Inline 2D/3D Cell Checks a B-Pillar Around Laser Cutting
Hot-formed parts such as a B-pillar reach a maximum inspection range of 1500 mm by 2800 mm, and they pass two inspection points around laser cutting. This cell pairs two 6-axis robots carrying 2D and 3D imaging modules, putting pre-cut 3D localization and millimetre-level edge shortage detection together with post-cut missing holes, incomplete cuts, slug adhesion, hole diameter and cut-edge verification in one inline station. The cycle is 42 seconds per part, subject to site acceptance results.
Cutting splits the problem in two
Before cutting, the questions are whether the part is in the correct position and whether the incoming edge is short of material. That is spatial localization and incoming quality control. After cutting, the questions are whether every hole was made, whether the cut went through, whether a slug stuck, and whether hole diameter and edge line are correct. That is a full check of the machining result. The criteria, imaging requirements and ownership differ.
Keeping them apart creates an attribution problem. If incoming edge shortage is not found before cutting, post-cut defects and incoming defects mix together and nobody can say whose problem it is.
Why both checks belong in one station
Moving the part between two processes loses time and invites handling damage. Compressing localization, edge shortage detection and post-cut inspection into one inline station means the part is located twice and judged within one framework, and the results merge by part into a single quality file.
Keeping precision across 1500 mm by 2800 mm
A single wide field sacrifices resolution. The robots carry the imaging module along planned paths and work close to the part in segments, turning the large area into several local fields, each with enough pixel density. The two robots split the sides of the part and reduce blind areas. Paths are preset per part type and switch with the template at changeover.
What 2D owns and what 3D owns
Hole diameters are extracted from 2D images at subpixel level, with accuracy coming from pixel density. Incomplete cuts and slug adhesion rely on 3D height information. The cut edge line is judged by contour continuity and position comparison. Pre-cut edge shortage compares the point-cloud contour with the nominal model, and a millimetre-level difference triggers an alarm. The two modalities are not stacked; they are assigned by task.
Changeover in hours is a software claim with a hardware caveat
Inspection programs are managed by variant. Changeover switches the inspection template and robot path parameters without touching hardware, and tuning plus a trial run completes at the hour level, supporting mixed-model production. The precondition is that part definitions and datum features are solid; otherwise changeover still needs extensive trial shots.
Where the two saved operators come from
With automatic inspection replacing manual sampling, one cell saves 2 operators, and people move from repetitive checking to anomaly handling and process improvement. Data goes to the quality management system, archived by batch and traceable to process and time. Anomalies are diverted by a reject mechanism, and the main line does not stop for them.
Questions that surface before purchase
What does pre-cut 3D localization actually establish?
Two things: the actual position and orientation of the part in the cutting station, which gives the laser its datum, and whether the incoming edge is short by a millimetre-level amount, which keeps defective stock out of cutting.
How are the five post-cut defects checked?
Missing holes and diameter come from 2D hole features compared against nominal positions. Incomplete cuts, slug adhesion and edge line use 2D and 3D together, with 3D confirming residual height and edge integrity while 2D verifies diameter and shape.
Does 42 seconds cover both stages?
Yes. The variant program completes pre-cut localization and edge shortage detection plus post-cut inspection within the takt, sharing robots and the judgment framework, with results archived inline.
When would changeover exceed the hour level?
When a new part type changes datum features, inspection items or robot reach, paths and teaching must be replanned. That should be confirmed by product family at selection.