Metal-Polymer Assembly Dimensional Evaluation by X-Ray Computed Tomography: An Experimental Approach Through Relative Intensity Intercomparison
- Open Access
- 01.09.2025
Erschienen in:
Aktivieren Sie unsere intelligente Suche, um passende Fachinhalte oder Patente zu finden.
Wählen Sie Textabschnitte aus um mit Künstlicher Intelligenz passenden Patente zu finden. powered by
Markieren Sie Textabschnitte, um KI-gestützt weitere passende Inhalte zu finden. powered by (Link öffnet in neuem Fenster)
Abstract
1 Introduction
The recent development of X-ray computed tomography (XCT) for industrial applications [1‐4] has supposed an innovation in the field of dimensional quality control of parts and assemblies. This technique is based on the acquisition of 2D X-ray images of the measurand around a 360º rotation, subsequently obtaining a reconstruction of the 3D volume. In this virtual representation, not only the external surface is possible to characterize, but also internal features, being able to evaluate hidden cavities and internal defects [5‐8], such as porosity [9].
The principle used in XCT for the image acquisition is the penetration of the measurand by X-ray beams, which are registered by a detector, creating a greyscale histogram. Grey value of each voxel depends on the attenuation of the X-ray in this certain part of the object, which it is directly related to two parameters: attenuation coefficient of the material and penetration length. In this terms, the main aspects of the attenuation coefficient are both density of the material and atomic number. Therefore, the denser is the material, the higher attenuation is registered, and more artifacts could appear in the reconstruction of the volume, creating difficulties in the metrological evaluation [10].
Anzeige
Several methods for the correction of common problems as beam hardening or ring artifacts have been developed [11‐14]; however, to overcome the production of artifacts and optimize the XCT evaluation of high-density materials (such as metals), the first step required is an adjustment of XCT settings (voltage, current, physical filter, etc.). Evaluation of high density materials require higher tube voltage and current to avoid or reduce artifacts [15]. A correct post processing of the data obtained is also fundamental to improve XCT quality [16‐18]. Calibration and traceability also play an important role on the accuracy of XCT evaluations, and specific reference standards are widely used to test the devices [19‐22].
This optimization on the evaluation procedure becomes more challenging in multi material parts [23, 24], as attenuation differs from one material to another; particularly, if the attenuation difference is significant, as is commonly found in metal-polymer assemblies as studied before [25]. Surface determination and extraction is a key aspect [26, 27], as this process is critical in order to reduce artifacts and, consequently, obtain the best metrological results possible. For this purpose, existing algorithms have been improved and new ones have been developed [28, 29], as well as an acceptance test [30] based on length measurement error test (E-test) and probing error test (P-test) described in standard ISO 10,360 [31, 32].
Here, an important indicator to consider is the relative intensity (I/I0) between the beam emitted and received by the detector. This attenuation ratio quantity is directly related to the energy of the beam source [33], the material path length [34, 35] and the attenuation coefficient of the material. Theoretically, the lower the relative intensity that reaches the detector is, the poorer the quality of the tomography is and, hence, more measurement inaccuracies will appear.
In this paper, developed in the framework of the first author’s thesis [36], an analysis of the attenuation ratio, defined by relative intensity, and its usage to define the expected quality variation of XCT measurements of metal-polymer assemblies is presented. Several studies have been done before regarding metal– polymer evaluation by means of XCT, in which the measurand of the reference standards were gaps between flat surfaces [29, 37, 38]. They are particularly useful to test how feasible it is to obtain a proper surface from both materials. For this study, the aim is to obtain similar attenuation in all projections to correctly monitor the relative intensity, and those previous reference standards were not suitable for this purpose; the solution has been to develop test workpieces with cylindrical metal elements which ensure similar levels of attenuation in every projection. With this premise, ad hoc test objects are designed to carry out the experiment, which include a polymeric base, polymeric internal cylinders, and several outer aluminium (Al) and steel (St) hollow cylinders with different thicknesses, to create a series of experiments with a controlled range of relative intensity cases for both materials. Experimental measurements by an XCT device have been performed after first XCT simulations for all the cases; the final aim of this study is to verify the level of correlation between relative intensity and XCT quality, which could help to simplify the setting adjustment in future metal-polymer multi material XCT measurements.
