Assessing Additive Manufacturing Processes with X-ray CT Metrology

Conference Paper (PDF Available) · May 2015with 2,092 Reads 
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
·
Conference: 2015 ASPE SPRING TOPICAL MEETING, At Raleigh, NC, Volume: VOLUME 60
Cite this publication
Abstract
In a recent review of CT metrology issues it was stated that, “Industry can no longer accept that intricate components produced by additive manufacturing or multi-material injection molding escape any geometrical and tolerance quality control for the only reason that there is no nondestructive method to measure the inner or internal geometry” [1]. Recently, with the development of X-ray computed tomography (CT) systems designed for applications of industrial metrology, CT has become the only commercially-available, non-destructive method to perform dimensional measurements of internal geometrical features [1, 2]. The aim of this study is to illustrate its applications for assessment of additive manufacturing (AM) processes. This paper presents a comparative study using CT for dimensional metrology (part to CAD comparison, wall thickness analysis, size) and for material inspection (void detection) to examine the outcome of two different AM processes that produced two identical flexure mechanisms. The flexures were fabricated by fused deposition modeling (FDM) and stereolithography (STL) methods at the Department of Mechanical Engineering from the University of South Carolina at Columbia. ABS (Acrylonitrile butadiene styrene) of elastic modulus 2.5GPa and a photopolymer material of elastic modulus 1.85GPa are used in with FDM and STL processes, respectively, and the two identical double-compound notch-type flexure mechanisms were fabricated by two different processes. Each leaf string is L25.0×W20.0×T1.0 mm3. Because the approach in this work focuses on the applications of CT metrology as a tool for assessment of manufacturing processes, particularly AM, with the aim of generalization and for pedagogical purposes, the FDM method will be referred to as AM1 and the STL method will be referred to as AM2. For reference, the FDM and STL refer to processes by which digital 3D design data is used to build a component in layers by depositing various polymeric or composite materials. In general, FDM and STL have a printing resolution of approximately 100 μm and 0.5 μm, respectively. In this research, the hypothesis is two-fold: CT technology can be used as an inspection/metrology tool for AM products, and AM technology can be used for precision engineering applications.
Figures - uploaded by Herminso Villarraga-Gómez
Author content
All content in this area was uploaded by Herminso Villarraga-Gómez
Content may be subject to copyright.
Full variance distribution for deviations of the CAD model in the outcome of the AM1 process. A cross section of the part-to- CAD comparison at a plane located at Z = −18.89 mm from the top part’s planar face is shown. Parts’ surface determination was performed on the original scan voxel data with VGStudio MAX 2.2 [5]. This software uses the CT reconstructed data to create a three-dimensional model of the volume that can be used in 2 or 3 dimensions. From the measurements depicted in Figure 2 for example, it can be seen that deviations from the nominal geometry (CAD model) rise up to ±0.25 mm and larger when using the AM1 process. In contrast, the AM2 process generated part-to- CAD deviations mostly between ±0.1 mm with a few exceptions particularly around the edges or corners of the part’s surface , see Figure 3. Deviations up to -4 mm were detected between measured and nominal dimensions of the CAD model of the part (flexure) for the AM1 process, Figure 2. These particular tolerance outcomes might be related to the printing resolution limits associated with the AM1 and AM2 processes, which have a printing resolution of approximately 100 μm and 0.5 μm, respectively. Individual components tolerances in the flexures can be tested, but the part-to-CAD comparison using the CT data offers the advantage of fully assessing the parts’ surface everywhere. This comprehensive assessment allows for decreased inspection time when comparing the AM1 and AM2 processes. The representation of CT data is based on the original voxel data to avoid measurement uncertainties caused by a second data representation (Standard tessellation language or STL conversion, for example, add inaccuracies to the surfaces during the triangulation process). This representation based on voxels provides geometric evaluations in 2D or 3D such as, nominal/actual comparisons and
… 
Advertisement
116 201 5 SPRING TOPICAL MEETING VOLUME 60
ASSESSING ADDITIVE MANUFACTURING PROCESSES WITH X-RAY
CT METROLOGY
Herminso Villarraga1, 2, ChaBum Lee4, Taylor Corbett4, Joshua A. Tarbutton4,
and Stuart T. Smith1, 2, 3
1Center for Precision Metrology, 2Department of Physics and Optical Science, and
3Department of Mechanical Engineering and Engineering Science
University of North Carolina at Charlotte, NC, USA.
4Department of Mechanical Engineering
University of South Carolina, Columbia, SC, USA.
In a recent review of CT metrology issues it was
stated that, “Industry can no longer accept that
intricate components produced by additive
manufacturing or multi-material injection molding
escape any geometrical and tolerance quality
control for the only reason that there is no non-
destructive method to measure the inner or
internal geometry” [1]. Recently, with the
development of X-ray computed tomography
(CT) systems designed for applications of
industrial metrology, CT has become the only
commercially-available, non-destructive method
to perform dimensional measurements of internal
geometrical features [1, 2]. The aim of this study
is to illustrate its applications for assessment of
additive manufacturing (AM) processes. This
paper presents a comparative study using CT for
dimensional metrology (part to CAD comparison,
wall thickness analysis, size) and for material
inspection (void detection) to examine the
outcome of two different AM processes that
produced two identical flexure mechanisms. The
flexures were fabricated by fused deposition
modeling (FDM) and stereolithography (STL)
methods at the Department of Mechanical
Engineering from the University of South Carolina
at Columbia. ABS (Acrylonitrile butadiene styrene)
of elastic modulus 2.5GPa and a photopolymer
material of elastic modulus 1.85GPa are used in
with FDM and STL processes, respectively, and
the two identical double-compound notch-type
flexure mechanisms were fabricated by two
different processes. Each leaf string is
L25.0×W20.0×T1.0 mm3. Because the approach
in this work focuses on the applications of CT
metrology as a tool for assessment of
manufacturing processes, particularly AM, with
the aim of generalization and for pedagogical
purposes, the FDM method will be referred to as
AM1 and the STL method will be referred to as
AM2. For reference, the FDM and STL refer to
processes by which digital 3D design data is used
to build a component in layers by depositing
various polymeric or composite materials. In
general, FDM and STL have a printing resolution
of approximately 100 µm and 0.5 µm,
respectively. In this research, the hypothesis is
two-fold: CT technology can be used as an
inspection/metrology tool for AM products, and
AM technology can be used for precision
engineering applications.
