A bend can open slightly when the forming force is removed. This elastic recovery is springback. Angle compensation allows for it during forming so that the unloaded part reaches the drawing angle.

What springback is

During bending, deformation has elastic and plastic components. When the forming load is removed, elastic recovery can open the bend. Over-bending can allow for this recovery, but the unloaded result must be measured to confirm the target.

Recovery depends on material, geometry and forming conditions. A changed result calls for checks of the material, setup and measurement; it does not by itself establish a machine calibration fault or a material cause.

How yield strength affects springback

With comparable material state, geometry and forming conditions, higher yield strength generally means greater springback. High-strength steel and mild steel bent to the same angle with the same tooling may therefore need different corrections; no general correction in degrees follows from this comparison.

Coils of the same grade can respond differently. Compare the yield strength documented on their certificates or test records alongside the angle measurements; identifying a strength change requires that material evidence.

How radius relative to thickness affects recovery

Under otherwise comparable conditions, a larger R/t generally increases springback sensitivity. Here R is the actual unloaded inside radius and t the actual thickness. If R is unmeasured, do not record the punch nose radius as a measured part radius. A program copied from a similar part with a different radius needs revalidation.

Compare material grade and state, actual t and R, tooling and forming method together to explain the response. Flat-length compensation addresses a separate question: Bend allowance vs bend deduction

Why a fixed offset drifts

Keep the verified offset with its material, radius and tooling conditions. After a coil or setup change, make a fresh trial and check the unloaded result before reusing the setting.

Compensation methods

MethodWhat it changesBest useLimit
Over-bendingPunch travels past the nominal angleSimple air bending with a known correctionCorrection drifts with material and radius
BottomingPunch presses the sheet into the dieReducing springback at the bendNeeds greater force and matching tooling; changes the process and machine/tooling conditions and must be reassessed before use
Angle measurement with closed-loop correctionA verified angle-measurement and controlled correction system can adjust from readingsProduction where the angle is criticalNeeds a measurement step and a controlled loop
Tooling with adjustable crowningCompensates deflection along the bend lengthLong bends where the angle varies along the lineDoes not replace material-based angle correction

For a particular machine, confirm which compensation and sensing systems are installed and whether their readings are taken under load or after unloading. On a long bend, crowning addresses lengthwise variation while a separately verified angle correction addresses the overall error.

Separate a global angle error from a lengthwise error

Observed patternCheck before adjustmentRetained result
Similar unloaded-angle error along the bendActual thickness, lot change, tool setup and angle conventionRepeat measurements with unchanged settings before a global correction
Centre differs from both endsSame instrument and support at marked positions; machine/tool alignment and deflection reviewKeep position-specific readings; a single angle offset cannot describe the profile
Large scatter on repeated partsSupport, measurement timing, material identity and instrument repeatabilityDo not chase each reading with a new offset

For a straight bend, β is the included angle and α the bend rotation at the same load state: Δβ=βunloaded−βloaded; α=180°−β. Δβ describes observed recovery, not a controller input. Record a loaded angle only under safety instructions or confirmation from the equipment manufacturer for that equipment and measurement method. Without a confirmed safe method, do not collect it or enter loaded tooling to complete a record. Mark it unmeasured; a programmed angle is no substitute. An unloaded reading alone can be compared with the target but cannot calculate Δβ. These relations supply no universal compensation angle. [A1]

For a coil-change comparison, retain the material grade, state and lot; measured thickness and inside radius; rolling direction and strength where known, with their sources; machine and tool IDs, V opening, forming method and program versions. Link each reading to a specimen, measuring position, instrument, support and time after unloading. Include the target angle, drawing tolerance, datum and recheck requirements. Keep measured values distinct from estimates and mark missing values as unknown. For material context see Sheet metal materials

A measure-first routine

  • Align the target unloaded β, drawing tolerance, datum, measurement method and production settings.
  • Make a new trial specimen from the production coil; measure the unloaded part with an appropriate instrument, not the machine display.
  • Separate global error, lengthwise variation and scatter between specimens using the checks above.
  • Within the approved process, confirm the adjustment separately for the actual control system; do not directly enter the unloaded-angle error or Δβ. Record the adjustment and label starting settings as awaiting verification.
  • Recheck on a new specimen with a different ID; retain settings and readings before and after adjustment.
  • After confirmation to the agreed requirements for this batch, save its program version. Revalidate at the next coil or lot change.

The unloaded-angle error indicates deviation from the target, not the controller correction. Confirm its sign, units and mapping for the actual control system. Repeatability evidence across batches is needed separately before treating a setting as a reusable starting point; it never becomes a universal value. Recheck counts and tolerances follow the drawing and agreement.

What not to conclude from one part

One conforming specimen does not prove stability. Compare different specimens within the batch and positions along long bends; repeatedly bending the same specimen is not independent repeatability evidence. A setting transferred to a machine with different deflection needs revalidation. Confirmation for the current batch is separate from evidence for reuse across batches. Full steps are in the trial-bend guide below. For combined symptoms see Defect atlas

Summary

Start with unloaded recovery, compare material and geometry under consistent conditions, then measure, adjust and verify on new specimens. Retain the conditions with the batch program and revalidate after coil changes. The bending process card below develops the radius and tooling questions.