Application Angle Detection Compensation in the Flexion Machine

A bending program fixes three things: ram depth, die opening, and punch radius. The material decides the rest. Yield strength, thickness, grain direction, and the amount of work hardening from earlier bends all shift the final angle. On a 2 mm 304 stainless part, a 0.1 mm thickness change can move the result by more than 1°.

Springback is the largest single error source. The sheet elastically recovers after the punch lifts, so an overbent 90° target may settle at 88.5°. Machine deflection adds to this. Under 60 tonnes of force the ram and bed flex, so the middle of a 2,000 mm part bends shallower than the ends.

A fixed-depth program assumes all of this stays constant. It does not. Coil thickness drifts within a single batch, and laser-cut blanks from different nests can vary in hardness. The operator then adjusts depth by hand, checks with a protractor, and adjusts again. That loop costs time and still leaves scatter.

The system measures the bend as it forms instead of after it is finished. Laser or camera modules sit on the ram and track the flange position across the bend line. Small deformation or contact sensors can also be used on tight geometries where optics cannot see the flange edge.

The controller takes that reading, compares it with the programmed angle, and calculates a correction. On a repeat part, the correction is applied to the next stroke rather than the current one. That is the key detail: the first part sets the baseline, and parts two onward run at the corrected depth.

Closed-loop correction typically brings a bending cell into the ±0.25° to ±0.5° range on mild steel and aluminum, assuming the tooling is in good condition. The exact figure depends on the machine, tooling, and how stable the incoming material is. An angle detection compensation system does not fix worn dies or a bent punch.

The industrial PC records every correction. That log is useful when a downstream weld fixture starts rejecting parts, because you can see whether the angle drifted or the position drifted.

• 1
  Sensor placementTwo modules cover most parts up to 2,000 mm; longer beds need a third.
• 2
  Loop speedMeasurement and correction complete inside a normal stroke cycle.
• 3
  Baseline partThe first piece of a run is usually scrapped or set aside for setup.
• 4
  Not a fix for toolingWorn dies shift the neutral axis; software cannot compensate that away.

High-volume parts with tight angle tolerance are the clearest case. If a bracket needs 90° ±0.5° and you run 5,000 pieces a month, manual adjustment produces scrap and consumes press time. The compensation loop removes most of the operator intervention.

Long parts benefit too. A 2,500 mm channel bends differently in the middle than at the ends because of ram deflection. A multi-point measurement catches that bow and corrects it, which hand gauging at the ends will miss.

Mixed-thickness nests also justify it. When a laser cuts blanks from 1.5 mm and 2.0 mm sheet in the same batch, the fixed program has one depth and the material has two behaviors. The loop adapts without a program change.

One-off and prototype work rarely justifies the setup. A single bracket checked with a digital protractor takes less time than calibrating the sensors and defining the measurement window. Job shops that run mostly one-off parts should spend the money on tooling instead.

Parts with very short flanges give the optics nothing to see. Flanges under roughly 8 mm to 10 mm, depending on the sensor head, fall outside the field of view. Hemmed edges and return flanges can block the line of sight as well. Those parts need a different strategy.

Very thick plate changes the calculus. Above about 6 mm in steel, bend radii and tooling loads dominate the result, and the angle scatters less from stroke to stroke. The compensation loop still works, but the gain over a well-set fixed program is smaller.

Material with heavy mill scale or a rough rolled surface can confuse laser triangulation. A quick test on the actual stock, not a sample from the shelf, is the only reliable way to confirm the sensor will hold signal.

State the angle and the tolerance separately. A note reading 90° ±1° tells the shop it can run a fixed program and inspect with a protractor. A note reading 90° ±0.3° tells the shop it needs in-process measurement or a very stable setup. Those are different quotes.

Call out the critical bend, not all of them. Most brackets have one or two angles that matter to the assembly and several that are cosmetic. Marking the critical ones lets the shop apply compensation where it counts and run the rest at normal speed.

Note the material grade and the incoming thickness range. If the drawing says 304 stainless without a thickness tolerance, the shop has to assume the full mill range, which is wider than most designers expect. Tightening that range usually costs less than adding compensation.

For machined-and-bent parts, say which operation sets the datum. A bend formed after milling has a different springback response than a bend formed from flat stock, because the machined section changes the stiffness. Mention the sequence on the RFQ and the shop can plan the setup.