The bracket came back from the laser shop 1.2 mm short. Not by a lot, but 1.2 mm on a part with four 90-degree bends meant the two mounting holes no longer lined up with the frame. The shop said the bend allowance was in the program. It was. The program used a K-factor of 0.5, the default in most CAD packages, and that was the whole problem.
The flat pattern length of a bent part is not the sum of the outside dimensions. When sheet metal bends, the neutral axis — the line where material neither stretches nor compresses — sits somewhere inside the thickness. Everything outside the neutral axis stretches. Everything inside compresses. The flat length equals the length along the neutral axis. If you assume the neutral axis is at the middle of the material (K-factor 0.5), you get the classic default. Real bends put the neutral axis closer to the inside surface, around 0.35 to 0.45 of the thickness for typical air bending.
The standard formula and where it bites
Bend allowance for a 90-degree bend, using K-factor:
BA = (π/180) x θ x (R + K x T)
For a 90-degree bend, θ = 90, so BA = 1.571 x (R + K x T).
Take 2 mm mild steel, inside bend radius 3 mm. With K = 0.5, the allowance per bend is 1.571 x (3 + 0.5 x 2) = 1.571 x 4 = 6.28 mm. With K = 0.38, which is closer to what air bending actually does, it’s 1.571 x (3 + 0.76) = 5.91 mm. The difference is 0.37 mm per bend. Four bends, and the part comes out 1.5 mm short. That matches the 1.2 mm the shop saw. The K-factor error stacked up across the bends.
The K-factor isn’t a material constant. It changes with:
- Bend radius to thickness ratio (R/T). Tight bends push the neutral axis inward.
- Bending method. Air bending and bottoming give different K-factors.
- Material. Stainless work-hardens differently than mild steel and springs back more.
- Tooling. The punch nose radius and die opening change the bend profile.
Why the CAD default is wrong more often than right
CAD software defaults to K = 0.5 because it’s a neutral starting point and it’s what the manual says. Real press brakes are built around tables that go back to before CAD existed. Most brake operators have a bend deduction chart taped to the machine, and it doesn’t come from the CAD package. The chart was built by measuring test bends on the actual machine with the actual tooling.
The reliable way to set up a sheet metal part in CAD:
- Cut a test strip of the actual material, say 100 mm long, in the actual thickness.
- Bend it 90 degrees with the tooling you’ll use.
- Measure the outside dimensions.
- Work backward to the bend deduction the machine actually produced.
- Enter that number, or the K-factor that reproduces it, into the CAD model.
It takes twenty minutes and it’s only valid for that material, thickness, radius, and method. Change any one and the number drifts. This is why shops that do the same bends for years have their own tables and why new shops fight dimensional fights on every first article.
What fixed the bracket
The bracket was 2 mm mild steel, air bent, with a 3 mm inside radius and four bends. We measured a test bend. The K-factor that matched the measured result was 0.36, not 0.5. The flat pattern was recalculated with that value. The part came out 0.1 mm from nominal on the next run, which the holes tolerated fine.
The other thing we changed: the holes. The original design had the mounting holes punched in the flat before bending. On a 1.2 mm-short part, the holes moved with the flat. Now the holes are punched after bending, using the bent part’s datum edges as references. That removes the bend allowance error from the hole position entirely. For parts with critical hole-to-hole distances that cross a bend, punching after bending is the only way to be sure.
There is one more trap worth mentioning. Some CAD packages offer both bend allowance and bend deduction, and they’re different numbers for the same bend. Using the wrong one in the wrong field silently offsets every bend. When the part is consistently off in one direction and the K-factor looks right, check that the field is actually the one the software is using. We’ve seen that one too.
The default K-factor of 0.5 is a guess, not a measurement. For 2 mm steel with a 3 mm radius, the real number is closer to 0.36. Test a bend, measure it, and put the actual value in the model. And if a part must hold dimension across a bend, punch after bending. The 1.2 mm short bracket wasn’t a laser problem. It was a 0.14 K-factor error multiplied by four bends.
