Composite Structures 93 (2011) 351359

Contents lists available at ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Use of material grading for enhanced buckling design of thin-walled composite rings/long cylinders under external pressure
Karam Y. Maalawi
Department of Mechanical Engineering, National Research Center, 12622 Dokki, Cairo, Egypt

a r t i c l e

i n f o

a b s t r a c t
This paper presents a mathematical model for enhancing the buckling stability of composite, thin-walled rings/long cylinders under external pressure using radial material grading concept. The main structure to be analyzed is built of multi-angle brous laminated lay-ups having different volume fractions of the constituent materials within the individual plies. This leads to a piecewise grading of the material in the radial direction. The objective is to maximize the critical buckling pressure while preserving the total structural mass at a constant value equal to that of a baseline design. The ber volume fractions are included among the standard design variables such as ber orientation angles and ply thicknesses, which are used by many investigators in the eld. The model employs the classical lamination theory, where an analytical solution that accounts for the effective axial and exural stiffness separately is given. The critical buckling pressure contours subject to the mass equality constraint are given for several types of anisotropic rings/long cylinders showing the functional dependence of the constrained objective function on the selected design variables. It is shown that material grading can have signicant contribution to the whole optimization process in achieving the required structural designs with enhanced stability limits. 2010 Elsevier Ltd. All rights reserved.

Article history: Available online 21 September 2010 The author dedicates this paper to his teacher; mentor and friend, late Professor Liviu Librescu of the ESM Department at Virginia Tech, USA Keywords: Buckling stability Structural optimization Fibrous composites Functionally graded material Laminated rings/cylindrical shells External hydrostatic pressure

1. Introduction Structural applications of composite materials are increasing in several areas where high stiffness/weight ratio and long fatigue life are most benecial [1]. A common application is the design of composite cylindrical shells under the action of external hydrostatic pressure, which might cause collapse by buckling instability [2,3]. Examples are the underground and underwater pipelines, rocket motor casing, boiler tubes subjected to external steam pressure, and reinforced submarine structures. The composite cylindrical vessels for underwater applications [4] are intended to operate at high external hydrostatic pressure (sometimes up to 60 MPa). For deep-submersible long-unstiffened vessels, the hulls are generally realized using multi-layered, cross-ply, composite cylinders obtained following the lament winding process [5]. Previous numerical and experimental studies have shown that failure due to structural buckling is a major risk factor for thin laminated cylindrical shells. Anastasiadis and Simitses [6] studied the buckling of long laminated cylindrical shells under external radial pressure using higher order deformation theory. Their formulation, however, was restricted to symmetric lay-ups with respect to the mid-surface, to eliminate the coupling terms, as well as constant-

Tel.: +20 002 0181472225; fax: +20 2 33370931.

E-mail address: NRC.Aero@gmail.com 0263-8223/$ - see front matter 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.compstruct.2010.09.007

directional pressure. More conservative results for a true uid pressure were given by Rasheed and Yousif [7,8] who applied standard energy formulation to derive the kinematics and equilibrium equations and the classical lamination theory to express the needed constitutive equations. They developed a generalized closed form analytical formula for calculating stability limits of thin anisotropic rings/long cylinders subject to hydrostatic pressure. Another rened treatment of the in-plane buckling of rings was given by Hodges [9] and Hodges and Harursampath [10]. Formulation was based on a non-linear theory for stretching and bending of anisotropic beams having constant initial curvature in their plane of symmetry with the only restriction of small strain in the pre-buckling state. Recently, the incorporation of material grading in which the physical and mechanical properties vary spatially can play an important role in the design optimization of a variety of structural systems [11]. The functionally graded material FGM concept originated in Japan in 1984 during the space project, in the form of proposed thermal barrier material capable of withstanding high temperature gradients. Since 1984, FGMs have been comprehensively researched, and are almost a commercial reality offering great promise in several applications. The basic knowledge on the use of FGMs and their wide applications can be found in [12]. Considering their applications in composite structures, Chen and Gibson [13] performed experimental and theoretical analyses to determine the in-plane ber distribution in unidirectional

