Dynamic bi-axial testing of woven composites - Center of Excellence ...

Materials Science and Engineering A317 (2001) 135– 139 www.elsevier.com/locate/msea

Dynamic bi-axial testing of woven composites J.D. McGee, S. Nemat-Nasser * Center of Excellence for Ad6anced Materials, Department of Mechanical and Aerospace Engineering, Uni6ersity of California at San Diego, 9500 Gilman Dri6e, La Jolla, CA 92092 -0416, USA

Abstract Thick panel glass/polymer composites are being considered as a structural component in ballistic armor. A new experimental technique has been developed to perform dynamic bi-axial testing on laminate composites. In-plane ultimate compressive strength of the composite is compared for various combinations of bi-axial stress and strain rate. Evolution of the damage in the composite and subsequent kink-band formation is studied by the use of high-speed photography and electron microscopy. © 2001 Elsevier Science B.V. All rights reserved. Keywords: Kink-band; Woven composite; Strain rate; Bi-axial; Experimental

1. Introduction

2. Experimental

Woven fiber-reinforced composites are being considered as structural backing in hybrid armor. In such a role, the composite undergoes dynamic loading due to the initial wave propagation and then subsequent dynamic structural loading, as the hybrid armor flexes in response to the momentum of the impact. The initial wave propagation introduces a multi-axial stress state in the composite, leading to extensive damage. Compressive loading in the plane of the weave can result in kink-band formation. An experimental technique using a split Hopkinson bar is developed to study the effects of multi-axial dynamic compression on woven composites. The experiments indicate that kink formation is sensitive to the magnitude of the confining pressure. It is also of interest to determine the damage evolution during the test. This is accomplished by three techniques: (1) high-speed photography of dynamic tests, (2) video of low strain-rate tests, and (3) electron microscopy of samples at various stages in the load history.

The objective of this experiment is to apply a constant confining stress in the through-the-thickness direction of the sample while loading the sample in the direction of the laminate. A 1.905 cm (0.75 in.) maraging steel split Hopkinson bar provides high strain-rate loading. A momentum trap on the Hopkinson bar is used to ensure that the sample is not re-loaded [1]. One of the assumptions of the Hopkinson bar technique is that the sample length is negligibly small. In practice, it is desirable that the sample size be limited to approximately the bar diameter. Longer samples tend to be impractical, as they may not attain dynamic equilibrium during the loading event [2]. For this testing, samples are 2.032 cm (0.800 in.) in length. For low strain-rate loading, a 20 kip Instron hydraulic load frame is used. The constant confining stress is applied through the thickness by a 7.62 cm (3 in.) diameter pneumatic cylinder. Confining pressures of up to 20 MPa (3000 psi) can be achieved with this compression fixture. Samples are first loaded through the thickness by the pneumatic cylinder and then loaded in the in-plane direction with either the Hopkinson bar or the Instron, depending on the desired strain rate. The crucial factor in achieving valid results with this test is the uniformity of the boundary conditions. For bi-axial loading, uniform boundary conditions on small

* Corresponding author. Tel.: + 1-858-5344772, fax: +1-8585342727. E-mail address: [email protected] (S. Nemat-Nasser).

0921-5093/01/$ - see front matter © 2001 Elsevier Science B.V. All rights reserved. PII: S 0 9 2 1 - 5 0 9 3 ( 0 1 ) 0 1 1 7 1 - 6

136

J.D. McGee, S. Nemat-Nasser / Materials Science and Engineering A317 (2001) 135–139

samples under dynamic loading are difficult to achieve due to the Poisson expansion in the sample during the test. The expansion causes portions of the sample surface to become unconfined or results in transverse loading of the confinement. We evaluated various loading configurations experimentally to determine suitable boundary condition configurations. Our criteria for an adequate configuration are: (1) The sample does not consistently fail at the corners during quasi-static testing; and (2) The ultimate strength of the composite at low confinements and low strain rates approaches the value achieved by more conventional uni-axial large sample compression tests [3]. For the first criterion, dynamic tests tend to be more forgiving of imperfect boundary conditions due to the time scale of the failure event, and thus the quasi-static test is more limiting. Three sample configurations are examined. The freeend configuration (Fig. 1a) was determined to be inappropriate for the laminate composites. As will be seen later, the strength of the composite is sensitive to the confining pressure. In this configuration, failure initiated at the ends of the sample due to lamina buckling at the unconfined corners. A free-edge configuration (Fig. 1b) is also evaluated. This configuration is found to be suitable for dynamic loading at high confinements. In quasi-static tests, however, damage initiated at the corners. The most uniform boundary conditions were achieved by decreasing the thickness of the sample and adding Teflon spacers (Fig. 1c). The low compliance of the Teflon allows for more uniform confinement in through-the-thickness of the sample throughout the test. In the primary load direction, the compliance and cross-sectional area of the Teflon is low enough that the Teflon transmits no significant force into the transmission bar. At higher confinement pressures of ca. 14 MPa (2000 psi), the Teflon flows out from between the sample and the compression fixture and pushes against the loading bars. For quasi-static loading, this is not significant since the Instron is used in displacement control that prevents the load column from separating from the sample. In the high strain-rate

