COVER SHEET Title: Experimental and Modeling Investigation of Blunt Impact to Stringer-Reinforced Composite Panels Authors: Zhi Ming Chen Hyonny Kim Gabriela K. DeFrancisci

ABSTRACT

Due to their high performance, carbon fiber composites are seeing increasing

use in aircraft applications. However, exposed composite structures are susceptible to delamination damage caused by transverse impacts. In particular, accidental impacts by ground service equipments (GSE) can involve high energy (over 1000 J) and can induce significant delaminations and even fiber failure in composite structures, without leaving visible signs that internal damage has occurred. The objectives of this research were to experimentally investigate blunt impact events by GSE on a composite fuselage, and to simulate the test results using finite element (FE) analysis. Blunt impact experiments were conducted on two stringer-reinforced composite skin panels that were similar to composite aircraft fuselage structures. One of the panels were quasi-statically indented and the other dynamically loaded at 0.5 m/s to produce significant damage consisting of shear tie radial delaminations, fiber failure/crushing, and also skin-to-stringer delaminations. For the dynamically-loaded specimen, stringer cracks and skin cracks also developed. A comprehensive FE model has been developed to predict different damage modes and damage extent created by these blunt impact events. Continuum shell elements were used for the panel with Hashin-Rotem failure criteria and surface-based cohesive interaction implemented to model in-plane ply failures and delaminations, respectively. The model results correlated well with the experimental data, closely matching the panel load vs. indentation data. The peak load was predicted within 5% accuracy, and the predicted delamination damage and crack formations coincided with the observed load drops. 1

Hyonny Kim, Professor Zhi Ming Chen, Graduate Student Researcher Gabriela K. DeFrancisci, Graduate Student Researcher Dept. of Structural Engineering, University of California at San Diego, La Jolla, CA 92093

INTRODUCTION

With the recent trend towards building lightweight and more fuel efficient aircraft, composites are increasingly used in primary structures such as the fuselage and wings [1]. While composites have high in-plane strength and stiffness, their out-of-plane properties are significantly weaker and are prone to being damaged by transverse impacts, which can often produce delamination and matrix cracking, as well as fiber breakage [2]. Accidental impacts by ground service equipment (GSE), in particular have been identified as a major impact threat source to aircraft. Specifically, 50% of major damage to commercial aircrafts is caused by damaging contact with baggage vehicles, while 60% of minor damage is caused by ground service vehicles [3]. The mass of these GSE can range from 2,300 kg to 3,000 kg for smaller GSE such as belt loaders, and even 10,000 kg for large cargo loaders. Upon collision, these heavy “projectiles” can impart large kinetic energy levels to the aircraft even though they are moving at low velocities of up to 0.5-1 m/s (1-2 mile/hr). Previous studies have shown that the size of local delamination damage on composite plates scales linearly with the initial kinetic energy of a hard impactor [4, 5]. However, GSEs are typically outfitted with soft rubber bumpers that deform during collision which creates very large contact area and reduced local stresses at the site of impact [6]. It was found in this present study that rubber bumpers can prevent the formation of local, visible cracks and delaminations, and can lead to widespread damage away from the impact site which is more difficult to detect.

For low velocity events, dynamic impact scenarios can often be experimentally represented using equivalent quasi-static tests. Equivalence was shown for fracture tests [7-9], for impact to composite plates [10-15], and composite shells [16, 17]. Quasi-static indentation tests can provide more insight to the damage progression and interaction of damage modes than dynamic impact tests since direct observations are more easily made. This allows damage mechanisms to be compared between specimens. In addition, vibrations that occur when a composite shell is impacted with a low velocity, high mass projectile are considered to be higher order effects and were found to have negligible influence on the panel when compared to the damage produced in quasi-static tests [16].

Composite aircraft fuselage structures consist of complex components, including skin, stringers, shear ties, C-frames, and connections between the C-frames and the floor beams. During impact the fuselage reacts to the applied load as a whole, and deformation of the components can cause contacts and interactions between them, and thus the individual failures often compete with each other. While experimental and analytical studies of the individual components have been conducted and published, few publications studied the fuselage at a larger scale containing two or more components. Therefore, the body of work that deals with the interaction of the components and their failure modes is mostly missing in literature. The research presented here is a part of a series of studies that aims to provide experiment data and establish analytical techniques for investigating the formation of low velocity blunt impact damage in multi-component composite fuselage structures.