FIGURE 1. Part to CAD comparison for the
outcomes of the AM1 and AM2 processes.
CT METROLOGY ANALYSIS
Figure 1 shows the variance analysis for the AM1
and AM2 flexures against the CAD model for the
parts. Here, the testing surface’s location for the
part is generated from a three-dimensional voxel
set of data (grey scales or absorption intensities)
obtained by the CT scanning of the parts with X-
rays . To allow dimensional measurements using
the X-ray CT technique, the parts were scanned
with a Zeiss Metrotom 800, CT machine [3, 4].
AMERICAN SOCIETY FOR PRECISION ENGINEERING 117
FIGURE 2. Full variance distribution for
deviations of the CAD model in the outcome of
the AM1 process. A cross section of the part-to-
CAD comparison at a plane located at Z =
18.89 mm from the top part’s planar face is
shown.
Parts’ surface determination was performed on
the original scan voxel data with VGStudio MAX
2.2 [5]. This software uses the CT reconstructed
data to create a three-dimensional model of the
volume that can be used in 2 or 3 dimensions.
From the measurements depicted in Figure 2 for
example, it can be seen that deviations from the
nominal geometry (CAD model) rise up to ±0.25
mm and larger when using the AM1 process. In
contrast, the AM2 process generated part-to-
CAD deviations mostly between ±0.1 mm with a
few exceptions particularly around the edges or
corners of the part’s surface, see Figure 3.
Deviations up to -4 mm were detected between
measured and nominal dimensions of the CAD
model of the part (flexure) for the AM1 process,
Figure 2. These particular tolerance outcomes
might be related to the printing resolution limits
associated with the AM1 and AM2 processes,
which have a printing resolution of approximately
100 µm and 0.5 µm, respectively. Individual
components tolerances in the flexures can be
tested, but the part-to-CAD comparison using the
CT data offers the advantage of fully assessing
the parts’ surface everywhere. This
comprehensive assessment allows for decreased
inspection time when comparing the AM1 and
AM2 processes. The representation of CT data is
based on the original voxel data to avoid
measurement uncertainties caused by a second
data representation (Standard tessellation
language or STL conversion, for example, add
inaccuracies to the surfaces during the
triangulation process). This representation based
on voxels provides geometric evaluations in 2D
or 3D such as, nominal/actual comparisons and
all further geometric analyses like wall thickness
analysis and dimensional metrology with
uncertainties of less than 1/10th of a voxel [5].
FIGURE 3. Full variance distribution for
deviations of the CAD model in the outcome of
the AM2 process. A cross section of the part-to-
CAD comparison at a plane located at Z =
18.89 mm (from the top) is shown.
A cross section of the parts at a plane located at
 = −18.89 mm (from the top part’s top planar
surface) is presented in Figure 4, showing
residual internal and spatial deformations
detected in the flexure AM1 (top image). On the
other hand, the manufactured part generated by
the AM2 process does not reveal the presence of
major deformations in the thin walled flexure leaf
structures.
FIGURE 4. Transversal cross section for the AM1
and AM2 parts at  = −18.89 mm. A sketch of
the CAD model is shown superposed on the AM2
part.
118 201 5 SPRING TOPICAL MEETING VOLUME 6 0
A 3D wall thickness analysis on the CT volume
data was also applied to locate the positions and
size of the individual areas with thinner wall
thicknesses, and the results can be visualized by
color-coding as shown in Figure 5 and Figure 6.
Due to the high density of porosity for the AM1
part, very thin wall thicknesses are generated
with the AM1 process, as can be seen from
Figure 5. Most walls for the manufactured part, in
that case, are in the region of 0.3 to 1.3 mm thick.
Only a small fraction of the total part’s surface
gets over the 2 mm thickness. On the other hand,
the wall thicknesses generated with the AM2
process (see Figure 6) for an identical part’s
design range between 1 to 22 mm making the
walls not only wider but stronger. Therefore, we
do not see evidence in the AM2 part of the same
kind of internal and spatial deformations detected
in the flexure AM1 (top image from Figure 4).
However, the method for wall thickness
detections is inaccurate at any corner of the part
under investigation [6], so those regions are not
assessed.
FIGURE 5. Full part’s wall thickness distribution
for the outcome of the AM1 process. A color
coded distribution for the cross section of the
AM1 part (at  = −18.89 ) is also shown.
FIGURE 6. Full part’s wall thickness distribution
for the outcome of the AM2 process.
CT DEFECT ANALYSIS
CT provides defect volume analysis that
determines voids within the manufactured parts,
their position, their size, volume and morphology.
Figures 7 and 8 show the porosity analysis or
defect volume distribution for the outcome of the
processes AM1 and AM2 respectively.
FIGURE 7. Defect volume distribution for the
outcome of the AM1 process, voids count in
function of their size. A transparent view of the
AM1 part depicting the internal defects is shown.
FIGURE 8. Defect volume distribution for the
outcome of the AM2 process.
Statistical information for the full defect volume
distribution within the whole part for each case
(AM1 and AM2) is presented. Particularly, a
higher density of porosity for the AM1 part is
revealed apparent in contrast to the AM2 part,
mainly distributed in planar layers stacked along
the Z-axis, which is the same direction along
which the manufacturing processes deposit the
layers of (ABS) thermoplastic material. Figure 7
shows an accumulation of voids for the AM1
process mostly in the range of 0.002 mm3 to 0.2
mm3 volumetric size (0.08 mm to 0.36 mm
radius), while for the AM2 process the majority of
voids (fewer than for AM1) have a size major
dimension of only between 0.002 mm3 to 0.02
mm3 (0.08 mm to 0.17 mm radius). These defects
(sizes) during the manufacturing processes might
be a reflection of the resolution limits of the AM
AMERICAN SOCIETY FOR PRECISION ENGINEERING 119
systems, or inability to produce very thin layers
(stacks) of material while building the part. Thus,
the printing resolutions of approximately 100 µm
for the AM1 process and 0.5 µm for the AM2
process can be associated with larger volumes of
voids or air bubbles inside the AM1 part than that
of an identical part built by the AM2 process.