Anzeige
2 Materials and Methods
This section outlines the methodology used for calculating I/I₀, the design principles of the test objects derived from the attenuation curve, and the experimental procedures employed throughout the study.
2.1 Attenuation and Relative Intensity
The primary objective of this study is to establish a correlation between I/I₀ and dimensional accuracy in XCT. To achieve this, the initial step involves designing an appropriate experiment that enables effective monitoring of this parameter.
It was necessary to obtain an attenuation curve for the materials involved in this experiment (Steel and Aluminium). Described by Beer-Lambert law [39], each material has its own attenuation curve, which is calculated by the following Eq. 1:
$$\:I={I}_{0} {e}^{-\mu\:x}$$
(1)
Where I is the intensity received by the detector, I0 is the intensity emitted, µ is the linear attenuation coefficient of the material and x is the penetration length.
XCOM database from NIST [40] has been utilized to obtain the mass attenuation coefficients (µm), which are used for the calculation of linear attenuation coefficients (Eq. 2):
$$\:\mu\:={\mu\:}_{m}/\rho\:$$
(2)
Where ρ describes density of absorbing material. Mass attenuation coefficient curves for the materials employed in this experiment are displayed in Fig. 1.
X-ray attenuation, and mass and linear attenuation coefficient, depend on X-ray energy as shown in Fig. 1. Characteristic energy spectrums have been obtained using software SPEKTR 3.0 [41], introducing the voltage and physical filter selected for each scenario (see Sect. 2.2) and calculating the corresponding Eav from the X-ray spectral fluence, using the X-ray spectrum obtained at each specified potential. An example of the attenuation curves calculated for all materials (St, Al and PA12– Pol) in two different values for Eav are shown in Fig. 2.
Fig. 2
Attenuation curve for St and Al for 2 examples of Eav values
Here, I/I0 represents the relative intensity and x represents the penetration length. It is shown that attenuation strongly varies for each material. It is also seen that path (penetration) length and average energy will also significantly affect relative attenuation.
Four theoretical values of I/I0 have been selected for the experiment. Based on the study of Villarraga et al. [39], I/I0 = 0.08 and I/I0 = 0.16 have been chosen as those values constitute a range in which the threshold for acceptable penetration of the X-rays through the object is found. As the objective of the experiment is to obtain results at different points of the curve, higher relative intensity scenarios have been included too, using multiples of the previously mentioned values (in this case, I/I0 = 0.24 and I/I0 = 0.48). Higher values than I/I0 = 0.48 have not been considered as they would not be realistic for the metals proposed, mainly in steel scenarios, as penetration length (and therefore, thickness) would be too small for the manufacturing of the cylinder.
2.2 Test Workpiece
As explained in Sect. 1, one of the requirements of the study is to maintain similar levels of attenuation in all projections. Although gap reference standards designed by other authors [29, 37, 38] are suitable to test material differentiation in multi-material parts, cylindrical elements are more adequate to keep this similar attenuation in every projection. With this first rule, an ad hoc test artefact has been designed to perform the experiment, consisting of a polymeric base made in polyamide PA12 by selective laser sintering (SLS), with four Ø6 mm polymeric cylinders placed in a square disposition (distanced 12 mm one to another) and various hollow metallic cylinders (outer cylinders) which are switched from case to case. Metals selected are common metal alloys– 6061 aluminium alloy and AISI 1040 steel. Thickness of the cylinder will vary according to the attenuation curve (modifying the outer diameter) and the I/I0 desired by noting the path length extracted from the intersection at the specific attenuation curve value. In Fig. 3, the rest of relevant dimensions of the assembly are presented.
Fig. 3
Here, x is the penetration length calculated from the attenuation curves obtained before. Eav and x are adjusted to refine each I/I0, modifying the dimensions of hollow cylinders (thickness) and the XCT settings (voltage, current and physical filter). In Table 1, a summary of penetration length and average energy selected for obtaining each I/I0 situation is displayed, along with the XCT settings used for obtaining the selected Eav.