DESCRIPTION OF EXPERIMENTS

Specimen Design

The two specimens tested were identical and they had geometry reflecting the construction of current carbon/epoxy composite aircraft fuselage design. They consisted of a skin of overall dimension 914 x 914 mm having a curvature of 2794 mm, two stringers (hat stiffeners), six shims (spacers between the shear ties and the skin), and six shear ties. Figures 1 and 2 show the specimen configuration. Both the stringers and the shims were co-cured to the panel’s skin in the autoclave. The stringers were oriented along the skin’s longitudinal direction at 305 mm spacing. The six shims were positioned along the hoop direction at 508 mm spacing as shown in Figure 1. The shear ties were then bolted to the skin post-curing using 6.35 mm shank diameter countersunk HiLok HL19 PB8-5 alloy bolts and HL70-8 aluminum collars. They were mounted directly on top of the shims and a portion of the stringer flange. Figures 3 and 4 provide additional detail of the stringer and shear tie geometry, as well as the stringer-to-skin and shear tie-to-skin connections. This specimen series was designated as StringerXX, as their primary stiffening components were the stringers. They were designed to observe aircraft fuselage response to impacts between the C-frames (mounted to shear ties).

Manufacturing of the specimens took place at the University of California at San Diego (UCSD), with curing at San Diego Composites. The specimen materials were carbon fiber with a toughened epoxy matrix supplied by Cytec Engineered Materials (aerospace-quality material): X840 unidirectional tape, and X840 3K and X840 6K woven fabrics. The skin layup was unidirectional [0/45/90/-45]2S with single layer of 3K fabric on each outer face. The stringers had a layup of unidirectional [0/45/-45/90/45/-45/0]S with the 0° direction oriented along the stringer main axis direction. The shear ties were all-fabric material (6K woven) for increased drapability into the compound curved regions of the shear ties. They had a layup sequence of [±45/0]3S.

Figure 1. Sketch of StringerXX test specimen (shims not shown).

Figure 2. Interior view of test specimen; impact load was applied from exterior skin surface.

Figure 3. Cross-section view of stringer-skin connection; portion of stringer co-cured to the panel

skin referred to as the stringer flange (dimensions shown in mm).

Figure 4. Cross-section view of shear tie and connection to skin (dimensions shown in mm).

Experimental Set-up

Two identical specimens (Stringer02 and Stringer05) were tested in UCSD’s Powell Structural Research Laboratory at different velocities. The same boundary conditions, shown in Figure 5, were used for both specimens. The shear ties were bolted directly to 6.35 mm thick steel fixtures in lieu of C-frames. Blunt loading was applied to both specimens by a D-shaped deformable rubber bumper (from TUG Belt Loader) mounted to the crosshead of the test apparatus, as shown in

Figure 5. The width of the bumper was 76 mm and the overall length was 198 mm, with an outer radius of curvature of approximately 102 mm. The specimens were loaded at the specimen center between the stringers and between the shear ties (see Figures 2 and 5). Additionally, a displacement potentiometer (POT) was attached to the interior face of the skin to track the indentation depth (i.e., deformation of the skin) underneath the bumper.

Figure 5. Test setup showing blunt loading and boundary conditions.

Specimen Stringer02 was quasi-statically indented at 15 mm/min using a 2700

kN SAETEC uniaxial tension/compression test machine. The quasi-static test speed allowed for better observation of failure initiation and progression, because the test could be stopped and unloaded at any point after various levels of damage has been generated, thereby allowing one to document the damage state. In this way, multiple loading cycles were applied to Stringer02 until significant damage had been accumulated. The Stringer05 test was designed to replicate real impact velocity of GSE observed LAX airport. It was dynamically loaded at 0.5 m/sec using a 980 kN MTS servohydraulic actuator. This test speed was chosen to match GSE braking speed within close proximity of an aircraft. Unlike the quasi-static loading of Stringer02, a single displacement-controlled loading stroke was applied to Stringer05 to generate significant damage during one test, as opposed to via multiple load cycles. Due to the faster loading rate, a Phantom v.7.3 high speed camera running at 6006 fps was used observe the interior side of the panel during the dynamic test. Post-test damage evaluation of the specimens was conducted visually and via non-destructive ultrasound scans using a Physical Acoustics Pocket UT portable system with 3 and 5 MHz transducers. The two sets of test results (quasi-static and dynamic) were compared to determine potential dynamic effects on how composite fuselage structures respond to this type of blunt loading. EXPERIMENT RESULTS Deformation of Rubber Bumper

During the experiments, the D-shaped rubber bumper deformed considerably as it flattens out before building up higher contact forces. The bumper conformed to

the panel surface, creating a larger contact area of approximately 8400 mm2, resulting in lower contact pressures relative to conventional metal indentors. The interlaminar shear stress around the periphery of the contact area was thus low, relative to typical metal-tip impacts, due to the much larger total contact perimeter length. When the bumper was fully flattened and the panel had undergone significant deformation, the bumper was pushing over the stringer-skin connections (i.e., stringer flanges), allowing a direct load path into the stringers walls, as shown in Figure 6.

