Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Explicit Finite-Element Analysis of 2024-T3/T351 Aluminum

Material under Impact Loading for Airplane Engine


Containment and Fragment Shielding
Murat Buyuk1; Steve Kan2; and Matti J. Loikkanen3

Abstract: Uncontained aircraft engine failure can cause catastrophic damaging effects to aircraft systems if not addressed in the aircraft
Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

design. Mitigating the damaging effects of uncontained engine failure and improving the numerical modeling capability of these uncon-
tained engine events are crucial. In this paper, high strain rate material behavior of one of the most extensively used materials in the
aircraft industry is simulated and the results are compared against ballistic impact tests. Ballistic limits are evaluated by utilizing explicit
finite-element 共FE兲 simulations based on the corresponding ballistic impact experiments performed at different material thicknesses.
LS-DYNA is used as a nonlinear explicit dynamics FE code for the simulations. A Johnson–Cook material model with different sets of
parameters is employed as a thermo-viscoplastic material model coupled with a nonlinear equation of state and an accumulated damage
evaluation algorithm for the numerical simulations. Predictive performance of the numerical models is discussed in terms of material
characterization efforts, material model parameters, mesh sensitivities, and effects of stress triaxiality. It is shown that mesh refinement
does not necessarily provide better results for ballistic limit simulations without considering and calibrating these interrelated factors.
Moreover, it is shown that current models that can only fit a specific function for damage evaluation as a function of stress triaxiality are
not always successful in predicting failure, especially if the state of stress changes significantly.
DOI: 10.1061/共ASCE兲0893-1321共2009兲22:3共287兲
CE Database subject headings: Nonlinear analysis; Impact; Projectiles; Penetration; Strain rates; Material properties; Finite element
method; Aircraft; Engines.

Introduction coordinated a research program associating industry, government


agencies, national research laboratories and universities to con-
Reliable high-speed impact computations are still a challenge. duct research to develop technologies to assist in the design of
Advanced computer codes are available and many material mod- fragment barriers to mitigate the impact from jet engine debris in
els exist to depict the behavior of metals under high-speed impact. the case of rotor failure. This research was focused on improving
However, the material data are often not publicly available or the understanding of metal fragments impacting and penetrating
airframe parts such as wings and fuselage, which are in close
may not comprise the needed high strain-rate behavior regime
proximity to the engines 共Johnson and Cook 1983; Lesuer 2000;
that is required for these ballistic impact studies. Also, critical
Kay 2003; Kay et al. 2007; Kelly and Johnson 2006; Gogolowski
details of the numerical modeling methodologies are usually not
and Morgan 2002兲.
very well defined. Consequently, there is a need for a reliable
This effort consisted of testing deformation and failure behav-
source that can offer objective finite-element 共FE兲 modeling
ior of Ti-6Al-4V titanium and 2024-T3/T351 aluminum for high
guidelines with generic material model parameters that can be strain rate material properties, ballistic impact testing, develop-
used with high-speed impact simulations. Having seen this need, ment of FE modeling techniques, and simulation of the test re-
high strain rate material characterization efforts has been granted sults. Material characterization data were obtained at high strain
by the Federal Aviation Administration’s 共FAA兲 Aircraft Cata- rates and large strains using the split Hopkinson pressure bar
strophic Failure Prevention Program as part of the research of 共SHPB兲 method. A new set of material constants for a Johnson–
uncontained engine failure events. FAA has commissioned and Cook 共J-C兲 material model for Ti-6Al-4V and 2024-T3/T351 was
found and compared to the existing data from the literature to
1
FHWA/NHTSA National Crash Analysis Center, George Washington evaluate the ability of the J-C material model in representing the
Univ., Washington, DC 共corresponding author兲. E-mail: mbuyuk@ncac. deformation and failure response under conditions relevant to
gwu.edu simulations of engine containment and the influence of uncon-
2
FHWA/NHTSA National Crash Analysis Center, George Washington tained engine debris on aircraft structures 共Johnson and Cook
Univ., Washington, DC. E-mail: cdkan@ncac.gwu.edu 1983; Lesuer 2000兲. In order to get more representative data for
3
The Boeing Company, Commercial Airplanes Division, P.O. Box aircraft skin, a series of ballistic impact tests were conducted,
3707, Seattle, WA 98124-2207. E-mail: matti.j.loikkanen@boeing.com
which were used to further refine the J-C parameters 共Gogolowski
Note. This manuscript was submitted on August 8, 2008; approved on
January 22, 2009; published online on June 15, 2009. Discussion period and Morgan 2002兲. Using these test data, a new set of J-C param-
open until December 1, 2009; separate discussions must be submitted for eters for Ti-6Al-4V titanium and 2024-T3/T351 aluminum was
individual papers. This paper is part of the Journal of Aerospace Engi- developed by Lawrence Livermore National Laboratory 共LLNL兲
neering, Vol. 22, No. 3, July 1, 2009. ©ASCE, ISSN 0893-1321/2009/3- 共Kay 2003兲. In a recent study, the J-C failure parameters for 2024
287–295/$25.00. T3/T351 aluminum were recalibrated in order to attain consis-

JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009 / 287

J. Aerosp. Eng. 2009.22:287-295.


Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 1. Ballistic limit tests for 2024-T3/T351 at different target material thicknesses

tency between simulations and the available ballistic limit mea- 254 mm free target area. The projectile was a 12.7 mm diameter
surements by LLNL 共Kay et al. 2007兲. Ballistic impact test data chrome 52100 alloy steel sphere. Ductile materials, like alumi-
used for this recalibration were conducted the University of Cali- num, have a sharp rise in the exit velocity plot just after the
fornia at Berkeley 共UCB兲 共Kelly and Johnson 2006兲 and consisted ballistic limit 共Zukas 1990兲. As it can be observed in Fig. 1, after
of thicker aluminum targets than previously tested. It was shown the ballistic limit, the residual velocity rises sharply and later
that the aluminum plates failed by dishing and petaling with slight follows a 45° slope, which is a characteristic property for many
plugging for the thinnest targets. As the thickness increased, the ductile materials. From these test results with the spherical pro-
amount of dishing and petaling decreased and the failure mode jectile, the ballistic limit for 1.5875 mm plate is found about
tended toward plugging. Therefore, the ballistic tests confirmed 122 m / s, for 3.175 mm it is about 213 m / s and for 6.35 mm it is
the transition of the failure mode with the change in thickness. In about 411.5 m / s.
a successive study by Boeing and George Washington University,
UCB tests were also modeled and simulated by using three-
dimensional 共3D兲 FE models while three sets of material model Numerical Modeling
parameters were evaluated according to their predictive perfor-
mance for the ballistic limits 共Loikkanen et al. 2005兲. It is shown All numerical models in this study are developed to be used with
that the different sets of parameters have different levels of suc- LS-DYNA as a nonlinear explicit dynamics FE numerical solver.
cess in predicting the ballistic limit depending on the change in
the thickness of the target material.
In this paper, practical FE modeling guidelines are developed Finite-Element Models
and presented for the aviation community to predict ballistic im- Three different quarter symmetric FE models are developed to
pact performance of aeronautical structures with reasonable accu- investigate their predictive performance. The mesh pattern and
racy and computational cost. Since the predictive capability of the size dependency are examined to find an optimized mesh for sta-
numerical models is directly related to the competence of the bility, accuracy, and efficiency of the impact analysis. These mod-
employed material model, the performance of one of the very els are cost effective 3D models, where a nearly perfect hit with
widely used material models is investigated with the available nonoblique projectile orientation can be modeled successfully.
experimental data. Since the target and the projectile are symmetric about both the
horizontal and vertical axes, only one quarter models of the test
target and the projectile are developed. Both the target and pro-
Ballistic Impact Test Results jectile are modeled with eight-node underintegrated hexagonal
solid elements. The FE model took advantage of the double sym-
The ballistic tests were all performed in the UCB Materials Labo- metry through the center. One quarter of the target and the pro-
ratory and the results are summarized by Fig. 1. The tests with jectile are modeled in the positive x-y quadrant with the center at
initial velocities below 305 m / s were performed with a pneu- the origin. Fully fixed boundary conditions are imposed on the
matic gun and a powder gun was used for higher impact veloci- outer boundary and the symmetry boundary conditions are im-
ties. The target plate thicknesses were 1.5875, 3.175, and posed at the other free surfaces. The mesh just under the projectile
6.354 mm plates were cut to 304.8 by 304.8 mm size and at- is chosen to be circular to assure that the failure occurs along the
tached to the 25.4 mm wide support frame, thus leaving a 254 by element boundaries. The mesh density under the impact zone is