Because those can affect material properties
(elastic modulus), that determine stiffness and
damping of the flexures. Moreover, fatigue and
long-term reliability issues can be raised due to
air-voids (or air bubbles).
FIGURE 9. Region of interest for the AM2 part
highlighting voids. 65 µm (up) and 20 µm (down)
CT scan voxel sizes (resolution) were used.
Void detection is limited by the CT scanner
resolution, and thus pores with a diameter of the
order of one voxel or less cannot be detected or
differentiated from noise in the CT image.
VGStudio MAX 2.2 actually allows a user defined
limit for void detection, which is recommended to
be in a minimum size of 2 voxels in diameter to
avoid false detection in the noise level of the data.
Therefore, a default parameter set to a volume of
approximately 8 voxels (=2x2x2) was applied for
the AM1 and AM2 parts studied here, which were
scanned with a voxel resolution of Vx=65 µm.
Voids with volumes in the order of (0.13 mm)3 =
0.002 mm3 or less cannot be assessed. Figure 9
shows a region of interest for the AM2 part
scanned with two different resolutions, 65 µm and
20 µm voxel sizes, and this shows how the 20 µm
scan not only makes the holes easier to detect
but additional holes were found.
A COMPLEX PART MANUFACTURED BY THE
AM2 PROCESS:
FIGURE 10. Wall thickness distribution for the
monolithic double compound-type flexure
mechanism manufactured by the AM2 process.
With the information acquired from CT analysis of
the AM1 and AM2 parts in the previous section,
the AM2 process was selected for manufacturing
a complex part called the monolithic double
compound-type flexure mechanism.Figure 10
shows the thickness distribution for the monolithic
double compound flexure with most of the walls
being approximately 2 mm thick. On the other
hand, part-to-CAD comparison for the same part
is shown in Figure 11. Again exhibiting tolerances
mostly around ±0.1 mm from the nominal
geometry of the CAD model. This is summarized
in Figure 12 which also reveals the existence of a
substantial portion of the flexures’ surfaces
outside of the ±0.1 mm tolerance. Those regions
are highlighted in Figure 11 (and Figure 12), with
the purple color representing deviations from
nominal geometry below the -0.1 mm, and a pink
color representing deviations above 0.1 mm.
120 201 5 SPRING TOPICAL MEETING VOLUME 6 0
FIGURE 11. Part to CAD comparison for the
monolithic double compound-type flexure.
FIGURE 12. Variance distribution for deviations
of the CAD model for the monolithic double
compound-type flexure mechanism.
Recalling the approximate 1/10th of a voxel size
for the uncertainty estimation of dimensional CT
metrology [5], since the monolithic double
compound flexure was CT scanned with a 91 µm
voxel size, an uncertainty of around 9.1 µm for the
geometrical evaluations presented here for this
part (the monolithic double compound flexure)
should be expected. However, this is a rough
estimation, and could be an overestimation since
the maximum permissible error (MPE) from the
CT machine certificate used for the research
presented here, is stated as  = (8 + /100)
µm with the length in mm [7]. The
approximation /2 might be a good
estimation of uncertainty coming from the CT
machine [8], but this estimation only accounts for
uncertainty coming from the repeatability of the
measurement instrument during its calibration
process. A more thorough uncertainty analysis to
include contributions coming from calibration,
statistics (repeatability of measurements),
systematic errors, and workpiece variations
needs to be done. This is beyond the scope of
this paper, and references [9, 10, and 11] provide
a more detailed analysis of these issues. Still, it is
important to note that for the AM tolerances
reported here, mostly approximately ±0.1 mm or
more, uncertainties in the reported
measurements in the order of 10 µm do not
change the conclusions in this paper.
CONCLUSIONS
AM is capable of manufacturing highly complex
3D structures, layer by layer, that can be light and
stable. CT provides a quantitative visualization of
objects layer by layer as well as in 3D. This
makes the CT technology particularly useful to
study the outcome of an AM process, not only for
inspection or material quality control but also for
dimensional metrology. In particular, tolerances
can be contrasted between different processes,
with non-destructive measurement of the additive
manufactured product. There is considerable
benefit in measuring the products in the
assembled state, not only for results comparison
(as in this paper), but also for tracking
dimensional changes after assembly of
modifications deliberately applied to a single
process or to a particular component. The
artifacts, AM1 and AM2, presented in this paper
could also represent two trials of a part that was
evaluated and a modified in an AM process.
From this work, it is evident that residual internal
deformations and spatial deformations were
detected in the manufacture of flexures with the
AM1 process (see Figure 4), and deviations up to
-4 mm were detected between measured and
nominal dimensions from the CAD model of the
flexure (see Figure 2). Improvement in the
manufacturing process for the part was observed
by using the AM2 process, reaching tolerances
down to ±0.15 mm (see Figure 3). These
particular tolerance outcomes might be related to
the printing limitations associated with the AM1
and AM2 processes, which in general have a
printing resolution of approximately 100 µm and
0.5 µm, respectively. Thus, as might be
expected, a larger volume of voids inside the AM1
part (Figure 7) compared to the defect volume
distribution for a similar part built by the AM2
process (Figure 8) was detected. However, void
detection is limited by the CT scanner resolution,
and thus pores with a diameter in the order of one
voxel or less could not be detected or
differentiated from noise in the CT images
presented here. For reference, AM1 and AM2
AMERICAN SOCIETY FOR PRECISION ENGINEERING 121
parts studied here were scanned with a voxel
resolution of Vx = 65 µm, while the monolithic
double compound flexure was CT scanned with a
91 µm voxel size. The density of defect
inclusions inside the AM1 and AM2 parts seems
to correlate to the printing limitations associated
with the AM1 and AM2 processes, which is also
reflected in the respective measurements of part
thickness distributions shown in Figure 5 and
Figure 6.