Figure 6. Illustration of the experiments after the D-shaped bumper has collapsed.

Observations of Stringer02

Figure 7 shows the force vs. indentation curves for Stringer02 for five separate loading sessions. Each successive re-loading passed through the point where the previous loading ended, showing that no change in the structure occurred during inspection. Inflections in the data plots were observed at low-level loads due to the stiffening behavior of the rubber bumper as its open “D” shape cavity deformed and collapsed. When the load reached approximately 2.89 kN (varies slightly with each loading), the bumper fully collapsed and essentially turned into a solid block, so additional load applied more directly translated into deformation of the panel.

Figure 7. Stringer02 contact force vs. indentation depth plots.

0

10

20

30

40

50

60

70

0 5 10 15 20 25 30 35 40

Con

tact

For

ce (

kN

)

Indentation Depth (mm)

1st Loading

2nd Loading

3rd Loading

4th Loading

5th Loading

The first two loadings created only shear tie failures. During the first loading,

low level clicking noises were observed starting ~35 kN of contact force, signaling potential matrix failures in the shear ties. The panel held a load of 44.84 kN and experienced an indentation of 27.7 mm directly under the center of the bumper a loud cracking noise was heard. When the panel was unloaded, the shear ties showed residual bending strains (nearly equal + and – values for strain gage measurements taken on the front and the back of the shear ties) and there was visible crushing damage in the shear tie corners. During the second loading, active clicking sounds were observed at 44.48 kN and large deformations in the panel produced opening moment deformations in the curved portion of the shear ties, leading to radial delamination at these locations. These failures at the shear tie level did not cause significant load drops, but they did produce measurable stiffness loss.

During the third loading, a load drop occurred at 57.96 kN at an indentation of 32.3 mm. The stringer flanges delaminated in the locations where the panel was bolted to the shear ties (i.e., at an internal structural feature located away from indentation site), despite there being no damage at the loaded zone. The size and location of the delamination areas can be seen in Figure 8.

Figure 8. Stringer02 after fourth loading. White boxes indicate area of delamination damage incurred

during 3rd loading. Dashed red box indicates area of delamination incurred during 4th loading.

Additional skin-stringer delamination occurred during the fourth loading (see Figure 8), at the load-applied zone, as well as at locations adjacent to the shear ties. Major load drop occurred at 61.33 kN at a displacement of 34.5 mm due to the large-sized delamination growth. However, no visible damage was found on the skin’s external surface after the panel was unloaded, despite the development of significant stringer delaminations. Figure 8 shows the exterior surface at this point, as well as the delaminated areas based on ultrasonic A-scan inspection. Additional crushing/fiber damage occurred in the curved region of the shear ties. Figure 9 shows the cumulative damage on one of the shear tie’s curved region. This damage occurred at the edge of the corner where the shear tie made contact with the stringer flange (i.e., in the primary load path). It should be noted that after the load drop at 61.33 kN, the panel still held a load of 47.06 kN.

During the fifth loading, asymmetrical loading of the panel was observed because the previously-delaminated stringer flange (see Figure 8) exhibited less stiffness than the intact stringer, thereby redirecting the load path towards the intact stringer. The panel held a load of 55.34 kN before extensive skin-stringer delamination occurred on the second stringer as well. The delamination of the second stringer extended to the free edge of the panel. It is hypothesized, however, that delamination of the stringer extended first from the shear ties location to the center of the panel (where loading is applied), because of the pre-existing delamination damage at the shear tie locations of that stringer flange. The pre-existing delamination could have acted as pre-cracks that propagate inward as the loading increased.

Figure 9. Stringer02 cumulative crushing damage (circled) of one of the shear ties after 4th loading.