Table 1
XCT settings and Attenuation parameters for each scenario, including the resolution of the tomographies defined by the voxel size (Vx). Obtained from [36]
I/I0 | Al | St | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
x (mm) | Eav (keV) | Voltage (kV) | Filter (mm) | Vx (µm) | x (mm) | Eav (keV) | Voltage (kV) | Filter (mm) | Vx (µm) | |
0.08 | 40 | 73 | 190 | Al 2.0 | 47.47 | 8 | 96 | 195 | Cu 1.0 | 47.47 |
0.16 | 30 | 73 | 190 | Al 2.0 | 47.47 | 6 | 96 | 195 | Cu 1.0 | 47.47 |
0.24 | 20 | 63 | 160 | Al 2.0 | 47.47 | 4 | 92 | 195 | Cu 0.75 | 47.47 |
0.48 | 10 | 60 | 160 | Al 1.0 | 47.47 | 2 | 88 | 190 | Cu 0.75 | 47.47 |
Same geometrical magnification (2.66X, considering source-to-object distance = 300 mm and source-to-detector distance = 800 mm), and therefore, same voxel size (Vx = 47.47 μm) has been used for all the scenarios; as well, different physical filters have been selected made by aluminium (Al) or copper (Cu), using Al for aluminium cases and Cu for steel scenarios as the similarity in the density between filter and workpiece allows for a better filtration. Penetration length and, consequently, thickness of hollow cylinders has been rounded to an integer number to optimize the manufacturing of the test parts; from this starting point, XCT settings were adjusted for their correspondent Eav (modifying voltage and physical filter) aiming to approximate as much as possible the real I/I0 to the proposed theoretical values. The two central scenarios are highlighted in the table for each material: after first simulations with all cases, I/I0 ≈ 0.16 and I/I0 ≈ 0.24 scenarios are selected for the experimental evaluation and verification. The intention is to obtain real data in cases with different attenuation but as close as possible to the threshold described by Villarraga et al. [39], which is I/I0 ≈ 0.16, to verify the accuracy of simulated values. In Fig. 4, manufactured hollow cylinders and polymeric base, an example of the assembly (Al 20 mm) and test object positioned in the XCT for the measurement is displayed.
Fig. 4
Details of the manufactured test object. (a) Individual parts (hollow cylinders (Al 20 mm, Al 30 mm, St 4 mm, St 6 mm) and base). (b) Assembly (Al 20 mm). (c) Test object positioned on the XCT device. Extracted from [36]
Tomographies with the workpiece completely straight could cause artifacts, as it is an inappropriate part position [42], therefore, as seen in Fig. 4c, test object is slightly tilted (≈ 15º). This, added to the 15º inclination of the cylinders in the polymeric base (Fig. 3b), ensures the correct geometrical conditions for the evaluation of the measurands. To obtain this base position, test object is placed as shown in Fig. 5. The effect of this inclination on the penetration length x and on the experiment is negligible, as (i) 15º only enlarge the penetration length in x/cos 15º = x/0.966 and (ii) in al the scenarios the variation is proportional. Effective time elapsed for the tomographies has been 31 min, resulting in a rotation velocity of the workpiece of ω = 0.003 rad/s.
Fig. 5
Schematic representation of the test object tilting in the positioning in the XCT system. Extracted from [36]
In addition to Al and St scenarios, a tomography of the base with no metallic cylinder (NM) is done as a reference, to intercompare the measurements with the other cases. Same magnification has been used for all the scenarios (Vx = 47.47 μm), selecting a voltage of 120 kV and physical filter of Al 1.0 mm, resulting in Eav = 54 keV and I/I0 = 0.78 considering the thicker element as the penetration length.
