1
Effect of reinforcement material on damage caused by low speed impact
of 2D braided composite plates
2nd February 2011
MPF Sutcliffe, C Monroy Aceves, WJ Stronge, RS Choudhry, AE Scott, L Lobo
Cambridge University Engineering Department, Trumpington Street, Cambridge, CB2 1PZ
Northwest Composites Centre, University of Manchester, Sackville Street, Manchester, M60 1QD, U.K
University of Southampton
Laser Optical Engineering Ltd
Summary
This report brings together various results for the Dowty braided fabrics impacted by gas guns at Cambridge following the formal completion of the project at the end of 2009. Three 2D-braided glass-carbon hybrids, with 50, 75 and 100% of carbon reinforcement, were tested using vibrations, static tests and gas-gun impact. [revise]
1. Introduction
Composite materials are widely used in applications requiring good specific stiffness and strength. However, there is a tendency for these materials to suffer more damage as a result of impact than equivalent metallic materials, either in the form of fracture of the relatively brittle fibres or due to cracking in the matrix [1]. Woven materials have traditionally been used in impact-critical applications due to their superior ability, relative to unidirectional composites, to maintain structural integrity under significant impact energies.
[needs augmenting…]
Impact damage results in a reduction of the composite in–plane compressive strength mainly due to delamination [?]. Delamination between plies occurs in the interlaminar (resin rich) regions where the extensional and bending stiffness differs due mainly to different fibre orientation between the layers or, in some cases, different materials [6, 7]. Further investigations have suggested that delamination is initiated by matrix cracks in both opening and shear modes [8, 9].
In general there are at least two phases of impact damage in composites: (1) dynamic compaction of the composite plate which is being compressed ahead of any colliding missile; and (2) delamination of plies in the neighbourhood of the impact site [2]. The exact nature of damage will depend on the composite weave architecture, resin properties and the properties of the colliding missile [?, 3]. However, for low speed impact, damage modes which are observed are bending damage on the distal surface and an approximately circular internal delamination, followed by fibre splitting and perforation or shear failure at high incident energies [4]. Moreover, the impact damage tends to initiate on the distal surface if the ratio of plate thickness to projectile nose radius is less than one (i.e. thin plates) and on the impact surface if the plate thickness is greater than the projectile nose radius (thick plates) [3, 5]. High speed impacts trigger different failure mechanisms because the deformation field induced by the impact induces higher frequency modes in addition to the fundamental (quasi-static) mode of deformation [6].
Tougher resins or three dimensional woven composites can be used to increase the delamination toughness and so improve the composite impact resistance. With tougher matrices and stronger interfaces, larger impact energies are required to initiate delamination [10]. Through thickness tows in three-dimensional weaves act as crack stoppers by altering the fracture paths from intra-tow mode to inter-ply mode; this approximately doubles the fracture toughness [11].
Hybrid materials containing a mixture of ductile glass fibres and relatively brittle carbon fibres can be used to increase the strain to failure of the composite and hence potentially increase the impact resistance, at the expense of increased weight and reduced stiffness. Previous work by the authors [4] has presented preliminary results for braided and 3D woven material. In this paper we explore the effect of the proportion of glass on the impact behaviour of 2D braided materials.
2. Materials
Dry fabrics were braided using a 2×2 bias braid from tows of glass and carbon1200 tex E-glass (Owens Corning) and 800 tex 12k HTA 40 carbon (Tenex). Densities for the glass and carbon, taken from data sheets, are 2550 and 1760 kg/m3, respectively. Three fabrics were woven, containing 50, 75 or 100% of carbon tows. The nominal volume percentage of carbon in the dry fabrics was estimated from the nominal fibre density and tex to be very close to the tow percentage.