With the information gained from this study, and
additional research on characterization of
polymers fabricated by AM processes [12], the
monolithic double compound-type flexure
mechanism was manufactured by the AM2
process. An initial characterization of the part by
CT is shown in Figures 10 through 12, and in the
future, additional dimensional metrology
(measurements of size) will be performed on the
features of interest in terms of functional
performance of the flexure mechanism. Lastly,
although a full uncertainty analysis has not been
completed for the measurement results
presented here, it is important to note that for the
AM tolerances reported, mostly around ±0.1 mm
or above, typical CT uncertainties in the order of
10 µm are a factor of ten lower, therefore
indicating that these tolerances are
predominantly those of the AM process.
ACKNOWLEDGMENTS
The authors would like to express their
appreciation to Carl Zeiss Industrial Metrology,
LLC, for providing the CT measuring machine,
and to Volume Graphics GmbH for providing
VGStudio MAX. This study could not have been
completed without their collaboration.
REFERENCES
[1] Kruth J. P., Bartcher M., Carmignato S.,
Schmitt R., De Chiffre L., and Weckenmann
A. Computed tomography for dimensional
metrology. CIRP Annals, Manufacturing
Technology 60 (821-842), 2011.
[2] Villarraga H., Morse E. P., Hocken R. J., and
Smith S. T. Dimensional metrology of
internal features with X-ray computed
tomography. Proc. of ASPE Annual Meeting,
Vol 59, 2014.
[3] Benninger R, METROTOM 800 computer
tomograph. Inovation Special Metrology 11
(6-7), The Magazine from Carl Zeiss, 2009.
[4] Lettenbauer H, METROTOM Measure in a
new dimension. Inovation Special Metrology
9 ((12-13), Magazine from Carl Zeiss, 2007.
[5] Reinhart C, Industrial computer tomography
A universal inspection tool. 17th World
Conference on Nondestructive Testing, Oct
2008, Shanghai, China.
[6] Reinhart C., Poliwoda T, Guenther T.,
Roemer W., Maass S., and Gosch C.
Modern voxel based data and geometry
analysis software tools for industrial CT. 16th
World Conference on Nondestructive
Testing, 2004, Montreal, Canada.
[7] Lettenbauer H., Georgi B., and Wei β D.
Means to verify the accuracy of CT systems
for metrology applications (in absence of
established international standards).
Symposium on Digital Industrial
Radiography and Computed Tomography,
2007, Lyon, France.
[8] Gupta S. V. Measurement Uncertainties:
Physical Parameters and Calibration of
Instruments, 2012, Berlin Heidelberg:
Springer-Verlag.
[9] Villarraga H., Thousand J. D., Morse E. P.,
and Smith S. T. CT measurements and their
estimated uncertainty: The significance of
temperature and bias determination. 15th
International Conference on Metrology and
Properties of Engineering Surfaces
proceedings, 2015, Charlotte, NC.
[10] Villarraga H., Clark D. O., and Smith S. T.
Strategies for coordinate metrology on
flexible parts: CMMs vs CT. 15th
International Conference on Metrology and
Properties of Engineering Surfaces
proceedings, 2015, Charlotte, NC.
[11] Villarraga H., Morse E. P., Hocken R. J., and
Smith S. T. Dimensional metrology of
internal features with X-ray computed
tomography. ASPE 2014 Annual Meeting,
pp. 684-689, Boston, MA.
[12] Corbett T., Kok T., ChaBum L., Smith S. T.,
Villarraga H., and Tarbutton J. A.
Identification of mechanical and fatigue
characteristics of polymers fabricated by
additive manufacturing process. ASPE 2014
Topical Meeting, Apr. 13-16, Berkeley, CA.
  • ... This is illustrated in Figure 7 ( Bauza et al., 2015). Other authors (Villarraga et al., 2015) used X-ray CT to evaluate AM parts made from ABS and photopolymer material using an FDM and a SLA process, respectively. CT scan results showed evidence of internal residual deformations and deviations from the nominal CAD model up to 4 mm and 0.15 mm for the FDM and SLA process, respectively (Villarraga et al., 2015). ...
    ... Other authors (Villarraga et al., 2015) used X-ray CT to evaluate AM parts made from ABS and photopolymer material using an FDM and a SLA process, respectively. CT scan results showed evidence of internal residual deformations and deviations from the nominal CAD model up to 4 mm and 0.15 mm for the FDM and SLA process, respectively (Villarraga et al., 2015). They also successfully mapped voids inside the printed components; however, they were not able to detect voids with a size on the order of one voxel or less due to limitations on the resolution of the CT scanner used in their work. ...
  • ... This is important because with a destructive method the dimensions extracted from the disassembled state of a product may differ from the actual geometrical dimensions in the original assembled state. At the present time, for example, X-ray CT based inspection is of particular relevance to the industry of additive manufacturing (AM) as a viable option to explore the internal structure of AM parts and potentially answer questions about their structural integrity, tolerance limits, residual stresses, and a variety of dimensional deviations and internal flaws (e.g., cracks or voids) that may affect the functional performance of AM devices or create risks of potential failures [179,190,191,192,193,194,195,196]. Tactile coordinate measurement machines (CMMs 1 ) or optical measuring instruments like a laser scanner can measure the external surface of a part, but not internal structures inaccessible to tactile or vision-based inspection. ...