2.3 Evaluation Methodology
Initial XCT simulations were carried out using the software aRTist 2.12 (BAM, Germany) to establish baseline references for each test scenario, including a metal-free case employed for dimensional intercomparison and calibration. Following these simulations, and as previously mentioned, central test specimens made of steel (St) and aluminium (Al) were fabricated. XCT measurements were then performed using the Zeiss Metrotom 800 G3/225 kV system (X-ray tube Viscom xt 9225D with a 7 μm focal spot, target material of wolfram, inherent filtration of 0.4 mm of aluminium and beam orientation of 50°; nominal source-to-object distance of 800 mm, detector pixel size of 127 μm with 1920 × 1536 pixels), with the Metrotom OS software utilized for configuring acquisition settings. Calibration of the device has been done according to normative VDI/VDE 2630 Part 1.3 [43], using automated procedures; therefore, no voxel size re-scaling has been necessary for the experiment.
Post-processing of both the simulated and experimental tomographic data was performed using VG Studio Max 3.4.2. A local gradient threshold surface determination was applied in Advanced– Multi-Material mode, with a search distance set to 4 voxels. No filters or corrections were applied to the reconstructed volumes. Regions of interest (ROIs) for each material were defined following the surface determination process.
Dimensional measurements– in this case, diameters, form deviations, and inter-cylinder distances– were extracted from the evaluation of the four polymeric cylinders. In addition, the contrast-to-noise ratio (CNR) was calculated to assess the quality of the tomographic data. This metric quantifies the contrast between different materials and the background, enabling an evaluation of the noise (in the form of scatter) introduced by the presence of metal in the grey value distribution of both the polymer and the surrounding background. CNR is calculated following the procedure described in [44], using Eq. 3:
$$\:CNR=\frac{\left|{A}_{Material}-{A}_{Background}\right|}{\sqrt{{\sigma\:}_{Material}^{2}+{\sigma\:}_{Background}^{2}}}$$
(3)
where AMaterial and σMaterial represent the mean and standard deviation of the grey values of the material ROIs, respectively; ABackground and σBackground represent the mean and standard deviation of the grey values of the background ROIs, respectively. CNR could be calculated from 2D slices of the volume or from the volume directly; here, in this experiment, volumetric ROIs have been defined as they are more representative of the attenuation of the complete part rather than single images. To calculate the material and background grey levels, a 3-voxel margin is added to the material zone for the background areas which are closer to the part, and therefore, more affected by noise (Fig. 6). CNR values have been calculated for both polymer and metal ROIs.
Fig. 6
ROI disposition for CNR analysis– 2D representation
3 Results
This section presents the evaluation of the test objects, including a comparative analysis of the CNR values and the dimensional measurements obtained.
3.1 CNR
CNR values have been calculated for both polymer and metal ROIs in each scenario, in simulated and experimental tomographies. In Fig. 7, a comparison of results considering I/I0 of each case is displayed, considering attenuation ratios and thicknesses of the cylinders shown previously in Table 1.
Fig. 7
CNR values obtained for each scenario and in each material. Extracted from [36]
Results show, in general, a correlation between higher I/I0 values and better CNR results. However, tendencies are different in Al and steel scenarios: in Al, linear correlation is more evident and consistent, both in polymer and metal ROI. However, St scenarios show lower contrast values and non-linear correlation, particularly in the polymeric area. Same tendencies are found in experimental and simulated results, with slight differences due to the possible errors associated with the simulation software and the uncertainty created by environmental conditions in real XCT. These similarities indicate that simulated tomographies can be used as a valid starting point.
As expected, the denser is the material, the better is the contrast (higher CNR); polymer ROI characterization, on the other hand, is more affected by artifacts and noise. In this aspect, it is observed that for low I/I0 values, CNR differences in polymer ROI are less significant, suggesting that a minimum difference polymer– air will be present no matter the attenuation of the metallic material. Polymeric attenuation, therefore, although is significantly lower than metallic, could not be considered negligible. To check the effect of the attenuation of the polymer on the total attenuation of the workpiece, tomographies of the metallic cylinders alone (without the polymeric base) have been done. Same XCT settings as indicated in Sect. 2.2 have been used, obtaining the CNR values of the metallic elements following the same procedure explained in Sect. 2.3.
