with a softer temper, will take a tighter bend along the grain. That is a material cost and a corrosion decision, not a forming one.

The cheapest of the three was the first, because it cost nothing but a drawing note and a re-nest. It was only cheap after the reject, which is the usual order of events.

Why 45° is worse than either extreme

A bend line at 45° to rolling is not a compromise between the two safe cases. It is the case where springback varies continuously along the bend, because the angle between the bend axis and the grain changes across the part if the blank is not perfectly square, and because the two ends of a long bend recover differently.

The practical result is twist. A 900 mm bend at 45° can come off the brake with one end 2° open and the other end closed, and the flange is no longer square to the web. Straightening it in a press leaves residual stress that shows up as oil-canning after coating. If a 45° bend is unavoidable, form it in a fixture that holds both ends, and expect to check squareness on a surface plate rather than with a square.

Measure the twist before arguing about it. Stand the part on the web, put a dial indicator on the top of the flange at each end, and read the difference over the bend length. A 900 mm flange with 1.5 mm of height difference end to end is about 0.1°, which is inside most general tolerances but visible as a gap when the panel is fixed to a frame.

What Ganesh runs

Cutting is a 12 kW fibre laser with a 40 ft × 16 ft bed, plus 6 kW and 3 kW support machines, cutting mild steel, CRCA, stainless, aluminium, brass and copper up to 20 mm on the laser. The bed size matters for grain because a long blank can be nested without splitting the part across two sheets, which is where orientation control usually breaks down.

Forming is a 200 T, 5-axis CNC press brake with a 4,450 mm bed and Delem control, to 10 mm. Tooling comes from the in-house tool room, so a punch radius that matches the table can be made rather than approximated from the standard set. Welding is three CLOOS robotic MIG cells and more than 30 manual MIG, spot and SPM stations, with fixtures built in-house. Coating is seven-tank pretreatment, an 18 ft × 14 ft powder booth and a curing oven, with DFT and cross-hatch adhesion checks.

Typical sheet metal tolerance is ±0.5 mm, tighter on critical features by agreement. PPAP and first article inspection are available, and the first article is where a grain-direction assumption gets proved or discarded. The bending capability page lists the brake and tooling, and the minimum bend radius guide gives the material floors that sit behind the table above.

What to put on the drawing

• A rolling direction arrow on the flat blank, with the text "RD" and the arrow pointing along the coil length.
• A note on each bend: "bend line perpendicular to RD" or, where the design cannot comply, the radius that applies.
• Inside radius per bend, with the direction-specific value where the bend runs along the grain.
• Whether the blank may be rotated in the nest, and if so, the minimum radius that rotation is allowed to create.
• Cosmetic face callout, because rotating the blank changes the visible polish or brush direction on stainless and aluminium.
• Material condition: annealed, half hard, H32, T6. The temper moves the radius more than the thickness does.
• Squareness or twist tolerance on long bends, not just the angle.
• A note that laser-cut edges are acceptable as bent, or that the edge must be deburred before forming.

The rolling direction arrow is the single highest-value note. Without it, the nesting decision is made by software optimising yield, and the first article becomes the test.

When this route is wrong

If a part needs a bend line along the rolling direction at a radius below the table, the answer is a different material condition, a stress relief after forming, or a redesign. We do not run in-house annealing or stress relieving, so a part that depends on a post-form heat treatment is outside the route here. A 6061-T6 bracket at 1t along the grain is the same case: it needs annealed stock or a different alloy, not a cleverer die.

Grain direction also stops being the controlling variable on heavily formed parts, where the tool geometry and the blank-holder force decide whether the material flows or tears. Those parts belong in a press tool, not on a brake. The deep drawing design rules cover that case.

Send the drawing with the grain arrow, the bend radii and the cosmetic faces marked to request a quote. We will confirm the nest rotation and the punch radius against the coil before the first blank is cut. More design notes are in the design and CAD hub, and the stainless bending guide covers the surface damage that comes with the radius decision.