Evaluation of Diameter Measurement Errors in Bores and Shafts Using X-ray Computed Tomography and Different Fitting Methods

As manufacturing tolerances become increasingly tight, non-contact measurement methods, such as X-ray computed tomography (XCT), are gaining importance in industrial dimensional metrology. This study evaluates the accuracy of diameter measurements for bores and shafts using XCT and compares three fitting strategies: Least Squares (LS), Maximum Inscribed (MI), and Minimum Circumscribed (MC). Aluminium test pieces with H7 reamed holes and POM shafts were scanned using a Tomoscope XS Plus system under standardized parameters, while reference measurements were obtained with a coordinate measuring machine (CMM). Measurement errors were analysed across different voxel sizes and nominal dimensions. Results show that LS fitting provides stable and reliable performance, with small positive errors (6–13 µm) and low variability, largely unaffected by voxel size. In contrast, MI fitting systematically underestimates bore diameters by 25–47 µm and exhibits higher sensitivity to voxel size. For shafts, LS fitting again demonstrated consistent results with minor negative errors, whereas MC fitting showed a trend of overestimation at fine voxel sizes and underestimation at coarse voxel sizes. These findings confirm that the choice of fitting method is the dominant factor influencing XCT measurement accuracy, with LS being the most robust for dimensional verification, while MI and MC remain relevant for functional fit requirements.