good practice guide ct€¦ · the objective of the good practice guide (gpg) is to provide a guide...
TRANSCRIPT
1
Good Practice Guide CT
Lonnie Andersen, DTI
Lorenzo Carli, Novo Nordisk
2
Content 1 Objective ...................................................................................................................................................... 3
2 Overview ...................................................................................................................................................... 4
2.1 Scanner (3.1 page 6) ............................................................................................................................ 4
2.2 Sample Preparation (3.2 page 6) ......................................................................................................... 4
2.3 Fixture Specifications (3.3 page 7)....................................................................................................... 4
2.4 Number of Parts (3.4 page 8) .............................................................................................................. 4
2.5 Scan Parameters .................................................................................................................................. 5
2.6 Measurements..................................................................................................................................... 5
2.7 Uncertainty .......................................................................................................................................... 5
3 Good Practice Guide Reference .................................................................................................................. 6
3.1 Calibration and Maintenance of Scanner ............................................................................................ 6
3.1.1 Calibration and Maintenance ...................................................................................................... 6
3.2 Sample Preparation ............................................................................................................................. 6
3.2.1 Temperature Acclimatization ...................................................................................................... 6
3.2.2 Sample Cleaning .......................................................................................................................... 7
3.2.3 Identification of Multiple Samples .............................................................................................. 7
3.3 Fixture Specifications ........................................................................................................................... 7
3.3.1 Fixture Material ........................................................................................................................... 7
3.3.2 Fixture Handling .......................................................................................................................... 8
3.4 Number of Parts .................................................................................................................................. 8
3.4.1 Height, Length and Thickness Ratio............................................................................................. 8
3.4.2 Magnification ............................................................................................................................... 9
3.4.3 Sample Interaction .................................................................................................................... 11
3.4.4 Position Dependency ................................................................................................................. 13
3.4.5 Fixture Level............................................................................................................................... 14
3.4.6 Fixture Angle .............................................................................................................................. 15
4 Uncertainty ................................................................................................................................................ 16
4.1 Assumptions and Approximations ..................................................................................................... 16
4.2 Task Specific Approach ...................................................................................................................... 17
4.3 Example: How to determine the uncertainty of the outer diameter measurement of an industrial
plastic item from Novo Nordisk A/S. ............................................................................................................. 18
4.3.1 Uncertainty Budget - General Procedure: ................................................................................. 18
4.3.2 Calculation of Uncertainty Budget ............................................................................................ 22
4.4 Comments ......................................................................................................................................... 25
3
1 Objective The objective of the good practice guide (GPG) is to provide a guide to achieve the best possible
accuracy in dimensional CT scanning, and thereby decreasing the uncertainty of the CT scan.
The guide deals with some of the aspects faced when using CT-scan in an industrial setting, trying to
optimize scan time, without decreasing the dimensional accuracy or at least considering the effect of
the choices made. This document does not provide a detailed uncertainty budget, but should rather
be seen as a guide as to which parameters to consider and investigate, when faced with increasing
the number of samples per scan to decrease the total scan time, and the parameters that could affect
the scan result in general.
The Guide is comprised of three sections:
Overview: A list of the parameters that should be considered when initializing a scan. Use this
guideline to ensure the best possible result for every scan.
Reference: A section with references to the different points in the overview if more detail is
needed.
Uncertainty: A simple and very general uncertainty budget. Can be used as a starting point
when estimating the uncertainty of the scans.
4
2 Overview The following points should be considered when performing CT scans of one or more samples.
2.1 Scanner (3.1 page 6) Scanner should be routinely calibrated and scale corrected.
2.2 Sample Preparation (3.2 page 6) Acclimatization of samples for at least 4 hours.
Clean Samples: o Metal samples: Alcohol on lint-free Kleenex o Polymer (etc.): Clean for dust with oil free compressed air.
Use markers to clearly identify sample position in fixture when scanning more than one part.
If proper acclimatization is not possible, systematic errors stemming from temperature should be corrected.
