Bench record / observed, repeated, explained
A mesh can look smooth, complete, and correctly shaped while carrying the wrong scale or an unreliable functional feature. First-scan verification therefore begins with defined measurements rather than a shaded rendering. The operator checks global scale, selected dimensions, holes, and edges in a sequence that keeps each result interpretable.
This workflow supports a shop acceptance decision. Formal product inspection requires a documented measurement procedure, suitable calibrated standards, uncertainty evaluation, and traceability appropriate to the tolerance. NIST emphasizes those elements for dimensional results from 3D imaging systems.
Protect the evidence
Keep the raw scan and original processed mesh read-only. Make a verification copy with a revision suffix. Record the scanner software version, processing settings, export format, stated export units, and the application used for inspection.
Import the copy into the inspection or CAD tool with units chosen explicitly. Autodesk Fusionâs Insert Mesh command asks the user to choose centimeters, millimeters, meters, inches, or feet. A correct-looking import dialog can still apply an unintended unit if the choice is assumed.
Leave repair, scaling, and smoothing until after the first measurements. Any correction should produce a new revision and a new result table.
Check gross scale first
Compare the mesh bounding box with the independently recorded overall spans of the physical part. Fusionâs mesh properties report bounding-box length, width, and height. Bounding dimensions can be misleading if the mesh includes a fixture fragment, isolated point, or tilted part, so inspect the extrema that create each bound.
A large uniform scale error across all axes usually points to an import-unit mismatch. A difference between inch and millimeter interpretation is large enough to diagnose before fine inspection. Correct the import units and repeat rather than scaling by eye.
If only one bound is wrong, look for incomplete coverage, a floating fragment, or orientation against the coordinate axes. A bounding box measures extents rather than a fitted engineering feature.
Define every dimension before selecting points
âWidthâ is ambiguous on a rough triangle mesh. The measurement record should say which surfaces or features define the result: distance between two fitted parallel planes, center distance between fitted cylinders, diameter of a fitted circular section, or distance between specific datum features.
Point-to-point picks are quick but sensitive to cursor position and individual triangles. Use them only for a coarse check. Functional dimensions deserve fitted features over a representative region.
Autodeskâs Measure tool reports different quantities according to selected geometry, including distance, length, angle, radius, diameter, area, and loop length. That variety makes selection type part of the measurement definition. Save a screenshot that shows both the selection and result.
Measure planes and external spans
Inspect broad planar regions under a facet or deviation display. Exclude rounded edges, obvious noise, fixture contact, and filled patches from the fitting selection. Fit a plane to each opposing surface if the software supports it, then measure the normal distance between planes.
Repeat the selection with independently chosen regions. A result that changes sharply when a few edge triangles are excluded indicates an unstable fit or damaged boundary. Record the spread and keep the displayed precision proportionate to the evidence.
Compare the mesh result with the physical reference reading and its instrument record. The difference is a bench observation for that feature and workflow. Scanner accuracy requires a broader, documented evaluation.
Inspect holes with sections
A bore is difficult for optical scanning because the camera sees only surfaces reached by the projected pattern and viewed by the sensors. The mouth may be complete while the deeper wall is sparse. A circle fitted to that partial surface can look convincing.
Create several cross-sections normal to the intended bore axis. Inspect whether each section contains enough observed wall and whether the fitted center remains stable. Fusion can create a mesh section sketch and fit lines, arcs, circles, ellipses, or splines to the section. Its documentation also exposes fit tolerance and maximum curve deviation, reminding the operator that the fitted curve is an interpretation of mesh data.
Record the section location, fitting method, included arc, diameter, and fit residual if available. Compare sections near the mouth and farther inside. A diameter that drifts with depth may reflect sparse coverage, smoothing, or a genuine taper. The physical part and another instrument decide which explanation is credible.
Treat fine edges as transition zones
A triangle mesh seldom contains the mathematically sharp intersection present in a CAD model. Sampling, fusion, and smoothing create a small transition. Picking the outermost vertex therefore rewards noise, while picking an inward vertex shortens the part.
Fit the neighboring surfaces and calculate their intersection when the edge is functionally important. For a rounded edge, fit the intended arc over a documented section. Check several locations along the edge rather than one attractive region.
Turn off smoothing display when possible and view actual facets. A visually crisp shaded edge may be produced by normals even when the geometry is rounded or sparse.
Separate observed holes from repaired holes
Use mesh analysis to locate open boundaries and separate shells. Fusionâs mesh properties distinguish closure, consistent orientation, and positive volume. Those properties describe mesh integrity; dimensional fidelity still needs feature measurements.
Mark every area created by automatic hole filling. A filled mounting face, bore, slot, or fine edge should be excluded from measurement unless the repair method and its uncertainty are explicitly accepted. Retain a copy without filling so the CAD recipient can see which surfaces were observed.
Repeat the measurement chain
Measure the same feature on the second reference scan created after remounting. Use the same definition but repeat the selection rather than copying fitted geometry. This reveals sensitivity to capture and analysis choices.
NIST notes that repeat measurements of uncalibrated workpieces alone are insufficient for a traceability claim because they do not assess measurement bias. A shop can still use repetition to assess stability. It should label the result accordingly and use calibrated artifacts where traceable dimensions are required.
Build a decision table
| Feature | Physical reference | Mesh result | Method | Decision |
|---|---|---|---|---|
| Overall span | Recorded instrument reading | Measured value | Fitted planes | Accept, investigate, or reject |
| Hole | Recorded bore reading | Fitted section result | Named section and fit | Accept, reference only, or rescan |
| Step | Recorded depth reading | Measured value | Plane to plane | Accept, investigate, or reject |
| Fine edge | Defined datum location | Derived intersection | Neighboring surface fits | Use, remodel, or rescan |
The tolerance column, if used, should come from the actual downstream task. A mesh acceptable as assembly context may be inadequate for manufacturing a mating component. The final disposition should say what the data can support: visual reference, design envelope, reverse-engineering sections, or measured inspection.
Release a verified revision
Save the verification report beside the raw project, original mesh, inspection copy, screenshots, and physical reference readings. State units in the filename or manifest, and include the coordinate orientation and any applied transform.
If scale is corrected, record the factor, reason, software command, and before-and-after check. If a feature fails, preserve that result and describe the required recapture or modeling treatment. A CAD handoff becomes trustworthy when the recipient can see both the useful surfaces and the limits of the evidence.