In-line Inspection in Roll Forming

1. Definition

In-line inspection (also written inline) means measuring product geometry, surface condition, or related process signals while the strip or profile is still moving through the production line—without stopping the mill solely to cut samples and walk them to a gauge room. In roll forming, the most common industrial form is non-contact laser profile measurement that digitizes the cross-section, compares it to a CAD template, and alarms when key dimensions leave control limits.

In-line inspection is a quality and process tool. It does not replace millwright skill or pass design; it shortens the feedback loop so coil-to-coil drift, setup error, and progressive wear show up before long scrap runs accumulate.

2. Why Offline Checking Is Not Enough

Traditional practice cuts a short sample, deburrs the cut edge, and checks dimensions on a height gauge, optical comparator, or CMM. That workflow is accurate for a snapshot, but it is slow relative to line speed. While the sample is being prepared, the mill may keep producing. Vendor literature on in-line profile systems for roll-formed metal notes that lot-to-lot coil variation in width, thickness, crown, camber, and mechanical properties often drives the profile out of specification during a run—exactly when offline sampling rate is lowest.

• Long lag between defect onset and discovery
• Operator bias in where and when samples are taken
• Cut-edge condition and sample handling can distort measured angles or gaps
• No continuous trend for SPC or coil-to-coil comparison

In-line systems do not eliminate first-article or audit checks; they change the economics so continuous dimensional surveillance becomes practical at production speed.

3. What “In-line” Means on a Forming Line

Industry explainers define inline measurement as sensors integrated into the process stream: lasers, cameras, micrometers, thermal sensors, and related electronics that report in real time. On a roll forming line the inspection frame is typically mounted after a defined station group—often near the exit, after a straightener, or before a final overbend station when the data will feed compensation.

For encyclopedia clarity: in-line inspection here covers sensing and decision support. Actuating a servo overbend from that signal is closed-loop control (separate entry).

4. Laser Profile Measurement

Commercial systems such as Starrett Bytewise Profile360 and Micro-Epsilon profileGAUGE-class frames use multiple line-laser (sheet-of-light) sensors around the profile. Each sensor projects a laser line and computes a height profile by triangulation from the deformed reflection. Software merges views into a 2D/3D point cloud of the cross-section.

Vendor materials describe the workflow as: digitize thousands of points → register to a CAD or golden template → extract named features (width, thickness, gap, radius, angle) → compare to tolerance bands → display pass / caution / fail and trends. Measurement rates on compact 360° frames are often on the order of tens to low hundreds of profiles per second depending on model and range—fast enough for continuous monitoring at typical roll forming speeds when properly synchronized.

4.1 Why lasers fit roll-formed sections

• Complex open sections (C, Z, Sigma, window/door profiles) need multi-view coverage
• Non-contact avoids scratching painted or galvanized surfaces
• Immediate feedback supports progressive small roll adjustments during setup
• DXF point export can feed die/CAD comparison for tooling development

5. Vision and Other Sensors

Laser profilometry is the workhorse for full cross-section geometry. Complementary in-line tools appear on many lines:

• Machine vision — hole presence/position after punching, logo/marking, edge damage, coating defects
• Laser micrometers — width or height on simpler sections
• Thickness gauges — incoming strip thickness for process correlation (not a substitute for profile shape)
• Length / encoder — cut-to-length verification and defect mapping along the coil
• Surface / roughness sensors — specialty lines (painted strip color, weld seam on tubular products)

Choose sensors by the failure mode that costs scrap. A line whose pain is angle drift needs profile lasers; a line whose pain is missing pierce holes needs vision after the punch press.

6. Typical Measured Features

Feature lists from industrial profile gauges commonly include:

Map each extracted dimension to a drawing balloon or customer CTQ list. Vague “looks OK” alarms do not help operators decide which stand to touch.

7. CAD Template Matching and Limits

Template matching aligns the measured cloud to a design section or a golden part captured from a known-good run. Control limits are then applied to extracted scalars. Best practice:

• Separate drawing tolerance from process control limits (often tighter for early warning)
• Version templates when tooling or customer drawings change
• Document which stations and straightener settings apply to the golden baseline
• Avoid over-constraining non-CTQ contour noise that only creates false fails

Pass / caution / fail indicators on the HMI reduce cognitive load: green for in-band, yellow for approaching limits, red for stop-and-check.

8. Setup Validation and Scrap Reduction

Vendor guidance for roll-formed metal emphasizes setup benefits: validate roll stands quickly, make small progressive adjustments and see results immediately, reduce dependence on offline checking during start-up, and shorten time to process stability. During a production run, continuous monitoring alarms when dimensions drift so operators isolate out-of-spec material sooner.

Scrap reduction is therefore both a start-up and a run-time effect. Coil changes that alter thickness or yield (and thus springback) show up as angle or height trends rather than as a surprise at the end of the coil.

9. Data Histories and SPC

Modern profile systems store histories, trend charts, and often export to third-party SPC or MES software. Practical uses:

• Compare today’s coil to last month’s best run on the same tooling
• Quantify improvement after a roll polish or straightener rebuild
• Support customer quality packages with continuous evidence, not only sample tickets
• Feed Six Sigma / capability studies with far denser data than offline sampling

Time-stamp and coil-ID association matter. A beautiful chart without coil identity is weak for claims analysis.

