Thickness tolerance and its effect on 304 stainless parts

A drawing that says "2.0 mm stainless" has not specified a thickness. It has named a nominal value inside a band that the mill is allowed to use in full, and every downstream number — flat length, weight, stiffness, weld gap — moves with the sheet. The parts that fail at assembly are rarely outside the drawing; they are inside it, at opposite edges of the band.

For 304 stainless thickness tolerance, the band is roughly ±0.09 mm at 2.0 mm in cold-rolled sheet. That sounds small until it is multiplied through a folding sequence. Each 90° bend develops 0.116 mm differently at the edge of the band, and a six-bend box accumulates 0.70 mm — more than the formed tolerance on most drawings.

The quality question is therefore not how to hold 2.00 mm exactly. It is which numbers the part depends on, and whether the thickness band has been converted into a flat-pattern and fit decision before the nesting run.

What the mill tolerance actually is

Two standard families govern the sheet. ASTM A480 sets thickness, flatness and edge tolerances for flat-rolled stainless in inch and metric practice; EN ISO 9445-2 covers continuously cold-rolled wide strip and sheet in Europe, with EN 10051 for hot-rolled material and EN 10029 for plate. In India, IS 6911 covers the same ground for plate, sheet and strip.

Each of them publishes a tolerance table indexed by thickness and width, and each offers more than one class. Normal tolerance is what arrives unless the purchase order asks otherwise. Special or tightened tolerance is a separate product, ordered with a lead time and a price, and it narrows the band rather than eliminating it.

Two properties of the mill product matter more than the number in the table. The first is crown: a cold-rolled sheet is slightly thicker at the centre than at the edges, so a measurement taken down the middle reads high and a measurement at the edge reads low. The second is that the tolerance applies to the thickness of the sheet as delivered, not to the thickness after blasting, pickling or coating, each of which removes or adds material.

The figures are the normal-tolerance band for the common width class at each gauge; the exact cell has to be read from the standard for the width and thickness actually ordered. Treat the table as the order of magnitude that a quality plan has to absorb, not as a substitute for the standard.

How the band moves the flat pattern

Bend allowance is the arc length at the neutral axis: angle in radians multiplied by (R + Kt). For 304 at a 2 mm inside radius, K is about 0.45. A 90° bend in a nominal 2.0 mm sheet gives 1.5708 × (2.0 + 0.9) = 4.555 mm. The same bend in a 1.91 mm sheet gives 4.491 mm, a difference of 0.064 mm.

The bend deduction moves further, because the outside setback carries the full thickness twice. Deduction is 2(R + t) minus the allowance: 3.445 mm at nominal and 3.329 mm at 1.91 mm, a 0.116 mm difference. Flanges of 40 mm and 30 mm therefore develop to 66.555 mm at nominal and 66.671 mm in the thin sheet.

If the flat is nested to the nominal 66.555 mm and the sheet arrives at 1.91 mm, the formed part comes out 0.116 mm undersize across those two flanges. Across a box with six bends the error reaches 0.70 mm, and the last bend carries all of it.

There is a second-order effect at the radius. In air bending, the inside radius follows the die opening until the punch tip radius exceeds the natural radius, which is roughly 0.16 to 0.20 times the V for 304. At a 16 mm V that is 2.6–3.2 mm, so a 2 mm punch sets the radius and the thickness variation changes the arc slightly without moving the radius much. At a 40 mm V the natural radius governs, and a thinner sheet in the same die does produce a smaller radius — which changes the allowance again.

Weight, stiffness and the load rating

304 weighs 8.0 kg per m² per millimetre. A 1,250 × 2,500 mm sheet is 3.125 m², so 50.0 kg at nominal 2.0 mm and 47.8 kg at the 1.91 mm low limit. The 4.5% difference is enough to change a two-person lift into a one-person lift, and enough to move the centre of gravity of a finished canopy.

Stiffness is the number that matters structurally. Bending stiffness scales with the cube of thickness, so a sheet 4.5% thin has (0.955)³ = 0.87 of the nominal stiffness, a 13% loss. Section modulus scales with the square, giving a 9% loss in bending strength. On a panel designed at 4 mm deflection, the thin sheet deflects about 4.6 mm.

That is why a load rating stated on a drawing is only meaningful with a thickness basis. A shelf rated for 150 kg at nominal 2.0 mm is not rated for 150 kg at the bottom of the band if the rating came from a nominal-thickness calculation with no margin.

Fit-up, weld gap and the coating stack

Thickness variation does not act alone. Three contributors stack in the same direction on an assembly.

The sheet contributes up to 0.18 mm across two parts at opposite edges of the band. Powder coating contributes 60–90 µm per face, so 0.12–0.18 mm where two coated faces meet. And the bend angle contributes its own error at the tip of a flange, which grows with flange length. A nominal 2.0 mm slip fit with 0.3 mm clearance can close completely on the worst combination.

Welding sees the same stack. MIG tolerates a gap up to about half the thickness on thin sheet before burn-through becomes likely; on 2 mm 304 that is roughly 1 mm, and a fit-up that varies by 0.2 mm across a seam is visible in the bead profile even when it passes. Fixtures close the gap rather than absorb it, which is why fixture design and thickness control are the same conversation.