Significant stiffness loss can be observed between each consecutive loading (see

Figure 7). This is due to the accumulated damage including delamination and crushing of the shear ties in the curved region, and also delamination of the stringer flanges following the third loading. The load level at which delamination growth occurred can be inferred by noting the gradual softening during the fourth and fifth loading in Figure 7. Past 25.4 mm of indentation, the rubber bumper reached a state of high compression so that its stiffness was much larger than the stiffness of the panel, and thus the load vs. displacement curve reflects panel deformations. Delamination growth in the stringers, to the degree that global stiffness is affected, is observable as a departure from the linear force vs. indentation displacement trend, indicating stiffness loss. By observing the last segments of the 4th loading curve just before the load dropped, it can be concluded that delamination growth of the first stringer’s inner flange initiated at a load level of ~56 kN and propagated until 61.65 kN due to the softening observed in this loading range. Similarly, an examination of the fifth loading curve shows that the delamination growth of the second stringer flange initiated at ~51 kN and propagated until 55.34 kN. Stringer05 Test Observations

The dynamically-loaded Stringer05 load vs. indentation plot is shown in Figure 10, along with the combined loading plot for Stringer02 obtained by compositing only the new indentation data from each successive loading. As can be seen from

this figure, the two panels show similar early response and peak loads. However, their initial failure modes and failure propagations were quite different.

Figure 10. Stringer02 and Stringer05 contact force vs. indentation depth. In Figure 10, two significant load-drops can be observed in the Stringer05 data

at 36.1 and 38.1 mm, with each load drop corresponding to failures occurring at one of the two stringer flanges that were being directly loaded. The first load drop, at 36.1 mm of indentation displacement, was surmised to be caused by skin-stringer delamination and stringer fiber failures on the stringer flange to the right of the bumper loading location. Although this stringer flange was not observed in the under-panel high speed video, carbon fiber debris was seen ejecting from that location during that time, indicating failure. Post-test examination of the specimen also confirmed this damage.

The second load drop, at 38.1 mm of indentation, was caused by the skin-stringer delamination and fiber failure of the stringer flange, at a location to the left of the bumper, as observed directly by high speed video. The high speed camera was used to monitor the stringer flange on the left side of the impactor, and had a view of the panel’s interior (stringer flange on the right of the impactor not visible). Figure 11 shows still-capture frames from the high speed camera, showing damage growth and corresponding image timing. Figure 11a shows the stringer-skin connection shortly after the bumper has contacted with the specimen, before any damage has been incurred. The metal rod in this figure is the displacement POT used to track the indentation (skin deflection) of the panel. Figure 11b shows the still frame at 199.0 msec after impact. Carbon fiber debris can be seen ejecting from the out-of-view stringer flange, indicating failure at that location. Figure 11c shows that delamination between the skin and stringer underneath the impactor occurred first at 199.6 msec after impact, and then the delamination grew towards the boundary supports (i.e., the shear ties). Afterwards, delamination of the stringer from the skin allowed the stringer to deform freely during loading, and stringer cracking (fiber failures) occurred within the next 4 msec (as can be seen in Figure 11d.) after the delamination damage has started. It should be noted that during this 3.4 msec between Figure 11b and 11d, the actuator has moved 1.7 mm (at 0.5 m/s speed).

0

10

20

30

40

50

60

70

80

0 5 10 15 20 25 30 35 40

Con

tact

For

ce (

kN

)

Indentation Depth (mm)

Stringer02 Combined

Stringer05

Figure 11. High speed video still frames during Stringer05 test at (a) t = 17.2 msec, (b) t = 199.0

msec, (c) t = 199.6 msec, and (d) 202.4 msec after impact. Post-test examination of the panel revealed that a through-thickness skin crack

formed along the skin-stringer joint. It is unknown when exactly the skin cracking occurred relative to the other failure events as that damage was located outside the high speed camera’s field of view. However, it is surmised that the skin crack occurred at roughly the same timeframe as the skin-stringer delamination and stringer cracking because high shear stress buildup. Figure 12 shows the post-test panel external surface with the through thickness skin crack. Despite the level of damage, there is almost no residual deformation of the skin. The only external sign of impact damage was the formation of skin crack. Figure 13 shows the post-test damage from both the exterior and internal views of the panel. The delaminated stringer flange area is highlighted by white hatch marks on the exterior view. Similar to the String02 panel, softening of the shear ties due to radial delamination and corner crushing was found. It is important to note that all failure modes were excited close to the skin-stringer connections (along the path primary load path).

Figure 12. Post-test state of Stringer05 showing surface cracks along the skin-stringer connection at

the impact location.

Figure 13. Center image: post-test A-scan map of Stringer05 (hatched areas are locations of skin-stringer Delamination); side images: crack formations along the stringer radii and on the flanges

viewed from the interior side.