2.3 Fixture Specifications (3.3 page 7) Use low-density polymer fixtures machined to fit the part: Insulation foam, dental wax, etc.
Avoid point compression of fixture – increases density -> noise
Use a fixture angle of at least 10°
2.4 Number of Parts (3.4 page 8) Use height/length/thickness ratio to utilize scan volume optimally while not compromising
the voxel size.
If more samples are scanned test the effect of:
o Decreasing the magnification (i.e. less voxels per sample).
o Sample interaction.
o Position dependency.
o Fixture level.
5
2.5 Scan Parameters Choosing the right scan settings can be very difficult, and for the purpose of this guide we
therefore assume that the operator has had some sort of training in how the scanner works
and how to set up the parameters for a good scan.
Most scanners are somewhat robust when parameters are kept within a reasonable range,
see Figure 17 p. 20.
Avoid hard filters when scanning polymer material.
The same parameters should always be used when scanning the same sample for industrial
purposes (not when checking for reproducibility).
Investigate the effect of beam hardening correction when scanning high density materials (do
reconstruction with and without beam hardening correction and compare to CMM
measurements)
2.6 Measurements Use a recommended software package, to ensure the highest possible accuracy.
Ensure that the alignment system, ie the coordinate system used for the registration, is
robust.
Make sure that the extracted features are robust (highest possible fit point density and
relevant fitting method).
If a measurement template is copied from one scan to another, inspect the automatically
generated fit points and if necessary adjust.
2.7 Uncertainty CMM measurements of the scanned part, or a calibrated work piece like reference is needed
to determine the uncertainty.
Reproducibility of scanner should be determined by carrying out at least 15 different scans
while:
o Changing the operator
o Scanning on different times during the day
o Varying the X-ray parameters.
Drift should be measured by doing continuous scans on a calibrated reference such as the
“Birthday cake”
Repeatability of measurements should be determined by carrying out at least 10 repeated
measurements on the same scan, repeating the entire measurement strategy each time.
6
3 Good Practice Guide Reference 3.1 Calibration and Maintenance of Scanner
3.1.1 Calibration and Maintenance The CT scanner should be scale corrected and the length scale in calibration control to
ensure optimal results. The specification of the calibration control depends on the
specifications from the manufacturer.
Recommendation
Regular calibrations should be performed, and the scanner specific calibration plan
should be followed.
3.2 Sample Preparation
3.2.1 Temperature Acclimatization The sample should be acclimatized to room temperature (20±1°C) to prevent
temperature variations. This is especially important when CT scanning is used for
metrology purposes due to material shrinkage. Most scanners used for metrology
purposes will have temperature control, minimizing the effect of temperature during
scanning.
Furthermore, many plastic parts will be affected by humidity, which should be
considered as well.
Recommendation
i. Measurements should be performed in a temperature and humidity
controlled environment if possible.
ii. At least 4 hours is recommended, for acclimatization, but it depends on the
sample material and size.
iii. If less time is allowed for acclimatization, or if environment does not have
sufficient climate control, material shrinkage or expansion can occur resulting
in a systematic error, which should be corrected for using the following
equation:
𝐿𝑐𝑜𝑟 = 𝐿𝑚𝑒𝑎𝑠 + ∆𝐿
where
∆𝐿 = 𝛼 × (𝑇 − 20℃) × 𝐿,
α is the thermal expansion coefficient of the sample, T is the temperature of
the sample during measurement (often this will be the laboratory
temperature), and L is the length measured. If the data is corrected, the
uncertainty of the correction must be included in the uncertainty budget.
7
3.2.2 Sample Cleaning To prevent noise on the scan the parts should be cleaned beforehand and handling
should be minimized.
Figure 1: Effect of dust on a polymer scan
Recommendation
If possible clean samples using pressurized air (air gun), ethanol etc. Gloves can be
used to avoid particle transference.
3.2.3 Identification of Multiple Samples Documentation of placement when more than one sample are scanned at a time is
important. Small identification numbers etc. may not be visible when data is loaded.
