IN-SITU DAMAGE COMPARISON BETWEEN FABRIC-FABRIC AND FABRIC … · different load steps with the unloaded specimen allows a determination of the influence of different damage types

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 1

IN-SITU DAMAGE COMPARISON BETWEEN FABRIC-FABRIC AND

FABRIC-UD BONDED CFRP

Christian Hannesschläger1, Stefanie Rauchenzauner2, Christian Gusenbauer1, Dietmar Salaberger1 and Johann Kastner1

1University of Applied Sciences Upper Austria, Stelzhamerstrasse 23, 4600 Wels, Austria, e-mail:

[email protected], [email protected], [email protected], johann.kastner@fh-

wels.at

2FACC, Fischerstraße 9, 4910 Ried im Innkreis, Austria, e-mail: [email protected]

Abstract

An interrupted in-situ X-Ray computed tomography (XCT) was used to investigate the influence of different Carbon fibre

reinforced polymers (CFRP) material systems on the crack progression and propagation of adhesive jointed CFRP plates. The

damage behaviour of adhesive jointed CFRP is decisive for the mechanical strength. For this paper the material systems (I)

CFRP fabric plate with another CFRP fabric plate and (II) a CFRP fabric plate with a unidirectional (UD) CFRP plate were in-

situ investigated with a high resolution X-Ray computed tomography (XCT) with a voxel size of (6 µm)³ at different load

steps. The damage volumes of both material systems were segmented and compared with each other. The results confirmed the

stress concentration at the end of the specimens [14]. In general the CFRP-cracks in the plates can be identified as major

damage mechanism. The different strength of the two material systems is caused by the missing damage in the UD-plate. This

lack of damage is caused by the high crack resistance of the UD-plies in respect to the force direction and results in a higher

strength of the system fabric-UD.

Keywords: x-ray tomography, in-situ, lap shear testing, adhesive bonded CFRP

1 Introduction

Carbon fibre reinforced polymers (CFRP) are widely used in aeronautic industries for their special properties such as light

weight, high specific stiffness, high specific strength and the high corrosion resistance [1]. An important manufacturing issue is

the bonding of CFRP plates. Therefore adhesives are widely used. A main advantage of the adhesives is the homogenous stress

distribution in comparison to other joining techniques [2].

The microstructure of the jointed area is decisive for the mechanical strength of adhesive jointed CFRP plates. In the past

several studies investigated the failure behaviour of bonded composites joints [3-6]. Therefore, light optical microscopy or a

visual inspection [7] was used to characterize the failures. With these methods a non-destructive 3D tracking of the crack

propagation is not possible. High resolution X-Ray computed tomography (XCT) as common non-destructive technique (NDT)

is used to characterise defects in CFRP specimens [8-9]. In the past in-situ XCT has already been used to investigate damage

mechanism in CFRP [10-11]. H. Kunz et al. showed a technique for particle tracking in polyurethane adhesive bonded CFRP

and steel with in-situ XCT while performing a lap shear test [12].

In this paper, the method in-situ XCT is used to determine the damage progression in adhesive during an adapted slotted single

lap shear test [13]. With interrupted in-situ XCT measurements under different load conditions it is possible to mechanically

test the adhesive jointed CFRP plates and to visualise and study the influence of different CFRP material combinations on the

crack progression and propagation.

2 Experimental Method

2.1 Material and Specimen

A supported epoxy adhesive was used to bond two different material systems: (I) CFRP fabric plate with another CFRP fabric

plate and (II) a CFRP fabric plate with a unidirectional (UD) CFRP plate. In both cases the adhesive layer is 0.2 mm thick. An

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 2

epoxy resin based adhesive, with a polyester fabric, was used for bonding. The adapted slotted single lap shear test specimens

were cut out of the bonded plates by jet cutting (red area in Figure 1).

Figure 1: (a) sketch of adapted slotted single lap shear test specimen (dimensions are given in mm, measurement area is indicated in red); (b)

cut section in detail, in which regions “top” and “bottom” are discussed in more detail by in-situ XCT slice images.

2.2 High-resolution XCT

The specimens were scanned with a Nanotom 180NF XCT device (GE phoenix | X-ray) with a voxel size of (6 µm)³ at

different load steps. This system consists of a sub-µ-focus X-ray tube and a 2304x2304 pixels flat panel pixel detector

(Hamamatsu). The scanning time was 80 min per load step. A beam hardening correction was applied to the data. The 3D

volume was analysed by using the software VGStudio MAX 3.0.

2.3 In-situ tensile stage

The experiment was performed with the CT5000 5 kN in-situ tensile stage (Deben) at room temperature (see Figure 2). Main

specifications of CT5000 stage can be found in Table 1.