352

K.Y. Maalawi / Composite Structures 93 (2011) 351359

reinforced composites. They considered distributions represented by polynomial functions, and applied Galerkins method to calculate the required coefcients from the resulting algebraic equations. Chi and Chung [14] studied the mechanical behavior of FGM plates under transverse loading, where a constant Poissons ratio and variable moduli of elasticity throughout the plate thickness was assumed. The volume fraction of the constituent materials were dened by simple power-laws, and closed form solutions using Fourier series were given for the case of simply-supported plates. Arshed et al. [15] applied RayleighRitz method to calculate frequencies of FGM cylindrical shells. A comparative study was given for different volume fraction laws, including polynomial, exponential and trigonometric functions. Boundary conditions were restricted to the case of simply supported edge condition. Recently, Saha and Maiti [16] also used power-law distribution to determine the natural frequency of axially loaded cylinder in which the properties are graded in the thickness direction. They employed Loves rst approximation shell theory where closed form solutions were given for the case of simply supported cylinder with several thickness/radius and length/radius ratios. In the eld of structural optimization, several papers appeared on the topic of buckling and stability optimization. Maalawi [17] presented a model for buckling optimization of elastic columns under mass equality constraint. He showed that the use of piecewise models in structural optimization gives excellent results and can be promising for similar applications. Another work by Maalawi and El-Chazly [18] dealt with both stability and dynamic optimization of multi-element beam type-structures. They formulated the associated optimization problems in a standard mathematical programming solved by the interior penalty function technique. In the eld of aeroelasticity, Librescu and Maalawi [11] introduced the underlying concepts of using material grading in optimizing subsonic wings against torsional instability. They developed exact mathematical models allowing the material physical and mechanical properties to change in the wing spanwise direction, where both continuous and piecewise structural models were successfully implemented. For brous laminated composite structures, the optimization of ply angles and thicknesses could allow the properties of the laminate to be tailored to a specic application. Chattopadhyay and Ferreira [19] performed a study to investigate the maximum buckling load of a cylinder subject to ply stress constraints using material and geometric design variables. A closed form shell equation was utilized for the buckling load calculation. Laminates were constrained to be symmetric, and the number of plies was included in the design variables. Results for graphite/ epoxy, glass/epoxy and Kevlar/epoxy models were found using the computer code CONMIN. Considering optimization of underwater cylindrical vessels, Tanguy et al. presented a paper [20] dealing with the optimal design of deep submarine vehicles. They developed a genetic algorithm procedure coupled with an analytical model to determine the laminate stacking sequences that maximizes the critical external buckling pressure. They also showed that the measured buckling pressures for glass/epoxy and carbon/epoxy cylinders appear to be in good agreement with numerical results and demonstrated the gains due to the optimized laminations. Little may be found in the literature that deals with buckling optimization of FGM rings/long cylinders under external hydrostatic pressure. The aim of the present study is, therefore, to incorporate the effect of changing the ber volume fraction in each lamina aiming at the achievement of enhanced stability limits of such shell-type-structures. Based on the mathematical concepts developed in a recent paper by the author [21], a useful optimization tool has been built for designing efcient congurations with improved buckling stability. This allows the search for optimal volume fractions that maximize the buckling pressure without violat-

ing the imposed mass constraint and manufacturing restrictions as well. The corresponding increases in the buckling pressures calculated with respect to a baseline design have been evaluated for several congurations, including cases of orthotropic, lament wound rings/long cylinders fabricated from different types of composite materials, namely; E glass/epoxy, S glass/epoxy, carbon/epoxy and E-glass/vinylester. It is assumed that the total structural mass is held at its reference value corresponding to the baseline design during the optimization process. Actually, substantial improvement in the overall stability level has been attained showing the usefulness of the proposed optimization model in arriving at the needed optimum designs for a variety of thin-walled anisotropic rings/long cylinders having arbitrary thickness/radius ratio. 2. Constitutive relationships with piecewise material grading in the radial direction The structural model used to represent the composite laminated shell-type-structures under study is schematically shown in Fig. 1. The 1, 2 and 3 are the principal directions of an orthotropic lamina, dened as follows: - Direction (1): Principal ber direction, also called ber longitudinal direction. - Direction (2): In-plane direction perpendicular to bers, transversal direction. - Direction (3): Out-of-plane direction perpendicular to bers; normal direction. 2.1. Elastic properties of a single lamina The reduced form of Hookes law for an orthotropic homogeneous lamina in a plane stress state may be written as [1]:

e11 Q 11 Q 12 0 r11 6 7 6 76 7 4 r22 5 4 Q 12 Q 22 0 54 e22 5 c12 0 0 Q 66 s12

32

3 1

where the square matrix [Q] is dened in terms of material properties as follows:

Q 11 Q 12

E11 ; 1 m2 E22 =E11 12 m12 E22 ; 1 m2 E22 =E11 12

Q 22

E22 ; 1 m2 E22 =E11 12 2

Q 66 G12

(2)

ds ds=Rd

(v) s (1)

s
d

x
(u)

z,(3)
(w)

Fig. 1. Laminated composite ring/cylindrical shell under external pressure (u displacement in the axial direction x, v in the tangential direction s,w in the radial direction z).

K.Y. Maalawi / Composite Structures 93 (2011) 351359 Table 1 HalpinTsai semi-empirical relations for calculating composite properties [22]. Elastic property E11 E22 G12 Mathematical formula* Em Vm + E1f Vf Em (1 + n gVf)/(1 gVf) Gm (1 + ngVf)/(1 gVf) m Vm + m12fVf

353

where uo(x, s), vo(x, s) and wo(x, s) are the displacements of a generic point (x, s) on the shell middle surface (z = 0) in x, s and z directions, respectively. The straindisplacement relations in terms of the middle surface strains and shell curvatures are given in the following:

g = (E2f Em)/(Ef2 + nEm) g = (G12f Gm)/(G12f + nGm)

m12

8 9 > exx > < = > :

Assuming no voids are present, then Vm + Vf = 1, where V denotes volume fraction. * Subscripts m and f refer to properties of matrix and ber materials, respectively.

ess eo z jss > > > ss > > cxs ; : co ; : jxs ; xs

8 o 9 > exx > < =

8 9 > jxx > < =

7
curvatures are [25]: 8 2 9 8 9 > @ wo > > @x2 > > jxx > < = < = vo @ @wo jss @s @s R > > > > @ @wo v o > > : : ; jxs ; 2 @x @s R