DISCUSSION

Deformation of the rubber bumper played a significant role in the damage formation for both tests. As the hollow D-shaped bumper was compressed and flattened, a more direct load path developed, allowing load to transfer directly into the adjacent stringer flanges, and subsequently into the shear ties and the steel boundary fixtures. Thus, damage was primarily found along this load path. Analysis of these experiments can be simplified to three-point beam bending of the stringers since the panel skin has much lower flexural rigidity compared to the stringer’s cross section. This concept is illustrated in Figure 14, with the shear force V(x) diagram of the stringer.

Figure 14. Side view of the panel, the impact effectively creates three-point bending load on the stringers; the shear force diagram is shown below this sketch.

As shown in the diagram, the stringer is effectively a 508 mm long beam

supported at its ends by the shear ties, with a pressure load applied over a small portion of the mid span by the rubber bumper. These shear forces produce transverse shear stresses, which will develop higher values at the geometrical transitions between the stringer and shear ties, thus making these locations most susceptible to initial failure. This simple viewpoint explains the locations of the initial skin-stringer delamination failure, and the locations of the stringer and skin cracks, as observed in the two tests.

To explain the differences between the two test results, the concept of dynamic localization of the deformation response is explored. In the Stringer02 (quasi-static) test, skin-stringer separation initiated at the shear tie support locations, whereas in the Stringer05 (dynamic) test, the same damage initiated locally under the impactor. The Stringer05 panel also suffered additional stringer and skin cracks at the impact zone. These damage modes indicate that dynamic localization can influence the panel deformation shape and stresses, even at a low impact speeds of 0.5 m/sec due to the large specimen size. Figure 15 illustrates the time-dependent deformation response of a beam undergoing dynamic three-point bending at various times during the impact loading.

Figure 15. Time-dependent deformation response of a beam. The first three sketches in Figure 15 show the initial deformation response of

the beam soon after initial contact, similar to that experienced by Stringer05. As can be seen in these sketches, this deformation response is influenced by inertial effects (i.e., higher order mode shapes) in which the central portion of the panel is mainly reacting against the dynamically applied loading. The boundary reactions do not balance with the loading, and localization of the loading does not allow equal shear force distribution along the “beam” length (as illustrated in Figure 14). Thus, higher stresses existed at the center of the stringer, thereby inducing failure there. As the time after initial contact increases (or for slower loading rate), deformation of the beam transitions into a quasi-static type response (like Stringer02), illustrated

by the last sketch in Figure 15. In this case, the global response of the panel balanced the reactions equally into the outer boundaries (shear ties). Also, since transverse shear stresses at the indented zone were kept low by the soft bumper, skin-stringer delamination was not initiated locally in Stringer02, and instead, the damage started at the locations of the outer boundaries where stiff shear tie contact would induce higher shear stresses (stress concentrations). Strain gauge data comparison between the two tests also supported this explanation of dynamic localization effect. D-Shaped Bumper Compression Test and FE Model

FE simulation of the StringerXX specimens requires the accurate modeling of the D-shaped bumpers used in these experiments because the local stresses in the impact zone strongly depend on the bumper material and deformation geometry. However, numerical simulation of the large deformations of a nearly incompressible material (i.e., the collapsing of the D-shaped rubber bumper) is known to introduce numerical instabilities due to excessive element deformations. To overcome the stability issues, the Ogden hyperelastic material model was used to define the rubber constitutive behavior since this model can achieve stable calculations for extremely high rubber strains [18]. For this model, the Ogden parameters were determined from a comprehensive test data set obtained by Treloar for a similar chloroprene rubber material [19]. The Ogden parameters are summarized in Table I. To assess the similarity between Treloar's rubber material and the rubber material used in GSE bumpers, uniaxial compression tests were conducted for a rubber bumper pushing against a rigid flat surface, as shown in Figure 16. The results of these experiments were compared in Figure 17 with the FE simulation of the bumper compression. As can be seen in the figure, the FE model closely matched with the experimental data.

TABLE I. OGDEN MATERIAL PARAMETERS WITH N = 2

i μi (MPa) αi Di (1/MPa)

1 0.459 3.564 0.01

2 3.409 -0.149 0.01

Figure 16. D-shaped bumper compression test setup (left) and fully collapsed bumper (right).

Figure 17. D-shaped bumper compression test data and FE result.