A displacement was applied with 1 mm/min. An interrupted slotted single lap shear test, with adapted dimension, was used to

identify damage at several load stages. Reference specimens of both material systems were tested to determine the quasi static

load steps for the in-situ experiment. The reference investigations showed a first detectable damage at ~2/3 of Fmax. Larger

material failures can be segmented at around 90 % of the maximum load.Table 2 shows the applied load steps in relation to the

maximum force of the reference specimen.

relative percentage of reference Fmax [%] stepi. breakage

Fabric-Fabric:

Force [N]

Fabric-UD:

Force [N]

0 1 - 0 0

66.6 2 - 1000 1550

90 3 - 1400 2100

95 4 - 1500 2200

~100 5 yes 1606 2205

Table 2: planned load steps for measurement

To ensure a stable scan without any motion artefacts, a relaxation time is needed (see Figure 3).

Specifications Tensile stage CT5000

Max. load 5000 N

Max. motor speed 1 mm/min

Modes tensile/compression

Max. extension 10 mm

Accuracy 1% of full scale range

Weight ~5.5 kg

Table 1: Specification of tensile stage CT5000

Z

Y

Ten

sile

sta

ge

X-r

ay s

ou

rce

mo

un

ted s

pec

imen

(a) (b)

(a) (b)

top

bott

om

adhesive

CFRP

CFRP

Figure 2: (a) Position of tensile stage within XCT-device; (b)

specimen position with respective coordinate system

Y

Z

X

Z

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 3

(a) (b)

Figure 3: (a) Examplaric diagram of all force cycles during an in-situ experiment over time (fabric-UD); (b) relaxation of the specimen: a

relaxation time of 6 minutes is used to ensure almost stable conditions during XCT scanning

2.4 Segmentation

The 32bit data was mapped to 16bit, to ensure that the average grey level of the air (10000) and adhesive (50000) peak is

always at the same grey value position. For noise reduction a median 3x3x3 filter was applied. The advanced surface

determination and the porosity analysis tool of VGStudio MAX 3.0 were used to segment cracks, debondings and pores. To get

rid of the surrounding air, the adaptive rectangular tool (ISO-90 Threshold) was applied. The resulting region of interest was

used for defect detection (region 1). To distinguish between the different failure modes, first of all the defects of the unloaded

specimens were segmented. For the pore segmentation the VGStudio pore analysis tool (only threshold: threshold ISO-50) was

used. Big pores are defined by size (larger than 1100 voxels) and sphericity (threshold of 0.36). The value 0.36 was determined

empirically .The sphericity threshold separates pores from cracks. In the system fabric-UD the biggest pore was calculated

separately without sphericity threshold, due to the low sphericity of this defect. To determine the cracks of the initial state an

inverted advanced surface determination (ISO-25) was used for segmentation (see Figure 4).

unloaded

1) pore segmentation on region 1 (step1)

2) subtract pores of region 1 = region 2 (step1)

3) region 2: invert surface (ISO-25)

4) invert surface = roi cracks of initial state 0N (region

cois(step1))

loaded

1) pore segmentation on region 1 (stepi)

2) subtract pores of region 1 = region 2 (stepi)

3) region 2: invert surface (ISO-25) = region 3 (stepi)

4) subtract dilated region cois (step1) of region 3 =

region 4 (stepi)

5) region 4: defect detection (threshold: 65 535)

6) region 4: sorting of defects by their x-coordinate

7) subtract defects inside the adhesive (x-coordinate) of

region 4 defects = CFRP cracks (region 5(stepi))

8) subtract region 5 of region 4 = debonding of

polyester fabric (region 6 (stepi))

9) subtract regions 5 and 6 of region 3 = cois (stepi)

region

Figure 4: Segmentation flow chart of unloaded (step1) and loaded (step3-4) specimen

measurement time

of an XCT scan

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 4

At the following load steps a separation of CFRP-cracks, debonding of polyester-fabric and cracks of initial state is necessary

because all three failure types are segmented by the advanced surface determination.The cracks of initial state can be easily

eliminated by dilating and subtracting the defect volume of 0 N initial state of the inverted advanced surface determination,

due to the fact that the volume of the initial state crack is not overlapping with the volume of any other defects. Afterwards

defect detection was applied on the remaining region. To separate debonding from CFRP-cracks, the defects were sorted by

their X coordinate position. Every defect inside the adhesive along the X-axis is detected as debonding of polyester-fabric.

After this step the CFRP-cracks and debonding were subtracted of the actual inverted advanced surface determination volume

to calculate the load step specific cracks.