288 / JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009

J. Aerosp. Eng. 2009.22:287-295.


Table 1. Details of FE Models
Total number of elements Number of elements
at the target through thickness Target mesh density In-plane
Mesh Symmetry through thickness mesh
type assumption 1.5875 mm 3.175 mm 6.35 mm 1.5875 mm 3.175 mm 6.35 mm 共mm/element兲 density
I 3D-1 / 4 1,674 3,348 6,696 2 4 8 0.79375 837
II 3D-1 / 4 3,348 6,696 13,392 4 8 16 0.396875 837
III 3D-1 / 4 6,975 13,950 27,900 3 6 12 0.529167 2,325

also finer than the surrounding. The in-plane mesh pattern is kept jectile. A more sophisticated material model is necessary for the
the same for all models while the element size through the thick- targets to simulate ballistic impact response, where a J-C material
Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

ness is varied. Each target thickness is modeled with three mesh model is employed.
densities while many more are experimented on during the cali-
bration runs. Simply refining the mesh does not necessarily im- Johnson–Cook Material Model
prove the accuracy since the material model parameters are The Johnson–Cook material model is used for the aluminum tar-
calibrated for a specific mesh size and mode of failure changes gets. J-C is a strain-rate and temperature-dependent 共adiabatic as-
during the simulation. Furthermore, there are no clear theoretical sumption兲 viscoplastic model. It is employed to describe the
guidelines on the required mesh density for the range of impacts response of 2024-T3/T351 aluminum. The J-C model represents
covered in this paper. Considering the softening effects and mesh the flow stress with an equation of the form 共Zukas 1990;
dependent failure algorithms, the only reasonable methodology to Hallquist 2003兲
find the most appropriate mesh size is trial and error while com-
paring the results against a controlled test data. Subsequently, it is ␴Y = 共A + B␧n兲共1 + C ln ␧˙ *兲共1 − T*m兲 共1兲
possible to arrive at an optimum after experimenting with differ- .
ent meshing and draw some guidelines for that particular case. where ␴Y = effective stress; ␧ = effective plastic strain; ␧*
For mesh patterns; Mesh-I and Mesh-II, the 1.5875 mm targets = normalized effective plastic strain rate 共typically normalized to
are modeled with two and four elements through the thickness a strain rate of 1.0 s−1兲; n = work hardening exponent; and A, B, C,
respectively, where 3.175 mm plate is modeled with four and and m = material constants, where their physical meanings are de-
eight elements through the thickness and the 6.35 mm plate is scribed in Table 2. The quantity T* is defined as
modeled with eight and 16 elements through the thickness cor-
respondingly. For the mesh pattern; Mesh-III, the concentric T − Troom
T* = 共2兲
circular layers of the target, are divided into three regions in a Tmelt − Troom
radial direction and the mesh density is gradually coarsening from
where Troom = room temperature; and Tmelt = melting temperature
the inner region, which is the potential impact region, to the outer
and is typically taken as the solidus temperature for an alloy. For
region. Mesh transition between regions is good enough to
high rate deformation problems, we can assume that an arbitrary
prevent stress wave reflections from the boundary of regions. The
percentage of the plastic work done during deformation produces
number of through thickness elements is chosen to be three for
heat in the deforming material. For many materials, 90–100% of
the 1.5875 mm in target as a baseline since implementation of
the plastic work is dissipated as heat in the material. Thus, the
reduced integration solid elements require at least three elements
temperature used in Eq. 共1兲 can be derived from the increase in
through the thickness to be able to capture accurate bending
deformation modes. A comparison of the size of the FE models
is summarized in Table 1. Fig. 2 illustrates closeups of the
1.5875 mm targets with the corresponding mesh patterns of
Mesh-I, Mesh-II, and Mesh-III, respectively, with a perspective
view.
A viscous based stabilization method is used during the simu-
lations to prevent hourglass modes of the reduced integration el-
ements. The amount of artificial energy is kept below 10% of the
kinetic energy throughout the simulation.
Contact behavior between the projectile and the target is at- Mesh-I Mesh-II
tained by using a penalty based single surface type contact algo-
rithm that uses a nodal constraint formulation with an element
erosion scheme.