E11 and E22 are the Youngs moduli in the 1 and 2 directions, G12, the shear modulus, and m12 the major Poissons ratio. Based upon the semi-empirical methods by HalpinTsai [22], Table 1 gives the mathematical formulas for determining the different elastic moduli for known type, properties and volume fractions of the ber and matrix materials. The factor n is called the reinforcing efciency and can be determined experimentally for specied types of ber and matrix materials. Experimental results fall within a band of 1 < n < 2. Usually, n is taken as 100% for theoretical analysis procedures, especially in case of glass and carbon composites. 2.2. Loaddeformation relationships For a generally orthotropic material, Eq. (1) must be transformed to reect rotated ber orientation angles:

where the middle surface strains and 9 8 o 9 8 @uo > > exx > > @x > < = > < = o @v o wo v o 2 ; 1 @wo ess @s R 2 @s R > o > > > : ; > @uo @ v o > : ; cxs @x @s

8
The resultant forces and moments per unit length applied at the middle surface are dened by the integrals:

Forces :

8 9 8 9 8 9 > Nxx > Z h=2 > rxx > n < = < = <r = X Z zk > xx > N rss dz rss dz > ss > > > h=2 > zk1 > : ; : : k1 Nxs sxs ; sxs ; 8 8 8 9 9 9 > M xx > Z h=2 > rxx > n < < <r = = = X Z zk > xx > M rss zdz rss zdz > ss > > > h=2 > zk1 > : : : ; k1 M xs sxs ; sxs ;

Moments :

10
where h is the total shell thickness and zk1 and zk, k = 1,2,. . .n are the coordinates of the kth lamina boundaries measured from the middle surface. Substituting for the stresses in terms of strains as dened in (3) and (7), we get:

exx rxx Q 11 Q 12 Q 16 6 7 6 76 7 4 rss 5 4 Q 12 Q 22 Q 26 54 ess 5 cxs sxs Q 16 Q 26 Q 66

32

3 3

The elements of the Kth lamina stiffness matrix, Q , which is now referred to the reference axes of the cylindrical shell (x, s, z), are given by:

Q 11 U 1 U 2 cos 2h U 3 cos 4h Q 22 U 1 U 2 cos 2h U 3 cos 4h Q 12 U 4 U 3 cos 4h Q 16 0:5U 2 sin 2h U 3 sin 4h Q 26 0:5U 2 sin 2h U 3 sin 4h Q 66 0:5U 1 U 4 U 3 cos 4h
The terms Ui are solely function of the material properties and, hence the volume fractions. They are no longer termed as invariant as has been cited by several investigators, and are dened by the following expressions [23]:

8 9 2 A11 > Nx > > > > >N > 6A > s > 6 12 > > > > > > <N > 6A = 6 xs 16 6 > M x > 6 B11 > > 6 > > 6 > >M > 4B > s > > > > 12 > > : ; M xs B16

A12 A22 A26 B12 B22 B26

A16 A26 A66 B16 B26 B66

B11 B12 B16 D11 D12 D16

B12 B22 B26 D12 D22 D26

9 38 B16 > eo > > xx > 7> eo > > > B26 7> ss > > > > > > 7> B66 7< co = xs 7 D16 7> jxx > > > 7> > > 7> > > D26 5> jss > > > > > : ; D66 jxs

11

where Aij are called the extensional stiffnesses given by:

Aij h

n X Q ij k ^k ^k1 z z k1

12a

Bij are called the bending-extensional stiffnesses given by:

Bij

2 n h X Q ij k ^2 ^2 zk zk1 2 k1

12b

U 1 3Q 11 3Q 22 2Q 12 4Q 66 =8 U 2 Q 11 Q 22 =2 U 3 Q 11 Q 22 2Q 12 4Q 66 =8 U 4 Q 11 Q 22 6Q 12 4Q 66 =8

Dij are called the bending stiffnesses:

Dij

3 n h X Q ij k ^3 ^3 zk zk1 3 k1

12c

For classical lamination theory, it is assumed that n layers of material are perfectly bonded together, with innitely thin, non-shear deformable boundaries. Using Kirchoff plate theory [24], which assumes that the in-plane displacements vary linearly through the thickness of the laminate, the displacements of a material point a distance z from the middle surface are:

^ where ^k zk =h is a dimensionless coordinate, and hk ^k ^k1 is z z z the dimensionless thickness of the kth lamina.

3. Analytical buckling model In order to restrict the time of calculation to acceptable values for the developed optimization tool, the analytical formulation shall be based on the derivation given by Rasheed and Yousif in their fruitful papers [7,8], which are based on the assumption of small hoop strain and rotation of circumferential elements. Such an approach provides good sensitivity to lamination parameters, and allowing the search for the needed optimal stacking sequences

@wo @x   @wo v o z s R/; ( 1 v x; s; z v o x; s z R @s R ux; s; z uo x; s z wx; s; z wo x; s;

354

K.Y. Maalawi / Composite Structures 93 (2011) 351359 Table 2 Buckling pressure formulas for limiting cases (w = 0 and a ( 1). Limiting cases Critical buckling pressure (pcr = 3D/R3) Bending stiffness, D Thin isotropic rings [25] Thin isotropic long cylinders [24] Thin orthotropic rings [6] Thin orthotropic long cylindrical shells with bers parallel to the shell axis x [6] Eh3/12 Eh3/12(1 m2) E22h3/12 E22h3/12(1 m12m21)

and volume fractions, which maximizes the buckling pressure in a reasonable computational time. The governing differential equations of anisotropic rings/long cylinders subjected to external pressure are similar to those of the isotropic case [24,25]. They are given as:

M 0ss

RN0ss

bNss bpR

M 00 RNss bNss 0 pwo v 0o pR2 ss

13

where the prime denotes differentiation with respect to angular position , and b mo w0o =R. Two possible solutions for Eq. (13) can be obtained; one for the pre-buckled state and the other termed as the bifurcation solution obtained by perturbing the displacements about the pre-buckling solution. For laminated composite rings and long cylindrical shells the only signicant strain components are the hoop strain (eo ) and the circumferential curvature (jss) of ss the mid-surface. The reduced form of Eq. (11) for the two cases can be shown to be:

E, m = isotropic modulus of elasticity and Poissons ratio. E22 = hoop modulus, m12 = Poissons ratio for axial load, m21 = m12E22/E11. (In cases with bers perpendicular to the shell axis, E22 should be replaced by E11).