Recommendation
Use the same setup in the scanner or use marks of high density on the fixture to
indicate the direction of numbering.
3.3 Fixture Specifications In order to get the best possible scan, or have the possibility of scanning multiple samples at once a
well-designed fixture is very important. Having a fixture, which is invisible on the scan, makes data
analysis much easier due to less noise. Furthermore, a well-designed fixture makes it possible to scan
multiple samples at once decreasing the total scan time.
3.3.1 Fixture Material The fixture material should have a very low absorption coefficient compared to the
scanned material (transparent) to avoid noise on the scan.
Recommendation
Insulation material, dental wax etc. are good candidates for plastic parts
8
3.3.2 Fixture Handling Avoid handling which could increase the density of the fixture material, i.e.
compression. Increasing the density by compression often result in noise on the scan,
which is difficult to remove afterwards.
Recommendation
Handle fixture with care.
3.4 Number of Parts
3.4.1 Height, Length and Thickness Ratio When scanning samples with one dominating dimension, the number of samples per
scan can be easily optimized without compromising the voxel size.
Figure 2: Increasing the number of samples without compromising the voxel size. h is the height of the sample, w is the width, and v is the tilting angle.
The image shows a case where the voxel size will be largely dependent on the height
of the sample. In this case adding more samples in the other directions, makes it
possible to scan more samples without compromising the voxel size.
When applying this, one should consider that subsequent clipping could be difficult
due to tilting of the samples (resulting in material overlap in the clipping box), and
one layer might be the best solution in this case.
Recommendation
Use the dominant measurement to optimize the number of samples in the less
dominant measurement.
Considerations:
9
i. If clipping is needed one layer of samples, (i.e. three in the above example) is
preferable. Avoid visual overlap between the samples due to tilting.
ii. Test effect of part interaction and placement in fixture.
3.4.2 Magnification When adjusting the scan magnification to scan one or more samples two things
should be considered:
i. If the effective volume of the samples is increased the magnification is
decreased and the voxel size increases – See Figure 3.
Figure 3: Relationship between voxel size and source-to-object distance (SOD) (magnification) for the Metrotom 1500.
ii. The highest magnification (i.e. the smallest voxel size) does not always
correspond to a better accuracy. In fact, geometrical errors of the axis, the x-
ray cone angle and other error sources related to the detector can affect the
3D reconstruction. Figure 5 below shows a measurement of an outer
diameter and inner diameter at different source-to-object distances (SOD)
(i.e. different magnifications), while keeping all other settings constant, to
illustrate the error.
0.0
20.0
40.0
60.0
80.0
100.0
120.0
0.0 100.0 200.0 300.0 400.0
Vo
xel S
ize/
µm
SOD/mm
Voxel size vs SOD
10
Figure 4: The sample used in the tests. Red arrow indicate the outer diameter (Dnom = 15.820 mm) and blue the
inner diameter (dnom = 4.377 mm) measured.
Figure 5: Test of the effect of magnification using two different dimensional measurements. The deviation from the nominal value is shown as the relative deviation (deviation divided by nominal
value)
It can be observed, that the relative deviations compared to CMM
measurements do not increase linearly with the voxel size. Thus, sub-voxel
accuracy is achieved (0.1-0.2 voxels at low magnifications). The largest
relative deviation is seen at the smallest voxel size, which could indicate a
calibration problem at the smaller voxel sizes. Such deviations in voxel size
should be investigated further to ensure the correct voxel size is used.
Recommendation
The effect of increasing the voxel size should be tested to determine how this effects
the uncertainty of the measurement.
0.000
0.050
0.100
0.150
0.200
0.250
0.300
40.0 50.0 60.0 70.0 80.0 90.0 100.0 110.0
Rel
ativ
e D
evia
tio
n
Voxel Size (µm)
Effect of Voxel Size
Outer Diameter D Inner Diameter d
11
Test three different magnification: magnification with only one part,
magnification with desired number of parts, and a magnification in the middle
of the two.
