Defining a tolerance in 3D scanning does not mean copying the accuracy figure from a scanner specification sheet. A real 3D measurement depends on the capture technology, part preparation, alignment, meshing, inspection software and operator skill. For a workshop, design office or quality department, the key is to connect the requested tolerance to a concrete decision: validating a part, rebuilding CAD, checking deformation, monitoring wear or documenting an assembly.
1. Distinguish resolution, accuracy and repeatability
Three concepts are often confused. Resolution describes the level of detail captured, for example the average distance between two points or the fineness of the mesh. Accuracy indicates the deviation between the measurement and a known reference. Repeatability measures the ability to obtain the same result several times under the same conditions. A part can be scanned with a very dense mesh without being accurate to one hundredth of a millimetre.
In practice, a realistic tolerance must be larger than the overall uncertainty of the process. If a scanner is specified at ±0.04 mm under ideal conditions, it would be risky to promise reliable inspection at ±0.05 mm on a glossy black 800 mm part scanned in a workshop. Environment, surface condition and alignment can quickly add several hundredths, or even tenths, of a millimetre.
2. Start from the function of the part
The useful tolerance depends on function. An aesthetic part, plastic cover, mould, dental implant, machining fixture or precision mechanical part will not have the same requirements. Before scanning, identify the critical areas: bearing surfaces, interfaces, support planes, centre distances, functional radii, sealing surfaces, wear zones or free-form geometries.
An effective method is to classify areas into three levels:
- Critical functional areas: require stable measurement, often with references or targets.
- Shape areas: useful for reconstruction or comparison, generally with wider tolerance.
- Non-critical areas: captured for context or documentation.
3. Account for size and geometry
The larger the part, the more error accumulation matters. A small 50 mm component can be measured with low uncertainty if the scanner is suitable. A 1.5 m part often requires markers, volumetric tracking, several positions and alignment control. Deep geometries, bores, undercuts and highly reflective surfaces also increase the risk of deviations.
For long or assembled parts, it is better to define tolerance by area rather than applying a single tolerance to the whole object. For example: ±0.05 mm on an accessible bearing area, ±0.15 mm on a shape surface, ±0.30 mm on the general envelope.
4. Include surface condition and preparation
Optical 3D scanners work best on matte and stable surfaces. Black, transparent, glossy, chrome or very dark parts may require matting spray. This preparation improves capture but adds a physical layer. Sublimating sprays reduce this issue, but the method must be documented when the goal is a quality measurement.
Temperature also plays a role. A freshly printed plastic part, a metal part just machined or a post-cured resin part may still change dimensionally. For serious inspection, let the part return to equilibrium and record the measurement conditions.
5. Build a simple uncertainty budget
An uncertainty budget does not need to be complex to be useful. It should add the main sources: scanner accuracy, calibration, alignment, surface noise, preparation, part stability and comparison method. The objective is to set a decision tolerance wider than the measurement noise.
| Source | Typical impact | Reduction action |
|---|---|---|
| Scanner calibration | Global offset | Regular calibration, stable environment |
| Multi-view alignment | Accumulated error | Markers, sufficient overlap, positioning strategy |
| Glossy/black surface | Noise or holes | Controlled matting |
| CAD reference frame | Incorrect interpretation of deviations | Datum alignment or controlled best-fit |
6. Choose the right alignment
A common mistake is to use automatic “best-fit” alignment for every inspection. Best-fit minimizes global deviations, but it can hide a functional defect. For quality inspection, it is better to align on the part references: plane A, axis B, point C, bearing surfaces or drawing-defined datums. Best-fit remains relevant for analyzing global deformation or an organic surface, but it must be clearly stated.
7. Define a decision rule
A tolerance must lead to a decision. For example: compliant if 95% of the inspected surface is within ±0.20 mm, with no functional area beyond ±0.10 mm. This rule prevents rejecting a part because of a few stray points or, conversely, accepting a part whose critical interface is out of tolerance.
Conclusion
A realistic 3D scanning tolerance is connected to the part, the scanner, the method and the intended use. To obtain defensible results, formalize the critical areas, preparation, alignment, uncertainty budget and acceptance rule. This discipline is what turns a point cloud into reliable industrial data.
