Beruflich Dokumente
Kultur Dokumente
December 1995
Available electronically at the following URL address: http://techreports.larc.nasa.gov/ltrs/ltrs.html
NASA Center for AeroSpace Information National Technical Information Service (NTIS)
800 Elkridge Landing Road 5285 Port Royal Road
Linthicum Heights, MD 21090-2934 Springfield, VA 22161-2171
(301) 621-0390 (703) 487-4650
Abstract
AAAAAAAA
Debonded regions
load L
Eb A
∫
2
U m = ---------- ε b dx
AAAAAAAA
(2a)
2
0
L
EbI
∫
2
Foam core U b = --------- ( w″ ) dx (2b)
2
Faceskin 0
Buckled faceskin tb L
Ω = P u′ dx ∫ (2d)
P P 0
2
z,w
lf ld lf
AAAAAAAAAAAAAAA
tb Eb , I
k
P P
tf
x, u
Symmetry
plane
L
where K is a constant and can be represented as Ef b/tf Introducing equations (8) and (9) into equation (4),
and Ef is the Young’s modulus of the sandwich foam assuming inextensional buckling (i.e., u = 0), and collect-
core. The variable ψ(x) is defined as ing the second-order terms gives the second variation of
the potential energy:
1 if 0 ≤ x ≤ l f
L
ψ ( x ) = 0 if l f ≤ x ≤ L – l f
∫ [ E b I ( w″ )
(6) 2 2 2 2
δ Π = + Kψw – P ( w′ ) ]dx (10)
1 if L – l f ≤ x ≤ L 0
For stability of a system in equilibrium, the second In the Rayleigh-Ritz method, the out-of-plane defor-
variation of potential energy needs to be positive mation at the onset of buckling w of the panel faceskin is
semidefinite, approximated as
2 M
δ Π[w] ≥ 0 (7) w( x) = ∑ Am φm ( x ) (11)
When the second variation of potential energy is zero, a m=1
neutral equilibrium configuration exists and indicates the where φm are selected functions that satisfy the geometric
onset of buckling. This point is characterized as the boundary conditions and Am represents their amplitudes.
bifurcation point (ref. 10). The Trefftz criterion states For simple support boundary conditions, a sinusoidal
that equation (7) is identically satisfied when the varia- function is used:
tion of the second variation is zero (ref. 11).
mπx
The expression for δ2Π is obtained by perturbing the φ m ( x ) = sin ----------- (12)
L
system in equation (4) such that
The Trefftz criterion used with the Rayleigh-Ritz
u → u0 + u method to determine the buckling load can be written as
(8)
w → w0 + w 2
∂δ Π [ w ]
----------------------- = 0 ( m = 1, 2, …, M ) (13)
∂ Am
where u0 and w0 represent the prebuckling configuration
and the variations u(x) and w(x) are infinitesimally small Dimensional parameters are nondimensionalized by
increments. For the prebuckling configuration, the length L as follows:
ξ = x/L
u 0 = – ---------- x
P
Eb A η = l f /L
(9) (14)
w0 = 0 µ = l d /L = 1 – 2η
3
The variable η is the foundation ratio, where a value the finite-difference method (ref. 10). The governing
of zero is equivalent to a faceskin (beam) with no foun- equations at the onset of buckling are
dation (unsupported) and a value of one-half is equiva-
lent to a fully supported faceskin (beam). The variable µ w
IV P
+ ------ w″ = 0
is the debond ratio and represents the fractional amount EI ( for debonded region )
of the strip which is debonded. The functions φm in equa- (18)
IV P k ( for bonded region )
tion (12) and ψ in equation (6) become φ̃ m and ψ̃ , w + ------ w″ + ------ w = 0
EI EI
respectively, after the nondimensionalized parameters in
equation (7) are used. Incorporating equation (13) gives In this method, N + 4 discrete evenly spaced gridpoints
the eigenvalue problem: x−1 to xN+2 are used with spacing ∆ as shown in figure 4.