Angle belongs in the same stack, and it is usually the larger term. The Delem control compensates springback, so what remains is variation rather than the nominal 2–3°: typically 0.2–0.5° per bend between parts. At a 100 mm flange that is 0.35–0.87 mm at the tip, which is more than the thickness contribution and explains why a box can be right at the first bend and out at the sixth.

A worked clearance check

A 2.0 mm 304 panel sits in a formed channel that is 2.4 mm deep inside, with both faces powder coated. At nominal thickness the panel is 2.0 + 0.12 = 2.12 mm with a 60 µm film per face, leaving 0.28 mm of clearance. At 2.09 mm the coated panel is 2.21 mm and clearance falls to 0.19 mm. At 1.91 mm it is 2.03 mm and clearance rises to 0.37 mm, which is fine for assembly but leaves the panel loose enough to rattle.

The fix is not a tighter sheet. Either the channel is formed to 2.6 mm to absorb the full band plus coating, or the panel is located by a feature that does not depend on thickness, such as a tab and slot. Both are drawing decisions, and both cost less than special-tolerance stainless.

Measuring thickness so the number is usable

A single micrometer reading in the middle of a sheet tells you the crown, not the sheet. The workable method is three readings per sheet, 25 mm in from each edge and one at the centre, recorded as a minimum and maximum against the heat number. A 0.01 mm resolution micrometer is adequate; an ultrasonic gauge is faster on a stack of sheets but needs a couplant and a calibration block of the same grade.

The mill certificate to EN 10204 3.1 gives the heat analysis and the tensile results. It does not certify the thickness of each sheet, because thickness is a rolling tolerance rather than a test result. If the drawing needs a thickness guarantee per piece, that is an inspection operation on receipt, not a certificate.

Record the range, not the average. The average hides the thin edge that produces the short flange, and the thin edge is what the brake and the welder actually see.

What the first article report should carry

The first article is the cheapest opportunity to correct the flat pattern. It should record the measured thickness at three points on the blank that became the part, the actual inside radius at the first and last bend, the formed outside dimensions across every bend in the chain, and the flat length before forming. With those four fields the K-factor can be solved back from the part rather than assumed from a table, and the corrected value goes into the programme before the batch runs.

A report that records only "dimensions within tolerance" throws that information away, and the second batch repeats the first batch's assumption.

What to put on the drawing

1. Thickness with a tolerance, or the mill standard and tolerance class — "2.0 ±0.09 mm to ASTM A480" or "2.0 mm, EN ISO 9445-2 normal".
2. Product form: cold rolled or hot rolled, because the tolerance and the surface differ.
3. Whether a 3.1 inspection certificate is required, and whether thickness is to be recorded per sheet.
4. Which dimensions are thickness-critical — weight, fit, stiffness — so the inspection plan can prioritise them.
5. Coating thickness and the faces it covers, so the fit stack includes the film.
6. Flat pattern basis and the K-factor assumed, so a thickness correction can be applied without renegotiating the drawing.
7. Flatness and twist limits over the part length, which are separate from thickness and are not covered by ISO 2768-1.

How Ganesh handles the band

Incoming 304 is checked at receipt against the purchase order and the certificate, with the thickness range recorded per heat. Nesting uses the actual gauge where a heat measures at the edge of the band, and the bend table is corrected from the first article rather than from the nominal sheet. First article inspection and PPAP are available, and the formed tolerance is typically ±0.5 mm with tighter features by agreement.

The route is the 12 kW fibre laser with a 40 ft × 16 ft bed, the 200 T 5-axis brake with a 4,450 mm bed to 10 mm, and three CLOOS robotic cells with more than 30 manual stations. Both plants hold ISO 9001:2015 and ISO 14001:2015, and manufacturing has run since 1992 across five units in Mumbai (Vasai) and Nagpur.

The sister articles cover the grade decision on 304 stainless sheet metal and the measurement of the formed result in bending tolerances and inspection. The materials and grades hub carries the equivalent thickness pages for the other grades.

When tightening the thickness is the wrong lever

A thickness callout is the wrong fix when the real variation is in the bend angle, the fixture or the weld sequence. A part that is 0.3 mm out at the flange tip after six bends is usually losing angle, not thickness, and ordering special-tolerance sheet adds cost to every kilogram without touching the cause.

It is also the wrong fix on a stiffness-driven part. If a panel deflects too far, the answer is a thicker nominal gauge, a rib or a folded edge, not a tighter band around a gauge that was never thick enough. The cube law means a 25% thickness increase buys almost twice the stiffness, which no tolerance class can match.

And it is the wrong fix where the assembly tolerance has already been allocated. Adding a special thickness tolerance to a part whose fit problem is a coating build of 0.18 mm solves the cheaper half of the problem.

Send the drawing with the thickness callout, the measured sheet range and the assembly it fits, and the review will say where the stack closes. Request a quote with the revision and the annual quantity.