A summary of the CNR results obtained is displayed in Table 2, as well as the percentual difference between tomographies of the complete assembly and the metallic cylinders alone.
Table 2
CNR comparison and deviation percentage
Scenario | CNR Assembly | CNR Only Metal | % Dif. Metal |
|---|---|---|---|
St 4 mm | 2,603 | 2,392 | -8,13% |
St 6 mm | 2,537 | 2,302 | -9,28% |
Al 20 mm | 4,677 | 4,295 | -8,17% |
Al 30 mm | 3,678 | 3,273 | -11,02% |
Minding that XCT settings have not varied, results show that contrast of the metallic cylinders does not decrease significantly due to the presence of the polymeric element; therefore, although attenuation of the polymeric base is not negligible, it does not contribute to the noise generated.
2D slices from tomographies of both cases with the same I/I0 value and their correspondent grey values histogram are shown in Fig. 8 (simulated scenarios) and Fig. 9 (real tomographies).
Fig. 8
Visual comparison of simulated Al and St cases with same I/I0. (a) Al 20 mm 2D slice. (b) St 4 mm 2D slice. (c) Al 20 mm grey values histogram. (d) St 4 mm grey values histogram. Figure from [36]
Fig. 9
Visual comparison of real tomographies in Al and St cases with same I/I0. (a) Al 20 mm 2D slice. (b) St 4 mm 2D slice. (c) Al 20 mm grey values histogram. (d) St 4 mm grey values histogram. Figure from [36]
Similar as seen in the CNR calculations (Fig. 7), contrast air– polymer observed is lower in 2D slices and peaks on the histogram are more diffuse for St scenarios, in which material definition is less clear. Beam hardening and scatter also affect steel cylinders in St scenario, blurring the inner area and creating higher dispersion of grey values. Histogram distribution in simulated tomographies (Fig. 8C and D) is significantly different than in real tomographies (Fig. 9C and D); the reason for this phenomenon lies in the limitations of simulation models, which cannot fully account for factors such as image noise, air density fluctuations, and environmental conditions (e.g., temperature and humidity variations) that are inherent to real XCT measurements. Nevertheless, the simulations, which include scattering effect occurred by the particle trajectories, exhibit the same general trends and yield comparable results, indicating that the simulation accuracy is sufficient for the purposes of this study as it has been also proved previously by other authors [45].
3.2 Dimensional Measurements
Diameters, form errors, and distances between features were extracted from measurements of the four cylinders in the polymeric base to obtain probing size, probing form and sphere distance errors. As previously mentioned, the initial XCT measurement of the polymeric base without metal (NM scenario) served as a reference for intercomparison with the four experimental configurations. An analogous procedure was applied to the simulated data, using the simulated NM scenario as a baseline for comparison with the four corresponding simulated metal cases. Deviations from NM scenario are displayed in Fig. 10 for simulations and in Fig. 11 for experimental XCT.
Fig. 10
Deviations from NM scenario in evaluation of polymeric cylinders (simulations). Extracted from [36]
Fig. 11
Deviations from NM scenario in evaluation of polymeric cylinders (real XCT). Extracted from [36]
Results are more stable in Al scenarios, where most of the deviations are in the same magnitude. Similar trend is followed both in simulated and experimental measurement, although in real tomographies results are more randomly distributed (mainly in St scenarios) due to non-modelled factors in simulations as stated in Sect. 3.1. The accuracy is better (less deviations from NM scenario) in scenarios with lower metal thickness– higher relative intensity– for both cases (Al and St), following same trends in scenarios with the same material. In Table 3, mean results of the deviations (in absolute value) in the real tomographies are summarized.
