Press brake crowning: what it corrects and when a machine needs it

A 2,500 mm bend in 3 mm CRCA comes off the brake at 90.0° at both ends and 90.6° in the middle. Nothing is worn and the tooling is new. The punch and the bed are straight with the machine idle; under 61 tonnes of bending force they are not, and the part records the difference to the tenth of a degree.

Press brake crowning is the compensation for that deflection. It raises the bed at mid-span by the amount the ram and bed sag under load, so the punch reaches the same depth into the die along the full length of the part. This article gives the deflection arithmetic, the three ways machines apply the correction, a test that takes twenty minutes with three coupons, and the cases where crowning is the wrong answer.

The bending and welded assemblies hub collects the rest of this cluster, from die selection to inspection.

What press brake crowning corrects, and why the ends are always right

In air bending, the angle is set by how far the punch tip travels below the line of the die shoulders. Whatever raises the die or lowers the punch at any station changes the angle at that station. Deflection does exactly that, and it does it symmetrically.

A press brake is a beam loaded in the middle and supported at the uprights. Load rises linearly with bend length, so a 500 mm bend pulls about 12 tf and a 4,000 mm bend on the same material pulls about 98 tf. Deflection follows the load on a fixed span, so the sag at mid-span grows with the length of the part. The punch sags down and the bed sags down; the tool gap at the centre grows by the sum, and the material there is bent less.

The size of the effect is set by the die opening, not by the machine's age. For a part resting on two die shoulders spaced V apart with the punch tip d below the shoulder line, the included angle between the legs is close to 2·arctan((V/2)/d). Differentiate that and the sensitivity at a 90° bend is near enough 2/V radians per millimetre of penetration. On a 24 mm V die that is 0.083 rad/mm, or 4.8° for every millimetre of stroke, and 0.48° for every 0.1 mm. On a 48 mm die it halves. That single number turns a deflection reading into an angle error, and it is the reason a wide die is more forgiving of a sagging bed.

Run the numbers for the opening example. A 2,500 mm bend in 3 mm CRCA with a tensile strength of 450 MPa on a 24 mm V die needs F = 1.42 × 450 × 2,500 × 3² / (1000 × 24) = 599 kN, or 61 tf. A brake of this size deflects roughly 0.12 mm at mid-span at that load. At 4.8° per mm, 0.12 mm is 0.58°, which is the 0.6° the part measured.

Bending length, tonnage and the angle spread it produces

The table works through one job family: 90° air bend, 3 mm CRCA at 450 MPa, 24 mm V die, a 4,450 mm bed with roughly 3,150 mm between the uprights. Deflection and angle spread are planning estimates for a machine of this class; the coupon test in the next section is what you measure on the machine you actually own.

Two readings of the table matter more than the rest. The first is that the tonnage stays comfortably inside a 200 T machine at every length, so capacity is not the constraint; the angle tolerance is. The second is the last row. A press brake is rated for its tonnage spread over the full bed, and squeezing 98 tf into 4,000 mm at mid-span loads the ram at close to the design limit per metre even though the total is half the machine's rating. Check the load per metre against the machine's chart, not the total against the nameplate.

The three ways a machine applies the correction

Manual wedge crowning

A row of wedges under the die rail is adjusted by hand, usually against a chart that plots wedge position against bending load. Resolution is typically 0.05 mm per step, which is enough for the 0.12 mm the example needs. The weakness is that the setting belongs to the operator and dies with the shift change, so the job card has to carry the wedge value, not just the angle.

Motorised and CNC wedge crowning

The same wedges, driven by motors and positioned by the control from the programmed tonnage, material and die. Resolution improves to about 0.01 mm and the value is stored with the program, so the next run repeats it. This is the version to specify on a new brake if the work includes enclosures or frames longer than 2 m with a tight angle callout.

Hydraulic crowning

Cylinders built into the bed push it upward against the load, and the pressure tracks the tonnage as the bend progresses. It compensates in real time, which matters when the material thickness varies across a sheet, but the correction is only as good as the load sensing that drives it. A hydraulic bed without a load cell is a manual system with a pump.

A fourth method is not crowning at all: a die machined with a crown ground into its underside, matched to one part length. It works for a single high-volume job and is a liability for everything else, because the crown is wrong the moment the length or the material changes.

Whichever system the machine has, the value it needs is a property of the job, not of the brake. Two jobs on the same bed, one at 61 tf over 2,500 mm and one at 98 tf over 4,000 mm, need different corrections, and a chart that lists one number per die opening is only accurate at one load. Build the correction table from the three-coupon test at two tonnages, plot the value against load, and store it with the program. That table is the difference between a crowning axis that is used and one that is left at zero because nobody knows what number to enter.