Derivatives of w are approximated by
F – 1
[
b Ũ ] + -----
- [ Ũ f [ Ω̃ ] { A } = λ { A }
] (15)
4
µ w n + 1 – 2w n + w n – 1
w″n = --------------------------------------------------
2
where the matrix elements are ∆
(19)
IV w n + 2 – 4w n + 1 + 6w n – 4w n – 1 + w n – 2
1 2 2 w n = -------------------------------------------------------------------------------------------------
d φ̃ i d φ̃ j 4
∆
Ũ b ij = ∫ ---------- ----------- dξ
dξ dξ
2 2
(16a)
0
with simply supported boundary conditions
1
w0 = 0
Ũ f ij = ∫ ψ̃φ̃i φ̃ j dξ (16b)
w–1 = –w1
0
(20)
1 wN + 1 = 0
dφ̃ i dφ̃ j
Ω̃ ij = ∫ -------- -------- dξ
dξ dξ
(16c) w N = –w N + 2
0
Substituting equation (19) into equation (18) for points
and x1 to xN along with the boundary conditions in equa-
EbI tion (20) provides N + 4 simultaneous algebraic equa-
1
λ = --- = ---------- (17a) tions for w−1 to wN+2. The nontrivial solution for the
P̃ 2
PL buckling load is represented by the lowest value of P for
which the determinant of the coefficients of the N alge-
4 braic equations vanishes (ref. 9).
K ld
F = --------- (17b)
EbI
Finite-Element Model
The terms P̃ and F are, respectively, the dimensionless
load and foundation stiffness parameters. Solving equa- Plane-strain finite-element analysis. Due to sym-
tion (15) for the largest eigenvalue λ yields the lowest metry about the panel’s half-length (x = L/2) and thick-
buckling load Pcr. ness, only a quarter of the debonded sandwich panel
cross section was modeled by 4-noded quadrilateral
Finite-difference method. An alternative method for assumed natural-coordinate strain (ANS) finite elements,
examining the elastic stability of a sandwich debond is as shown in figure 5 (refs. 12 and 13). The element is
AAAAAAAAAAA
Springs
P
4
Faceskin
Core
Symmetric
boundary
conditions
Debonded region
Faceskin
Core
based on a Mindlin/Reissner shell formulation. Eight A linear stability analysis was performed using
layers of 4-noded quadrilateral finite elements are used NASA Langley Research Center’s COmputational
to model the faceskin and sandwich foam regions. MEchanics Testbed (COMET) code (ref. 12) to compute
Duplicate nodes for the faceskin and core at their com- the buckling load of the debonded panel. The linear elas-
mon boundary along the debonded region created the tic stability analysis in COMET is formulated using the
debonded region between the faceskin and core. (See concept of adjacent equilibrium. Buckling occurs when
fig. 6.) These duplicate nodes are allowed to translate the membrane strain energy is converted to bending
independently. strain energy in the adjacent equilibrium configuration.
5
The linear elastic stability analysis reduces to an eigen- changes in debond length, core moduli, and Poisson’s
value analysis to find the bifurcation from the membrane ratios affect the buckling load. Tables 1 and 2 summarize
state to the bending state. The eigenvalue satisfies the data used in the analyses for the Rayleigh-Ritz and
finite-difference methods and the plane-strain finite-
Kφ i + λ i K g ( σ )φ i = 0 ( i = 1, 2, … ) (21) element analyses. The subscripts (i, j = 1, 2) for Ebij and
where Efij are along the x- and z-directions, respectively. In gen-
eral, the aluminum faceskin and foam core are treated as
K assembled linear elastic stiffness matrix isotropic materials. However, to simulate the Winkler
Kg(σ) assembled geometric stiffness matrix foundation theory using finite-element analysis, the
φi ith eigenvector or mode shape Poisson’s ratios and the foam longitudinal stiffness
(x-direction) were given reduced constitutive properties.