Modeling of Stringer05 Experiment

A FE model of the Stringer05 (dynamically impacted) panel test has been developed in Abaqus/Explicit. Due to the symmetry of the panel, only half of the panel and the impactor have been modeled. The Abaqus CAE drawing of the panel model is shown in Figure 18, with the plane of symmetry highlighted in red. As can be seen in the figure, details of the stringer flange and the shear tie radius were included to increase the fidelity of the model. First, the step joint of the skin-stringer connection has been modeled with the individual ply drops in stringer flange. This was necessary in order to predict the panel’s stiffness since the stringer flange is in the primary load path that transfers load from the bumper to the stringer section, and ultimately to the shear ties. A model created without the ply drops in the stringer flange would overestimate the panel’s stiffness. Also, the step joint affects the transfer of transverse shear stresses between the skin and stringer, so having the correct geometry for the step joint is important in correctly predicting the shear stress state in the skin-stringer connection. The skin and the stringer are meshed separately to allow for cohesive surface interaction between their contacting surfaces. Secondly, the shear ties were also modeled with four stacked layers of elements in the through-thickness direction so that cohesive surface interactions could be applied between the layers to simulate shear tie delamination (induced by radial tension stresses). As discussed in the test results section, radial delamination of the shear ties have been observed to reduce the rotational stiffness of the shear ties, allowing the panel to soften before other failure modes were excited. The shear ties were separated into only four individual layers in the model, as opposed to twelve individual plies, for lower computation cost reasons.

0

50

100

150

200

250

0 20 40 60 80 100

Contact Force (kN

)

Displacement of Back of Bumper (mm)

Bumper Test 1

Bumper Test 2

Bumper Test 3

Bumper Test 4

FEA Ogden Model

Figure 18. Solid model geometry for Stringer05 model.

The fully meshed Stringer05 FE model is shown in Figure 19. In this figure, the

rubber bumper, attached to a steel plate is shown in white, and the composite panel is shown in blue. The 8-node continuum shell (SC8R) elements were used for the panel, and the 8-node continuum solid elements (C3D8I) was used with hyperelastic material for the D-shaped bumper (as described in the previous section). The built-in Composite Layup Manager in Abaqus CAE interface was been used to define the sectional properties of the skin, stringers, shims, and shear ties. Carbon/epoxy material properties were obtained from the manufacturer’s data sheets and supplemented by the closest matching material published by Daniel and Ishai [20]. To model the softening behaviors seen in the test, the Hashin-Rotem failure criterion has been implemented to all of the composite sections to simulate in-plane ply failure, and cohesive surface interactions have been implemented to simulate delaminations (skin-stinger interface and in shear tie radius region). The carbon/epoxy properties and the cohesive properties used in this model are summarized in Table II to IV.

Figure 19. Half-symmetric FE model of Stringer05.

TABLE II. ELASTIC CARBON/EPOXY LAMINA PROPERTIES

Material ρ (g/cm3)

E1 (GPa)

E2 (GPa)

ν12 G12

(GPa) G13

(GPa) G23

(GPa)

Tape 1.63 168 10.3 0.27 6.89 6.89 3.72

Fabric 1.61 80.0 80.0 0.06 6.48 5.10 4.07

TABLE III. HASHIN-ROTEM IN-PLANE MATERIAL FAILURE PARAMETERS

Material F1t (MPa)

F1c (GPa)

F2t (MPa)

F2c (MPa)

F6 (MPa)

Tape 2799 1620 57.2 227 75.8

Fabric 993 772 855 896 71.0

TABLE IV. COHESVE SURFACE INTERACTION FAILURE PARAMETERS

Material K

(GPa/mm) Tn

(MPa) Ts

(MPa) Tt

(MPa) Gnc

(N/m) Gsc

(N/m) Gtc

(N/m)

Tape (Skin-to-Stringer)

170 45.9 77.2 77.2 403 1629 1629

Fabric (Inside Shear Tie)

170 45.9 77.9 77.9 771 3152 3152

Additional interactions, constraints, and loads were chosen to best simulate the contact conditions observed the Stringer05 experiment. A surface-to-surface contact interaction was been defined between the panel skin and bumper surfaces. For the interaction property, the penalty contact constraint was chosen over the kinematic contact constraint as it is more effective at reducing contact chattering. Furthermore, a simple surface-to-node displacement tie constraint was defined for the shear tie and skin connections, representing the bolts used in the test specimen. This is a valid simplification because the bolt connection did not experience any failures. Similarly, fixed displacement boundary conditions were applied to the ends of the shear ties where they were bolted to the steel fixtures in the physical panel. Finally, the steel plate and the bumper were commanded to displace at a constant velocity of 0.5 m/sec towards the panel for 175 mm. Model Results