&

Nss M ss

'

Aani Bani Bani Dani

!&

eo ss jss

' 14

It is to be noticed here that the formula given in (19) for calculating pcr is only valid for thin rings/cylinders with thickness-to-radius ratio (h/R) 6 0.1. The reduced forms of Eq. (19) for some limiting cases where w = 0 and a ( 1 are given in Table 2.

The matrix elements Aani, Bani and Dani in Eq. (14) are to be determined depending on whether they are related to the case of thin rings or long cylinder shells. 3.1. Case of thin rings In the case of thin rings the axial and shear forces (Nxx, Nxs) must vanish along the free edges. The bending and twisting moments (Mxx, Mxs) may also be neglected. Therefore, the rst, third, fourth and sixth rows of (11) are solved for the strains and curvatures in terms of (eo ; jss ) to give the following matrix relation: ss

4. Optimization problem statement The associated optimization problem shall seek maximization of the critical buckling pressure pcr (i.e. minimization of pcr) while maintaining the total structural mass constant at a value equals to that of a reference baseline design. Optimization variables include the ber volume fraction (Vfk), thickness (hk) and ber orientation angle (hk) of the individual kth ply, k = 1, 2,. . .n (total number of plies). Side constraints are always imposed on the design variables for geometrical, manufacturing or logical reasons to avoid having unrealistic odd-shaped optimum designs. 4.1. Denition of the baseline design

8 o > exx > > < o

cxs eo S1 1 S2 ss > jxx > jss > > > > : ; jxs
2 A11 A16 A66 B16 B66 6A 6 16 where S1 6 4 B11 B16

9 > > > =

&

' 15

B11 B16 D11 D16

B16

A12

B12

6A B66 7 7 6 26 7 and S2 6 4 B12 D16 5 D66 B26

B26 7 7 7 D12 5 D26 16

Substituting (15) back into (11), it can be shown that:

Aani Bani

Bani Dani

!
ring

A22 B22

! B22 S2 T S1 1 S2 D22

It is convenient rst to normalize all variables and parameters with respect to a baseline design, which has been selected to be a unidirectional, orthotropic, single layer ring/long cylinder with the bers parallel to the shell axis x and with equal volume fractions of ber and matrix materials, i.e. Vf = Vm = 50%. Optimized shell designs shall have the same total structural mass, properties of the matrix and ber materials, mean radius R and total shell thickness h of the baseline design. Therefore, the preassigned parameters, which are not subject to change in the optimization process, ought to be the type of material of construction, mean radius and total thickness of the shell. 4.2. Optimization model Coupling the analytical buckling shell model to a standard nonlinear mathematical programming procedure can perform the search for the required optimized lamination. The design variable vector, ~d , which is subject to change in the optimization process, x is dened as:

17

3.2. Case of long cylinders For the case of a long cylinders, the out-of-plane displacements are restrained (that is: eo co jxs 0). Therefore, the only xx xs strains to be taken into considerations are the in-plane hoop strain (eo ) and the circumferential curvature (jss). Accordingly, the ress duced matrix of Eq. (14) takes the following form:

^ ~d V fk ; hk hk k 1; 2; . . . n x

20

Aani Bani

Bani Dani

cylinder

A22 B22

B22 D22

18

The nal closed form solution for the critical buckling pressure is given by the following mathematical expression [8,21]:

pcr 3

Dani R3

!"

# 1 w2 =a ; 1 a 2w

   1 Bani ; R Aani

  Dani Aani R2 1 19

where the dimensionless thickness of the kth lamina is dened by ^ hk hk =h. The total structural mass M is kept equal to the baseline ^ design mass M0, so the dimensionless mass M M=M0 equals 1. Since the ber volume fraction of the baseline design Vf0 equals 50%, a feasible design must satisfy the constraint equation Pn ^ k1 V fk hk 0:5, where Vfk is the ber volume fraction of the kth lamina. Therefore, the buckling optimization problem considered herein might be cast in the following standard mathematical programming form:

