-
DEVELOPMENTS IN
LAMINATED
N93-30857IMPACT DAMAGE MODELING VO'l_
COMPOSITE STRUCTURES 1
Ernest F. Dost, William B. Avery, and Gary D. Swanson /The
Boeing Company, Seattle, Washington
and
Kuen Y. Lin
University of Washington, Seattle, Washington
Introduction
Damage tolerance is the most critical technical issue for
composite fuselage structures studied inATCAS. The ATCAS program
goals in damage tolerance include the characterization of
impactdamage, models for impact damage simulation, and
understanding the behavior of notches anddelaminations.
The characterization of potential impact damage states in
fuselage is being accomplished through test.Configured structure
will be impacted in different locations with a number of different
impactorvariables. Thc damage states will be assessed both
nondestructively and destructively.
An approach for predicting the post-impact compressive behavior
of laminated composites has beendeveloped at Boeing over the past
several years. Dr. K.Y. Lin and Dr. Z.Q. Chen at the University
ofWashington will be enhancing and generalizing this approach to
account for the different potentialdamage states and failure modes
found in the test program described above.
Tension damage tolerance is currently being addressed through a
test program and analysisdcvclopmcnt by Dr. F.K. Chang at Stanford
University. Future work with Dr. P.A. Lagace and Dr.M.J. Graves at
Massachusetts Institute of Technology will address dynamic fracture
including pressureeffects.
Objectives
The objective of the work being presented is to understand both
the impact damage resistance andresidual strength of laminated
composite fuselage structure. An understanding of the different
damagemechanisms which occur during an impact event will (a)
support the selection of materials andstructural configurations
used in different fuselage quadrants and (b) guide the development
ofanalysis tools for predicting the residual strength of impacted
laminates. Prediction of the damagestate along with a knowledge of
post-impact response to applied loads will allow for
"engineered"stacking sequences and structural configurations;
intelligent decisions on repair requirements will alsoresult.
] This work is being funded by Contract NAS1-18889, under the
direction of J.G. Davis and W.T.Freeman of NASA Langley Research
Center.
721
-
Potential Impact Damage States
A schematic diagram classifying characteristic damage states
(CDS) that have been observed in flatlaminates following
low-velocity impact by spherical objects is shown. Planar and
cross-sectionalviews of CDS are given in the figure. Three classes
of CDS consisting of symmetric damage throughthe laminate cross
section are shown in this figure. Damage size and type (fiber,
matrix, or combined)depend on variables such as delamination
resistance and impact energy. The most common damageobserved in
experiments with a stacking sequence used for material screening
tests (i.e.,[45,0,-45,90,]nS) was matrix damage [1, 2].Plate
boundary conditions, laminate thickness, and material form are
among the variables which maysuppress delamination, causing damage
dominated by fiber failure. Fiber damage, when present, tendsto
concentrate at the impact site. Matrix damage is also centered at
the impact site, but tends to radiateaway from this point to a size
dependent on delamination resistance. The most general
classificationof symmetric damage involves both fiber and matrix
failure.
Many factors can affect the CDS symmetry. Test observations have
indicated thin laminates andheterogeneous stacking sequences tend
to have unsymmetric CDS with damage concentrated oppositethe
impacted surface. Very thick laminates are also expected to have
unsymmetric damage, but withdamage concentrating closer to the
impacted surface. Work by the current authors has indicated
thatdelamination resistant materials have a stronger tendency for
unsymmetric CDS than brittle materialstested with the same impact
variables [3].
Impact Event
Impact Damage
.P_lamat_Yie,w_
Potential Impact Damage States
= f ( Material, Laminate, Structural, and Extrinsic
Variables)
@ Fiber Damage Matrix Damage Fiber & Matrix Damage
Cross-sectional View
SymmetricDamage
UnsymmetricDamage
Ill I ===
722
-
Damage Modeling Applications
The panel shown below is a carbon fiber reinforced plastic
(CFRP) wing gauge panel impacted on thestiffener cap. The impact
damage located on this stiffener cap was nonvisible. An identical
panel hadbeen impacted on the stiffener attachment flange edge.