λi ith eigenvector or buckling load factor The Poisson’s ratios of the faceskin and core, νb and νf,
were set to 0, and Young’s modulus of the sandwich
Results and Discussion foam, Ef 11, was set to approximately 0 (10−4) psi. The
Three analytical methods were used to calculate the reduction of constitutive properties allowed evaluation of
.buckling load of the debonded sandwich panel strip, the effects of the simplifying assumptions of the elastic
which consisted of an aluminum faceskin and a foam foundation approach. In addition, the tests evaluated the
core. Parametric studies were performed to study how effects of transverse shear.
(a) Set 1
(b) Set 2
6
Effect of Sandwich Foam Stiffness on Buckling required for a convergent solution. As a result, conver-
Load gence is slower than the Rayleigh-Ritz method. In addi-
tion, the solution procedure of this method was
Rayleigh-Ritz method. The Rayleigh-Ritz method inefficient because the buckling load could not be solved
was used to perform a parametric study of variations in for directly as in the Rayleigh-Ritz method. Instead, the
the foundation stiffness of the core. The dimensionless buckling load was determined by using successive itera-
foundation stiffness F varied from 0 (no core stiffness) to tions of the bisection method until a root of the polyno-
a typical foam core stiffness of 30.0 (K = 12500 psi) for a mial equation was found (ref. 14). A change of sign of
specific value of the debond ratio µ. The debond length the determinant of the coefficients in the algebraic equa-
was fixed at 2.0 in., or µ = 0.067 in. The results from this tions indicates a root (buckling load) between two suc-
study are presented in terms of the relative buckling cessive iterations. Due to the length of the computational
load Pcr /PE where PE denotes the Euler buckling load time for this analysis, 360 gridpoints were used as the
(π2Eb I/L2) of the debonded faceskin. final solution despite the lack of full convergence.
A plot of Pcr /PE versus F in figure 7 shows the con-
vergence of the Rayleigh-Ritz method as the number of Finite-element method. The 2-D plane-strain finite-
shape functions is increased. As expected, the buckling element model shown in figure 5 was analyzed using
load of the debonded sandwich panel strip increases as 4-noded quadrilateral elements. The finite-element
the foundation stiffness increases. Twenty-five terms in model consisted of 1280 elements and 1385 nodes for a
the Fourier sine series were sufficient to obtain a con- symmetric model. Duplicate nodes are used to model the
verged solution. debonded region. Two eigenvalue analyses were
performed using this model. In the first analysis, the
Finite-difference method. The finite-difference aluminum faceskin and foam core were assumed to be
method was also used to predict the buckling load. Fig- isotropic with the properties shown in table 2. The finite-
ure 8 illustrates the convergence of buckling load as the element results from the isotropic case (at F = 30.0)
number of gridpoints is increased when using the finite- are about 20 percent higher than those from the
difference method. Because the finite-difference method Rayleigh-Ritz and finite-difference analyses, as shown in
discretely models the elastic foundation as a series of figure 9. This difference increases as the foundation
closely spaced springs, a large number of gridpoints are stiffness increases.
300
Eb = 1.0 × 107 psi
F 0.2 in.
AAAAA
2.0 in.
0.8 in.
30.0 in.
Relative buckling load, Pcr /PE
200 10 terms
25 terms
50 terms
100
0 10 20 30
Dimensionless foundation stiffness, F
7
200
60 divisions
120 divisions
240 divisions
360 divisions
AAAAA
F 2.0 in. 0.2 in.
AAAAA
60 divisions
120 divisions
50 0.8 in. 240 divisions
360 divisions
30.0 in.
0 10 20 30
Dimensionless foundation stiffness, F
150
AAAAAA
0.8 in.
30.0 in.
Relative buckling load, Pcr /PE
100
Rayleigh-Ritz
Finite difference
50 Finite-element analysis (isotropic)
Finite-element analysis (υb = υf = 0, Ef 11 ≈ 0)
0 10 20 30
Dimensionless foundation stiffness, F
Figure 9. Comparisons of relative buckling load results from different methods for variations in foundation stiffness.