    Article
    Full-text available
    X-ray computed tomography (also referred to as X-ray CT, CAT scan, or simply ‘CT’) is a technological advancement with expanding applications, from medical imaging and nondestructive evaluation to, more recently, dimensional metrology. The CT technique is now used to measure a specimen’s geometrical dimensions (of both internal and external features). As a result, CT presently contributes to dimensional inspection and geometric analysis for technology companies spanning a variety of industries such as aerospace, automotive, electronics, medical devices, plastic components, metalworking, and additive manufacturing (one of the main drivers presently pushing the use of CT for dimensional measurement). For medical diagnoses or other qualitative analyses that depend mainly on feature recognition dimensional accuracy is not necessary. In contrast, for precision engineering applications accurate dimensional measurement is the essence of X-ray CT metrology. This article describes the development of X-ray CT metrology beginning with a historical overview that spans the discovery of X-rays to the invention of CAT scan and focuses with greater detail on its expansion toward industrial dimensional measurements. Following this overview is a brief review of the current state of the art of the technology—specifically focused on issues of metrology—and of the present standardization efforts in the design of acceptance tests for evaluating metrological performance of X-ray CT. As of writing, the CT metrology technique is still evolving with several technical issues yet to be resolved, in particular, to find better ways of expressing uncertainties associated with CT dimensional measurements. Supported by data indicating a growing commercial/industrial market, this technology appears to be in an ‘early adoption’ phase. (Go to: https://doi.org/10.1016/j.precisioneng.2019.06.007)
  • ... In actuality, measuring and characterizing each internal defect from the CT data can be a daunting task, mainly due to the sheer quantity of data that resides in high-resolution, volumetric CT datasets. Fortunately, many tools are available to assist in the measurement and analysis of large quantities of internal defects, e.g., see [34][35][36][37], and automated defect recognition algorithms are increasingly being implemented in practice, including tools for image processing and statistical analysis techniques that catalogue and summarize defects inside a particular volume. In the end, whether or not to pass or fail an AM component will need to rely upon comprehensive knowledge of the defects and their relationship to material properties and input process parameters. ...
    Conference Paper
    Full-text available
    Additive manufacturing (AM) technologies—or 3D printing, as they are popularly known—show promise as a way to transform traditional production manufacturing because they can produce highly complex geometries and customized parts directly from the part design model without dedicated tooling. However, many questions about the structural integrity of 3D printed parts—tolerance limits, layer defects, residual stress, and material inclusions—remain unanswered because AM process parameters and disruptions during the material layering (generally in powder form) may induce a variety of dimensional deviations and internal flaws (e.g., cracks or voids) in the final product. These flaws might affect the performance of AM devices and create risks of potential failures, so the support of metrology and non-destructive testing (NDT) techniques for better assessment of AM parts is needed [1-5]. One of the challenges to the assessment of AM parts is that many will have internal features, these are generally inaccessible from the outside to vision and contact-based inspection techniques for quality control. While destructive methods can be used to extract measurements, dimensional information from the disassembled state of a product may differ from the actual geometrical dimensions in the original assembled state. This paper describes some of the main issues associated with the measurement of AM parts and some future trends for the development of alternative techniques for measuring complex shaped AM parts. Along with a brief discussion of limitations of traditional inspection technologies, such as coordinate measuring machines (CMMs) and optical-based systems, the particular case of X-ray computed tomography (CT) as a technology to support AM inspection and development is discussed. The benefits of X-ray CT for the assessment of the structural integrity of AM parts and deviations typically encountered in AM dimensional geometry when compared to reference/nominal geometry will be considered.
  • ... An investigation into the X-ray CT scanning methods for a flexure mechanism fabricated by Additive Manufacturing offers insight into the capabilities of the process[2]. Whilst this research was in relation to polymeric products, the principles and benefits apply equally to metal AM applications. ...
    Article
    Full-text available
    X-ray Computed Tomography (micro CT) is just one option for the inspection of metal AM parts. Other options include using eddy current, ultrasonic technology, white-light interferometry and non-interferometric optics. However, given recent developments, it is micro CT that has the most potential in view of its unique capability for the inspection of complex internal structures and geometries without destroying the part. The capabilities of this inspection method are presented here in a brief overview. The article is available online, for free and in full in the following link: http://www.metal-am.com/wp-content/uploads/sites/4/2017/06/MAGAZINE-Metal-AM-Summer-2017-PDF-sp.pdf
  • ... 18 Similar CT metrology and porosity analyses of additive manufactured parts are presented. 19 The method has an American Society for Testing and Materials (ASTM) standard for choice of scan parameters and analysis methods. 20 A recent review of nondestructive testing of additive manufactured components discusses various aspects, including the use of X-ray tomography. ...
    Article
    Full-text available
    Quality control of laser additive manufactured medical implants is of interest, especially if nondestructive quality control can be performed on parts before implantation. X-ray micro-computed tomography (microCT or CT) can be used for defect/porosity analysis as well as for comparing the part surface with its computer-aided design (CAD) file. In both cases, the limited use of CT is partly due to the variation in scan types and the quality of scans that can occur. We present a simple method demonstrating the use of a light metal casting as a reference porosity sample, to confirm good CT image quality and to quantify minimum detectable pore size for the selected CT scan settings. This makes a good comparison for additive manufactured parts, since castings generally contain more porosity. A full part-to-CAD comparison shows how the part is compared with its CAD file, as a second-quality control. The accuracy of the CAD variance is given by the minimum detectable pore size. Finally, the part is sectioned and scanned at two higher resolution settings showing small porosity (10–50 lm diameter) present but well distributed, as expected.
  • ... Dimensional metrology with CT allows for size and form determination, wall thickness analysis, and part-to-CAD or part-to-part comparison against a standard (with color mapping). Measurement of voids/porosity morphology, and distribution is also possible [2,42,43]. In addition, CT enables reverse engineering by converting the 3D reconstructed model of a part to an STL model, which can be used for 3D printing or additive manufacturing, and computer-aided manufacturing [2]. ...