K.Y. Maalawi / Composite Structures 93 (2011) 351359

355

Minimize subject to
n X k1

^ pcr
N X k1

5. Results and discussions The given approach outlined above shall be applied to several cases of study of thin-walled, anisotropic, radially graded rings/ long cylinders subjected to external hydrostatic pressure. The functional behavior of the candidate objective function, as represented ^ by maximization of the dimensionless buckling pressure pcr (i.e. ^ minimization of pcr ) without mass penalty, shall be thoroughly investigated in order to see how it is changed with the optimization variables in the selected design space. Table 3 gives the different properties of the chosen composite materials, which have favorable characteristics and are desirable for the manufacturing of cylindrical shell-type-structures [4]. The variation of the longitudinal modulus E11, transverse modulus E22 and the in-plane shear modulus G12 with the ber volume fraction Vf for the different composite materials is shown in Fig. 2. The reinforcing efciency factor n has been taken as 100% in the subsequent investigation and numerical calculations. Using the formulas given in Tables 1 and 2, appropriate values for the orthotropic properties and critical buckling pressure (pcro) of the baseline design have been determined and given in Table 4. They depend upon the type of composite material utilized and the shell thickness-to-radius ratio (h/R) as well. 5.1. Long cylinders optimization The rst case study to be examined herein is a long thin-walled cylindrical shell fabricated from E-glass/epoxy composites with the lay-up made of only two plies (n = 2) having bers parallel to the xaxis (i.e. h1 = h2 = 0). Considering the case with no side inequality constraints imposed on the design variables, Fig. 3 shows the ^ developed pcr -level curves, augmented with the mass equality con^ straint, in V f 1 h1 design space. It is seen that such a constrained objective function is well behaved in the selected design space having the shape of a tent with its ceiling formed by two curved lines, above which the mass equality constraint is violated. Their zigzagged pattern is due to the obliged turning of many contours, which are not allowed to penetrate the tents ceiling and violate the mass equality constraint. The curve to the left represent a 100% ber volume fraction of the outer ply, Vf2, while the other curve to the right represents zero ^ volume fraction, that is Vf2 = 0%. Two local minima with pcr near a value of 0.90 can be observed: one to the lower left zone near ^ the design point V fk ; hk k1;2 0:15; 0:25; 0:6165; 0:75 while the other lies at the upper right zone close to the point (0.625, 0.745), (0.135, 0.255). This represents degradation in the stability level by about 10.6% below the baseline value. On the other hand, the unconstrained absolute optimum value of the dimensionless critical buckling pressure was found to be 1.7874 at the design point (1.0, 0.145), (0.415, 0.855). A more realistic optimum design has been obtained by imposing the side constraints: 0.25 6 Vfk 6 0.75, k = 1, 2. The attained solution is

^ V fk hk 0:5 21

^ hk 1

V L 6 V fk 6 V U ^ hL 6 h 6 hU ;
k

hL 6 hk 6 hU

k 1; 2; . . . n

^ where pcr pcr =pcro is the dimensionless critical buckling pressure, pcro is the critical buckling pressure of the baseline design and the subscripts L an U denote the lower and upper bounds imposed on the various design variables. In a real-world manufacturing process, the lament-winding angles hk must be chosen from a limited range of allowable lower (hL) and upper (hU) values according to technology references. This optimization problem may be thought as a search in a 3n-dimensional space for a point corresponding to the minimum value of the objective function and such that it lie within the region bounded by subspaces representing the constraint functions. 4.3. Optimization method A method, which has a wide applicability in engineering applications, is the penalty function method [26,27]. In this method the original objective function F~d is augmented with terms, called x penalty terms, such that as ~d approaches a constraint surface x one term increases indenitely. Since the algorithm seeks to minimize the value of the objective function then it will try not to penetrate any constraint surface. Thus all constraints are taken into consideration by representing them by penalty terms in the objective function expression. The most commonly used interior penalty function is cast in the form [26]

U~d ; r F~d r x x

M X j1

1 Gj ~d x

22

where U~d ; r is the modied objective function, Gj ~d is the Jth x x constraint function and r is a multiplier. A sequence of unconstrained minimization problems is solved with successively decreasing values of r. The MATLAB Optimization Toolbox [27] offers routines named fmincon and fminsearch implementing both constrained and unconstrained formulations. The MATLAB facilities are used for interacting with the routines which calculate the required numerical values of the original objective function and constraints. The two equality constraints of Eq. (21) can be used to discard any two variables of the whole set of design variables dened in Eq. (20), reducing the dimensionality of the optimization problem to 3n2.

Table 3 Properties of selected composite materials. Property* Fiber E-glass Youngs moduli (GPa) Shear moduli (GPa) Poissons ratio Mass density (g/cm3)
*

Matrix S-glass 86 86 35 0.23 2.49 Carbon (AS-4) 235 15 27 0.20 1.81 Graphite (GY-70) 520 9 28 0.15 1.86 Epoxy (3501-6) Em: 4.30 Gm: 1.60 mm: 0.35 qm: 1.27 Vinyl-ester 3.50 1.30 0.35 1.15

E1f: 73 E2f: 73 G12f: 30 m12f: 0.23 qf: 2.54

Data are taken from [1]. For a nal structural design, however, experimental verications are necessary.

356

K.Y. Maalawi / Composite Structures 93 (2011) 351359

Fig. 2. Variations of Youngs and shear moduli with ber volume fraction.

