Tidimensional Scanning Detection CNC: Reading Transverse Section Radian on Machined Parts

• 1
  How tidimensional scanning detection CNC builds a radian value
• 2
  Where the accuracy of a scanned radian actually lands
• 3
  What we check before trusting a transverse section radian
• 4
  Which machined features suit scanning and which do not
• 5
  Scanning, CMM, and comparator on a transverse section radian

A transverse section is the cut you get when you slice a part perpendicular to its main axis. The radian of that section is the local curvature along the cut outline. On a shaft fillet, a turbine blade root, or a cam profile, that curvature controls fit and stress. The old way to get it: cut the part, place it on an optical comparator, trace the silhouette, then fit a circle by hand. Slow, and the cut destroys the part.

Tidimensional scanning detection CNC skips the cut. A structured-light or laser scanner sweeps the surface and produces a dense point cloud, often millions of points on a part the size of a hand. Inspection software re-slices that cloud into a plane you choose, normal to the axis or normal to a local surface direction. The software then fits a circle or a spline to the outline and reports a radius, a diameter, or a deviation from nominal.

The word tidimensional here just means three-dimensional. Nothing exotic. What matters is that the data set is volumetric, so you can re-slice the same scan many times without touching the part again. That is the real shift from a comparator: one capture, many sections.

The reported radian is only as good as the fit. A five-point circle fit on a rough outline will drift by tens of micrometres. A least-squares fit over 300 points on the same arc is stable. Always record which fit method produced the number, because two labs can measure the same arc and disagree purely on fit choice.

• 1
  Point cloudDense XYZ data captured optically, no contact
• 2
  Re-slicingAny section plane can be cut from the same scan
• 3
  Fit methodLeast-squares vs. minimum-zone changes the result

Optical scanning is not a magic replacement for a CMM. On a clean, matte, well-lit surface, a good structured-light system resolves form to roughly ±0.02 mm over a 100 mm field. On a shiny aluminium fillet under the wrong angle, the same system can be off by 0.1 mm or more, because the sensor reads the reflection rather than the surface.

Curvature amplifies that error. A radius value comes from a second derivative of the point positions, so a small normal-direction error becomes a larger radius error. On a tight fillet of R2 mm, a 0.02 mm surface error can move the fitted radius by 0.05 mm. On a gentle R50 mm sweep, the same error barely shows. Rules of thumb: the tighter the radius, the more the scan struggles.

Surface finish matters just as much. A bead-blasted or anodized matte face scatters light evenly and scans cleanly. A mirror-polished face at Ra 0.2 μm reflects the pattern away from the sensor and creates holes in the cloud. In our shop we scan after bead blasting where the drawing allows it, and we keep polishing for the final step.

The honest summary: tidimensional scanning detection CNC is excellent for verifying that a radian is close, for catching a wrong tool path, and for full-field comparison against CAD. It is weaker when you need a certified number on a tight radius with a mirror finish. For those parts, a touch probe on a CMM remains the reference.

• 1
  Matte surfaceScans cleanly, low noise in the point cloud
• 2
  Mirror finishDrops points, needs spray or coating first
• 3
  Tight radiusCurvature magnifies surface error

Every scan needs a datum. If the part is fixtured loosely, the software aligns the cloud to CAD by best fit and the radian drifts with the alignment. We clamp on a machined face or use a fixture plate with known pin locations. The alignment error then stays under 0.01 mm instead of wandering with every re-scan.

We also scan the same section twice with different exposure settings. If the two fitted radii differ by more than 0.03 mm, the surface is fighting the sensor and the number is not ready to ship. That check costs a few minutes and has caught more bad data than any software setting.

For parts we machine to ±0.005 mm, the scan is a screening tool, not the final word. It confirms the 5-axis tool path produced the intended fillet and that no gouge or step sits on the arc. The certified dimensional report still comes from contact measurement, and we issue it on request.

Scan data also feeds back into programming. When a fillet scans consistently undersize across a batch, that is a tool wear signal, not a one-off defect. We log the deviation trend and adjust the offset before the next run. Over 127 CNC machines and 3 plants, that feedback loop is how a scanning report turns into a process change rather than a filing cabinet entry.

• 1
  Fix the datumClamp on machined faces, not on rough stock
• 2
  Re-scan onceTwo exposures, compare the fitted radii
• 3
  Scan as screenContact measurement certifies the tight calls

Scanning pays off on free-form surfaces: impeller blades, turbine vanes, organic brackets, and any part where a print calls out a profile tolerance band rather than a single dimension. Those features are painful to probe point by point and easy to capture in one optical pass. The resulting colour map shows the whole surface against CAD in a way a probe report never does.

It also pays off on thin-walled parts. A touch probe at low force still deflects a 0.5 mm aluminium wall, so the measurement changes the part. A non-contact scan leaves the wall alone. For sheet-metal-like geometry and thin ribs, that is a real advantage, not a preference.

Scanning is a poor fit for deep bores, cross-holes, and internal features with limited line of sight. The sensor simply cannot see the floor of a Ø20 mm bore 80 mm deep. It is also a poor fit for hard, sharp edges where the optical footprint rounds the corner and the reported edge radius grows.

A practical split: use the scan for external curvature and full-surface comparison, use the CMM for internal geometry and certified tight-tolerance calls, and use a roughness tester for Ra. One method, one job. Trying to make the scan do all three is how inspection reports become arguments.

• 1
  Good fitBlades, vanes, brackets, free-form external surfaces
• 2
  Poor fitDeep bores, hidden internal features, sharp edges
• 3
  Thin wallsNon-contact avoids probe deflection