    Conference Paper
    Full-text available
    Industrial X-ray computed tomography (CT) systems have the ability to map internal and external structures simultaneously in a non-destructive way with high imaging resolution. Recently, there has been an increase of surveys in the field of dimensional metrology referring to CT as a tool for nondestructive dimensional quality control (i.e., traceable measurement and geometrical tolerance verification of industrial components). This increase in surveys runs parallel to the growth of commercial markets for industrial X-ray CT technologies and research institutes as well as metrology-governing bodies' growing interest in creating standarization. Currently, there is a lack of international standards that provide comprehensive procedures and guidelines for dealing with the verification of CT systems' dimensional metrology performance and developing task-specific measurement uncertainty budgets in compliance with the Guide to the Expression of Uncertainty in Measurement (GUM). To overcome this, some CT manufactures have opted to design their own calibration methods so that they can provide an estimate of maximum permissible error (MPE) for the measurements obtained with systems dedicated to metrology tasks. Essentially, the traceability of the instrument to the meter is provided with an expanded uncertainty upper-bounded by the MPE. In an effort to clarify some of these concepts, this paper gives a brief review of the use of X-ray CT for dimensional metrology with an update on the international attempt to create standards for metrological testing and uncertainty assessment with this technique. An example of in-house calibration is presented, which found deviations in the range-4.4 m to 3.5 m between CT measurements and calibrated references obtained at the National Institute of Standards and Technology (NIST), and this is contrasted to the MPE limits pre-established for CT measurement. A particular emphasis is made in the understanding of the terms " trueness " , " precision " , " accuracy " , and " uncertainty " , so the main metrology-related terminology is revisited with reference to international standards and other guidelines. It is concluded that while in-house calibrations might suffice, international standards are still needed, not only to reach homogeneity in the commercial market but also to avoid misinterpretations. In addition, users and manufacturers from the industry of measuring equipment need to better understand the terms " accuracy " and " uncertainty " , which are often misused and interchanged.
  • Article
    Full-text available
    X-ray computed tomography (CT) is increasingly recognized as a promising measuring technique for dimensional metrology. Various methods are being developed to improve measurement accuracy. Tests of new methods for such applications include accuracy evaluation with the use of calibrated workpieces; however, the internal algorithms of image acquisition and data processing might influence the experimental error, and then also the comparison of methods at different CTs. The accuracy of the results of tomographic measurements is influenced by many factors, one of which is the setting of the threshold value. The article presents the results of an attempt to use Monte Carlo simulated images to estimate deviations to determine threshold values to improve measurement accuracy and additionally, to estimate the impact of data processing. The differences of the results obtained from the simulated images were up to 4 % larger than those from tomographic images. It was caused by degradation of the image contrast by scattered radiation.
  • Thesis
    Full-text available
    X-ray computed tomography (CT)—more commonly known as CAT scan—has recently evolved from the world of medical imaging and nondestructive evaluation to the field of dimensional metrology; the CT technique can now be used to measure a specimen’s geometrical dimensions (of both internal and external features). As a result, CT presently contributes to the areas of dimensional inspection and geometric analysis for technology companies that produce manufactured parts for a variety of industries such as automotive, aerospace, medical devices, electronics, metalworking, injection molding plastics, composite materials, ceramics, and 3D printing or additive manufacturing. While dimensional accuracy is not crucial for medical diagnoses or other qualitative analyses, accurate dimensional quantification is the essence of X-ray CT metrology. Despite increasing advances in this technology, the current state of the art of CT metrology still confronts challenges when trying to estimate measurement uncertainties, mainly due to the plethora of influencing factors contributing to the CT measurement process. Gradual progress has occurred over the last decade toward a better understanding of some of these influencing factors that were illuminated by a series of collaborative research initiatives between a collective of several universities and institutions (predominantly located in the European Union) committed to the advancement and development of industrial CT scanning as a measuring technology. In an effort to further understand phenomenologically the role of variables affecting the precision and accuracy of CT dimensional measurements, this dissertation presents a series of experimental studies that evaluate the performance of cone-beam CT measurements, and their uncertainty estimates, in comparison with reference measurements generally obtained from tactile coordinate measurement machines (CMMs). In some cases, the results are contrasted against simulations performed in Matlab software (to compute fan-beam projection data) and an additional simulation tool called “Dreamcaster” (for ray casting and Radon-space analysis). The main CT variables investigated were: temperature in the X-ray CT enclosure, number of projections for a CT scan, workpiece tilt orientation, sample image magnification, material thickness influences, software post-filtration, threshold determination, and measurement strategies. For dimensions of geometric features ranging from 0.5 mm to 65 mm, a comparison between dimensional CT and CMM measurements, performed at optimized conditions, typically resulted in differences of approximately 5 µm or less for data associated with dimensional lengths (length, width, height, and diameters) and around 5 to 50 µm for data associated with measurements of form, while expanded uncertainties computed for the CT measurements ranged from 1 to over 50 µm. Methods for estimating measurement uncertainty of CT scanning are also assessed in this work. Special attention is paid to the current state of measurement comparisons (in the field of dimensional X-ray CT) by presenting a comprehensive study of metrics used for proficiency testing, including rigorous tests of statistical consistency (null-hypothesis testing) performed with Monte Carlo simulation, and particularly applied to results from two recent CT interlaboratory comparisons. This latter study contributes to the knowledge of methods for performance assessment in measurement comparisons. In particular, it is shown that the use of the En-metric in the current state of CT interlaboratory comparisons could be difficult to interpret when used to evaluate performance and/or statistical consistency of CT measurement sets.