Both panels had significant reductions instrength from their
undamaged strength. Analytical tools developed to predict the
post-impactresponse of CFRP structure must have enough generality
to account for different failure modes whichoccur during impact.
The approaches presented take into account both stress
redistribution andchanges in panel postbuckling response due to
sublaminate buckling [ 1-3]. The prediction ofpost-impact response
due to local fiber failures was presented by Cairns [4].
Impact Damage Discrete Modeling
Wing Structure with Integral StiffenersImpacted on Free
Flange
.._ L-7
////
_LIII
'///,r/rji/j!!! _
Jtli'jcl#[[/3.. I II ," i I I I ! ;
'/L/1/i,'////rliil[llllll// I [F// t[ l r/1// f/l/I/ il
/Ill
'///,1#1t'lll
/fill
'/.ill I '
t..lll'/II-IllJ
IIIf'111
11111111111"//I////IIIlillllllll --
llllllllj/ --'ll/ll/ , l l --,/,,///M.--'I,'III/tl#_t # Ii117
--r1711111/ if# /illI /v t i11_
Global/Local Nonlinear Post-Buckling Analysis
723
-
Experimental Data ShowingPost-impact CompressionPerformance as a
Function or Laminate Stacking Sequence
Compression after impact (CAI) data axe shown as a function of
the drop weight impact energy (i.e.,drop height x drop weight).
Data scatter for each type of laminate layup was small compared to
thetotal range of results. This indicated that laminate stacking
sequence is critical to CAI. Note that allother material, laminate,
structural, and extrinsic variables were held constant for the
tests. The oneexception was for the [452,902,-452,0212S laminate
which had 32 plies instead of 24 plies. Despite theadditional
thickness, this laminate did not have the highest CAI for a given
impact energy, againindicating the importance of stacking sequence.
These data illustrate that models to predict residualstrength of
impacted laminates must include stacking sequence dependent CDS
details [3].
Experimental Data Showing Post-Impact CompressionPerformance as
a Function of Laminate Stacking Sequence
...... t_ 638 MPa50O
400_-_:_----- --_
_ + .... :...Q_
u) 300 -LUrrt--03UJrr
x
_< 200 LL
+
V
100O
A
V VX X v
z_ xv
+
&
Legend .[
[45,90,-45,0] 3S[452 ,902 ,-4F;2 ,02] 2S[453'903"453 '03] S
[3o,so,se,-so,-3o,o]2s
[3o,6o,9o,-3o,-6o,o] 2s[45,(90,-45)3 ,(0,45)2,0] S
[45,(0,-45)3,(_,45)2,90]Sl
Material: IM7/8551-7
Ply Thickness = 0.188 mm
Specimen Width = 10.2 cm
xv
A
v v
v
o
AA d A
+ L_
+ *
V
Note
I ....... Marks Undamaged Specimen Stability Limit forStacking
Sequence With Associated SymbolI
22.6
(200)Drop Weight Impact Energy, J (in Ib)
!
45.2
(400)
724
-
SUBLAMINATE STABILITY/REDUCED STIFFNESS CAIMODELING
The CDS for a set of impact variables used in material screening
tests was described in earlier work[ 1,2,5]. These tests use a
[45,0,-45,90]nS laminate stacking sequence. As discussed earlier,
any fiberdamage caused by impact tends to concentrate at the core
of the CDS. A network of matrix cracks anddelaminations comprise
the remainder of the CDS. Delaminations at each ply interface are
connectedto those at neighboring ply interfaces by transverse
matrix cracks. In a planar view, double-lobeddelaminations formed
at each interface. These delaminations are wedge shaped due to the
re/4difference in orientation of neighboring plies, breaking the
CDS into octants. The ply orientationangles increase in n/4
increments from the impacted surface to the center. The stacking
sequence andCDS is reflected at the center. Ply orientations
decrease by n/4 with each ply from the center to theback side. This
pattern causes a CDS with interconnected delaminations spiraling
toward the center,reversing direction, and proceeding out toward
the back side.