^ pcr max 1:2105 at the design point (0.75, 0.215), (0.4315, 0.785), showing that good shell designs with higher stability level ought to have a thinner inner layer with higher ber volume fraction and a thicker outer layer with less volume fraction. To see the effect of the ply angle, another case of study has been considered for a cylinder constructed from two balanced plies (h) with equal thicknesses and same material properties of E-glass/epoxy composites. This type of stacking sequence is widely used in lament wound circular shells since such a manufacturing process inherently dictates adjacent (h) layers. Fig. 4 shows the developed isomerits in the (Vf1 h) design space which are well behaved, monotonic and symmetric about the horizontal line of zero ply angles. A local minimum can be observed near the design point (Vf ^ 1, h) = (0.375, 0.0) with pcr 0:9985, indicating a degradation in the stability level below the baseline design. It is also seen that the absolute maximum occurs at the design points (Vf1, h) = ^ (0.5, 90) with pcr max 3:45766, which means that the dimensional critical pressure, pcr = 3.45766x2.865 = 9.906 x(h/R)3 GPa. Fig. 5 depicts the nal global optimum designs of cylinders constructed from adjacent (+h) and (h) plies for the different types of the selected composite materials. All shall have the same opti^ mal solution V fk ; hk k1;2 0:75; 0:215; 0:4315; 0:785, indepen-

dent upon the shell thickness-to radius ratio (h/R), a major contribution of the given formulation. Two distinct ranges can be observed; 0 < h < 30 and 30 < h < 90. In the former, the glass brous composites are better than the carbon ones in resisting buckling, while in the second range the buckling pressure of carbon composite is much exceeding that of the glass types, reaching a remarkable value of 32.686 (h/R)3 GPa for hoop wound cylinders. A comparison with the results obtained by Rasheed and Yousif [8] for a balanced (63) angle-ply cylinder made of E-glass/vinyl

M <1

M >1
Vf1 h1 =0.5 (Vf2=0%)

(1-Vf1) h1 =0.5 (Vf2=100%)

Table 4 Material properties and critical buckling pressure of the baseline design (pcro). Material type Orthotropic mechanical properties* E11 (GPa) E-Glass/epoxy S-Glass/epoxy Carbon/epoxy E-Glass/vinyl ester
*

pcro x (h/R)3 GPa

E22 (GPa) 11.18 11.404 7.60 9.325

G12 (GPa) 4.21 4.285 4.16 3.50

m12
0.29 0.29 0.275 0.29

Rings 2.795 2.851 1.90 2.33

Cylinders 2.865 2.913 1.909 2.379

38.65 45.15 119.65 38.25

Volume fractions: Vf = Vm = 50%. Reinforcing efciency factor n = 100%.

^ Fig. 3. The optimum tent-like design space containing pcr isomerits augmented ^ with the mass equality constraint M 1:0. Case of two-layer, E-glass/epoxy cylinder with bers parallel to cylinder axis (h1 = h2 = 0).

K.Y. Maalawi / Composite Structures 93 (2011) 351359

357

^ Fig. 4. pcr Isomerits in (Vf1h) design space under mass equality constraint. Case of long cylinder constructed from two balanced, E-glass/epoxy layers.

Max. buckling pressure x (h/R)3, GPa.

35 30
Carbon / Epoxy

25 20
S-glass/Epoxy

15
E-glass/Epoxy

10 5 0 0 10 20 30 40 50 60 70 80 90

E-glass/Vinylester

called a lumped-lay-up construction with the inner half of its wall composed of 90 hoop layers and the outer half made of 20 helically wound layers. The second type has different stacking sequence where the 20 layers are sandwiched in between outer and inner 90 hoop layers. Optimum solutions are given in Table 6 indicating substantial increase in the critical buckling pressure as compared with the non-optimal solutions presented in Refs. [7,8]. It is seen that good designs shall have thicker hoop wound layers with higher volume fraction of the bers near the upper limiting values imposed by the manufacturers. On the other hand, the sandwiched helically wound layers are seen to be thinner and have less ber volume fractions. 5.2. FGM ring optimization ^ Fig. 6 shows the developed pcr isomerits in (Vf1Vf2) design space for a two-layer, unidirectional ring made of E-glass/epoxy ^ ^ composites. Using the equality thickness constraint, h1 h2 1, one of the ply thicknesses can be eliminated. Another variable can also be discarded by applying the mass equality constraint, ^ M 1:0, which further reduces the number of design variables to only any two of the whole set of variables. The level curves shown in the gure represent the dimensionless critical buckling pressure augmented with the equality mass constraint. It is seen that the function is well behaved and continuous everywhere in the selected design space, except in the empty regions of the rst and third quadrants, where the mass equality constraint is violated. The cross lines Vf1 = 0.5 and Vf2 = 0.5 represent the level curves of the baseline design. Full similarity can be seen between the contours in the second and fourth quadrants, where the contour ^ with pcr 1:0 splits the region into two distinct zones, each of which contains both the absolute minimum and maximum design ^ ^ points, the former having pcr min 0:9105 located at V fk ; hk k1;2 0:25; 0:2505; 0:5835; 0:7495 in the second quadrant and (0.5835, 0.7495), (0.25, 0.2505) in the fourth quadrant. The abso^ ^ lute maxima pcr max 1:209 occur at V fk ; hk k1;2 0:4315; 0:785; 0:75; 0:215 in the second quadrant and (0.75, 0.215), (0.4315, 0.785) in the fourth quadrant. To examine the effect of ply angle, the same case is reconsidered by generating the associated level curves in (Vf1h) design space for a balanced (h) lay-up, as depicted in Fig. 7. It is seen that the contours are fully symmetric about the horizontal line h = 0 where

Ply angle, degrees.

Fig. 5. Variation of the absolute maximum buckling pressure with ply angle for balanced (h) cylinders with structural mass preserved constant.