The deformed shape of the model at peak contact force is shown in Figure 20, and a comparison of the load vs. skin indentation curves between the experimental data and the FEA model is shown in Figure 21. As can be seen in Figure 21, there is good matchup between finite element model and the test panel in terms of their response up to 36.7 mm corresponding to the failure load. The FEA model’s peak load was 68.6 kN, which was 2.3% higher than the experimental peak load at 67.0 kN. In the Striner05 experiment, local skin-stringer delamination, stringer and skin

fiber failures, and shear tie radius delaminations were observed. The FE model captured each of these failure events with varying degrees of success. The occurrence of each failure event, as predicted by the FE model, is indicated in Figure 21. Crack formation was determined in the model by examining the Hashin-Rotem ply failure parameters (i.e., fiber tension, fiber compression, matrix tension, and matrix compression) at each integration point through the element thickness. If all of the integration points have experienced failure in one of the criteria, then a crack has formed at the element. Likewise, delamination in the model was determined by examining the cohesive surface damage parameter on the relevant joined surfaces.

Figure 20. Stringer05 Model at peak contact force.

Figure 21. Comparison of Stringer05 experimental data and FE model result.

Figure 22 shows the FE-predicted failure events (plotted in terms of Abaqus'

failure parameters) at the loads levels indicated in Figure 21 (labeled points a to d). Figure 22a shows the cohesive surface damage parameter in an interior surface of a shear tie. While it is difficult to pinpoint when shear tie radial delamination occurred during the Stringer05 dynamic test, it was observed from Stringer02 quasi-

0

10

20

30

40

50

60

70

80

0 10 20 30 40 50 60 70 80

Contact Force (kN

)

Indentation (mm)

Stringer05 Experiment

Stringer05 FE Model

b c

aa. Shear Tie Radial Delaminationb. Formation of Skin Cracksc. Stringer Wall Bucklingd. Skin-Stringer Delamination

d

static testing that shear tie failures occurred at ~35 kN. The Stringer05 FEA model predicted radial delamination to initiate at ~31 kN.

In addition, the model predicted the final failure modes to be stringer local wall buckling, formation of skin surface cracks, and skin-to-stringer delamination. The sequence and indentation depth corresponding to each of these events in the simulation were similar to those found in the experiment. Figures 22b and 22c show the Hashim Rotem fiber damage parameter at the peak load (68.6 kN) on the skin and the stringer, respectively. Figure 22d shows delamination on the skin-to-stringer connecting surface initiating during the load drop at ~65 kN. In the simulation, the skin-to-stringer delamination was found to develop during the load drop, as opposed to occurring at the peak load, as observed in the experiments. Also, there is discrepancy in the predicted degree of skin crack formation. At the end of the loading, only the outermost plies have failed in the model, whereas a complete through-thickness crack was found on the panel skin in the experiment. These discrepancy can be attributed to the fact that the formation of skin cracks, stringer cracks, and delamination are all competing failure modes, reflecting different ways in which impact energy can be absorbed by the panel. The extent of the skin crack is affected by the extent and timing of the delamination failure. Since the delamination damage is delayed in the model, the development of skin cracks is thus also different.

Figure 22. Stringer05 model damage predictions at load levels shown in Figure 21, red and grey colors indicate areas of damage: (a) shear tie radial damage at ~31 kN, (b) skin surface crack at

peak load 68.6 kN, (c) stringer crack at peak load 68.6 kN, and (d) skin-stringer delamination during load drop at ~65kN.

CONCLUSIONS

Quasi-static indentation and dynamic transverse impact experiments were performed on two identical stringer-reinforced composite panel specimens using a D-shaped rubber bumper. The damage observed included shear tie radius delaminations/crushing and skin-to-stringer delamination. Also, stringer and skin cracking (fiber failures) developed for the dynamically-loaded specimen. These failures were excited along the primary load path, namely through the stringers out to the shear ties. The quasi-statically indented specimen initially experienced skin-to-stringer delaminations at the support locations away from the loading zone. Whereas the dynamically impacted specimen first experienced this mode of damage locally, near the loading zone. The time-dependent deformation response of the panel therefore can strongly affect the locations of the failure initiation. For the slowly-loaded quasi-static loading rate, the damage did not leave externally visible indications of damage being present. This suggests that accidental impacts happening under very slow loading rates can produce less detectable damage may be located away from the impacted site. Finite element modeling approaches have been demonstrated which can predict the behavior of such composite structures under dynamic impact. The peak load and indentation of the dynamically impacted panel were predicted within 5% accuracy. Furthermore, the formation of shear tie radius delamination, skin-to-stringer delamination, and fiber failures were predicted using Abaqus’ built-in cohesive surface interaction and the Hashin-Rotem failure criteria. However, once failures have been initiated, the interactions between the failure modes were found to drive the extent of each failure, and caused difficulties in predicting the exact final failure size. ACKNOWLEDGEMENTS Sponsorship of this research was provided by the Federal Aviation Administration Joint Advanced Materials and Structures Center of Excellence (FAA JAMS CoE). Program manager Curt Davies and technical monitor Lynn Pham are gratefully acknowledged for their support of the project. The authors would also like to acknowledge Dr. Larry Ilcewicz of the FAA for his support. REFERENCES 1. Roeseler, W. G., B. Sarh, and M. U. Kismarton, 2007. “Composite Structures: The First 100