Use CMM measurements to determine deviation.
3.4.3 Sample Interaction When scanning several samples at a time, sample interaction should be considered.
The number of parts will increase the amount of material that the X-ray must
penetrate. This means that the X-ray energy must be increased, which in turn can
result in noise effects on the scan.
Figure 6 show the difference between the outer diameter D and the inner diameter d
when the sample is scanned alone in the four different positions (D1 to D4 and d1 to
d4) and together with three other samples. Measuring more parts shows a systematic
underestimation of all measurements except d2 and d4, when compared to reference
measurements due to mutual interaction. However, the deviation is within the
repeatability of the scanner.
Figure 6: The effect of scanning one vs four samples. The result show the effect of scanning one sample alone in four different positions compared to scanning it with three other samples in the same positions. D1-D4 is the outer diameter measured in the four positions; d1-d4 is the inner diameter. All measurements are compared to the CMM value and shown as the relative deviation (Figure 5).
The same tendency was seen on the roundness measurement, where the roundness was
higher when scanning one sample vs four (Figure 7), however this could not be compared to
the CMM measurements.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
D1 D2 D3 D4 d1 d2 d3 d4
Rel
ativ
e D
evia
tio
n
Sample Nr and Measurement
One Sample vs Four
1 Sample 4 Samples
12
Figure 7: The effect of scanning one sample vs four. Measurements are done on the large diameter.
Recommendation
When deciding to scan several samples the effect of one sample versus several should
be tested. If a systematic error is present this should be corrected for, and the
uncertainty of the correction should be included in the uncertainty budget, see
section 4.3.1 p. 18.
0.028
0.029
0.030
0.031
0.032
0.033
0.034
0.035
0.036
R1 R2 R3 R4
Ro
un
dn
ess /
mm
Sample Nr and Measurement
One Sample vs Four
1 Sample 4 Samples
13
3.4.4 Position Dependency Scanning more samples can introduce position deviations due to different detector
areas, position relative to the X-ray beam, and height in the fixture due to the tilting
angle. If the detector is good then ideally there should not be a dependency, however
if the detector is not that good then there can be an effect.
Figure 9 show the relative deviations from reference values of four samples scanned in
four different positions in the fixture. No apparent systematic error is seen, at this
detector level.
Figure 8: Top view position of samples in fixture, showing rotation of samples to test for position dependency.
Figure 9: Investigation of the effect of fixture positions. Each sample is compared to the respective CMM measurement. The deviation is shown as the relative deviation compared to the CMM value.
Recommendation
The quality of the detector and how it affects the deviation from fixture positions
should be investigated when scanning more than one sample. The above example
0
0.02
0.04
0.06
0.08
0.1
0.12
1 2 3 4
Rel
ativ
e D
evia
tio
n
Position Nr
Effect of Position
Item 1 Item 2 Item 3 Item 4
14
(Figure 8 and Figure 9) can be used as an inspiration to such a test setup. If a
systematic error is present this should be corrected for, and the uncertainty of the
correction should be included in the uncertainty budget, see section 4.3.1 p. 18.
3.4.5 Fixture Level Different vertical positions use different areas of the detector (see Figure 10). Figure 11
shows the difference between using a fixture level where the detector is known to be
“good” (middle 2,4) and where it is known to be “bad” (upper 1,3).
Upper position
Middle position
Figure 10: Different detector areas used.
Figure 11: Effect of Level corresponding to different detector areas.
Recommendation
The effect of the height level should be investigated to rule out detector errors. If
there are certain levels where the detector does not perform optimally, these should
be avoided when scanning.
0.000
0.010
0.020
0.030
0.040
0.050
1 2 3 4
Dev
iati
on
/ m
m
Position
Effect of Fixture Level
Middle position Upper position
15
3.4.6 Fixture Angle Tilting the samples reduces the noise in the scan from flat surfaces and limits the
longest distance the X-ray must travel through the material in unsymmetrical
samples. The graph in Figure 12 shows the results of 4 different items, where the
outer diameter (D) was measured with a CMM. The samples were then scanned with
a fixture tilted in different angles. The effect of the tilting angle does not seem to be
very large on the sample scanned, however, this is something that will depend on the
geometry of the sample, and should therefore be tested.
