See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/375718498

Dimensional Measurements of Additive Manufacturing Parts: A Comparison

of X-ray Computed Tomography and Coordinate-Measuring System Using
Physical and Simulation Approaches

Conference Paper · September 2023

CITATION READS

1 183

10 authors, including:

Weidong Liu Xiao Chen

University of Huddersfield University of Huddersfield
14 PUBLICATIONS 69 CITATIONS 10 PUBLICATIONS 40 CITATIONS

SEE PROFILE SEE PROFILE

Wenhan Zeng Wenjuan Sun

University of Huddersfield KU Leuven
113 PUBLICATIONS 1,284 CITATIONS 65 PUBLICATIONS 810 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Weidong Liu on 17 November 2023.

The user has requested enhancement of the downloaded file.

 Joint Special Interest Group meeting between euspen and ASPE
Advancing Precision in Additive Manufacturing
KU Leuven, Belgium, September 2023
www.euspen.eu

Dimensional Measurements of Additive Manufacturing Parts: A Comparison of X-ray

Computed Tomography and Coordinate-Measuring System Using Physical and
Simulation Approaches

Weidong Liu1, Xiao Chen1,2, Wenhan Zeng1, Wenjuan Sun2, David Gorman2, Alan Wilson2, Qunfen Qi1, Paul J.
Scott1, Xiangqian Jiang1, Shan Lou1
1EPSRC Future Metrology Hub, University of Huddersfield, Queensgate, Huddersfield, UK
2National Physical Laboratory, Materials & Mechanical Metrology Department, Teddington, Middlesex, UK

[email protected]

Abstract
The purpose of this study is to assess the effectiveness of X-ray Computed Tomography (XCT) in measuring dimensions of additively
manufactured (AM) parts, and compare it to a well-established tactile coordinate-measuring system (CMS). The study investigates
three factors: XCT beam hardening, XCT voxel size scale error, and CMS mechanical filtering effect. The performance of XCT and CMS
in measuring a half-smooth and half-rough AM hollow cylinder is compared. To complement physical measurements, simulation
methods are used to investigate the individual impacts of XCT beam hardening and CMS mechanical filtering effect. The physical
experimental results show that the elimination of XCT beam hardening aids in obtaining accurate internal dimensions but deteriorates
external dimensions. XCT voxel size scale error can be compensated by either the two-sphere calibration or the CMS normalization
method. However, it is essential to ensure that these methods are based on an accurate reconstruction volume. The physical
measurement results suggest that the deviation between XCT and CMS measurements of the rough cylinder part is mainly due to the
mechanical filter effect of CMS, which is further confirmed by simulation measurements.

Keywords: Additive manufacturing, Dimensional metrology, X-ray Computed Tomography, Coordinate-measuring system, Simulation