This discrepancy may have been induced by model- discrepancy is caused by the simplifying assumptions
ing the core material as a Winkler foundation in which of the elastic foundation approach, a second finite-
the elastic springs represent the core. To verify that the element analysis was performed with reduced constituent
8
properties. The Poisson’s ratios of the aluminum face- Fifty terms in the Rayleigh-Ritz method and
skin and foam core, νb and νf, were set to 0, and Young’s 360 gridpoints in the finite-difference method were used
modulus of the foam, Ef 11, was set to approximately 0 in this study. In the finite-element analysis, two sets of
(10−4) psi. This reduction of the material properties elim- constituent properties were again used. Set 1 used isotro-
inates the effects of the Poisson’s ratios and the longitu- pic properties and Set 2 used reduced constituent proper-
dinal stiffness of the core. Figure 9 shows a comparison ties (νf12 = νb12 = 0, Ef11 ≈ 0). The results in figure 10
of the second set of finite-element results with the show that the normalized buckling load decreases as µ
Rayleigh-Ritz and finite-difference methods. Excellent increases, indicating that the buckling load for the unsup-
agreement is observed between the finite-element analy- ported skin decreases as the debonded length increases.
sis using the reduced material properties and the results The results from the isotropic plane-strain finite-element
from the Rayleigh-Ritz and finite-difference methods. analysis are about 10–20 percent higher (depending on
the debond length) than those predicted by the Rayleigh-
Effect of Debond Length on Buckling Load Ritz and finite-difference methods. This suggests that the
In a second parametric study, the debond ratio µ was simplifying assumptions on Poisson’s ratio and the longi-
varied from 0 (fully bonded) to 1 (fully debonded) for a tudinal stiffness of the core can substantially affect buck-
fixed value of F = 30.0 (K = 12500 psi). The limiting ling load predictions. However, excellent agreement is
value of µ = 0 yields the buckling load for a sandwich observed once these properties are reduced.
panel strip with no debonds, and a value of µ = 1 yields
the Euler buckling load PE for a fully debonded faceskin. Effect of Transverse Shear on Buckling Load
The results are presented in figure 10 in terms of the nor- Transverse shear effects were found to be insignifi-
malized buckling load Pcr /PFULL, which is the ratio of cant for the studies performed in this paper. From equa-
the buckling load divided by the buckling load of a beam tion (4.76) in reference (ref. 10), the buckled fully
fully supported by an elastic foundation (ref. 10): bonded faceskin is calculated to have a wavelength
2 of 2.68 in. for the largest K considered in this study
2 π EbI
2
kL (12500 psi). This corresponds to a ld/tb ratio of 13.4,
P FULL = n --------------- + ----------- ( n = 1, 2, … ) (22)
L
2
n π
2 2 which is estimated to contribute less than 2-percent error
for the buckling load of the debonded faceskin using ref-
where n is the number of half sine waves required to erence (ref. 15). Therefore, transverse shear effects were
give the lowest buckling load, depending upon the not considered in the Rayleigh-Ritz and finite-difference
parameters. analyses.
1.2
Eb = 1.0 × 107 psi
AAAAA
F 0.2 in.
1.0 ld
Normalized buckling load, Pcr /PFULL
.8
AAAAA
AAAAAA
0.8 in.
30.0 in.
.6 Rayleigh-Ritz
Finite difference
Finite-element analysis (isotropic)
Finite-element analysis (υb = υf = 0, Ef 11 ≈ 0)
.4
.2
0 .2 .4 .6 .8 1.0
Debond ratio, µ
Figure 10. Comparisons of normalized buckling load results from different methods for various debond ratios and F = 30.0.
9
Conclusions 2. Hwu, Chyanbin; and Hu, Jian S.: Buckling and Postbuckling
of Delaminated Composite Sandwich Beams. AIAA J., vol. 30,
Three numerical methods were used to analyze the no. 7, July 1992, pp. 1901–1909.
buckling of a debonded aluminum faceskin on a sand-
wich panel with a foam core. The Rayleigh-Ritz and 3. Webster, John D.: Flaw Criticality of Circular Disbond
finite-difference methods were used to study a 1-D Defects in Compressive Laminates—1980–81 Interim Report.
beam-on-elastic-foundation model that represents a strip NASA CR-164830, June 1981.
of a sandwich panel with a through-the-width debond.