Material Models
The steel projectile is modeled by using the linear elastic material
model. The model does not allow plastic yielding but can have
large deformations. Mesh-III Perspective View of the Mesh Patterns

Plasticity is unnecessary for this component, because neither


the tests nor early runs showed any signs of yielding in the pro- Fig. 2. Mesh patterns for 1.5875 mm target

JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009 / 289

J. Aerosp. Eng. 2009.22:287-295.


Table 2. J-C Material Model Parameters
Literature LLNL-1 LLNL-2 LLNL-3
Parameter Notation 共Johnson and Cook 1983兲 共Lesuer 2000兲 共Kay 2003兲 共Kay et al. 2007兲
共a兲 Strength parameters
Density 关kg/ m 兴
3
␳ 2,770 2,770 2,770 2,770
Poisson’s ratio ␯ 0.33 0.33 0.33 0.33
Modulus of elasticity 关MPa兴 E 73,084 73,084 73,084 73,084
Static yield limit 关MPa兴 A 265 369 369 369
Strain hardening modulus 关MPa兴 B 426 684 684 684
Strain hardening exponent n 0.34 0.73 0.73 0.73
Strain rate coefficient C 0.015 0.0083 0.0083 0.0083
Thermal softening exponent m 1.0 1.7 1.7 1.7
Reference temperature 关°K兴 Troom 294 294 294 294
Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

Melting temperature 关°K兴 Tmelt 775 775 775 775


Specific heat 关J / kg- ° K兴 cP 875 875 875 875
共b兲 Damage parameters
D1 0.13 0.13 0.112 0.31
D2 0.13 0.13 0.123 0.045
D3 −1.5 −1.5 −1.5 −1.7
D4 0.011 0.011 0.007 0.005
D5 0 0 0 0
共c兲 Mie–Grunesien EOS parameters
S1 1.338 1.338 1.338 1.338
S2 0 0 0 0
S3 0 0 0 0
␥0 2 2 2 2

temperature according to the following expression: is compressed or expanded. The Mie–Gruneisen EOS with cubic


shock velocity-particle velocity defines pressure for compressed
␣ materials as
⌬T = ␴共␧兲d␧ 共3兲

冋 冉 冊 册
␳c
␥0 a
where ⌬T = temperature increase; ␣ = percentage of plastic work ␳0Csp
2
␮ 1+ 1− ␮ − ␮2
2 2

冋 册
transformed to heat; c = heat capacity; and ␳ = density. In this p= + 共␥0 + a␮兲Eint
study, the entire plastic work is converted into heat for the mate- ␮2 ␮3
1 − 共S1 − 1兲␮ − S2 − S3
rial of consideration. ␮+1 共␮ + 1兲2
Damage in the J-C material model is based on a cumulative 共6兲
damage law
and for expanded materials as
⌬␧
D= 兺 ␧f
共4兲 p = ␳0Csp
2
␮ + 共␥0 + a␮兲Eint 共7兲