tests however, the Teflon pushes the incident and transmission bars apart which creates gaps between the bars and the sample, causing experimental errors during the dynamic loading. Thus, the free-edge configuration is used for the dynamic high-confinement loadings. To study damage evolution in the samples, a highspeed camera (Hadland Imacon), is used to record the dynamic tests and video recording is used for the quasi-static tests.

3. Material The samples are prepared from a 45-layer balanced twill-weave S2 glass in a vinyl ester C350 matrix. The individual plys are oriented at 0°. The University of Tuskegee, using the vacuum assisted resin-transfer molding technique, manufactures the composite. The void fraction of the finished material is approximately 8%, as determined by microscopy. The thickness of the panel is 2.03 cm (0.8 in.) and varies by 1.25 cm m − 1 (0.15 in. ft. − 1) across the span of the panel. Samples are cut from the panel and end-milled to final tolerances. The bi-axial test samples are 2.032 cm (0.800 in.) in length and 0.762 cm (0.300 in.) in width. The samples used in the compliant configuration are 1.397 cm (0.550 in.) in height and those used in the free-edge configuration are in the as-received height of approximately 2 cm. The length of the sample corresponds to approximately two wavelengths of fiber undulation.

4. Experimental results

4.1. Strain-rate/confinement sensiti6ity The ultimate strength of the composite is used to characterize the sensitivity to confining pressure and strain rate. S2/vinyl ester samples are tested at strain rates varying from 0.001 to 1000 s − 1 and at confining pressures up to 13.8 MPa (3000 psi). Fig. 2 illustrates

Fig. 1. Bi-axial loading configurations: (a) Free-end configuration results in buckling of the lamina at the edges and premature failure of the sample, (b) Free-edge configuration is used for dynamic loading and higher confining pressure, and (c) Compliant confinement with Teflon spacers is used for quasi-static testing and dynamic testing up to 14 MPa (2000 psi) confinement.


137

while the confining pressure prevents catastrophic failure. In addition to changes in the ultimate strength, the nature of the kink band changes with increasing confinement. At low confinement levels, failure tends to be by axial splitting or small angle kink bands. As the confining stress increases, the kink band rotates to larger angles (Figs. 3 and 4).

4.2. Damage e6olution

Fig. 2. Effects of strain rate on composite response.

Fig. 3. Effects of strain rate and confining stress on ultimate strength of composite. The kink band rotates towards the confinement, as the confining pressure increases.

the strain-rate sensitivity of the composite. The ultimate strength increases dramatically at higher strain rates. Variations of the ultimate strength over a range of strain rates and confining pressures are summarized in Fig. 3. The strength of the composite is also sensitive to changes in confinement when the confinement level is low. This sensitivity drops off, as the confinement increases, approaching a plateau. Large strains are achieved due to frictional sliding along the kink band,

Our observations indicate that kink formation is preceded by extensive micro-mechanical damage that tends to initiate at defects (e.g. voids), although some distributed damage also occurs, with no obvious correlation with the pre-existing defects. Kink formation is not the initial failure mechanism, but rather kinks emerge and propagate across the sample after a critical amount of micro-mechanical damage has accumulated, i.e. kink band formation is the final stage of the failure process. To gain insight into the failure process, three techniques are used: video of quasi-static tests, high-speed photography of dynamic tests, and electron microscopy of tested samples. A combination of these photographic techniques and microscopy indicate that kink formation occurs after damage initiation in the sample. Some evidence of the initiation and propagation is available from the high speed photography of high strain rate loading. Photographs taken at 10mS intervals show that damage can be visually observed when the stress in the sample nears the peak value, and visible damage propagates across the sample (Fig. 5) as the sample unloads. At this point in the loading the kink band is fairly discreet and is characterized by cracking in the out of plane tows and microbuckling of the in-plane tows (Fig. 6a). Although the kinks observed in the photographs are at the surface, microscopy indicates that damage in the kink band extends across the sample (0.3 in.) and that surface damage is a good indication of the internal damage, albeit, the surface

Fig. 4. Kink band orientation for (a) unconfined, and (b) 10.3 MPa (1500 psi) confinement.