The CDS described above splits the laminate into separate
sublaminates. These sublaminates areconnected in a fashion similar
to a spiral staircase, but are conceptualized as circular disks to
simplifythe analysis. The sublaminates near the outer surfaces vary
in thickness from 2 to 5 plies. The next setof sublaminates are 4
plies in thickness with stacking sequence varying stepwise around
the damage.This type of sublaminate can repeat several times,
depending on the number of plies in the stackingsequence. Damage
that occurs approaching both sides of the laminate midplane results
in twodiscontinuous sublaminates and a symmetric core sublaminate
that varies in thickness from 2 to 8plies. The total number of
sublaminates for a [45,0,-45,90]n s laminate stacking sequence is
(2n+1).This can be generalized for other repeating stacking
sequences which increment by either decreasingor increasing ply
angles if a sum of the difference between adjacent angles in the
repeat element equalszero (i.e., [ct, lg,t_..... 0]n S where
{13-o_}+ {t)-_ }+ ... + {ix-0 } = 0.0). Absolute values of each
differenceshould also not exceed 90 .
The analysis method used for comparison with experiments is
documented in [1,2]. In summary, fivebasic steps are followed in
applying the method. First, the CDS is identified and simulated as
asublaminate with ply stacking sequence and thickness representing
an average of those appearing inthe real CDS. Second, a sublaminate
stability analysis is performed using damage diameter as
anindependent variable characterizing the planar size of the CDS.
This is done using a modification tothe buckling analysis method
described in [6]. The modification accounts for sublaminates
withunsymmetric ply stacking sequences [ 1,2]. Third, effective
reduced stiffness of the impact damagezone is calculated using
results from sublaminate stability analysis. Fourth, the inplane
stressconcentration associated with the reduced stiffness is
determined. Finite elements are used for thisstep in order to
account for specimen width/damage size interactions. Finally, a
maximum strainfailure criteria is applied to predict CA1.
Note that steps 2 through 4 of the sublaminate stability
analysis method should be modified for themost general CDS in which
sublaminate parameters (e.g., diameter, thickness and stacking
sequence)vary significantly through the laminate thickness. The
more general model is currently beingdeveloped.
725
-
Experimental Determination of Sublaminate Bucklingand Strain
Distribution of Impacted Laminates
Sublaminate stability and subsequent load redistribution of
compressively loaded impact damagedcoupons are being examined
experimentally. Moire interferometry was employed to measure
bothin-plane and out-of-plane displacements of impacted coupons as
a function of load. A micro-Moiregrid (600 lines/millimeter) used
to measure inplane displacements was applied to the tool side
whilethe other side used shadow Moire (60 lines/cm) to measure
out-of-plane displacements. A typicalMoire fringe pattern
displaying out-of-plane displacements for a [45,0,-45,90] 3S
specimen with adamage diameter of 1.28" is shown. The in-plane
u-displacements are shown next to it. Byexamination of the in-plane
displacement contours, one can discem that an inplane strain
concentrationoccurs near the damage area.
Out-of-Plane Displacement Contours In-Plane Displacement
Contours
726
-
E Sublaminate Stability/Reduced Stiffness andxperimental Results
for AS6/3501-6, (45,0,-45,90)ss
This figure shows good comparisons between predictions and
experimental results using agraphite/epoxy material (AS6/3501-6).
The undamaged compressive strength was measured as 501MPa (72.7
Ksi). Damage was created by both static indentation and drop weight
impact. Finitespecimen width becomes important as damage diameters
increase. As shown, the model accuratelypredicted CAI throughout
the range studied. The CAI lower limit for infinitely wide coupons
wouldcorrespond to the maximum stress concentration of three for a
quasi-isotropic laminate (i.e., 167 MPa).
Sublaminate StabilityReduced Stiffness Predictionsand
Experimental Results for AS6/3501-6, (45,0,-45,90) 5s
vv
_Jn