Table 3
Absolute mean values of the deviations (in µm) from reference case
Scenario | I/I0 | Diameters (µm) | Form Errors (µm) | Distances (µm) |
|---|---|---|---|---|
St 4 | 0.24 | 8,152 | 5,129 | 9,467 |
St 6 | 0.16 | 9,779 | 17,351 | 12,968 |
Al 20 | 0.24 | 3,249 | 2,029 | 5,746 |
Al 30 | 0.16 | 5,815 | 6,349 | 4,335 |
In general, higher errors are found as attenuation increases. This effect is particularly significant in form error, where deviations are up to 3 times bigger; as studied before [25], this feature is more sensitive to the noise created by metal. These values obtained in the dimensional measurements are directly related to the results from the CNR analysis, where St scenarios present higher levels of noise and worse contrast both in metal and polymer ROIs, and hence, higher dimensional deviations. The latter suggests that: (i) relative intensity is a relevant factor to consider in metal-polymer assemblies, and (ii) results cannot be extrapolated from one metal to another as the intrinsic properties of the materials affect the quality of the tomography differently.
This point can be highlighted for the particular case of the cylinder nº3 (Cyl3). The behaviour of the measurements obtained from this element seem to be different in St scenarios. This is mainly caused by its position in the assembly (see Fig. 12), as it is the closest cylinder to the metal part and therefore the most affected by the noise created by the steel. However, in Al scenarios its measurements follow the same trend as the rest of the cylinders; again, it is caused by the different noise created by both metals. In conclusion, additionally to the attenuation of the material, the distance from the metal elements highly influences the dimensional results.
Fig. 12
2D slice showing the position of Cyl3
To confirm the traceability of the measurements, uncertainty calculations have been performed for the experimental results following normative VDI/VDE 2630 − 2.1 [46], using Eq. 4:
$$\:{U}_{XCT}=k*\sqrt{{u}_{cal}^{2}+{u}_{p}^{2}+{u}_{w}^{2}+{u}_{b}^{2}+{u}_{drift}^{2}}$$
(4)
where k is the coverage factor (k = 2 for a 95% confidence), ucal is the standard uncertainty of measurement due to the calibration uncertainty of the XCT machine, as stated in its calibration certificate, up is the standard uncertainty of measurement due to the measurement process, (standard deviation of the repeated measurements), uw is the standard uncertainty of measurement due to variations in materials and production (due to variations in, e.g., coefficient of expansion, form errors, roughness, elasticity and plasticity), ub is standard uncertainty of measurement of the correction of the systematic error b and udrift, which is the standard uncertainty of measurement due to the change (drift) in workpiece shape since the calibration referred to.
Here, udrift has been consider negligible as all measurements have been done with no time lapse. For the calculations of ucal, maximum permissible error (MPE) in a rectangular distribution (\(\:\text{M}\text{P}\text{E}/\sqrt{3}\)) has been used, as suggested by Villarraga-Gómez et al. [47]. A MPE = 4 μm for form errors and MPE = 4 + L/100 µm for size and distances measurements has been considered, according to datasheet provided by the manufacturer of the XCT device. For up values, five iterations have been done for each case, without removing the part from the tomograph. Results of the uncertainty calculations are shown in Fig. 13.
Fig. 13
Comparison of uncertainty values in experiments for each scenario and each feature type. Obtained from [36]
Consistent with the trend observed in dimensional deviations, measurement uncertainties increase in scenarios with higher attenuation (i.e., lower I/I₀ values). However, this increase is markedly more pronounced in the steel (St) cases.
All uncertainty values registered are over 4 μm as two of the main contributors (MPE and resolution of the device) are common for all the measurements.
In Table 4, mean values for expanded uncertainty (UXCT) and standard deviation (σ) of the repeated measurements for each scenario and each feature is shown.
Table 4
Summary of mean values for expanded uncertainties and standard deviations
Diameters | Form errors | Distances | ||||
|---|---|---|---|---|---|---|
UXCT (µm) | σ (µm) | UXCT (µm) | σ (µm) | UXCT (µm) | σ (µm) | |
NM | 4.77 | 0.17 | 4.63 | 0.18 | 4.81 | 0.35 |
Al 20 | 4.85 | 0.41 | 4.83 | 0.69 | 4.96 | 0.68 |
Al 30 | 7.34 | 2.77 | 5.01 | 0.91 | 5.53 | 1.37 |
St 4 | 5.30 | 1.04 | 5.23 | 1.03 | 5.47 | 1.28 |
St 6 | 7.50 | 2.72 | 5.44 | 1.34 | 8.67 | 3.57 |
As shown in the table, the differential factor here is the repeatability of the process, which is numerically described as the standard deviation (σ) of the results obtained for each single measurement. Therefore, as measuring repeatability worsen, uncertainties increase.