Figure 12: Test for influence of tilting angle.
a. Recommendation
Tilting effect is highly dependent on the sample. The general recommendation is to
use a tilt angle of at least 10°, and if scanning very unsymmetrical samples, or flat
samples, the tilting angle should be investigated (Figure 12).
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0° 10° 30° 45°
Rel
ativ
e D
evia
tio
n
Tilting Angle
Effect of Tilting Angle
Sample 1 Sample 2 Sample 3 Sample 4
16
4 Uncertainty
There are numerous factors and parameters that affect CT uncertainty (see Figure 13), making its estimation quite challenging. For this reason, a task specific approach is recommended in industry where some assumptions and approximations are needed to simplify the problem. Figure 13 is color coded according to the contributions to the uncertainty budget (see table 1).
Figure 13 Ishikawa diagram illustrating factors influencing the uncertainty in CT.
4.1 Assumptions and Approximations The color coding in Figure 13 is used to illustrate how the many contributions to the uncertainty
budget of CT measurements can be simplified. These general approximations are made:
1. Green: The parameters in the green area are pooled into the contribution from
reproducibility (up).
2. Purple: The parameters in the purple area are considered in the measurement uncertainty
(um) related to the method
3. Red: The parameters in the red area are considered as a part of the uncertainty from the
temperature fluctuations (utemp). If there are known environmental factors which are not
controlled, these should be included as a separate contribution.
4. Blue: The parameters in the blue area can be included in the uncertainty due to drift (udrift).
5. Yellow: The parameters in the yellow area will depend on the object scanned. In this case
these parameters are not considered as an individual contribution, however if the object is of
17
high density or is made from multiple materials this effect should be included in the
uncertainty budget.
It is assumed that the number of parts, filter, fixture, scan parameters etc. are chosen
beforehand following the suggestions described in the guideline, and they will be kept
constants whenever performing repeated CT scan of the same item over time.
4.2 Task Specific Approach The estimation of measuring uncertainty for CT scanning depends on the specific task under
consideration and cannot be generalized to any item, strategy, software etc.
Systematic errors must be corrected choosing one of the following approaches:
1. Perform reference measurements on different equipment (CMM, optical scanner,
profilometer etc.).
2. Scan a reference object under similar experimental conditions as the parts under
considerations.
In both cases, use the results to compensate for CT measurements.
Random errors must be estimated and included in the uncertainty budget. The following error
sources should be taken into account (at least):
Table 1: Random errors that should be included in the uncertainty budget.
Symbol Uncertainty Description Distribution Quantification
ucal Calibration Uncertainty of reference
measurements or calibration certificate
Gauss,1 𝑈𝑐𝑎𝑙
𝑘
udrift Drift Drift in calibrated work
piece between calibration dates.
Gauss,1
𝑆𝑇𝐷𝑐𝑎𝑙 𝑟𝑒𝑓
utemp Temperature
(object+environment) Temperature effects U*, 2
𝛼 × 𝐿 × ∆𝑇
√2
up Reproducibility Reproducibility of the
scanner.
Rectangular
, 3
𝑆𝑇𝐷𝐶𝑇 𝑅𝐸𝑃
√3
um
Software repeatability (alignment +
measurement)
STD of repeated measurements on the
same scan. Gauss, 1 𝑆𝑇𝐷𝑅𝐸𝑃 𝑀𝐸𝐴𝑆
ub Bias Reference measurements Gauss, 1
18
* The temperature has been chosen as a U-shaped distribution, due to known fluctuation. Using a U-
shaped distribution the extreemes are taken into account (worst case scenario) and the contribution
is not underestimated. A rectangular distribution could be used if this is not the case.