Flat plates were resin transfer moulded from 9 layers of the braided material with a ±45° lay-up, using an epoxy resin ( Huntsman Araldite LY 564 with Aradur 2954 hardener), to give a plate thickness of 4.34±0.02 mm. The panels were C-scanned (see details below) to identify and exclude any poor quality regions. Square samples with side length approximately 130 mm were cut out for testing with tows running across the plate diagonals. Plate thicknesses were measured from at least three samples, taking four measurements per sample. The fibre volume fraction was estimated from the pre-form areal densities, the nominal dry-fabric density (to give the volume of dry fibre per unit area) and the final sample thickness. For all three materials the fibre volume fraction was estimated as 0.54. Samples were measured and weighed, and from this the mass density was deduced to be 1668, 1587 and 1461 kg/m3 for the 50, 75 and 100% carbon plates, respectively. The densities were very consistent, within 1.5% of the mean value for the three samples per panel measured.
3. Plate elastic property measurement
3.1 Vibration measurements
The elastic properties of the plates were estimated from their vibration response, using the method described in [5?]. The 130×130 mm samples were suspended from two corners using light thread and then lightly tapped with an instrumented impact hammer. The response of the samples was measured using a light accelerometer attached to the plate.
Table 1 gives the resonant frequencies for the undamaged samples. The first three resonances were used to extract plate stiffness parameters, using the methodology described by McIntyre and Woodhouse [?]. Table 1 gives the plate stiffness constants, with xy properties referred to the plate edges (along the bias direction) and 12 properties referred to the tow directions. Standard rotation matrices are used to relate these properties.
To measure mode shapes the 50% carbon undamaged specimens were lightly tapped on a 5×5 grid, centred on the centre of the specimen, with a nominal 25 mm spacing between grid points. Q factor results were taken from averaging these grid data (excluding outliers associated with loading on nodes). Mode shapes for the first three modes are given in Fig. 1, corresponding to cross (X), George cross (+) and ring (o) modes.
% carbon Mode / 1 / 2 / 3 / 4 / 550 / 723 / 904 / 1281 / 1977 / 2720
712 / 902 / 1268 / 1972 / 2678
701 / 878 / 1222 / 1916 / 2628
75 / 757 / 1023 / 1394 / 2188 / 3024
100 / 752 / 1153 / 1540 / 2391 / 3167
Table 1. Undamaged sample resonant frequencies (Hz)
%carbon / Ex (GPa) / Gxy (GPa) / yx / E1 (GPa) / G12 (GPa) / 1250 / 14.9 / 17.4 / 0.66 / 38.8 / 4.50 / 0.11
14.9 / 17.8 / 0.66 / 39.3 / 4.47 / 0.11
15.7 / 18.5 / 0.65 / 40.4 / 4.78 / 0.094
75 / 15.7 / 21.4 / 0.68 / 46.2 / 4.66 / 0.078
100 / 14.3 / 24.9 / 0.75 / 53.2 / 4.07 / 0.068
Table 2. Plate elastic properties
Figure 1. Mode shapes; (a) mode 1 – Cross (690 Hz), (b) mode 2 - George cross (856 Hz), (c)mode 3 - ring (1206 Hz).
3.2 Static tests
Static centrally-loaded tests were undertaken on 50% and 100% carbon samples, using a screw driven universal testing machine at a slow speed of 0.5 mm/min. The deflection was measured from the bottom surface, so that local deformation at the contact was excluded. The plate was supported along a square of side length 100 mm, either simply-supported or using the same clamping conditions as for the impact test.
The plate stiffness k = /F, where is the central displacement and F the force, was calculated from the slope of the force-displacement curve. Values for the free and clamped conditions are given in Table 3. An effective elastic modulus can be derived using the plate bending formulae from Roark [?], who gives the stiffness of a centrally-loaded square plate of side length L and thickness t, made of isotropic material with an elastic modulus E and Poisson’s ratio 0.3 as
(8)
The corresponding values of are 0.0611 and 0.1267 for fully clamped and simply-supported cases. Taking L = 100 mm, t = 4.357 mm (assumed the same for both plates), gives the effective moduli included in Table 3. The higher effective stiffness for the simply supported plates indicates that the clamped condition is not perfect. The values of 22.9 and 28.8 GPa for effective moduli for the simply-supported 50% and 100% carbon plates, respectively, are between the corresponding values inferred from the vibration tests (Table 2) for Exy and E1 along the bias and tow directions of 14.9 / 39 and 14.3 / 53 GPa for the 50% and 100% carbon plates, respectively.