Regarding the research on XCT voxel size scale error, there are
1. Introduction various methods for determining the voxel size, with the most
common one being based on the distance from the X-ray source
X-ray Computed Tomography (XCT) is a non-destructive to detector, and the distance from X-ray source to measurement
imaging technique that enables the restoration of internal object [4]. However, this method is often affected by
geometries. In comparison to well-established Coordinate temperature, drift effects, distance errors of source-object-
Measuring Systems (CMS) technology, XCT offers the advantage detector, leading to voxel size scale error [5]. To address this,
of contactless evaluation of overall structures. Moreover, it several correction methods are available, including spherical
remains unaffected by mechanical filtering effects on the disk, computer-aided accuracy (CAA) database, and additional
surfaces of rough additive manufacturing (AM) components. compensation strategies using CMS [4]–[6].
However, despite its significant advantages in dimensional This study employs a combination of physical experimentation
measurement, XCT still faces challenges in terms of traceability and simulation techniques to investigate the influence of beam
and reliable calibration, which are influenced by various factors hardening and voxel size on dimensional measurements in XCT.
such as XCT beam hardening and XCT voxel size scale error. Specifically, a novel approach for correcting voxel size scale error
Regarding the study of XCT beam hardening, previous research is explored, obviating the need for an additional XCT scan and
conducted by Townsend et al. [1] revealed that beam hardening mitigating the effects of material differentiation. To facilitate
affects XCT measurements of the internal and external accurate XCT and CMS simulations, a virtual part is synthesised
diameters of smooth cylinders. Where the internal diameter was using authentic AM surface data. Furthermore, the CMS
underestimated and external diameter was overestimated. simulation algorithm is enhanced by incorporating 2D
Through local iterative surface determination and global voxel morphological operations to account for the influence of
size calibration, it was possible to reduce the disparity between neighbouring surface topography. Figure 1 shows the flowchart
XCT and CMS measurements to less than 1% for both internal diagram of the general methodology.
and external diameters. Lifton et al. [2] investigation
encompassed scattering and beam hardening in XCT simulation,
which demonstrated that local iterative surface determination
effectively mitigated scattering, but could not eliminate beam
hardening errors. Furthermore, Yang et al. [3] highlighted that
the reliability of different surface determination methods varies
depending on the specific circumstances. To address this issue,
Yang’s study proposes the utilisation of a watershed algorithm
Figure 1. Outline of the proposed methodology.
to enhance robustness under beam hardening conditions.
2. Material and Methods (including both internal and external surfaces); (c) 6 sets of cylinder
circumference profile data (including both internal and external
2.1. Manufacture of Physical Part surfaces).
A hollow cylinder composed of 316L stainless steel powders
was fabricated by selective laser melting (SLM) using a laser For the XCT measurements, three repeat measurements were
power of 110 W, a scan speed of 500 mm/s, a hatch spacing of conducted using a Nikon Metrology XCT H225M with voltage
110 μm, and a layer thickness of 50 μm. Subsequently, half of 150 kV, current 150 μA, 0.5 mm copper pre-filter, 3141
the cylinder was subjected to surface treatment to achieve a projection images and voxel size 20 μm. The data were
smooth surface texture (see Figure 2 (a)-(c)). Given the negligible reconstructed with three beam hardening correction strengths
influence of mechanical filtering effects on CMS for such smooth (1, 2, and 3), using the local iterative surface determination.
surfaces, the dimensional measurements obtained from the Voxel size scale error was corrected using two separate
smooth-half cylinder were employed as reference values for the approaches. In the first approach, NPL’s two-sphere reference
computation of normalisation factors to standardise XCT object was scanned to derive the centre-to-centre distance for
measurements. global voxel size scale error [7]. And the second approach with
CMS normalised for internal and external separate. Table 1 lists
the nomenclature for XCT results.

Table 1 Nomenclature of different calibration types and different beam

hardening elimination.

BHE BHE1 BHE2 BHE3

Voxel
correction
Raw BHE1 Raw BHE2 Raw BHE3 Raw
CMS BHE1 BHE2 BHE3
Normalised Normalised Normalised Normalised
Two-sphere BHE1 BHE2 BHE3
corrected Corrected Corrected Corrected

3.2. Simulation Measurement Setup

To simulate CMS measurements, a morphological
closing/opening filter was applied to the corresponding
circumference profiles of the virtual sample, utilising 2 mm
probe tip diameters. The morphological closing/opening filter
emulates the movement of a ball rolling, similar to the scanning
motion of a CMS probe [7]. However, a limitation of performing
morphological filter solely on the 2D circumference profiles is
the lack of consideration for neighbouring surface topography.
Figure 2. Hollow cylinder part: (a) raw sample; (b) sample with its half
To address this limitation, a compensation method is
side turned into a smooth surface; (c) nominal dimensions of the design; proposed to enable a 2D disk structuring element to achieve
(d) construction of the virtual sample. equivalent results to those obtained with a 3D sphere [8]. This
method aims to capture sufficient surface topography data
2.2. Synthesis of the Virtual Part necessary for evaluating the interaction between the 3D probe
The virtual sample was generated by synthesising planar tip and the surface. Refer to Figure 4 for a visual representation.
surface data acquired using an Alicona G4 Focus Variation
System (FVM) and subsequently transforming it into a cylindrical
geometry. Moreover, smooth surfaces were incorporated into
the virtual part to mimic the turned surfaces observed on the
physical counterpart, as depicted in Figure 2 (d).

