The complaint is familiar on any air bend longer than 3000 mm: the two ends of the workpiece measure close to the programmed angle, while the middle reads one to two degrees different, and the deviation is reproduced shift after shift. The controller shows a single Y value, both cylinders report arrival, and the part still comes out curved in angle. This article separates the four mechanisms that can produce that result and gives a measurement route that identifies which one is active on a specific machine.

1. The practical problem

On a 3200 mm air bend the angle error is almost never uniform. It is either symmetric, with the deviation peaking at mid-span, or asymmetric, with one end high and the other low. Symmetric deviation points to deflection and compensation; asymmetric deviation points to ram parallelism, a Y1/Y2 offset, or a load that is not centred. A third pattern, a random scatter with one bad spot, points to tooling. Because a single angle gauge reading at one end of the bed cannot distinguish these, the first diagnostic step is always a five-point measurement across the bend.

2. Failure mechanism

In air bending the punch load enters at the two side frames while the die reaction is distributed along the bed. The upper beam and lower bed therefore behave as beams loaded across a span and deflect toward each other at mid-span, closing the punch-to-die gap where the workpiece is in the middle of the machine. Extra penetration at mid-span means a smaller included angle there. The second mechanism is geometric: if the two cylinders do not sit at the same height, the ram is a straight line that is not parallel to the bed, so the whole workpiece is bent progressively from one end to the other. The third is a change in effective die geometry, from a worn punch tip or a damaged die shoulder, which alters the required penetration locally.

3. Key parameters

Four numbers govern the size of the effect: the span between side frames, the bending force per metre, the beam section stiffness in the vertical plane, and the die opening. Deflection rises with the cube of the span, so the same tonnage per metre produces much more sag on a 4000 mm bed than on a 2500 mm bed. The die opening matters twice, because it sets both the required force and the angle sensitivity near the end of the stroke. Crowning setting is the parameter that cancels the calculated deflection; if it is left at a value carried over from another job, the machine will bend the ends correctly and the middle incorrectly.

4. Diagnosis and calculation method

• Bend a 1000 mm test strip at the same die opening. If the short strip is straight, deflection dominates; if it is already curved, examine parallelism.
• Measure the 3200 mm part at five points with a digital protractor zeroed on a surface plate, and convert the readings to a deviation from nominal.
• Estimate required crowning from the machine's own deflection data, or from beam theory using the published frame section; do not reuse a value from another job.
• Bring the ram down onto gauge blocks at both ends with the pump idling and compare indicated position to physical gap; this separates feedback error from structural deflection.
• Repeat the same bend with the load centred and 300 mm off centre; a change of more than about 0.5 degrees indicates a synchronisation or eccentricity contribution.
• Log oil temperature at the start and end of the test to rule out thermal drift before adjusting any mechanical parameter.

5. Prevention

For workpieces beyond 3000 mm, three specification choices matter more than the control brand: linear scale feedback on both cylinders, a crowning system that can be set as a function of load rather than in fixed steps, and a beam section designed for the span. Our 160T/4000 synchronised servo press brake is built with dual-scale feedback and a load-related crowning table for this reason, and the same architecture is available across our 160 tonne hydraulic press brake range. In production, the preventive routine is simpler than the diagnosis: record the first-off five-point measurement for every long job, keep the crowning table by die opening and material, and repeat the test after any tooling change.

6. FAQ

Is a two-degree spread over 3200 mm normal?

No. On a machine with working crowning and good parallelism, the spread across a 3200 mm air bend should be a fraction of a degree, with the exact figure set by the machine's published accuracy and the material. A spread of two degrees indicates an uncompensated mechanism, not the limit of the process.

Can the operator correct this with the angle correction function?

Only if the error is uniform. A correction shifts the whole bend, so it fixes a parallel offset and does nothing for a curved deviation, and applying one to a curved part usually makes the ends worse while the middle comes good.

Does a thicker workpiece reduce the problem?

It changes it. Thick plate requires higher tonnage and therefore more deflection, but the angle sensitivity to penetration is lower, so the net result depends on the die opening used. Both effects must be checked against the machine's tonnage curve rather than assumed.