10. Sensor Placement and Environment

Frames are built for factory floors (often stainless housings, sealed optics), but roll forming still challenges sensors:

• Oil mist and scale dust on lasers—plan wipe schedules and air purge if specified
• Vibration and strip flutter—stabilize the pass line with guides before the frame
• Specular glare on bright galvanized or stainless—sensor angles and laser wavelength matter
• Temperature drift of the frame versus the mill foundation—follow vendor calibration routines
• Cut-off shock and flying shear vibration—mount away from high-impact stations when possible

Placement after the last forming station measures finished geometry. Placement before a final adjustable station enables compensation. Placement after punching measures formed-and-pierced product as the customer receives it.

11. From Inspection to Closed-Loop Control

In-line inspection alone alerts humans. Closed-loop systems use the same measurement stream to adjust actuators (final overbend, side rolls, straightener settings) within safe limits. Research and industrial practice on HSS/UHSS roll forming describe measuring geometry near the exit and adapting the last pass. Treat closed-loop as a second project: safety interlocks, actuator authority limits, and material-model assumptions must be engineered explicitly. See the Closed-loop Control encyclopedia entry.

12. Limits and False Alarms

• Occlusion: deep channels or overlapping flanges may leave shadowed zones even with four lasers
• Motion: high strip bounce increases scatter—fix stability first
• Template mismatch: wrong CAD revision causes systematic “fails”
• Thermal expansion of long aluminum or thin open sections between stations
• End-of-coil fish-tail / end flare regions that are not representative of steady-state product

Train operators to distinguish sensor health alarms from true geometric alarms. A dirty window is not a bent flange.

13. Boundaries

This page covers in-line dimensional and related quality sensing on roll forming lines. It does not specify brand purchase decisions, guarantee measurement uncertainty for a given profile, or claim that in-line systems replace destructive tests or coated-product corrosion testing. Machine kW, line speed ratings, and commercial lead times are out of scope for this encyclopedia.

14. Buyer / Engineer FAQ

Do we still need offline gauges if we buy in-line lasers?

Yes. Keep calibrated offline methods for first article, customer disputes, and periodic verification of the in-line system itself. In-line and offline should agree within a documented bias budget.

Where should the frame sit on a new line?

Default for quality monitoring: after finishing stands and straightener, before packing. Default for springback compensation R&D: just upstream of the adjustable final station. Some plants install both a monitor frame and a control frame.

Can one system handle every profile we run?

Often one frame covers a family of similar envelopes if measuring range and FOV fit. Large envelope changes (tiny window bar vs wide racking upright) may need different sensor configurations or dual recipes. Confirm measuring range and accuracy class with the vendor against your largest and smallest sections.

Is vision enough without lasers?

Vision excels at 2D features (holes, print, edge damage). Full 3D section control usually needs line lasers or equivalent profilometry. Many lines combine both.

How does in-line inspection relate to Industry 4.0?

It is a primary data source: dense dimensional streams for MES, SPC, and eventual closed-loop algorithms. Without trustworthy sensors, digital dashboards only visualize guesses.

What about painted or oily strip?

Non-contact lasers avoid scratch risk, but oil films and high gloss can degrade signal quality. Process lubrication practice and sensor optics care must be planned together (see Lubrication page).

Will in-line inspection find bow and twist?

A single cross-section frame measures section shape at one longitudinal station. Bow and twist over length need multiple stations, length-wise sensors, or offline straightness checks. Cross-section OK does not prove longitudinal shape OK.

15. Practical Implementation Checklist

1. List CTQ dimensions from the customer drawing and map each to a sensor feature
2. Define control limits vs drawing limits and alarm behavior (caution vs stop)
3. Stabilize pass line and lighting/optics environment before chasing microns
4. Create golden templates per tooling set and revision-control them
5. Train setup crews to use live data during roll adjustment, not only at end of shift
6. Correlate coil certificates (thickness, yield) with dimensional trends
7. Schedule optic cleaning and calibration verification
8. Decide whether data feeds only HMI alarms or also closed-loop actuators

16. Related Encyclopedia Topics

• Springback Compensation — why angles drift and how overbend is applied
• Closed-loop Control — actuating from in-line measurements
• Bow / Camber, Twist, Edge Waviness, Fish Tail / End Flare — defect vocabularies that offline or multi-sensor checks still own
• Roll Forming CAE — predicted shape vs measured shape
• Material Certificate (MTC) — coil property inputs that explain dimensional shifts
• Machine Accuracy — mechanical capability limits of the mill itself

17. Summary for Specifiers

Specify in-line inspection when coil-to-coil variation, long start-up scrap, or customer CTQ density makes offline sampling too late. Prefer multi-view laser profilometry for complex open sections; add vision for punched features. Keep offline metrology for verification. Treat continuous data as a process asset: histories, SPC, and—when engineered—closed-loop compensation. Sensors report facts; flower design and roll setup still decide whether those facts stay green.