Digital Twin and Simulation in Roll Forming Tooling Design
I'm Iris Xu, a Sales Manager in ZTRFM.
There is a myth in our trade that a new profile has to be cut in steel before anyone knows if it will form. It was true twenty years ago, when the first article was the test. It is not true now. The whole process can be simulated before a single roller leaves the CNC, and the shops that still skip simulation are paying for it in scrap and late deliveries.
What a digital twin of the line is
A digital twin here is not a pretty 3D model. It is a working copy of the line, the strip, and the tooling, linked to real data, that you can run on a computer. You feed it the coil grade, the thickness, the flower, and the stand geometry, and it returns how the metal will move, where it will thin, and where it will fight back. Change a roll angle and the twin shows the consequence in minutes instead of after a toolroom overnight.
How the simulation actually works
The tool most roll form designers know is COPRA, from data M Sheet Metal Solutions in Germany. COPRA RF handles the roll design and the flower, the step by step bend sequence from flat strip to finished section. COPRA FEA RF takes that design and runs a nonlinear elastoplastic finite element analysis, which is the part that predicts how the sheet actually behaves as it is bent.
A finite element analysis splits the strip into a mesh and solves the stress at each node as the section passes through the stands. From that you read the things that wreck a first article:
- Longitudinal strain, which tells you if one pass is doing too much work and will crack the corner.
- Wrinkles and edge wave, which appear when the material is pushed faster than it can flow.
- Springback, the bounce the section makes after the last roll lets go, which decides your final tolerance.
- Bowing and twist, the global shape errors that show up only after several stands.
COPRA is the standard tool for cold roll formed sections in more than 50 countries and is available in 10 languages, so when a customer sends a COPRA project the handover is clean. We use the same family of tools to check a flower before steel is cut.
Virtual flower optimization
The flower is the heart of the process. Too few passes and the strip folds; too many and you waste stations and floor length. The simulator lets you slide the bend distribution and watch the strain map change, then lock the version that keeps every pass inside a safe strain band. That is the difference between a line that forms quietly and one that screams at stand four.
What it saves: scrap and lead time
The two numbers that matter to a buyer are first article scrap and lead time. The table shows the typical swing we see when a profile moves from trial and error to simulated design.
| Stage | Trial and error | Simulated first |
|---|---|---|
| First article scrap | several coils | one coil or less |
| Tooling revisions | 2-4 loops | 0-1 loop |
| Lead time to good part | weeks | days |
| Risk of line stop at launch | higher | lower |
On a thick or high strength section the saving is larger, because a wrong roll there costs real steel and real days. The twin also gives the customer a documented flower, which helps when the line moves to a second plant and someone has to repeat the result.
Where the limits are
A simulation is only as good as its inputs. The mesh needs real material data: yield, hardening, thickness, and the actual coil batch. Plug in a generic steel grade and the springback number drifts. The software also will not catch a worn bearing or a misaligned stand, those are physical faults a model cannot see. So we still prove the first article on the floor. The point is that the floor proof should be a confirmation, not a discovery.
For the manufacturing side there is also ISO 23247, the ISO digital twin framework for manufacturing, which sets out how a digital twin of a production system should be described and linked to the real one. It is the standard buyers can ask for when they want the model to be more than a vendor demo.
The material data is where buyers can help. Send the real grade, the actual thickness range from your mill, and the hardening curve if you have it. A generic catalogue value for yield strength can move the springback prediction by a degree or two, which is enough to push a tight tolerance section out of spec at stand six. The twin is only as honest as the coil you describe to it.
There is also a difference between the tooling twin and the line twin. The tooling twin checks the flower and the rolls; the line twin adds the drives, the uncoiler, the shear, and the straightener so you can watch the whole process, not just the bending. For most new profiles the tooling twin is enough, and it is the cheaper one to run. Ask which twin a supplier is offering, because the word gets used for both.
A practical caveat for buyers comparing quotes: a supplier who reports a clean simulation but will not share the strain map is asking you to trust a number you cannot read. Ask for the map and the springback report as deliverables. If they are part of the offer, the design was almost certainly done properly; if they are a favour afterwards, the part may still be a guess.
Practical takeaway
When you order a new profile, ask for the simulation, not just the drawing. Request the strain map and the springback report, and confirm the material data used matches your coil. A supplier who can show the virtual flower before cutting steel will almost always deliver a cleaner first article and a shorter wait.
At ZTRFM we build the roll forming machines and the tooling, and we run the forming simulation as part of the design so the line reaches you with the flower already proven on screen. A part that forms right the first time is cheaper for everyone than a line that needs three tries to learn.