3. Physical and Simulation Measurement setup

3.1. Physical Measurement Setup

For the CMS measurements, a Zeiss Contura G2 HTG fitted
with a VAST XXT 2 mm probe head was used to measure the test
piece. The turned end was taken as the benchmark to measure
the diameters at six depths on the cylinder (repeat
measurement for each depth five times) (see Figure 3.).
Figure 4. Deformation of sectioned ring surface: (a) surface section,
sphere probe; (b) deformation of surface section, cylinder probe.

XCT scan is simulated by aRTist with: 316L stainless steel with

density 8 kg/m3, monochromatic and polychromatic X-ray,
voltage 125 kV, current 1000 μA, no pre-filter, 3141 projections,
point focal spot type, voxel size 20 μm, and local iterative
surface determination.
Figure 3. CMS measurement procedure: (a) physical artefact mounted in
a three-jaw chuck; (b) different diameter measurement depths
4. Result and Discussion when employing polychromatic X-ray simulation, a higher
deviation is observed in the internal diameter (4.9 µm)
4.1. Physical Measurement compared to the external diameter (0.2 µm).
It is noted that the internal diameter of the corrected data is
almost the same as the CMS normalised result (presented in Table 2 Diameters resulting from 2 mm probe CMS and XCT simulation
Figure 5(b)). This implies that the two-sphere correction with measurements of the AM as-built and smooth surface (Unit: µm).
BHE3 applied reduced the voxel size scale error and beam
hardening error on the measurement of internal diameters; Smooth AM as-built
however, a deviation of 15.4 μm is noticed on the external Internal External Internal External
diameter. It is probably because the BHC provided by CT Pro is a Nominal 5989.3 9907.6 4926.2 10112.0
global correction that applies to the internal and external
surfaces. While the deviation of the internal geometry was XCT Mono
5991.9 9904.7 4925.3 10116.3
reduced, more departures appeared for the external geometry. [+2.6] [-2.9] [-0.9] [+4.3]
This is also evidenced in the BHE1 results. The deviation between 5994.2 9907.8 4925.1 10115.4
XCT Poly
the XCT corrected external diameter and the CMS diameter is [+4.9] [+0.2] [-1.1] [+3.4]
only 0.8 μm, (refer to Figure 5 (a)), which indicates that the
impact of beam hardening on external diameter is trivial. The 5989.3 9907.6 4859.4 10174.3
CMS
[0] [0] [-66.8] [+62.3]
results of BHE2 seem unexpected. For the external surface, the
[]: difference from nominal
diameter deviation (3.0 µm) is larger than BHE1 (0.8 µm) but
smaller than BHE3 (15.4 µm). However, for the internal diameter,
the deviation of BHE2 (1.5 µm) is the largest after correction. It 5. Conclusion
is speculated that BHE2 enhances noise without sufficiently
reducing beam hardening image artefacts resulting in the To evaluate the dimensional measurement capabilities of XCT
observed increase in the deviation. under different beam hardening elimination and voxel size scale
For the AM as-built surfaces, the XCT measurement results error correction techniques, a comparative analysis was
generated after the two-sphere correction and the CMS conducted by comparing XCT measurements with those
normalisation are close to the BHE1 external diameters and the obtained using CMS. This assessment involved examining
BHE3 internal diameters, respectively. The difference between physical and virtual hollow cylinder parts with AM and turned
them is 0.6 μm and 0.1 μm, respectively, (refer to Figure 5 (b)). surface textures, the key findings of this investigation are
In all cases, larger deviations at tens of micrometres are presented below:
observed between XCT and CMS measurements. This significant
deviation can be attributed to the mechanical filtering effect of 1. XCT can achieve comparable result to CMS for internal and
CMS. external dimensions with smooth and rough surface
textures in the correct data configuration.
2. Tactile CMS suffers from a significant mechanical filtering
effect in the case of AM rough surface texture. The
external diameter would overestimates, but
underestimates the diameters of the internal diameter.
3. Beam hardening elimination could result in a more
accurate determination of internal surfaces. However, it
may also diminish the accuracy of external-surface
determination (depending on the correction strength).
Inappropriate correction strength can produce negatively
affect results.