The resulting Task Specific uncertainty is:
𝑈𝐶𝑇 = 𝑘√𝑢𝑐𝑎𝑙2 + 𝑢𝑑𝑟𝑖𝑓𝑡
2 + 𝑢𝑡𝑒𝑚𝑝2 + 𝑢𝑝
2 + 𝑢𝑚2 + 𝑢𝑏
2
4.3 Example: How to determine the uncertainty of the outer diameter
measurement of an industrial plastic item from Novo Nordisk A/S.
The following example shows how to use the above uncertainty parameters (Table 1) in the
uncertainty estimation of a specific item.
CMM and CT measurements are performed on eight industrial items from Novo Nordisk A/S, the
outer diameter “D” (see Figure 14) is measured, systematic errors are considered, and if possible,
corrected for and the uncertainty budget is determined.
Figure 14 Drawing of the item used in the uncertainty budget calculation. The blue arrow shows the outer diameter D, with a nominal value of 15.820 mm
4.3.1 Uncertainty Budget - General Procedure: The following steps were performed in the determination of the uncertainty budget.
Step 1: CMM measurements were performed for all the items under consideration, in this case
8 items (see Figure 18).
Step 2: CT scan and measurements were performed following the instructions of this guideline
(see Figure 18). Four samples were scanned at the same time (see Figure 15.)
19
Step 3: The reproducibility was tested having 3 different operators make 5 repeated scans (see
Table 2). The outer diameter was measured from each scan (see Figure 17), and the
standard deviation was calculated for each person. The largest standard deviation was
used as the uncertainty contribution due to reproducibility.
Table 2: Scan settings in reproducibility test.
Current (µA) Voxel Size (µm) Voltage (kV)
Operator 1
233 49 180
283 44 160
290 49 180
284 44 160
259 44 160
Operator 2
381 52 190
306 49 180
224 44 160
401 55 200
290 49 180
Operator 3
330 49 180
300 49 180
309 47 170
308 47 170
306 47 170
Figure 15 Four items scanned at the same time.
Figure 16 CAD model of the piece.
20
Figure 17: Reproducibility results for the outer diameter based on five different scans by each of the three different operators.
Step 4: Comparison of CT and CMM measurements to compensate for systematic errors (see
Figure 18). The difference between the mean value of the CMM measurements and the
mean value of the CT measurements can be used to correct for the systematic error.
When doing this the uncertainty of this correction must be included in the budget (ub).
Figure 18 CMM and CT measurements comparison.
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
Operator 1 Operator 2 Operator 3
Rel
ativ
e D
evia
tio
n /
mm
-1
Reproducibility
Serie1 Serie2 Serie3 Serie4 Serie5
15.770
15.775
15.780
15.785
15.790
15.795
15.800
15.805
15.810
1 2 3 4 5 6 7 8
Mea
sure
d v
alu
e /m
m
Item
CT versus CMM
CMM CT AVG CMM AVG CT
21
The 8 CMM values were plotted against the 8 CT values and a “least square fit” was
performed. The slope of the curve was used to compensate the systematic errors. The
contribution from part to part variation i.e. the form error of the work piece will also be
taken into account in the bias.
The data was corrected using the following equation:
𝐶𝑇𝑐𝑜𝑟 = 𝐶𝑇 ×𝐶𝑀𝑀
𝐶𝑇,
where 𝐶𝑀𝑀̅̅ ̅̅ ̅̅ ̅ and 𝐶𝑇 are the mean values of the CMM and CT values respectively. The
results of the correction can be seen in Figure 19 and Figure 20.
Figure 19: CMM and corrected CT data.
15.780
15.785
15.790
15.795
15.800
15.805
15.810
1 2 3 4 5 6 7 8
Mea
sure
d v
alu
e /
mm
Item
Corrected CT Data vs CMM Data
CMM CT
22
Figure 20: The CMM data plotted against the CT data. The goodness of fit shows that the correction factor (1.008) fits the data well.
Step 5: Calculate the uncertainty budget for CT measurements.