50 % carbon / 100 % carbonk (MN/m) / E (GPa) / k (MN/m) / E (GPa)
Simply supported / 1.49 / 22.9 / 1.87 / 28.8
Clamped / 1.61 / 12.0 / 2.08 / 15.4
Table 3. Plate stiffness k and effective modulus E for centrally loaded static test.
4. Strength test methodologies
4.1. Impact testing
The samples of size 130×130 mm were mounted in a steel support frame with a square window of 100×100 mmm for impact testing. The edges of the sample were clamped onto the support frame using a thick steel plate, which also containing a window and additionally a small gap to allow a side-on view of the impact site. A steel hemispherical-ended impactor of tip radius 6.25mm and mass of either 12.5 or 44.5g was projected normally at the plates using a gas gun. Energies up to 50 J were used, with corresponding peak velocities of 83 and 46 m/s for the light and heavy impactors, respectively. High speed photography was used to measure indirectly the impact and rebound speeds as well as the contact force. The trailing edge of the impactor was identified from the camera images using an image detection algorithm to track the position of the impactor as a function of time. Velocity and acceleration were taken by numerically differentiating this displacement. The contact force was inferred from the deceleration of the impactor. It is assumed that the contact time is sufficient to allow stress waves to propagate through the impactor and for it to reach equilibrium. For the longer impactor a typical contact time of 300 µs corresponds to stress waves travelling 31 times the length of the longer impactor, so that this assumption is reasonable.
4.2 Compression after impact testing
Braided panels containing 50:50 glass/carbon were tested in compression after being impacted. Because of the relatively small specimen sizes (130×130 mm) compared with the extent of the damage it was not considered appropriate to cut the panels down to allow testing along the fibre direction (recall that the panels were cut with their sides along the bias direction). Hence these specimens were loaded along the bias direction. The loading faces were ground flat and parallel before testing. A purpose-built jig was constructed to support the edges of the panels and prevent macro-buckling, in a similar way to the standard compression after impact specimen []. The specimen was compressed in a screw-driven Instron testing machine at a loading rate of 1 mm/min with the load introduced using flat platens. The end compression of the specimens was calculated from the cross-head displacement. A relatively small correction was made for the machine compliance, measured using a massive steel block in place of the specimen.
5. Impact damage assessment methods
A range of methods, both destructive and non-destructive, were used to assess the damage incurred in the impact. As well as providing useful and complementary information about the damage, this spread of tests provides the opportunity to compare and correlate the observations obtained with the various test methods.
5.1 Ultrasonic C-scanning
Ultrasonic C-scanning was undertaken using a MIDAS NDT water-jet inspection system, using an unfocussed 5MHz probe in transmission mode and a 50mm/s scan rate. The damage area was found from the region above a suitable intensity threshold.
5.2 Contact profilometry
Contact profilometry was performed using a Taylor Hobson Form Talysurf 120 to measure the maximum crater depth at the impact site.
5.3 Visual observation
A standard flat bed optical scanner was used to record the visual appearance of the specimens on their impact and distal faces at a resolution of 300 dpi.
5.4 Sectioning
Selected specimens were sliced using a Struers Accutom-50 to make a series of sections through the damage area. The samples were polished, and imaged using a BX51 Olympus microscope with an automatic stage.
5.5 Micro-CT X-ray imaging.
[Needs completing – Anna] Two 75% carbon samples, which had been impacted at the same nominal energy of 35 J but with the different masses, were examined using X-ray tomography. These samples had C-scan damage areas of 1744 and 798 mm2 for the low and high mass impactors respectively, to give corresponding effective c-scan damage diameters of 47 and 32 mm.