in which effective plastic strain at failure is defined as where Eint = internal energy; Csp = intercept of the vs-vp curve;
S1 – S3 = coefficients of the slope of the vs-vp curve; ␥0
␧ f = 关D1 + D2 exp共D3␴*兲兴关1 + D4 ln ␧˙ *兴关1 + D5T*兴 共5兲 = Gruneisen gamma; a = first-order volume correction to ␥0; and ␮
where ⌬␧ = increment of effective plastic strain during an incre- is given as
ment in loading; and ␴* = mean stress normalized by the effective ␳
stress, which is often referred to as triaxiality. The parameters D1, ␮= −1
␳0
D2, D3, D4, and D5 = fracture constants. Failure of elements is
assumed to occur when D = 1. The failure strain ␧ f and thus the
accumulation of damage is a function of stress triaxiality, strain Selection of Set of Material Model Parameters
rate, and temperature. Failed elements are then removed from the Four different sets of material model parameters that are listed in
FE model with an element erosion algorithm. The J-C material Table 2 are found in the literature 共Johnson and Cook 1983兲 and
model is used in conjunction with Mie–Gruneisen equation of LLNL material characterization tests 共Lesuer 2000; Kay 2003;
state 共EOS兲 model. Four different sets of material model param- Kay et al. 2007兲. The flow surfaces are illustrated in Fig. 3 for
eters are listed for 2024-T3/T351 in Table 2. these sets of parameters. The flow surface for LLNL-1, LLNL-2,
and LLNL-3 use the same strength constants. The damage model
Mie–Gruneisen Equation of State constants are also listed in Table 2 and the differences in fracture
The Mie–Gruneisen EOS model defines the pressure-volume re- locus are illustrated in Fig. 4 with respect to stress-triaxiality at a
lationship in one of two ways, depending on whether the material nominal strain rate of 1 / s. LLNL-2 and LLNL-3 differ from the

290 / JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009

J. Aerosp. Eng. 2009.22:287-295.


Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 3. Differences in flow surface data for 2024-T3/T351

rest by considering a recalibration of the parameters with the of damage constants do not work for the whole range of material
ballistic impact test results from LLNL and UCB. Since it is thickness 共Kay et al. 2007; Kelly and Johnson 2006; Gogolowski
reported that the type of failure mode is subject to change from and Morgan 2002兲.
petaling to plugging due to the change in thickness, the same set Optimization and calibration are necessary to find the best set
of parameters that can work reasonably well for the other thick-
nesses. The predictive performance of these parameters is given
in the proceeding sections.

Ballistic Impact Simulation Results

Ballistic impact response for three different material thicknesses


is examined with different meshing options and by employing
different material model parameters. The responses are illustrated
by generating initial velocity versus residual velocity graphs,
where they can be compared to the UCB ballistic test data. The
“Reference” line that is overlaid in each graph represents the
condition if impact and exit velocities are the same, which indi-
cates no loss in the energy. Therefore, the distance between the
reference line and the data line can be accepted as the velocity or
energy lost during the penetration.
Fig. 4. Differences in fracture locus data for 2024-T3/T351

(a) (b) (c)

Fig. 5. Penetration through 1.5875 mm target and petal formation: 共a兲 simulation; 共b兲 test-front face; and 共c兲 test-rear face

JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009 / 291

J. Aerosp. Eng. 2009.22:287-295.


Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 6. Ballistic limit predictions for 1.5875 mm targets