4 Discussion
As expected, both the CNR analysis and the dimensional evaluation of the test object confirm a correlation between higher relative intensity (I/I₀) and improved XCT quality, in both simulated and experimental datasets. The uncertainty calculations for diameters, form errors, and distances measured on the polymeric cylinders support this trend, showing increased uncertainty with rising attenuation. This is primarily attributed to reduced repeatability, reflected in higher standard deviations (σ) observed for repeated measurements of the same features. The underlying cause is the elevated noise level associated with greater attenuation, which introduces random variations that hinder accurate metrological assessment. Significant differences are found in the dimensional evaluation, where both scenarios with I/I0 ≈ 0.16 presented errors of up to 3 times bigger than scenarios of the same material with I/I0 ≈ 0.24. This is an important increment considering that attenuation is only 33% higher.
However, the effect of attenuation varies depending on the material causing it. Specifically, results obtained for equivalent attenuation levels (i.e., identical I/I₀ values) differ between steel and aluminium scenarios. In steel cases, noise and imaging artifacts are noticeably more pronounced. Visual inspection of 2D slices from the steel scenarios reveals reduced contrast between the polymeric base and the background, while greyscale histograms exhibit poorer material peak separation and greater dispersion in the metal peak—primarily due to beam hardening effects associated with steel. These observations are corroborated by CNR calculations, which consistently show significantly higher contrast in aluminium scenarios across all evaluated regions. From a dimensional perspective, steel scenarios demonstrate reduced repeatability, with greater standard deviations in measurements and larger deviations relative to the reference (non-metal-covered) evaluation. This is further supported by the increased measurement uncertainty observed in the steel cases.
5 Conclusions and Future Work
In this study, an analysis of the attenuation ratio, defined by relative intensity, and its usage to define the expected quality variation of XCT measurements of metal-polymer assemblies is presented. An experimental study has been conducted using an ad hoc designed test object to normalize the attenuation level in every projection of the tomography. The workpiece includes a polymeric base with four polymeric cylinders used as a measurand surrounded by hollow cylinders made by two metals: aluminium and steel, with different thicknesses. CNR calculations of the 3D volumes obtained and dimensional evaluation of the measurands have been performed, both in simulations and real tomographies.
Results show that there is a correlation between lower attenuation (higher relative intensity) and better XCT quality. Scenarios with the reference attenuation threshold [39] of I/I0 ≈ 0.16, present significant deviations of up to 3 times bigger than cases with slight lower attenuation (≈ 33%); this suggests that, in presence of elements with higher radiopacity than this limit, measurements on low– density parts should be taken with caution as important dimensional errors could be found.
Apart from this, values obtained differ significantly between aluminium and steel scenarios, as noise level is considerably higher in St scenarios than Al scenarios with equivalent attenuation. This suggests that extrapolation of results from one material to another is challenging, as not only the relative intensity should be considered. Indeed, this parameter is a good indicator for noise level and XCT quality predictions but its usage alone should be limited to assemblies in which the material configuration is the same. This study stands as a starting point for a generalisation in attenuation effect on XCT evaluation, and it is highly probable that other factors related to the attenuating material would affect to the quality of the tomography, apart from the attenuation ratio registered.
With this knowledge, further experiments will be done with other materials to adequately characterize their behaviour, considering other factors such as the internal structure of the material itself.
Acknowledgements
This work was supported by project PID2021-127134O-B-I00 funded by MCIN/AEI/ https://doi.org/10.13039/501100011033 and by ERDF A way of making Europe.
Declarations
Ethical Approval
All ethical COPE guidelines have been considered.
Consent to Participate
All authors consent to participate in this work.
Consent for Publication
All authors consent to publish this work.
Competing Interests
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.