3 Results and Discussion

The progression of damage could be characterised by registering the individual load steps on each other. Different damage

types such as pores, debonding of polyester fabric and CFRP cracks were identified and segmented. A comparison of the

different load steps with the unloaded specimen allows a determination of the influence of different damage types on the final

breakage. Figure 5 shows the obtained results of fabric-fabric.

Figure 5: Fabric-fabric: comparison of unloaded (a) and broken status (b) shows a final fracture mainly in the CFRP. The volumes top and

bottom regions are visualized in detail in Figures 6-8.

Figure 6 shows a slice images (Z-X) of the fabric-fabric system at top and bottom region. From the beginning of damage

initiation to 95% Fmax,reference the defects can be observed only within these regions. Defect propagation can be seen at certain

load steps. The area of defect origin is highlighted by a red cross. It can be observed that the defect starts by a debonding of the

polyester fabric within the adhesive (see Figure 7). After this initiation the observed cracks are growing mainly through the

CFRP (see figure 6: 1400 N and 1500 N). It can be noticed that the CFRP cracks are expanding in load direction (z-axis) after

they have reached the first CFRP-layer. The defects can be detected only in one CFRP plate at the positions top and bottom.

These stress concentrations at the end of the overlap is a well-known problem of single lap-joints [14]. The final breakage goes

primarily through the CFRP. It is noticeable, that the debonding failure in the bottom region is closed after the final breakage

(see Figure 6: bottom 1606 N).

carbon fibre bundle CFRP-matrix

polyester fabric adhesive to

p

bo

tto

m

CFRP-cracks

top-region bottom-region

top-region bottom-region

1 mm 1 mm

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 5

0 N 1000 N 1400 N 1500 N 1606 N

Figure 6: Fabric-Fabric slice images comparison (top/bottom region Z-X) of the defect propagation at certain load steps; failure mode:

debonding of polyester fabric (green) and CFRP-cracks (red). Region (a) is shown in detail in figure 7.

Figure 7: Debonding of the polyester fabric within the adhesive in detail at load step 1000 N

Figure 8 shows the comparison of slice images (Z-X) of the fabric-UD system at top and bottom position. The pictures of

column 0 N show CFRP cracks in the initial state. These initial cracks are not changing throughout all load steps. At 1550 N

the first force induced defects are visible (debonding, CFRP-crack). It is obvious that no damage can be detected in the UD-

plate at any load step.

(a)

to

p

CFRP-cracks

500 µm

bo

tto

m

Z

X

100 µm

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 6

0 N 1550 N 2100 N 2200 N 2205 N

Figure 8: Fabric-UD slice images (top/bottom region Z-X) of the defect propagation at certain load steps; failure mode: debonding of

polyester fabric (green), CFRP-cracks (red) and initial cracks (blue)

The CFRP-cracks are expanding in load direction after they have reached the first ply. No further crack deflection can be

observed. The same has been seen before at the system fabric-fabric. The CFRP-cracks are the dominating failure leading to

the final fracture.

A comparison of minimum intensity projections at top and bottom position shows that the cracks are growing in both material

systems only in the CFRP fabric layers next to the adhesive. (see Figure 9). The observed damage initiation occurs mainly in

the adhesive layer (see Figure 9: column ~2/3 Fmax). Afterwards the cracks are continuously growing through the adhesive

layer and the CFRP with increasing load. It is noticeable, that there are no defects in the UD-plies. A growth of initial cracks

and pores cannot be observed.

to

p

bo

tto

m

Z

X

Debonding

of polyester

fabric

CFRP-cracks

init

ial

crac

ks

500 µm

init

ial

crac

ks

deb

on

din

g o

f p

oly

este

r fa

bri

c

CFRP-cracks

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 7

Top: ~2/3 Fmax,ref Top:95% Fmax,ref Bottom: ~2/3 Fmax,ref Bottom: 95% Fmax,ref

Top: ~2/3 Fmax,ref Top:95% Fmax,ref Bottom: ~2/3 Fmax,ref Bottom: 95% Fmax,ref

Figure 9: Comparison of different load steps of top and bottom regions: damage propagation is visualised minimum intensity projections (thickness: 300 µm) of (a) fabric-fabric and (b) fabric-UD; failure mode: debonding of polyester fabric (green), CFRP-cracks (red), initial

cracks (blue) and pores (violet).

In particular four different defect types can be classified that are colour coded. To understand the different breakage behaviour

of the two different material systems, a segmentation and quantification of each load step is essential.