Target Thickness 1.5875/ mm III is found to provide the best results and Mesh-I is found to be
the least successful overall. This can be explained by the fact that
The penetration through a 1.5875 mm aluminum target is illus-
using two reduced integration elements through the thickness is
trated in Fig. 5, and the comparison between the measurements
not enough in predicting bending behavior. Also, Mesh-III has
and computations is shown in Fig. 6.
more in-plane elements that can provide better representation of
A petaling type of deformation mode is observed. There is
the petaling than Mesh-II.
significant local bending in the target and material fails mostly in
tension. It should be noted that the plug is pushed out in front of
the sphere, and there are three radial cracks in the target. The Target Thickness 3.175 mm
failure mode, the plug formation, and the radial cracks all corre- The penetration through 3.175 mm aluminum target is shown in
spond well to the test results. Fig. 7, and the comparison between the measurements and com-
It can be observed that the LLNL-2 material data gave excel- putations is shown in Fig. 8.
lent correlation, where the rest of the material model parameters The projectile penetrates through the target in mixed mode:
with different flow surface and failure parameters did not agree as partly petaling and partly plugging. There is some local bending
well with the test results for either mesh patterns. This deviation and some local shearing in the target; hence, the material fails
from the test results while using the LLNL-3 material model pa- partially in shear and partially in tension. Note that a plug is
rameters can be explained by the fact that these parameters were pushed out in front of the sphere and the size of the radial cracks
adjusted for thicker targets and may not necessarily be appropri- in the target gets smaller. The failure mode and plug formation
ate for thinner samples. The differences in the flow surface can correspond with the test results. The LLNL-2 parameter simula-
simply be explained by the fact that all LLNL data are developed tions show a relatively poor correlation with the test for all three
by using high strain rate material characterization. Therefore mesh patterns. However, the LLNL-3 data, with failure param-
LLNL flow surface is expected to be more accurate in simulating eters for thick plates, agree better with the test results. It can be
high-speed impact simulations. seen that the employed material model parameters provide dras-
Even though the effects of material model parameters are tically different results: where LLNL-2 parameters underpredict
found to be more significant than the effects of mesh size, Mesh- the ballistic limit while LLNL-3 overpredicts.

PLUG

(a) (b) (c)

Fig. 7. Penetration through 3.175 mm target and mixed-mode failure formation: 共a兲 simulation; 共b兲 test-front face; and 共c兲 test-rear face

292 / JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009

J. Aerosp. Eng. 2009.22:287-295.


Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 8. Ballistic limit predictions for 3.175 mm targets

Target Thickness 6.35 mm was tuned for the thicker targets. However, as the material thick-
ness increases, the accuracy when compared to experimental re-
The penetration through 6.35 mm aluminum target is shown in
Fig. 9 and a comparison between the measurements and compu- sults is no longer as good. Fig. 11 illustrates the reason why a
tations is shown in Fig. 10. single set of material model parameters is not able to keep the
The projectile penetrates through the target in plugging mode. same predictive capability during the simulation of various thick-
The material fails entirely in shear. Note that a plug is pushed out nesses. As the thickness changes 共as with the increase in ballistic
in front of the sphere, and there are no radial cracks in the target. limit兲, the type of deformation and failure changes drastically
The failure mode and the plug formation correspond to the test from petaling to plugging. Bending and necking types of defor-
results. The LLNL-2 material data give a poor match with the mation patterns change into shearing and spalling types of defor-
test, but the LLNL-3 data are much better since they were spe- mation. The failure fully depends on multiaxial state of stress, and
cifically calibrated for thicker targets. There is not a significant therefore no single set of material model parameters is capable of
mesh dependency observed, which can be explained by the fact covering the whole fracture locus as illustrated in Fig. 4. Fig. 11
that all mesh sizes were small enough to predict a shear domi- shows the deviation in the failure strain as a function of stress-
nated impact zone. triaxiality for three representative finite elements at the failure
zone that have eroded and demonstrates the lack of predicting
failure strain accurately while using a damage model that employs
Ballistic Limit Summaries and Discussion a predetermined failure locus based on curve fitting. It is shown
that the J-C damage model, which uses an exponential function to
Evaluation tasks accomplished within the framework of this effort construct the failure locus based on stress-triaxiality, is not always
included the assessment of four different sets of material model successful in predicting accurate strain at failure.
parameters while predicting the results of the ballistic impact Material model parameters are usually calibrated for a particu-
tests. The predictive performance of the simulations for lar material gauge length and are not adjusted or tuned dynami-
1.5875 mm material is found promising for all parameters except cally according to the changes and transitions in deformation and
for the LLNL-3 set, which was expected since this particular set failure. Therefore, they are expected to have some compromise in

PLUG

(a) (b) (c)

Fig. 9. Penetration through 6.35 mm target and plug formation: 共a兲 simulation; 共b兲 test-front face; and 共c兲 test-rear face

JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009 / 293

J. Aerosp. Eng. 2009.22:287-295.


Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

Fig. 10. Ballistic limit predictions for 6.35 mm targets

the predictive performance if the transition happens. It is also with four different sets of parameters. It is found that a single set
shown that the in-plane or through thickness mesh sensitivities of material model parameters may not necessarily predict the bal-
also depend on the material model parameters and therefore a listic impact performance of a material accurately if there is a
regularized material model is needed, where failure should be significant change and transition occurs in the deformation mode.
mesh size independent. It is observed from the simulation results that the effect of the
change in thickness from thinner to thicker targets causes the
change in the type of deformation and failure from petaling to
Conclusions plugging, respectively.
It is found that the set of material model parameters that are
The ballistic impact tests are simulated by using LS-DYNA as an tuned for a specific element size gives better accuracy. Otherwise
explicit dynamics FE code incorporating a J-C material model increasing the element density does not improve the performance

1.5875 mm 3.175 mm 6.35 mm

Petaling Plugging
Bending‐Necking Shearing‐Spalling
Fig. 11. Transition of failure mode

294 / JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009

J. Aerosp. Eng. 2009.22:287-295.


significantly. Consequently, it is necessary to recalibrate the J-C Johnson, G. R., and Cook, W. H. 共1983兲. “A constitutive model and data
damage parameters in order to attain a better consistency between for metals subject to large strains, high strain rates and high tempera-
simulations and the available ballistic limit measurements. How- tures.” Proc. 7th Int. Symp. on Ballistics, The Hague, The Nether-
ever, transition of the failure mode makes it impossible to capture lands, 541–547.
the change in deformation and failure by using a single set of Kay, G. 共2003兲. “Failure modeling of titanium 6Al-4V and aluminum
parameters. The current material law does not allow adjusting the 2024-T3 with the Johnson–Cook material model.” FAA Rep. No.
material model parameters dynamically. DOT/FAA/AR-03/57, Federal Aviation Administration, Washington,
D.C.
Kay, G., Goto, D., and Couch, R. 共2007兲. “Statistical testing of alumi-
Acknowledgments num, titanium, lexan and composites for transport airplane rotor burst
fragment shielding.” FAA Rep. No. DOT/FAA/AR-07/26, Federal Avia-
The writers would like to acknowledge Mr. William Emmerling, tion Administration, Washington, D.C.
Mr. Donald Altobelli, and Dr. Chip Queitzsch of the Federal Avia- Kelly, S., and Johnson, G. 共2006兲. “Statistical testing of aircraft materials
tion Administration 共FAA兲 for their technical and financial sup- for transport airplane rotor burst fragment shielding.” FAA Rep. No.
Downloaded from ascelibrary.org by Selcuk Universitesi on 01/21/15. Copyright ASCE. For personal use only; all rights reserved.

port. DOT/FAA/AR-06/9, Federal Aviation Administration, Washington,


D.C.
Lesuer, D. R. 共2000兲. “Experimental investigations of material models for
References Ti-6A1-4V titanium and 2024-T3 aluminum.” FAA Rep. No. DOT/
FAA/AR-00/25, Federal Aviation Administration, Washington, D.C.
Gogolowski, R. P., and Morgan, B. R. 共2002兲. “Ballistic experiments with Loikkanen, M. J., Buyuk, M., Kan, C. D., and Meng, N. 共2005兲. “A
titanium and aluminum targets.” FAA Rep. No. DOT/FAA/AR-01/21, computational and experimental analysis of ballistic impact to sheet
Federal Aviation Administration, Washington, D.C. metal aircraft structures.” Proc., 5th European LS-DYNA Users’ Conf.,
Hallquist, J. O. 共2003兲. LS-DYNA keyword user’s manuel, Livermore Livermore Software Technology Corp., Birmingham, U.K.
Software Technology Corporation, Livermore, Calif. Zukas, J. A. 共1990兲. High velocity impact dynamics, Wiley, New York.

JOURNAL OF AEROSPACE ENGINEERING © ASCE / JULY 2009 / 295

J. Aerosp. Eng. 2009.22:287-295.

You might also like