5.6 Shearography
Braided samples of the three materials impacted either with the 45 g impactor at 12 J or with the 12.5 g impactor at 35 J were examined for damage using shearography. The extent of damage was estimated by thresholding the raw signal. The maximum of the impact and distal areas was chosen as characteristic. [Needs more detail - Leon]
5.7 Vibration measurements
More extensive measurements of frequency and Q factor, similar to the measurements for elastic properties as described in section ?, were made on the 50% carbon material, for both undamaged and impacted samples, to explore the effect of damage on the vibration response. The damaged samples had been impacted by a gas gun (see section ?) with an impactor of either 12.5 or 44.5 g. None of the plate measurements were ‘before-and-after’ comparisons, making results prone to differences in plates, e.g. tow orientation differences.
6. Impact damage assessment results
6.1 C-scan results
There was a clear division in the C-scan intensity between the undamaged regions of the plate and a roughly circular region around the impact site. Figure 3 plot the measured damage area as a function of impact energy for the three plates and two impactor masses. The damage area increases roughly linearly with the impact energy for a given material and impactor mass. However there are significant differences between sets of data, with the damage area being significantly smaller for the heavier impactor at the same energy. The 50% carbon material showed the smallest damage area, with the 75 and 100% carbon braids having roughly similar damage. Only one sample was perforated, the 100% carbon braid with the highest impact energy of around 50 J, impacted by the heavier impactor.
Fig. 3. Effect of impact energy on measured damage area using c-scanning
6.2 Profilometry
Figure 4 plots the dent depth, measured by contact profilometry, as a function of impact energy. The dent depth increases roughly linearly with impactor energy with similar behaviour for different materials and impactor masses, with the exception of two 100% carbon high mass samples and one high energy, high mass 75% carbon sample. These samples suffered much more extensive denting than the other tests (e.g. severe visible damage on the distal surface at nominal impact energies of 35 and 50 J), and the 100% carbon high-mass highest-energy sample was perforated.
Fig. 4. Effect of impact energy on measured damage depth using profilometry.
6.3 Observations: visual images, sectioning and X-ray imaging
[Needs modifying in light of CT images] Figure 6 compares cross-sections of two 100% carbon braids under the impact site, for the same impact energy, but with different impactor masses. The impact site was near the top of each figure. Damage with the heavier impactor Fig. 6 (a), although smaller in c-scan area, is characterized by more severe tow damage, while the lighter impactor, Fig. 6(b), causes more widespread delamination damage. This is consistent with the dent measurements, showing in some cases a deeper dent with the heavier impactor at corresponding energies.
Figure 6. Sections at impact site for 35J impact, 100% carbon braid,(a) heavy impactor (b) light impactor.
Figure 7 compares the C-scan results and the scanner visual images of the distal surface of 50% carbon braids, with the two impactor masses at an impact energy of 50 J. With the glass tows, significant matrix damage, probably in the form of matrix debonding, is clearly seen as a significant whitening. Corresponding damage which would presumably also be present in the carbon tows is not visible. This whitening appears to be more intense for the heavier impactor, at comparative impact energies, but not as widespread as for the lighter impactor. The area detected as damage by c-scanning (approximated by the circles superimposed on the figures), corresponds reasonably closely to the observed areas of whitening. The threshold used for the C-scan images has excluded some areas where whitening is seen around the edges of the damage patch.
Figure 8. Comparison of optical (distal face) and c-scan images, 50 % carbon material, nominal impact energy of 50 J:
left - optical, right – C-scan, upper - 12.5g impactor, lower - 44.5 g impactor.
All images are to the same scale. The circles superimposed on the images have an area equal to that derived from the C-scan thresholding procedure.
6.4 Shearography
Figure 3 compares the damage area identified by shearography with that measured by c-scanning, for the six samples impacted either with the 45 g impactor at 12 J or with the 12.5 g impactor at 35 J. There is a good correlation between the two methods. It seems likely that the damage seen in the scanner images of the distal surface (see Fig. ?) corresponds to that detected by shearography.