First visual cracks can be seen at a load of ~2/3 Fmax. These first CFRP-cracks and debonding failures are too small for

segmentation. Nevertheless a quantitative comparison of the load steps 0, 90 and 95% of Fmax,reference was performed. The

volume of each defect type at every position represents the damage evolution. Figure 10 shows colour coded defect volume

progression of the system fabric-fabric (pores = violet, debonding of polyester fabric = green, CFRP-cracks = red). It is

obvious that the initial pore volume is nearly constant over all load steps. The shape of the debonding is defined by the

polyester fabric (see Figure 11: row (a)). CFRP-cracks can be identified as the main defect type. It can be observed that they

are growing from top and bottom, in an angle of 45° to the y-axis, to the specimen centre. This behaviour is expected, since the

orientation correlates to the ply orientation.

0 N 1400 N 1500 N

Figure 10: Fabric-Fabric system colour coded defect type 3D comparison at different load steps (pores = violet, debonding of polyester fabric

= green, CFRP-cracks = red). The regions (a) and (b) are visualized in detail in Figure 11.

bo

tto

m

to

p

Z

Y

(b

) fa

bri

c-U

D

(a)

fab

ric-

fabri

c

Y

X 1500 µm

(b)

(a)

CF

RP

-cra

cks

deb

on

din

g o

f p

oly

este

r fa

bri

c

init

ial

crac

ks

po

res

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 8

Figure 11 shows the visualized failure types, debonding and CFRP cracks in detail.

unsegmented slice segmented slice segmented: 3D

Figure 12 shows the defect volume progression of the fabric-UD. In the unloaded stage, larger pores and smaller cracks in the

fabric-plies are visible. A propagation of these initial cracks cannot be detected. It also can be observed that these initial

defects are not connected with the CFRP-cracks and debonding volume. Due to this fact, an influence of these defects as a

crack origin can be neglected. Furthermore, a growth of debonding volume from load step 2100 N to 2200 N can be observed.

It is noticeable that CFRP-cracks can only be detected in the fabric in the bottom region. The orientation of these cracks

correlates with the orientation of the fabric plies.

0 N 2100 N 2200 N

Figure 13 shows a defect volume comparison between the investigated material systems over the segmented load steps. It can

be clearly seen that there is no noticeable increase of initial cracks and pores at the final breakage. In general the CFRP-crack

volume is always larger than the debonding volume. At the fabric-UD system a nearly parallel progression of debonding and

CFRP-cracks can be observed. In opposite to this, a much higher increase of CFRP-crack volume than debonding volume can

be observed at the fabric-fabric system, from load step 1400 to1500 N. Also the volume proportion of debonding to CFRP

cracks is at the system fabric-fabric higher than at the system fabric-UD. The difference can be explained by the missing cracks

in the UD-plate.

Z

Y

Z

Y

Figure 11: Visualization (slice- and 3D-images) of the segmented defects: (a) debonding and (b) CFRP-crack

Figure 12: Fabric-UD system colour coded defect type 3D comparison at different load steps (initial cracks = blue, pores =

violet, debonding of polyester fabric = green, CFRP-cracks = red)

Z

X

UD Fabric

300 µm

300 µm

(a)

(b)

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 9

Figure 13: Defectvolume comparison of the systems fabric-fabric and fabric-UD over the segmented load steps

In both material systems the crack initiation occurs mainly in the adhesive layer (see Figure 9). But only in the fabric-fabric

system CFRP cracks are in both plates visible. This lack of CFRP-cracks in the UD-plate explains the lower CFRP-crack

gradient (step from 1400 to1500 N) at the fabric-UD system in comparison to fabric-fabric system. It can be concluded that the

stress is not high enough for a crack initiation in the UD-plies. The main advantage of UD in this case is that the force and the

plies are orientated in the same direction.

5 Conclusion and Outlook

In general the interrupted in-situ XCT measurements allowed a detailed look on the defect propagation of adhesively bonded

CFRP during an adapted slotted single lap shear test. It has been seen shown that the influence of the substrate on the damage

mechanism can be determined in a qualitative and quantitative manner. The results of the interrupted in-situ scanned fabric-

fabric specimen were compared with the fabric-UD specimen. The stress concentration at the end of the specimen [14], in the

overlapping region, can be seen in both materials. A voxel size of (6 µm)3 is sufficient to detect visually the first defects at ~2/3

of Fmax,reference. At a load step of 90% of Fmax,reference the defects could be segmented and quantified. Four different failure types

could be observed. An influence of initial defects (pores, cracks) from the unloaded step, on the final breakage cannot be seen.

