Deviation analysis

Deviation analysis is the comparison of the actual shape of a manufactured part with its nominal geometry, most often of 3D scanning data against the CAD model (also called nominal/actual comparison or part-to-CAD comparison). The result is a color map of deviations across the entire surface and a numerical evaluation of selected dimensions and geometric tolerances.

The first and decisive step is alignment, that is, registering the measured data to the CAD model. A best-fit alignment looks for the position with the smallest overall deviations. A datum-based alignment mimics the drawing or the fixturing: the part is aligned, for example, to its seating plane, a hole axis and a side face (the 3-2-1 principle), or to the reference points of a car body (RPS). The same data can give very different results with different alignments.

The software then calculates, for each point, the shortest signed distance to the nominal surface: a positive deviation usually means excess material, a negative one missing material. The color scale is set according to the tolerance, so areas within tolerance are green, areas with excess material usually yellow to red and areas with missing material blue. Deviation callouts at points, cross sections with plotted deviations and feature evaluation complete the picture: planes, cylinders or circles are fitted to the data to check diameters, positions, flatness and other GD&T tolerances.

Two scans can also be compared, for example a part before and after heat treatment, loading or service, as can a reverse-engineered CAD model and the scan it was built from.

Deviation analysis is used for first article inspection of parts from a new mold or process, when tuning injection molds and die casting dies (where and by how much to correct for shrinkage and warpage), on castings and forgings to check machining allowances, on sheet metal parts to assess springback and on 3D printed parts to evaluate warping. It also helps track down the causes of assembly problems, settle complaints and monitor wear.

Unlike point-by-point measurement, it gives an overview of the entire surface. Tight tolerances on functional dimensions, such as fits and hole positions in hundredths of a millimeter, are measured more reliably by a coordinate measuring machine, which is why the two methods are often combined.

A best-fit alignment spreads the error over the whole part: a twisted part then looks half plus and half minus, and a systematic error stays hidden. To assess conformity with the drawing, ask for alignment to the drawing datums, and use best fit where only the shape matters.

A color map without a legend and a known scale range says nothing, because a suitably chosen scale can make the same part look excellent or like scrap. The report should state the CAD model revision, the alignment method, the scale range and tolerances, the measuring conditions and numerical values for critical dimensions. Where data is missing (holes, shadowed areas), the map shows nothing.

Flexible parts such as sheet metal and thin-walled plastics deform differently when free than when assembled. Agree whether they will be measured free or in a fixture, and allow for scanner uncertainty: deviations of a few hundredths of a millimeter are at the limit of what a scan can reliably resolve.