Conformance and nonconformance in segmentation-free X-ray computed tomography geometric inspection

Additive Manufacturing (AM) is changing the manufacturing paradigm as it makes it possible to generate complex geometries that are impossible using conventional technologies. However, conventional GPS/GD&T practices are inadequate both at specifying and verifying geometric tolerances. In both cases, they lack the required flexibility. Applying volumetric instead of surface representations helps to solve the problem of specifying tolerances and coheres with topological optimization. The verification paradigm must be modified, too, as AM allows an increase in part complexity without a corresponding increase of cost. Among measurement techniques, only X-ray computed tomography (XCT), which is volumetric, is capable of easily measure complex parts. Leaving the discussion of volumetric tolerance specifications to the future, the aim of this work is exploring a part geometric accuracy verification by direct comparison between its nominal geometry and geometric tolerance volumetric representation, and an XCT volumetric image of it. Unlike the conventional use of XCT for geometric verification, this is a segmentation-free verification. The method is based on the “mutual information” of the two, i.e. information shared by the measured and nominal representations. The output is a conformance statement that does rely on a measurement but nor on a specific measured valuenot rely on a measurement result. This makes defining a decision rule considering consumer’s and producer’s risks difficult: uncertainty does not exist in this case. Statistic and simulation techniques make it possible to estimate these risks, defining a numerical model of the distribution of the gray values in a specific portion of the XCT image. Finally, an additive manufacturing case study validates the methodology.