Table 5 Buckling pressure comparison for (63) E-glass/vinyl ester cylinders (in MPa). pcr (MPa) h/R 1/50 1/25 1/20 1/15 1/25 1/20 1/15
*

pcr,max = 7.784 (10 h/R)3 (MPa) Rasheed and Yousif [7] 0.058 0.462 0.902 2.137 0.462 0.902 2.137 Optimal designs* 0.0623 0.4982 0.9730 2.3064 0.4982 0.9730 2.3064 Increase (%) 7.41 7.84 7.87 7.93 7.84 7.87 7.93

^ V fk ; hk k1;2 0:75; 0:215; 0:4315; 0:785.

ester and varying thickness-to-radius ratio (h/R = 1/50, 1/25, 1/20, 1/15) is given in Table 5. The actual dimensional values of the design variables can be determined by multiplying by the actual baseline cylinder thickness and radius (e.g. R = 127 mm, h = 2.54, 5.1, 6.35, 8.45 mm, as given in Ref. [7]). Other cases of study include optimization of multi-layered cylinders made of AS-4 carbon/epoxy composites. Two different constructions proposed by Rasheed and Yousif [7] will be considered herein. The rst one is

358

K.Y. Maalawi / Composite Structures 93 (2011) 351359

Table 6 Optimum buckling design of multi-layered, AS-4 composite cylinders. [90/20] Lay-up cylinder pcr,max = 9.37x(10h/R)3 (MPa) Ref. [7] 1/50 1/25 1/20 1/15 0.064 0.516 1.013 2.418 Optimum 0.075 0.596 1.171 2.776 Gain (%) 17.19 15.50 15.60 14.81 [90/20/90] sndwich-layup cylinder pcr,max = 36.634x(10h/R)3 (MPa) Ref. [7] 0.231 1.848 3.608 8.552 Optimum 0.293 2.344 4.579 10.854 Gain (%) 26.84 26.84 26.91 26.92

(h/R)

Optimum solution: ^ Two Helical layers : V f ; h; h 0:250; 0:225; 20  ^ Two Hoop layers : V ; h; h 0:705; 0:275; 90
f


^ V f ; h; h 0:2925; 0:235; 20  ^ V ; h; h 0:6835; 0:265; 90

f

Domain of mass constraint violation

M >1

Domain of mass constraint violation

M <1

^ ^ Fig. 6. pcr isomerits in (VfVf2) design space, M 1:0. Case of two-layer, E-glass/ epoxy ring with unidirectional lay-ups.

^ ^ Fig. 7. pcr isomerits in (Vf1h) design space (M 1:0). Case of E-glass/epoxy ring constructed from balanced (h) angle-ply with equal thicknesses.

the design point at the middle corresponds to the baseline design ^ ^ pcr 1:0. The absolute maxima, pcr max 3:458, occur at the design points (0.5, 90), corresponding to a hoop wound lay-up, which is an expected natural result. The region in the middle resembles a buttery bounded by congurations having the same ^ critical buckling pressure of the baseline, i.e. pcr 1:0. It contains two local minima at the design points (0.365, 27) and (0.635, ^ 27), where pcr min 0:951 representing about 5% degradation in the stability limits below the baseline design. Optimum solutions for both E-glass/epoxy and carbon/epoxy composite rings ^ with V fk ; hk k1;2 0:75; 0:215; 0:4315; 0:785 and different ply angles are shown in Fig. 8.

Another case of study, reported by Rasheed and Yousif [7], dealing with orthotropic carbon AS-4 rings cut from 12 in. tubes is considered herein. Due to the lament winding process, the tubes were not true hoop wound but rather manufactured with the bers forming (88) with the tube axis. Comparisons of the achieved optimum solutions with the results given in Ref. [7] are presented in Table 7. It is seen that an increase in the critical buckling pressure of about 24% has been attained, where the obtained ^ optimum design point was found to be V fk ; hk k1;2 0:75; 0:1375; 0:46; 0:8625. This means that good ring designs ought to have thinner inner layer with an increased ber volume fraction and thicker outer layer with lesser volume fraction.

K.Y. Maalawi / Composite Structures 93 (2011) 351359 Table 7 Results of the critical buckling pressure for balanced (88), AS-4 rings. Composite ring designation ADC-28 ADC-29 ADC-30 (h/R) pcr (MPa) Rasheed and Yousif [7] 1.2368 9.3965 30.0580 pcr,max (MPa) = 31.996x(10h/R)3 1.5385 11.6603 37.3180 Increase (%)