  • Presentation
    Full-text available
    X-ray computed tomography (CT)—more commonly known as CAT scan—has recently evolved from the world of medical imaging and nondestructive evaluation to the field of dimensional metrology; the CT technique can now be used to measure a specimen’s geometrical dimensions (of both internal and external features). As a result, CT presently contributes to the areas of dimensional inspection and geometric analysis for technology companies that produce manufactured parts for a variety of industries such as automotive, aerospace, medical devices, electronics, metalworking, injection molding plastics, composite materials, ceramics, and 3D printing or additive manufacturing. While dimensional accuracy is not crucial for medical diagnoses or other qualitative analyses, accurate dimensional quantification is the essence of X-ray CT metrology. Despite increasing advances in technology, the current state of the art of CT metrology still confronts challenges when trying to estimate measurement uncertainties, mainly due to the plethora of influencing factors contributing to the CT measurement process. Gradual progress has occurred over the last decade toward a better understanding of some of these influencing factors that were illuminated by a series of collaborative research initiatives between a collective of several universities and institutions (predominantly located in the European Union) committed to the advancement and development of industrial CT scanning as a measuring technology. In an effort to further understand phenomenologically the role of variables affecting the precision and accuracy of CT dimensional measurements, this dissertation presents a series of experimental studies that evaluate the performance of cone-beam CT measurements, and their uncertainty budgets, in comparison with reference measurements generally obtained by tactile coordinate measurement machines (CMMs). In some cases, the results are contrasted against simulations performed in Matlab software (to compute fan-beam projection data) and an additional simulation tool called “Dreamcaster” (for ray casting and Radon-space analysis). The main CT variables investigated were: temperature in the X-ray CT enclosure, number of projections for a CT scan, workpiece tilt orientation, sample magnification, material thickness influences, software post-filtration, threshold determination, and measurement strategies. For dimensions of geometric features ranging from 0.5 mm to 65 mm, a comparison between dimensional CT and CMM measurements, performed at optimized conditions, typically resulted in differences of approximately 5 µm or less for data associated with dimensional lengths (length, width, height, and diameters) and around 5 to 50 µm for data associated with measurements of form, while expanded uncertainties computed for the CT measurements ranged from 1 to over 50 µm. Methods for estimating measurement uncertainty of CT scanning are also assessed in this work. Special attention is paid to the current state of measurement comparisons (in the field of dimensional X-ray CT) by presenting a comprehensive study of metrics used for proficiency testing, including rigorous tests of statistical consistency (null-hypothesis testing) performed with Monte Carlo simulation, and particularly applied to results from two recent CT interlaboratory comparisons. This latter study contributes to the knowledge of methods for performance assessment in measurement comparisons. In particular, it is shown that the use of the E_n-metric in the current state of CT interlaboratory comparisons could be difficult to interpret when applied to evaluate performance and/or statistical consistency of CT measurement sets.
  • Conference Paper
    Full-text available
    Additive Manufacturing (AM), more commonly known as 3D printing, is the process of building a product or part by applying materials or powders in very thin layers, until the final product has been built. This capability allows for a variety of new design possibilities for AM such as hinge-based mechanisms, shock absorbing castings, integrated gaskets, and so on [1-3]. The process starts with a 3D model in an STL format, which is imported into the AM software. The AM software converts its geometry into horizontal slices in the form of multiple layers with varying thicknesses. Many processes include the use of UV, laser, or thermal-associated technology; such as Selective Laser Sintering (SLS), Fused Deposition Modeling (FDM), and Stereolithography (SLA) [1]. Materials commonly utilized in AM range from resins and polymers to steel. Recently, a number of AM-based devices for precision sensing and actuation applications have been introduced [2,3]. However, the stability characteristics of devices fabricated by AM processes are not well understood and there is very little to indicate what the dominant manufacturing parameters are, how creep and fatigue are affected by different processes and materials, and how to design for stability. For the general aspects of polymer fatigue, the interested reader is referred to a few excellent reviews and articles that adequately cover much of the recent polymer fatigue research on life time [4,5], fatigue crack propagation [5-7], cyclic softening [8,9], thermo-mechanical effects [10,11], fracture surface morphology [12], processing variables [13,14] and cumulative damage models [15]. Here the physical testing for identifying mechanical and fatigue characteristics including tensile strength and high and low cycle fatigue is conducted in accordance with the applicable ASTM standard.
  • Conference Paper
    Full-text available
    This paper addresses the significance of accurate temperature determination using X-ray CT for dimensional metrology and the importance of measurement bias quantification in the CT results in both the CT measurements and their estimated uncertainty. If the CT and the CMM systems are located on the same shop floor, it is frequently assumed that the same controlled environmental conditions apply; however that is not necessarily true for the temperature, not even for CT machines located in (thermally) isolated measurement laboratories. Local and temporal temperature gradients from heating of the X-ray sources in the machine can occur inside the CT scanner. For example, measurements in two different samples when the system was turned on showed variations in temperature of 2.5 to 5.2 ˚C near to the x-ray source, a corresponding change of 0.5 to 1.0 ˚C in the specimen temperature after time delays of around 3 hours while the ambient temperature fluctuations in the laboratory environments only varied in a range of 0.2 ˚C. If the workpiece undergoes thermal expansion while placed inside of the tomographic system, the CT dimensional uncertainties will increase. Measurements obtained in this work showed variations in geometry of up to 3.78 µm for temperature changes of 0.5 ˚C with a Poly(methyl methacrylate) (PMMA) thermoplastic sample with major dimensions 20 mm X 40 mm X 60 mm.-----Please see a more thorough and updated version of this research in Chapters 3 and 5 of my dissertation thesis: https://www.researchgate.net/publication/327070768_Studies_of_Dimensional_Metrology_with_X-ray_CAT_Scan
  • Conference Paper
    Full-text available
    This paper illustrates strategies for metrology using X-ray computed tomography (CT) to perform dimensional measurements on the interior structures of parts. Of particular interest is understanding how the current CT technology enables measurement of the internal features that are normally inaccessible to tactile coordinate measuring machines (CMMs). To evaluate these measurements, they are compared to reference values obtained by a destructive method. Conventional measurement techniques using CMMs or an optical device like a laser scanner can only measure the exterior surface of a part, but not interior structures. CT is a non-contact measurement technique and the only commercially available non-destructive method to perform dimensional measurements on internal geometry. The National Institute of Standards and Technology (NIST) has designed an artifact with internal geometry (see Fig. 1) to be characterized with CT technologies and to compare the measured dimensions with reference calibration values obtained using a Moore M48 CMM [1]. In this paper we present details of the coordinate measurement analysis completed for the determination of the artifactʼs dimensional measurements with data from an Xray CT machine. The main purpose is to guide the reader through a typical process of analysis usually applied on CT tomography to obtain dimensional metrology information. Results obtained by this method and measurements taken by conventional CMM techniques is presented to provide both a comparison of the two sets of data from these two measurement processes and observations about the capabilities of CT technologies as coordinate measurement systems.-----Please see a more thorough and updated version of this research in Chapter 4 of my dissertation thesis: https://www.researchgate.net/publication/327070768_Studies_of_Dimensional_Metrology_with_X-ray_CAT_Scan
  • Article
    Full-text available
    INTRODUCTION Additive Manufacturing (AM), more commonly known as 3D printing, is the process of building a product or part by applying materials or powders in very thin layers, until the final product has been built. This capability allows for a variety of new design possibilities for AM such as hinge-based mechanisms, shock absorbing castings, integrated gaskets, and so on [1-3]. The process starts with a 3D model in an STL format, which is imported into the AM software. The AM software converts its geometry into horizontal slices in the form of multiple layers with varying thicknesses. Many processes include the use of UV, laser, or thermal-associated technology; such as Selective Laser Sintering (SLS), Fused Deposition Modeling (FDM), and Stereolithography (SLA) [1]. Materials commonly utilized in AM range from resins and polymers to steel. Recently, a number of AM-based devices for precision sensing and actuation applications have been introduced [2,3]. However, the stability characteristics of devices fabricated by AM processes are not well understood and there is very little to indicate what the dominant manufacturing parameters are, how creep and fatigue are affected by different processes and materials, and how to design for stability. Here the physical testing for identifying mechanical and fatigue characteristics including tensile strength and high and low cycle fatigue is conducted in accordance with the applicable ASTM standard.