138


Fig. 5. High-speed photography of kink band formation and propagation.

Fig. 6. Comparison of damage in (a) a propagating kink band and (b) fully developed sliding kink band.

impression tends to be wider and more diffused due to out of plane deformations and buckling. This out of plane deformation actually has the effect of magnifying internal damage into damage that can be observed visually. After the initial kink formation and propagation, a plateau is reached on the stress-strain curve and the kink band is fully developed. This is similar to kink formation observed in unidirectional composites [4] and in woven composites of different material systems [5]. At this stage, the kink tends to act as a large frictional crack, as the two sides of the sample slide past each other and the kink band widens and accumulates more damage (Fig. 6b). This damage includes large rotations of particles, buckled fiber tows, and extensive matrix cracking. For some dynamic tests, multiple crossing and parallel kink bands are observed.

Samples that are loaded to before failure and just after failure provide a good indication of the micromechanical damage that occurs in the material. Some samples with partially developed kink bands are examined in detail. In addition to the fiber buckling, damage is seen to include cracks in the tows that are perpendicular to the loading plane. These cracks consist of fiber-matrix debonding and matrix cracking. They remain confined to the fiber tows (Fig. 7) until the kink band is fully developed and begins to slide. These cracks continue across the width of the sample, following the undulation of the tows. They are also present at the tip of the kink band. Microscopy of samples loaded to before failure indicates that these cracks are present prior to buckling of the fiber tows and mostly occur around voids, but are also dispersed through out the sample.


139

6. Conclusion As either the through-the-thickness confinement or the strain-rate increases, the ultimate compressive strength of the composite increases dramatically. The failure is initiated by micro-cracking, leading to damage localization and kink-band formation. As confinement increases, the localized failure zone transforms from axial splitting to the kink-band formation. Our experiments indicate that micromechanical damage prior to failure is the critical component of the failure initiation. Our results therefore provide guidance for establishing realistic micromechanically-based failure models for this class of materials.

Acknowledgements Fig. 7. Fiber/matrix debonding and matrix cracking are limited to the cross tows until the kink band is fully developed.

Funding for this project was provided by the Army Research Office under grant cDAAH04-95-01-0369 to UCSD. References

5. Discussion Failure of composites under in-plane axial loading tend to occur by axial splitting or kink formation. The failure mode is dependent on the specific material system involved. Furthermore, variations of material properties can change axial splitting/delamination into kink formation. Some of these parameters that can affect the failure mode include defect density [5], fiber volume fraction [6], fiber-matrix interface properties [7], and temperature [8]. The results presented above indicate that strain rate and through-the-thickness compression can also alter the failure mode. As pointed out by Grape and Gupta [8], this behavior is similar to the transition from axial splitting to shear-faulting in rock as through-the-thickness compressive stress is increased [9– 11]. .

[1] S. Nemat-Nasser, J.B. Isaacs, J.E. Starrett, Proc. R. Soc. Lond. A435 (1991) 371 – 391. [2] Follansbee, P. The Hopkinson Bar. Metals Handbook, ASM, Metals Park, OH, Vol. 8, 1985, pp.198 – 203. [3] A. Haque, H. Mahfuz, C. Ingram, S. Jeelani, Compos. Eng. 4 (6) (1994) 637 – 651. [4] S. Sivashanker, N.A. Fleck, M.P.F. Sutcliffe, Acta Mater. 44 (7) (1996) 2581 – 2590. [5] V. Gupta, K. Anand, M. Kryska, Acta Mater. 42 (3) (1994) 781 – 794. [6] A.M. Waas, S.H. Lee, Dynamic compressive behavior of glass fiber reinforced unidirectional vinyl ester composites, 35th Annual Technical Meeting of the Society of Engineering Science, 27 – 30 September 1998, Washington State University, Pullman, Washington. [7] S.R. Swanson, G.E. Colvin, J. Eng. Mater. Technol. 115 (1993) 187 – 192. [8] J.A. Grape, V. Gupta, Mech. Mater. 30 (1998) 165 –180. [9] S. Nemat-Nasser, H. Horii, J. Geophys. Res. 87 (B8) (1982) 6805 – 6821. [10] H. Horii, S. Nemat-Nasser, J. Geophys. Res. 90 (B4) (1985) 3105 – 3125. [11] H. Horii, S. Nemat-Nasser, Phil. Trans. R. Soc. Lond. 319 (1986) 337 – 339.