Future work will investigate more capable beam hardening

elimination methods and surface determination which are
expected to yield good results for both external and internal
dimensions.

Figure 5. Deviation of XCT measurement from CMS measurement in Acknowledgements

each region with repeatability error, (a) turned external; (b) turned
From the University of Huddersfield, the authors gratefully
internal; (c) AM external; (d) AM internal.
acknowledge the support of EPSRC (EP/S000453/1) and
(EP/P006930/1), and also the 3M Buckley Innovation Centre via
4.2. Simulation Measurement
3M BIC Fellowship. W Liu grateful acknowledges the support of
Table 2 presents the internal and external cylinder diameters
the scholarship under the China Scholarship Council
obtained from CMS and XCT simulations. While CMS
(202208890009). From the National Physical Laboratory, the
measurements are unaffected by mechanical filtering on ideal
authors gratefully acknowledge the funding from the UK
smooth surfaces, the rough texture of the AM surface introduces
government's DSIT through the National Measurement System
substantial deviations. Specifically, the AM internal surface
programmes at the National Physical Laboratory.
exhibits a significant deviation of 66.8 μm, while the external
surface shows a deviation of 62.3 μm due to the mechanical
References
filtering effect. In comparison, the deviations observed in the
XCT simulation measurements are considerably smaller than
[1] A. Townsend et al., “An interlaboratory comparison of X-ray
those in CMS. Monochromatic X-ray simulation does not computed tomography measurement for texture and
produce the beam hardening effect, resulting in minimal dimensional characterisation of additively manufactured
deviations for both the external and internal diameters, parts,” Addit Manuf, vol. 23, pp. 422–432, Oct. 2018, doi:
measuring only 2.6 µm and 2.9 µm, respectively. Conversely, 10.1016/j.addma.2018.08.013.
 [2] J. J. Lifton and S. Carmignato, “Simulating the influence of
scatter and beam hardening in dimensional computed
tomography,” Meas Sci Technol, vol. 28, no. 10, 2017, doi:
10.1088/1361-6501/aa80b2.
[3] X. Yang, W. Sun, and C. L. Giusca, “An automated surface
determination approach for computed tomography,” NDT
and E International, vol. 131, Oct. 2022, doi:
10.1016/j.ndteint.2022.102697.
[4] S. Carmignato, Industrial X-Ray Computed Tomography.
[5] P. Müller, J. Hiller, Y. Dai, J. L. Andreasen, H. N. Hansen, and L.
De Chiffre, “Quantitative analysis of scaling error
compensation methods in dimensional X-ray computed
tomography,” CIRP J Manuf Sci Technol, vol. 10, pp. 68–76,
2015, doi: 10.1016/j.cirpj.2015.04.004.
[6] W. Sun, S. Brown, and R. Leach, “NPL REPORT ENG 32 An
overview of industrial X-ray computed tomography Sun W,
Brown S B and Leach R K JANUARY 2012,” Measurement, no.
January, 2012.
[7] W. Sun, S. Brown, N. Flay, M. McCarthy, and J. McBride, “A
reference sample for investigating the stability of the imaging
system of x-ray computed tomography,” Meas Sci Technol,
vol. 27, no. 8, 2016, doi: 10.1088/0957-0233/27/8/085004.
[8] W. Liu, “Verification of x‐ray computed tomography for
additive manufacturing dimensional & surface texture
metrology and deployment to internal geometry
measurement,” University of Huddersfield, 2022.

View publication stats