  • Article
    Computer Tomography has become a well recognized tool in industrial quality control. Modern computer tomography systems ranging from micro-CT to huge multi MeV systems allow us to generate more and more detailed views of the inner of nearly any object. With the scan resolution becoming smaller and smaller, and at the same time image matrices becoming larger and larger, we are able to localize smallest defects even in large scale objects. At the same time even with the same data set we are able to measure the outer and inner geometry of an object with a measurement point density never known before from classical tactile or optical techniques. However, scanning objects in high resolution generates huge amounts of data, easily exceeding two GByte per scan. These huge amounts of data have caused a major drawback of a wider acceptance of CT technology in industrial use. Either no software tools have been available at all or available software process chains haven't been able to process these amounts of data in reasonable time. This presentation will introduce a new generation of 3D visualization and analysis software tools for industrial CT users. Interactive visualization of huge data sets with several Gbyte in size has become possible on a standard PC. Automatic wall thickness analysis and internal defect/porosity analysis can be done within minutes. In addition this presentation will also demonstrate the latest generation of software tools for highly accurate 3D geometry analysis based on voxel data. Introduction: More than one hundred years ago X-ray technology started its triumphal procession when Wilhelm Conrad Roentgen discovered a new kind of radiation in his laboratory in Wuerzburg, Germany in the year 1895. In 1901 Prof. Roentgen received the first Nobel Prize ever consigned. Over the years many inventions driven by the X-ray technology followed as well as many more Nobel Prizes, especially in the medical application field. More then seventy years later the computer tomography has been invented by G. N. Hounsfield in 1971, and again the Nobel Prize for medicine was assigned in 1979 to G. Hounsfield and A. Cormack for their pioneer work on computer tomography. Up to this moment most of the developments on X-ray technologies and computer tomography have been focused on special medical applications. Another twenty years later computer tomography (CT) has become a powerful, well accepted tool in industrial applications as well. Today industrial CT is on its way to become a major tool of industrial quality control in high-tech branches, not only for material testing but for geometry analysis as well. This article will describe the evolution of industrial CT over the last few years and it will present the most modern software tools for voxel data based image and geometry analysis. Results & Discussion: THE EVOLUTION OF INDUSTRIAL CT In the late eighties CT became popular as a tool for non destructive testing in industrial applications. At that time CT was used to create two-dimensional cross sections of an object under investigation. The engineers performed visual inspections of a rather small number of slices, e.g. they looked for inclusions within material probes or other internal defects. A typical image size at that time has been in the range of 256 x 256 pixels. In the later 1990s industrial CT became a true 3D imaging technology. Stacks of continuous scanned slice images were created, so that the full volume of an object under investigation became digitized. Therefore 3D quantitative analysis became possible, for instance the measurement of the volume of an object, its surface area or just distance measurements in an arbitrary orientation. At the same time the accuracy of the CT scans became better and better. Today objects with the size of 50 cm in length are scanned at a resolution of 0.2 to 0.3 mm. All these enhancements in CT performance were possible due to more stable and more powerful X-ray sources, better detector systems and enhancements in the complete CT system design. On the other hand the amount of data within a single data set became larger and larger by scanning a larger number of images; in addition the images itself became larger. After a long period where the standard CT image size has been 512x512 pixels, the 1024x1024 pixels image are now a standard size.
  • Article
    Full-text available
    X-ray computed tomography (CT) reconstructs an unknown object from X-ray projections and has long been used for qualitative investigation of internal structures in industrial applications. Recently there has been increased interest in applying X-ray cone beam CT to the task of high-precision dimensional measurements of machined parts, since it is a relatively fast method of measuring both inner and outer geometries of arbitrary complexity. The important information for the user in dimensional metrology is if measured elements of a machined part are within the defined tolerances or not. In order to qualify cone beam CT as an established measurement technology, it must be qualified in the same manner as established measurement technologies such as coordinate measurement machines (CMMs) with tactile or optical sensors. In international standards artefacts are defined that are calibrated by certified institutions. These artefacts are defined by certain geometrical elements. CT measurements are performed on the reconstructed object volume, either directly or using an intermediate surface-extraction step. The results of these measurements have to be compared to the values of the calibrated elements; the level of agreement of the results defines the accuracy of the measurements. By using established methods to define measurement uncertainty a very high level of acceptance in dimensional metrology can be reached for the user. Only if results are comparable to standards of the established technologies the barriers of entry into metrology will be removed and all benefits of this technology will be available for the user.
  • Article
    The paper gives a survey of the upcoming use of X-ray computed tomography (CT) for dimensional quality control purposes: i.e. for traceable measurement of dimensions of technical (mechanical) components and for tolerance verification of such components. It describes the basic principles of CT metrology, putting emphasis on issues as accuracy, traceability to the unit of length (the meter) and measurement uncertainty. It provides a state of the art (anno 2011) and application examples, showing the aptitude of CT metrology to: (i) check internal dimensions that cannot be measured using traditional coordinate measuring machines and (ii) combine dimensional quality control with material quality control in one single quality inspection run.