4. Chai, Herzl; Babcock, Charles D.; and Knauss, Wolfgang G.:
The results were compared with the buckling loads cal- One Dimensional Modelling of Failure in Laminated Plates by
culated from a 2-D finite-element analysis using 4-noded Delamination Buckling. Int. J. Solids & Struct., vol. 17,
quadrilateral plane-strain elements. Two parametric stud- no. 11, 1981, pp. 1069–1083.
ies were performed to evaluate the effects of the elastic
foundation stiffness and debond length. Results from the 5. Chai, Herzl; and Babcock, Charles D.: Two-Dimensional
three analytical methods were compared and discussed. Modelling of Compressive Failure in Delaminated Laminates.
As expected, the results showed that the buckling load J. Compos. Mater., vol. 19, Jan. 1985, pp. 67–98.
increases with increasing core foundation stiffness and
6. Somers, M.; Weller, T.; and Abramovich, H.: Influence of Pre-
decreases with increasing debond length. determined Delaminations on Buckling and Postbuckling
The Rayleigh-Ritz method was shown to be superior Behavior of Composite Sandwich Beams. Compos. Struct.,
to the finite-difference method in obtaining the buckling vol. 17, no. 4, 1991, pp. 295–329.
loads because it requires less computational time and
7. Vizzini, A. J.; and Lagace, P. A.: The Buckling of a Delami-
exhibits fast convergence. However, both the Rayleigh- nated Sublaminate on an Elastic Foundation. J. Compos.
Ritz and finite-difference methods assumed a Winkler Mater., vol. 21, Dec. 1987, pp. 1106–1117.
core foundation and, thus, ignored the Poisson’s ratio and
longitudinal stiffness effects of the core material. Results 8. Kim, W. C.; and Dharan, C. K. H.: Facesheet Debonding
from the Rayleigh-Ritz method and the finite-difference Criteria for Composite Sandwich Panels Under In-Plane
method, which are based on the elastic foundation Compression. Eng. Fract. Mech., vol. 42, no. 4, July 1992,
approach, show lower buckling loads than those pre- pp. 643–652.
dicted by the 2-D plane-strain finite-element analysis,
9. Brush, Don O.; and Almroth, Bo O.: Buckling of Bars, Plates,
which models the sandwich foam core as an isotropic
and Shells. McGraw-Hill, Inc., 1975.
material. When the core is forced to behave as a Winkler
foundation so that the Poisson’s ratios of the core are set 10. Chen, Wai-Fah; and Atsuta, Toshio: Theory of Beam Columns.
to 0 and its longitudinal stiffness is reduced to be infini- Volume 1—In-Plane Behavior and Design. McGraw-Hill, Inc.,
tesimally small, the finite-element analysis agrees with 1976.
the results from the Rayleigh-Ritz and finite-difference
methods. Therefore, to accurately predict the buckling 11. Langhaar, Henry L.: Energy Methods in Applied Mechanics.
load of a debonded sandwich panel with sandwich foam John Wiley & Sons, Inc., 1962.
cores, a finite-element model that includes the Poisson’s 12. Stewart, Caroline B.: The Computational Structural Mechan-
ratio and longitudinal core stiffness should be used. ics Testbed User’s Manual. NASA TM-100644, 1989.
10
Form Approved
REPORT DOCUMENTATION PAGE OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instructions, searching existing data sources,
gathering and maintaining the data needed, and completing and reviewing the collection of information. Send comments regarding this burden estimate or any other aspect of this
collection of information, including suggestions for reducing this burden, to Washington Headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson
Davis Highway, Suite 1204, Arlington, VA 22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188), Washington, DC 20503.
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
December 1995 Technical Memorandum
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
Buckling Analysis of Debonded Sandwich Panel Under Compression
WU 510-02-12-01
6. AUTHOR(S)
David W. Sleight and John T. Wang
Unclassified–Unlimited
Subject Category 01
Availability: NASA CASI (301) 621-0390