How to test whether a machine needs it

The test costs three coupons and twenty minutes, and it beats any specification sheet.

1. Choose the job with the highest tonnage and the longest bend in the next month's schedule.
2. Cut three coupons 150 mm long from the same sheet, and mark them left, centre and right.
3. Bend all three in one cycle at the same station positions on the bed, using the production die and the production angle.
4. Measure the included angle of each with a digital protractor on a granite surface, and record the material thickness at each coupon.
5. Compare the spread. Under 0.5°, run the job as it is. Between 0.5° and 1°, correct at the control if the machine has a crowning axis. Over 1°, or over half the drawing tolerance, the bed needs compensating before the job runs.
6. Repeat the test after any tooling change, because the correction value is tied to the load and the die, not only to the machine.

A second check isolates the machine from the tooling. Mount a dial indicator on a magnetic base on the bed, bring the punch down onto a test block at the centre at the working tonnage, and read the bed movement. Compare with the deflection the crowning chart assumes. If the two disagree, the chart is for a different tooling setup and the operator is compensating for the wrong number.

What Ganesh runs on long bends

Long work at Ganesh runs on a 200 T, five-axis CNC press brake with a 4,450 mm bed and a Delem control, forming material to 10 mm. The sheet metal bending capability describes the machine and its tooling, and the 4,450 mm bed article covers what the bed length opens up for long panels and frames.

The brake is the constraint on bend length, not on part length. A 4.5 m panel bent across its width is routine; a 6 m part that needs a single continuous bend is not, because the bed runs out before the part does. For those, the route changes to a segmented bend with a joint, a rolled section, or a design that moves the bend to a shorter edge.

Where the job's angle spread exceeds the drawing tolerance, the correction is applied at the machine and verified on a first article measured at three positions across the part, which is the same three-point check the coupon test uses. First article inspection and PPAP documentation are available on request, and the ±0.5 mm typical tolerance we quote for formed sheet metal features applies to position, not to bend angle; the angle tolerance is whatever the drawing calls out.

What to put on the drawing

The angle callout is the field that decides whether crowning is needed, and it is the one most often left to a general note.

• Included angle with an explicit tolerance, for example 90° ±0.5°, stated as included rather than leg angle so the inspector measures the same feature the toolmaker set.
• Bend length and position of the bend line. A 400 mm bend and a 2,500 mm bend on the same part need different corrections.
• Inside radius and the V die you expect, because the radius and the die set the tonnage and the sensitivity.
• Grain direction, if the leg must bend without cracking, so nesting does not rotate the part.
• Straightness of the leg if it matters, since ISO 2768-2 class K allows 0.8 mm on a leg between 1,000 mm and 3,000 mm and that is usually looser than the appearance requirement.
• The reference face for the angle measurement, usually the outside of the bend.

A drawing that says "bends 90°" and leaves ISO 2768-1 to govern has called up ±5 arcminutes, because the medium class gives ±0°5′ for an angle whose shorter side exceeds 400 mm. That is 0.083°, roughly a seventh of the error this article opened with, and no press brake holds it on a 2.5 m bend without a correction system and a controlled first article. The standard covers the case where nothing is indicated; it is not a tolerance anyone can hold by default on formed sheet.

When crowning is not the fix

Crowning compensates a machine. It does not compensate a process, and three cases get misdiagnosed as deflection.

If the part is tight at one end and open at the other, the bed is not sagging symmetrically. Suspect a ram that is out of level, unequal tool heights or a sheet with a thickness wedge across its width, and fix that first. If the legs are wavy along their length rather than the angle being wrong, the cause is the bend line running across a laser-cut edge with a hard spot, or a punch that is narrower than the bend, and the long panel guidance covers the sequence that avoids it.

And if the bend is short but the angle is inconsistent from part to part, deflection is not the variable. Material tensile strength varies between coils and along a coil, and a 30 MPa spread against a 450 MPa nominal changes the springback of a 90° air bend by about 0.2–0.3° at the same stroke. That is small against a ±1° callout and decisive against a ±0.1° one, and it is a material and setup control problem rather than a machine one; the press brake buying guide explains which control features address it.

Send the bend that will not hold its angle, with the material, thickness, length, radius and the angle tolerance, and the engineering team will say which machine it runs on and where the correction comes from. A quotation request with the drawing attached is enough to start.