7 Cybelec Modeva 10S Secrets to Mastering Precision Bending

• 1
  Cybelec Modeva 10S material data drives springback
• 2
  Adaptive bending cycle corrects the angle in real time
• 3
  Offline programming removes the trial bend
• 4
  Macros turn part families into one program
• 5
  Tooling and clamping set the real accuracy ceiling
• 6
  Network data closes the quality loop
• 7
  Calibration goes past the basic setup routine
• 8
  When each Cybelec Modeva 10S technique earns its keep

Every metal recovers part of its bend when the punch lifts. Yield strength sets how much, thickness and inner radius set how far the recovery reaches into the section. A generic profile for "304 stainless" is a guess. The Modeva 10S material library stores tensile strength, yield point and elastic modulus per material, and the controller uses those numbers to compute an over-bend angle instead of letting the operator add a few degrees by feel.

The library only works if the numbers are real. Pull the mill certificate for the heat you are running, then bend three test coupons at the same thickness and grain direction. Measure the recovered angle with a probe or a laser sensor, back-calculate the correction factor, and write it into the profile. Two or three iterations usually settle it. After that the same job repeats without a trial bend.

Grain direction is the part most shops skip. A 90° bend across the rolling direction behaves differently from one along it, often by 0.5–1.5° on 304 or 5052. Store both directions as separate entries. The extra row costs a minute and removes a rework loop.

When material data is wrong, the symptom is consistent: first part good, tenth part drifting, or the same program behaving differently after a coil change. Fix the data before you touch the ram offsets.

• 1
  Store per-heat valuesTensile, yield and modulus from the mill certificate, not a handbook average.
• 2
  Split by grain directionSeparate entries for with-grain and cross-grain bends.
• 3
  Re-verify after coil changeA new heat can shift springback by a full degree.

The adaptive cycle measures the sheet while it is still in the die. A laser or mechanical probe reads the actual angle at a set penetration depth, the controller predicts the finished angle from that reading, and the ram position is trimmed before the bottom of the stroke. The correction happens inside the cycle, not after the part comes out.

This matters on long parts. A 3,000 mm bend in 6 mm stainless will show a crown effect, because the ram and bed deflect in the middle. Without compensation the center reads 2–3° open compared to the ends. Adaptive control with a multi-point probe corrects the ram position across the length, and the finished part lands within a few tenths of a degree.

Tool wear and temperature also move the angle over a shift. A die that has run 20,000 strokes sits differently from a new one. Adaptive correction absorbs that drift automatically, which is why long runs stay stable without an operator checking every twentieth part.

The limit is measurement geometry. If the probe cannot see the flange, or the part is too short for a stable reading, the adaptive cycle has nothing to work with. Below roughly 30 mm of flange height, fall back to stored correction factors and inspect more often.

Programming at the machine ties up the brake and invites scrap. The Modeva 10S accepts machine code generated offline from a 3D model, so the bend sequence, tool selection, collision check and backgauge positions are settled before the sheet is loaded.

The simulation step is the useful part. A box bend with four flanges has a correct order; get it wrong and the punch hits a previously formed wall. Offline software flags that collision in seconds. It also checks whether the flange will clear the tooling and whether the backgauge fingers can reach the reference edge.

Bend deduction and K-factor come from the model, not from a table on the wall. That keeps flat patterns consistent between the press brake and the laser or punch that cut the blank. When the flat pattern and the bend program disagree, the part is out of tolerance no matter how good the controller is.

Offline programming pays back on repeat families, not on one-offs. If a bracket runs four times a year with three bends, the setup cost of a full simulation may not be worth it. If it runs monthly with eight bends and tight flanges, it is.

A part family shares tooling, bend sequence and tolerance, and differs only in a few dimensions. User-defined macros let you write the sequence once with variables for length, flange height and angle, then call it with a parameter list.

The gain is not typing speed. It is consistency. When every variant runs the same logic, a change to the springback factor propagates to the whole family in one edit. Without macros the same fix has to be applied to a dozen programs, and one gets missed.

Macros also carry the checks. A conditional can reject a parameter set that would collide with the tooling or exceed the ram stroke, and it does so before the operator loads the sheet. That is cheaper than a crash.

Keep the macro library small and documented. A folder of forty undocumented macros is worse than no macros, because nobody trusts them and everyone writes a new one.

The controller can only correct what the tooling allows. A worn punch tip with a 0.2 mm radius error will not produce a sharp, repeatable inner radius, and the controller has no way to know the tool is worn unless you tell it.

Match the die opening to the material thickness. Air bending in a die opening of 8× to 10× thickness keeps the required tonnage reasonable and the inner radius predictable. Go below 6× and the tonnage climbs fast while the radius becomes sensitive to every variation in thickness.

Segmented tooling needs shimming discipline. A 0.05 mm height difference between segments shows up as an angle step at the joint. Measure the assembled tool line before a tight job and record the shim stack so it can be repeated.

Clamping matters more than most operators expect. If the punch is not seated fully, or the die is not centered under the ram, the bend is skewed along its length. The controller cannot compensate for a mechanical offset that changes with every setup.

• 1
  8×–10× thickness die openingPredictable radius and manageable tonnage for air bending.
• 2
  Check the tool lineMeasure assembled segments; record shim stacks.
• 3
  Seat and centerA skewed die cannot be fixed in software.

When the brake is on the network, every job writes a record: program number, material heat, tooling set, measured angles, operator. That record is what turns a controller into a quality system, because it links a finished part back to the conditions that made it.

The practical use is drift detection. If measured angles on a family start trending 0.3° high over two weeks, the data shows it before parts go out of tolerance. Without records, the same drift shows up as a customer complaint.

It also shortens setup on repeat work. The next operator pulls the last run's parameters instead of rebuilding them, and the first part is usually good. On a job with ten bends, that can save most of a shift over a year.

Keep the data honest. A recorded value that came from an operator override is not a measurement, and mixing the two makes the trend useless.

The startup routine checks the obvious: ram position, backgauge repeatability, angle sensor response. That is the floor, not the ceiling. Precision bending needs a schedule tied to how hard the machine works.

Check ram parallelism under load, not just at zero. A brake that reads level unloaded can deflect unevenly at 80 percent of rated tonnage. Put a dial indicator or a laser across the bed at working load and record the profile.

Verify the backgauge over its full travel. Repeatability at one position says nothing about accuracy 800 mm away. Check three positions along each axis and log the deviation.

Recalibrate the angle sensors against a certified reference at a fixed interval, and after any crash, tooling change or controller update. The interval depends on duty cycle; a brake running three shifts needs it more often than one running one shift.