The crack initiation event can be determined as the debonding of the polyester fabric. After this first damage step, the cracks

are growing mainly through the adhesive nearest fabric ply until the final fracture occurs. This major influence of the CFRP

cracks is determined by the high defect volume ratio of CFRP cracks to debonding of polyester. It can be concluded that the

higher strength of the fabric-UD system is caused by the higher crack resistance of the UD in comparison to the 45° plies of the

fabric. All plates show a potential crack initiation behavior by debonding of the polyester fabric, but only the strength of the

UD-plate is high enough to prevent a growth of those early cracks. The lack of damage in the UD plates leads to a significant

higher strength of the system fabric-UD. In a further step, an improved system of fabric-fabric with one UD-layer next to the

adhesive and the system UD-UD should be investigated. This single UD-layer may increase the crack resistance of the plate.

To sum up in case of single lap shear strain the system fabric-UD shows a better resistance against crack growth. With the

usage of fabric-UD instead of fabric-fabric the maximum force can be increased by ~37 percentages.

Acknowledgements

The authors like to acknowledge FACC Operations GmbH and CoLT Prüf und Test GmbH for the manufacturing of the

specimen. This work was supported by the K-Project for “non-destructive testing and tomography plus” (ZPT+) and by the

COMET program of FFG and the federal government of Upper Austria and Styria and supported by the project “multimodal

and in-situ characterization of inhomogeneous materials” (MiCi) and by the European Regional Development Fund (EFRE) in

the framework of the EU-program IWB2020.

References

[1] B. Jahn, D. Karl. Composites-Marktbericht, 2012

[2] G. Habenicht. Kleben: Grundlagen, Technologien, Anwendungen. Springer Berlin Heidelberg New York, 2006

[3] K. S. Kim, J. S. Yoo, Y. M. Yi and C. G. Kim, Failuremode and strength of uni-directional composite single

lap bonded joints with different bonding methods, Compos. Struct., vol. 72, no. 4, pp. 477–485, 2006.

[4] J. K. Kim, H. S. Kim and D. G. Lee, Investigation ofoptimal surface treatments for carbon/epoxy composite

adhesive joints., Journal of adhesion science and Technology, vol. 17, no. 3, pp. 329–352, 2003.

8th Conference on Industrial Computed Tomography, Wels, Austria (iCT 2018)

www.3dct.at 10

[5] T. Cheuk and L. Tong, Failure of adhesive bonded composite lap shear joints with embedded precrack, Compos. Sci.

Technol.,vol. 62, no. 7-8, pp. 1079–1095, 2002.

[6] K. D. Potter, F. J. Guild, H. J. Harvey, M. R. Wisonm and R. D. Adams, Understanding and control ofadhesive crack

propagation in bonded joints between carbon fibre composite adherends I. Experimental, International Journal of

Adhesion Adhesives, vol. 21, no. 6 ,pp 435-443,2001

[7] D 5573-99, Standard Practice forClassifying Failure Modes in Fiber-Reinforced-Plastic (FRP)Joints,2005

[8] A. Amirkhanov, A. Amirkhanov, D. Salaberger, J. Kastner, M. E. Gröller, C. Heinzl

Visual analysis of defects in glass fiber reinforced polymers for 4DCT interrupted in situ tests Computer Graphics

Forum 35, pp 201-210, 2016

[9] S. Senck, D. Salaberger, C. Gusenbauer, B. Plank, B. Plank, G. Rao and J. KastnerTalbot-Lau grating interferometer

XCT for the quantitative characterization of damage in polymers after impact and static tensile testing, Proceeding 19th

World Conference on Non-Destructive Testing (WCNDT2016), 2016

[10] A. E. Scott, M. Mavrogordato, P. Wright, I. Sinclair, and S. M. Spearing, In situ fibre fracture measurement in carbon-

epoxy laminates using high resolution computed tomography, Compos. Sci. Technol., vol. 71, no. 12, pp. 1471–1477,

2011

[11] R. Böhm, J. Stiller, T.Behnisch, M. Zscheyge, R. Protz, S. Radloff, M. Gude, W. Hufenbach, A quantitative comparison

of the capabilities of in situ computed tomography and conventional computed tomography for damage analysis of

composites, Compos. Sci. Technol., vol. 110, pp. 62–68, 2015.

[12] H. Kunz, E. Stammen & K. Dilger, Local displacement measurements within adhesives using particle tracking and In

Situ computed tomography, Journal of Adhesion, 2015.

[13] EN 2243-1, Aerospace series - Non-metallic materials - Structural adhesives- Test method - Part 1: Single lap

shear,2005

[14] M. B. Banea and L. F. M. da Silva, Adhesively bonded joints in composite materials: An overview, J. Mater.: Des.

Appl., 223,pp. 1-18,2009