359

Acknowledgements The author dedicates this paper to his teacher, late professor Liviu Librescu of the ESM department at Virginia Tech. USA. Professor Librescu provided me, during our collaborative research at VT, with much fruitful scientic help in the eld of composite structures and aeroelastic stability. This help gave me the motivation to pursue the present research topic. The author also wishes to acknowledge the helpful discussions and papers from Professor Hayder Rasheed of the Department of Civil Engineering, Kansas State University, USA. References
[1] Daniel IM, Ishai O. Engineering mechanics of composite materials. 2nd ed. New York: Oxford, University Press; 2006. [2] Simitses GJ. Buckling of moderately thick laminated cylindrical shells: a review. Composites Part B 1996;27b:5817. [3] Sridharan S, Kasagi A. On the buckling and collapse of moderately thick composite cylinders under hydrostatic pressure. Composites Part B 1997;28b:58396. [4] Davies P, Chauchot P. Composites for marine applications part 2: underwater structures. Dordrecht: Kluwer Academic Publishers; 1999. [5] Graham D. Composite pressure hulls for deep ocean submersibles. Compos Struct 1995;32:33143. [6] Anastasiadis JS, Simitses GJ. Buckling of pressure-loaded, long, shear deformable cylindrical laminated shells. Comput Struct 1993;23:22131. [7] Rasheed HA, Yousif OH. Buckling of thin laminated orthotropic composite rings/long cylinders under external pressure. Int J Struct Stabil Dynam 2001;1(4):485507. [8] Rasheed HA, Yousif OH. Stability of anisotropic laminated rings and long cylinders subjected to external hydrostatic pressure. J Aerospace Eng 2005;18(3):12938. [9] Hodges DH. Non-linear inplane deformation and buckling of rings and high arches. Int J Non-Linear Mech 1999;34(4):72337. [10] Hodges DH, Harursampath D. Inplane buckling of anisotropic rings. In: 15th ASCE engineering mechanics conference. Columbia University, New York, June 25; 2002. [11] Librescu L, Maalawi KY. Material grading for improved aeroelastic stability in composite wings. J Mech Mater Struct 2007;2(7):10114. [12] Suresh S, Mortensen A. Fundamentals of functionally graded materials. Cambridge University Press; 1998. [13] Chen WH, Gibson RF. Property distribution determination of non-uniform composite beams from vibration response measurements and Galerkins method. J Appl Mech 1998;65:12733. [14] Chi SH, Chung YL. Mechanical behavior of functionally graded material plates under transverse load part I: analysis. Int J Solids Struct 2006;43:365774. [15] Arshad SH, Naeem MN, Sultane N. Frequency analysis of functionally graded material cylindrical shells with various volume fraction laws. J Mech Eng Sci 2007;221(12):148395. [16] Saha R, Maiti PR. Vibration characteristics of axially loaded rotating functionally graded cylindrical shell, 49th AIAA/ASME/ASCE/AHS/ASC structures, structural dynamics and materials, 710 April 2008, Schaumburg, IL, USA; 2008. [17] Maalawi KY. Buckling optimization of exible columns. Int J Solids Struct 2002;39:586576. [18] Maalawi KY. El-Chazly NM. Global optimization of multi-element beam-type structures. In: The 2nd international conference on advances in structural engineering and mechanics, ASEM02, Busan, South Korea, August 2123; 2002. [19] Chattopadhyay A, Ferreira J. Design sensitivity and optimization of composite cylinders. J Compos Eng 1993;3:16979. [20] Tanguy M, Mariusz P, Bernard G, Pierre C. Optimal laminations of thin underwater composite cylindrical vessels. J Compos Struct 2002;58(4):52937. [21] Maalawi KY. Optimal stability design of anisotropic rings/long cylinders under external pressure. J Mech Mater Struct 2008;3(4):77593. [22] Halpin JC, Tsai SW. Effects of environmental factors on composite materials, technical report AFML-TR-67-423, Dayton, OH; 1967. [23] Reddy JN. Mechanics of laminated composite plates and shells: theory and analysis. 2nd ed. NW: CRC Press LLC; 2004. [24] Simitses GJ. An introduction to the elastic stability of structures. Englewood Cliffs, NJ: Prentice Hall, Inc.; 1976. [25] Brush DO, Almroth BO. Buckling of bars, plates and shells. USA: McGraw-Hill Inc.; 1975. [26] Vanderplaats GN. Numerical optimization techniques for engineering design with applications. New York: McGraw-Hill; 1994. [27] Venkataraman P. Applied optimization with MATLAB programming. New York: John Wiley; 2002.

(1/27.5) (1/14.0) (1/9.50)

24.4 24.09 24.16

35

Max. Pcr x (h/R)3, GPa.

30 25 20 15 10 5 0 0 10 20 30 40 50 60 70 80 90

Carbon/epoxy E-glass/epoxy

Ply angle, degrees

Fig. 8. Variation of the absolute maximum buckling pressure with ply angle for balanced (h) rings under mass equality constraint.

6. Conclusions An efcient mathematical approach for enhancing the buckling stability limits of thin-walled anisotropic rings/long cylinders with radial material grading has been developed. The formulation of an optimal lamination design against buckling has been thoroughly investigated, where useful design charts are given for several types of anisotropic rings/long cylinders showing the functional dependence of the critical buckling pressure on the ber volume fractions, ply thickness and stacking sequence. Side constraints are imposed on the design variables in order to avoid having oddshaped optimized congurations with unrealistic values of the volume fractions and ply angles. An analytical buckling model has been implemented, which provides good sensitivity to lamination parameters, and allows the search for the needed optimal design in an acceptable computational time. The proposed model deals with dimensionless quantities in order to be applicable to thin shells with arbitrary thickness-to-radius ratios, which is a major contribution of this work. Results have indicated that the optimized laminations lead to signicant increases, always exceeding several tens of percent, of the buckling pressures with respect to the reference or baseline design. Types of composites considered include E-glass/epoxy, S-glass/epoxy, carbon/epoxy and E-glass/vinylester. It has been shown that the overall stability level of the laminated, radially graded composite shell structures under considerations can be substantially improved by nding the optimal ply thicknesses and ber volume fractions without violating both the mass equality constraint as well as any of the imposed side constraints. The stability limits of the optimized shells have been substantially enhanced as compared with those of the reference or baseline designs. Future aspects shall consider buckling optimization of cylindrical shells with material grading in both the circumferential and axial directions.