Years,” presented at the 16th International Conference on Composite Materials, July 8-13, 2007. 2. Abrate, S., 1991. "Impact On Composite Materials," Appl. Mech. Rev., 44(4): 155-190. 3. International Air Transportation Association, 2005. "Ground Damage Prevention Programme

Targets 10% Cost Reduction, In Industry Times," Annual Report 2005, International Air Transportation Association

4. Wu, E. and K. Shyu, 1993. "Response Of Composite Laminates To Contact Loads And Relationship To Low-Velocity Impact," J. Comp. Mtrls., 27(15): 1443-1464.

5. Liu, D., 1988 "Impact-induced delamination - a view of bending stiffness mismatching," J. Comp. Mtrls., 22: 674-92.

6. Heimbs, S., 2012. "Rubber Impact on 3D Textile Composites," Appl. Comp. Mtrls., 19(3-4):

275-295. 7. Maikuma, H., J.W. Gillespie, Jr., and D.J. Wilkins, 1990. "Mode II Interlaminar Fracture of the

Center Noch Flexural Specimen Under Impact Loading," J. of Comp. Mtrls., 24(2): 124-149. 8. Crosley, P.B. and E.J. Ripling, 1998. "Significance of Crack Arrest Toughness(Kla) Testing," in

Crack Arrest Methodology and Applications, G.T. Hahn, M.F. Kanninen ASTM STP 711: pp. 301-337.

9. Murthy, P.L.N. and C.C. Chamis, 1986. "Dynamic Stress Analysis of Smooth and Noched Fiber Composite Flexural Specimens," in Composite Materials: Testing and Design (Seventh Conference). J.M. Whitney, ASTMSTP 892: pp. 368-391.

10. Jackson, W. and C. Poe, 1992. "The Use of Impact Force as a Scale Parameter for the Impact Re-sponse of Composite Laminates," NASA Technical Memorandum 104189.

11. Lee, S. and P. Zahuta, 1991. "Instrumented Impact and Static Indentation of Composites," J. Comp. Mtrls., 25(2): 204-222.

12. Elber, W., 1983. "Failure Mechanics in Low-Velocity Impacts on Thin Composite Plates," NASA Technical Paper 215.

13. Kwon, Y. and B. Sankar, 1993. "Indentation-Flexure and Low Velocity Impact Damage in Graphite Epoxy Laminates," J. Comp. Tech. Res., 15(2): 101-111.

14. Sjoblom, P., T. Hartness, and T. Cordell, 1988. "On Low-Velocity Impact Testing of Composite Materials," J. of Comp. Mtrls., 22(30): 30-52.

15. Tan, T. and C. Sun, 1985. "Use of Statical Indentation Laws in the Impact Analysis of Laminated Composite Plates," J. Appl. Mech., 52(7): 6-12.

16. Wardle, B. and P.A. Lagace, 1998. "On the Use of Quasi-Static Testing to Assess Impact Damage Resistance of Composite Shell Structures." Mech. Adv. Mater. Struc., 5(1): 103-121.

17. Meyer, P.I., 1988. "Low-Velocity Hard-Object Impact of a Filament-Wound Kevlar/Epoxy Composite," Compos. Sci. Technol., 33(4): 279-293.

18. Kim, B., S.B. Lee, J. Lee, 2012. "A Comparison amoung Neo-Hookean Model, Mooney Rivlin Model, and Ogden Model for Chloroprene Rubber," Int. J. Precis. Eng. Man., 13(5): p. 759-764.

19. Treloar, L.R.G., 1940. "Stress-Strain Data for Vulcanised Rubber under Various Types of Deformation, " Trans. Faraday Soc., 40: 59–70.

20. Daniel, I.M., O. Ishai, 2006. Engineering Mechanics of Composite Materials, 2nd Ed., Oxyford University Press, pp. 377-380.