Closed-loop Control in Roll Forming

1. Definition

Closed-loop control in roll forming means measuring a product property (typically bend angle or section geometry) during production and automatically adjusting actuators so the measured property tracks a target. The loop closes when sensor data continuously corrects the process—unlike open-loop setup, where roll positions are fixed after trial-and-error or model-based prediction alone.

In metal-forming research this sits under closed-loop control of product properties: the controlled variable is the workpiece state (angle, thickness track, shape), not only motor speed or torque. In-line inspection provides the eyes; closed-loop control provides the hands—within engineered limits.

2. Open-Loop vs Closed-Loop

Most industrial mills still run open-loop with skilled setup. Closed-loop systems appear where high-strength scatter, flexible forming, or automotive CTQs make fixed overbend insufficient.

3. Why Feedback Matters for HSS/UHSS

High and ultra-high strength steels show larger springback and greater sensitivity to yield variation than mild steel. Fixed overbend that was perfect for yesterday’s coil can miss today’s. Groche and co-workers framed inline compensation as a way to correct springback without requiring full knowledge of every material and process parameter at each coil change—measure the shape just before a calibration/overbend station and act.

CIRP reviews of product-property control note the same uncertainty pattern as sheet bending, amplified by the continuous nature of roll forming: you only see the finished angle after material has traveled through the mill.

4. Control Architecture

A practical architecture has four layers:

1. Reference — target angle(s), width, or contour from drawing / recipe
2. Plant — roll stands, overbend unit, gap actuators, strip transport
3. Sensors — lasers, cameras, optionally load/thickness/yield proxies
4. Controller — computes error, applies filtered command within authority limits, accounts for delay

Optional observers estimate unmeasured states (e.g., inferred springback) from noisy measurements. Outer loops may adapt setpoints by coil grade; inner loops trim actuators at high rate.

5. Sensing Layer

Common product sensors in published roll-forming control work:

• 2D laser scanners / laser triangulation for flange angle and local section
• Multi-view profile frames for full cross-section CTQs (see In-line Inspection)
• Cameras combined with lasers in some inline compensation setups
• Process sensors: roll load, incoming thickness, experimental continuous yield monitors in research lines

Sensor placement defines the delay: measuring after the actuator observes the result; measuring just upstream of the overbend station predicts the correction needed for the material about to enter that station. Flexible-forming studies mount 2D lasers on forming stands to track springback along variable-depth rails.

6. Actuation Layer

Actuators used in literature and industrial concepts include:

Authority must be limited: unbounded automatic gap change can crush tooling or create wave defects. Encode soft limits (recipe window) and hard stops (mechanical).

7. Transport Delay and Smith Predictors

Roll forming is continuous. The strip location where you actuate is not the same place you measure a moment later. That transport delay is approximately distance divided by line speed and is known if both are known. CIRP summaries describe closed-loop calibration systems that use a Smith Predictor to compensate for this known delay between actuation and measured response—otherwise a naive PID can oscillate or hunt.

Practical implications:

• Speed changes change delay—controller must update delay estimate with encoder speed
• Acceleration/deceleration at coil start/end needs special modes (hold last command, freeze loop)
• Filtering measurement noise before the predictor reduces chatter on actuators

8. Literature Benchmarks

Published trials illustrate what closed-loop can achieve under laboratory or pilot conditions—not a universal plant guarantee:

• Groche et al. on prediction and inline compensation of springback in HSS/UHSS roll forming: online calibration independent of full material/process parameter knowledge, using measurement ahead of the compensation station.
• CIRP review citing closed-loop calibration on a high-strength steel U-profile: open-loop model-based design left springback deviations on the order of tens of degrees in the cited case; closed-loop with Smith Predictor reduced deviation to a small fraction of a degree relative to the target angle in those trials.
• Flexible roll forming of variable-depth automotive rails (e.g., DP780 studies): 2D laser monitoring of inhomogeneous springback along length, with planned variable overbending derived from measured springback distribution.

Cite these as evidence that feedback works in principle for springback-dominated CTQs. Replicate capability on your mill with your coils and your actuator stiffness before promising customers the same numbers.

9. Flexible / Variable-Section Forming

Flexible roll forming changes depth or width along the part length. Springback then varies longitudinally, so a single fixed overbend is structurally wrong. Research approaches measure springback after forming steps and synthesize a smoothed overbend function for a subsequent flexible overbend pass. Closed-loop (or at least measurement-driven open-loop trajectories) becomes almost mandatory for automotive rail accuracy.

