1. Laboratory X-ray CT Applied to the
Characterisation of Tubular Composite Specimen
Fabien Léonard and Philip J. Withers
Henry Moseley X-ray Imaging Facility, The University of Manchester
Abstract
This poster presents the assessment of a tubular composite
specimen by X-ray computed tomography (XCT). Standard
measurements such as void size distribution and void content
have been performed as well as innovative measurements
to obtain a full 3D description of the specimen defect structure.
The advanced data analysis was correlated to the 3D visual-
isation to further the understanding of the damage within the
specimen. The methodology presented here on a cylindrical
specimen can be extended to any thin- and thick-walled rod
and tubular geometry.
Introduction
The past few decades have seen the increase use of composite ma-
terials for high performance engineering applications [1], led mainly
by the aerospace industry [2]. The development of composites for
other mainstream applications has also been linked to the develop-
ment of new manufacturing techniques (such as 3D weaving, tape
winding and braiding) producing near net shape fibre preforms
for manufacturing of complex 3D components [3]). This study
demonstrates the potential of XCT for standard and advanced 3D
measurements for damage characterisation on a tubular composite
specimen. Additionally, the methodology developed here could be
applied to any rod or tubular component.
Experimental
• Specimen manufacturing
The specimen was a tape winded polyether ether ketone (PEEK)-
carbon fibre composite tube having a 6 mm inner diameter,
an 11 mm outer diameter and a 50 mm length.
• X-ray computed tomography
The specimen was scanned on the Nikon Metrology 225/320 kV
Custom Bay system (Figure 1) equipped with a 225 kV static
multi-metal anode source (Cu, Mo, Ag, W and a minimum focal
spot size of 3 µm) and a PerkinElmer 2048 × 2048 pixels 16-bit
amorphous silicon flat panel detector.
Figure 1: Nikon Metrology 225/320 kV Custom Bay system.
The scanning was performed with the copper target using a voltage
of 60 kV and a current of 170 µA. The data acquisition was car-
ried out with an exposure time of 2000 ms, and no filtration. The
number of projections was set to 3142 and the number of frames
per projection was 1. The entire volume was reconstructed at full
resolution with a voxel size of 25.0 µm along the x, y, and z dir-
ections. The data were then loaded into VGStudio MAX software,
and converted from 32 bits to 8 bits with the grey scale remapped
from [0,52] to [0,255]. The data processing was performed with
Avizo Fire 7.0.1 software.
References
[1] Beukers, A. (2001)
Polymer Matrix Composites: Applications.
Encyclopedia of Materials: Science and Technology (Second Edition).
Elsevier, 7384-7388.
[2] Rawal, S.P. and Goodman, J.W. (2000)
Composites for Spacecraft.
Comprehensive Composite Materials. Pergamon, 279-315.
[3] McClain, M., and Goering, J. (2012)
Overview of Recent Developments in 3D Structures.
Conference SME Manufacturing with Composites 2012.
Results and Discussion
Examples of 2D slices from the 3D volume are presented in Figure 2. Defects can be clearly identified in all orthogonal planes
(XY , XZ, and Y Z) with some of the defects running over significant distances along the vertical length of the specimen.
(a) XY plane (b) XZ plane (c) Y Z plane
Figure 2: 2D orthoslices from 3D reconstructed volume of specimen.
Standard data processing has been applied to the voxels labelled as voids in order to obtain the void equivalent diameter
distribution and the evolution of the global void volume fraction along the vertical axis (Figure 3). The void volume fraction
results obtained from XCT (3.1 ± 0.3 %) were consistent with optical microscopy results (2.9 % porosity, XY plane).
(a) labelled individual pores
µ
µ
µ
(b) individual pore size distribution
±
(c) global void volume fraction along z axis
Figure 3: Standard void data processing.
Advanced data analysis has been performed to obtain the distribution of voids both vertically and radially (Figure 4). Although
there is very little variation along the vertical axis (4a), 3 peaks can be observed on the radial distribution graph (4b): a low
intensity peak between 0 and 0.5 mm from the inside surface and two high intensity peaks, respectively between 0.5 and 2 mm;
and 2.5 and 3.6 mm from the inner surface of the composite cylinder. The combination of both vertical and radial distributions
is obtained through a correlation histogram (4c) that highlights in 3D the locations with the most damage.
(a) vertical distribution (b) radial distribution (c) correlation histogram
Figure 4: 3D spatial void distribution.
The voxels corresponding to the 3 peaks obtained in Figure 4b have been separated into 3 different labels (Figure 5). The first
selection (5b), corresponding to the voids closest to the inner surface of the tube, appears to be cracks that propagate along
the vertical axis of the specimen. The other 2 selections (5c) are similar to one another and are composed of large cracks
propagating along the circumference of the cylinder.
(a) full damage (b) radial cracks from inner surface only (c) circumferential cracks only
Figure 5: Separation of damage based on spatial distribution.
From the damage morphology, it has been proposed that the mechanisms involved in their formation were different. The
vertical cracks have been linked to the removal of the specimen from the metallic mandrel after manufacturing whereas the
circumferential cracks have been linked to thermal stresses developing during manufacturing.
Conclusion
• This study presents original results on laboratory X-ray computed tomography applied to the characterisation of damage
within a cylindrical composite specimen.
• It has been demonstrated how standard data processing can be applied to obtain structural information such as void size
distribution and void volume fraction.
• Examples of advanced data processing and visualisation have been presented, highlighting the strength of CT as a
3–dimensional tool for damage characterisation of cylindrical composite specimens, but also more generally for any thin-
and thick-walled rod and tubular component.
Acknowledgments
The authors acknowledge funding from the EPSRC
(grants EP/F007906/1 and EP/F001452/1).
Contact details
fabien.leonard@manchester.ac.uk
+44 (0)161 306 3608