4.3.2 Calculation of Uncertainty Budget With reference to Table 1 and the resulting uncertainty budget, the 6 different contributors must be
evaluated:
ucal: Uncertainty of reference measurements for the 8 items measured on a CMM (Ucal).
udrift: In this example there is a calibrated workpiece, and the contribution from drift can
therefore be neglected.
However, if there is no calibrated workpiece, the following should be considered. The
contribution from drift considers the uncertainty of the CT calibration by scanning the
reference artefact three times. This is basically the uncertainty of scale error correction, and
in this case this is performed by the person performing the calibrations (i.e. Zeiss).
o If calibrations are done in house by the operator, the drift can be calculated by doing
regular scans of a calibrated reference artifact (such as the “birthday cake”). The
standard deviation of these measurements can then be used as a measure of the
drift.
utemp: Depends on the material, the temperature and the dimension being measured. In this
case the nominal dimension is 15.820 mm and the material is PC/ABS. The thermal expansion
coefficient is α = 70.2x10-6 m/(m K). In this case it is assumed that the temperature variation in the
y = 1.0008xR² = 0.7076
15.792
15.794
15.796
15.798
15.800
15.802
15.804
15.806
15.808
15.810
15.778 15.780 15.782 15.784 15.786 15.788 15.790 15.792 15.794 15.796
CM
M m
easu
rem
ent
/ m
m
CT measurement / mm
CT vs CMM
23
laboratory is no more than ± 1K, making ΔT = 1 K (half width). The uncertainty contribution can be
easily calculated using the equation in Table 1. If the temperature fluctuations are measured to be
larger than this the uncertainty contribution from temperature will be larger as well.
up: Estimated from 3 different operators performing 5 different scans by following this guideline.
Scanning parameters were adjusted every time to achieve optimal image quality. Voltage,
current and voxel size where therefore slightly different for all the 15 scans (see Table 2).
um: For a given point cloud, the post-processing operations were repeated 12 times to
investigate the influence of the alignment to the CAD model and measuring strategy.
ub: The standard uncertainty of the correction of systematic errors (i.e. goodness of the fit
for systematic error correction).
The bias was calculated as the standard deviation of the difference between the 8 CMM
values and the corrected CT values (see Fig. 18)
24
Table 3: Uncertainty Contributions
Contributor Calculation
(μm) %
contribution
ucal 6𝜇𝑚
2 3.0 50.1
utemp 1.1𝜇𝑚
√2 0.8 3.4
up 2.9𝜇𝑚
√3 1.7 15.6
um 0.5𝜇𝑚 0.5 1.4
ub 2.3𝜇𝑚 2.3 29.5
UCT 2√3.02 + 0.82 + 1.72 + 0.52 + 2.32 8.5
Expanded Uncertainty (k=2)
Figure 21: The percentage uncertainty contribution from each of the six factors.
u_cal u_temp u_p u_m u_b
50.1
3.4
15.6
1.4
29.5
Un
cert
ain
ty C
on
trib
uti
on
%
UNCERTAINTY CONTRIBUTION
25
4.4 Comments Looking at the percentage contribution of each error source (Figure 21 and Table 3), it can be seen that in this
specific case there is not a dominant error contributor. If the expanded uncertainty should be reduced, the
following improvement can be considered:
1. The systematic error compensation, and the resulting bias were based on measurements of eight
items. By increasing the number of measurements a better estimate of the systematic error can be
achieved, thus reducing the bias.
2. The operator was free to choose the optimal scanner parameters (voltage, current, magnification i.e.
voxel size) to perform the task. By finding the optimal parameters for this specific task the variability
of the up contributor can be decreased.
Last but not least, it should be noted that the resulting expanded uncertainty is task specific, and only applies
to measurements of this specific item, using a specific fixture, and only at the experimental conditions
described in the introduction of this section.
When measuring other plastic parts this general approach to uncertainty is still valid, but the results from this
budget cannot be transferred to a different part.
When measuring metal parts for example, the influence of filters, image artefacts like beam hardening, and
effects of scattering and absorption might highly influence the resulting uncertainty.
26