Three overbend timing strategies appear in that literature family: overbend early after each pass; incremental overbend after the top-hat is finalized; or full overbend after the final forming pass. Choice depends on accessibility of actuators and how springback accumulates.

10. Material Monitoring and Data-Driven Paths

Recent industrial research continuously monitors incoming material properties and thickness, combines them with roll load and inline shape, and trains data-driven models to predict shape quality under real coil fluctuation. That path can support:

• Feed-forward hints (adjust before the bad angle appears)
• Hybrid loops (model + feedback)
• Operator decision support when full closed-loop authority is not yet approved

FEA datasets (e.g., COPRA RF/FEA-class studies) expand the training space beyond what a plant can sample in a week of coils. Still, plant validation on live material remains non-negotiable.

11. Authority Limits and Safety

• Clamp actuator travel to recipe windows derived from tooling design
• Rate-limit motion so stands cannot slam
• Watchdog: freeze actuators if sensor health fails or strip is absent
• Manual override always available to millwrights
• Log every automatic move with coil ID for quality traceability
• Separate “monitor only” mode from “control enabled” mode during commissioning

Safety PLCs and forming PLCs should agree on who owns the actuator. Ambiguous ownership causes dangerous fights between automatic and manual commands.

12. Implementation Checklist

1. Prove in-line measurement repeatability on the CTQ (In-line Inspection first)
2. Choose one primary controlled variable (often flange angle)
3. Select an actuator with enough stiffness and resolution for that CTQ
4. Measure transport delay vs speed; design predictor / delay compensation
5. Commission in monitor-only; compare recommended vs operator moves
6. Enable limited authority; expand window only after capability data
7. Define freeze behavior at splice, acceleration, and E-stop
8. Train operators: when to trust the loop, when to open it

13. Limits and Failure Modes

• Wrong CTQ — controlling angle while the customer rejects bow/twist
• Saturated actuator — springback exceeds overbend travel
• Coupling — fixing angle worsens edge wave or twist
• Sensor occlusion / oil — false error drives wrong correction
• Model mismatch after tooling polish or station rebuild
• End effects — fish-tail / end flare regions confuse the loop

MIMO (multi-input multi-output) control for several CTQs at once is research-heavy. Most first installations close one well-instrumented loop.

14. Boundaries

This page explains closed-loop product-property control concepts for roll forming. It does not specify PLC code, guarantee published trial accuracies on commercial lines, or quote machine power/speed. Prices and lead times are out of scope. Related pages: In-line Inspection, Springback Compensation, Pass Design, Machine Accuracy.

15. Buyer / Engineer FAQ

Is closed-loop the same as a good PLC?

No. Line PLCs already close loops on motors, brakes, and cut length. Product-property closed-loop specifically regulates formed geometry using product sensors.

Do we need closed-loop if we already have lasers?

Lasers without actuators are inspection. Many plants gain most scrap reduction from alarms and faster setup. Add closed-loop when coil scatter still exceeds drawing limits after good setup discipline.

Will feedback fix a bad flower?

No. It can trim residual springback within a controllable band. Redesign passes if the section cannot be held.

What about AHSS springback compensation pages?

Springback Compensation covers strategies (overbend, FEA, flower). Closed-loop is the automatic execution of compensation using live measurements.

How do we validate for automotive PPAP?

Run capability with control enabled and disabled; document sensor calibration, authority limits, and freeze logic; include coil identity in control logs.

Can one loop handle width and angle together?

Possible with coordinated actuators, but start with the CTQ that fails most often. Coupled loops need careful decoupling or model-predictive methods.

16. Related Encyclopedia Topics

• In-line Inspection — sensing without actuation
• Springback Compensation — why overbend exists
• Pass Design — controllability of the section
• Roll Gap Adjustment — manual/automatic gap as an actuator family
• Roll Forming CAE — models used in feed-forward and hybrid control
• Yield Strength / MTC — disturbance variables the loop fights

17. Summary for Specifiers

Specify closed-loop control when product CTQs are dominated by coil-to-coil springback or flexible geometry that fixed setup cannot hold. Require: proven sensors, a dedicated actuator, delay-aware control, authority limits, and a monitor-only commissioning path. Treat published U-channel trial accuracies as motivation, not a contractual template. Closed-loop turns in-line measurement into continuous correction—it does not excuse weak tooling or poor pass design.