Você está na página 1de 15

Composite Structures 143 (2016) 272286

Contents lists available at ScienceDirect

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

Exact sensitivity analysis of stresses and lightweight design


of Timoshenko composite beams
Qimao Liu
Department of Civil Engineering, Aalto University, Espoo 02150, Finland

a r t i c l e

i n f o

Article history:
Available online 16 February 2016
Keywords:
Composite beam
Lightweight design
Exact sensitivity
Failure
Optimization

a b s t r a c t
The paper describes the novel optimization techniques for lightweight design of composite beams. The
optimization model is to find width and depth of composite beams to minimize the mass of beams under
the stiffness, strength and delamination failure constraints. The exact formulae for displacements, stresses and their sensitivities with respect to width and depth are derived using Timoshenko continuous
beam theory. The analytical stiffness, strength and delamination failure functions, and their gradients
are obtained using the exact expressions of displacements and stresses, and their sensitivities. The mass
and its gradient are also expressed analytically. The standard gradient-based nonlinear programming
algorithms are employed to solve lightweight design problems of composite beams. The lightweight
designs of composite beams are performed using the proposed optimization techniques. Three standard
gradient-based optimization methods (SQP, interior-point and active-set) using exact derivatives converge to the same lightweight design. However, gradient-based algorithms using finite difference sensitivities may not lead to optimal lightweight designs. It is necessary to develop exact sensitivity analysis
method instead of the difference methods for gradient-based algorithms.
2016 Elsevier Ltd. All rights reserved.

1. Introduction
Composite structures are gaining more and more successful
applications in aerospace, aircraft, automobile, train, naval, and
defence industries because of their high performance characteristics, such as high strength-to-weight ratio, high stiffnessto-weight ratio, superior fatigue properties and high corrosion
resistance [1,2]. For example, 50% of Boeings 787 Dreamliner has
been manufactured of advanced composites, specifically epoxy
and carbon fiber [3]. As the research, knowledge and confidence
on the composite structures increase, the composite structures
are gradually acting as the main load-carrying components, not
only as the secondary load-carrying components in the
engineering application. The beams are the major transversal
load-carrying member in the engineering structure systems. For
example, aircraft wing, helicopter blade, wind turbine blade, robot
arm, and space antenna are typical composite beam structures. The
floor beams of Boeings 787 are made of composite material. It is
the first commercial airplane to use composite floor beams.
From the viewpoint of the mechanics models of a composite
laminated beam, the composite laminated beams can be classified
Tel.: +358 (0) 503502879; fax: +358 (0) 947023758.
E-mail address: qimao.liu@aalto.fi
http://dx.doi.org/10.1016/j.compstruct.2016.02.028
0263-8223/ 2016 Elsevier Ltd. All rights reserved.

into EulerBernoulli beam model [4,5], EulerBernoulli beam


model with torsional rotation [6,7], Timoshenko beam model [8]
and Timoshenko beam model with torsional rotation [9,10].
EulerBernoulli beam model is only applied to the composite laminated beams with a big ratio of span to thickness. However,
Timoshenko beam model is applied to the composite laminated
beams with a small or big ratio of span to thickness. From the
viewpoint of the number of degree of freedom of a composite laminated beam, the composite laminated beams can be treated as
continuous composite beam models (infinite degree of freedom)
and discrete models (finite degree of freedom). The continuous
models can achieve the exact solution of the composite beams.
The discrete models (finite element models) can only obtain the
approximate solution of the beams. The finite element models
are believed to be easier implemented than the continuous composite beam models. It is often tedious and difficult to derive the
exact solution of a continuous composite beam model.
Many researchers have developed the analysis and optimal
design methods for the composite beam structures using discrete
models. For example, Blasques and Stolpe [11] performed the
maximum stiffness and minimum weight optimization of
laminated composite beams using finite element approach. The
fiber orientations and layer thicknesses are design variables.
Cardoso et al. [12,13] used finite element technique to deal with

273

Q. Liu / Composite Structures 143 (2016) 272286

design sensitivity and optimal design of composite thin-walled


laminated beams using torsion-bending beam model. Liu [14,15]
developed the analytical sensitivity analysis method for the composite structures based on the finite element method and lightweight design method using the analytical sensitivity. Neto et al.
[16] performed the sensitivity analysis and optimal design of composite beam structures using finite element solver FEAP. Sedaghati
et al. [17] developed a finite element model to study the mechanical and electrical behavior of laminated composite beam with
piezoelectric actuators and a design optimization methodology
by combining the finite element model and the sequential quadratic programming technique. Hamdaoui et al. [18] investigated an
optimal design approach for choosing the most suitable material
for high damping and low mass for a sandwich beam. Valido and
Cardoso [19] implemented the optimal design of the various geometrically nonlinear composite laminate beam structures, which
is based on finite element analysis and sensitivity analysis model.
Kim et al. [20] designed and manufactured the hybrid glass/carbon
composite bumper beam via the design optimization process combined with the impact analysis. Belingardi et al. [21] optimized
beam section profile and beam curvature for crashworthiness
using commercial finite element software ABAQUS. Blasques [22]
developed a methodology for simultaneous topology and material
optimization in optimal design of laminated composite beams with
eigenfrequency constraints. Blasques and Stolpe [23] described a
novel framework for simultaneous optimization of topology and
laminate properties in structural design of laminated composite
beam cross sections. The disadvantage of finite element analysis
is that the approximate numerical solution and sensitivity
obtained by the finite element models heavily depend on the mesh
schedules [14]. On the other hand, the finite element approaches
have much lower efficiency than the analytical approaches of the
continuous composite beam models.
Many researchers have developed the analytical approaches to
achieve the exact responses and sensitivities for the continuous
composite beam models. For example, the exact vibration frequencies of the continuous composite beams are achieved using the
EulerBernoulli beam model [47] and Timoshenko beam model
[810]. The analytical analysis methods and optimization design
of the continuous composite models using the different non
gradient-based algorithms (particle swarm algorithm and genetic
algorithm) for the thin-walled composite box-beam helicopter
rotor blades have been investigated [2426]. Liu [5] derived the
exact solutions and sensitivity of the first four frequencies using
the continuous composite model and developed the gradientbased algorithm to achieve the lightweight design of the solid composite laminated beams. Lentz and Armanios [27] described a
gradient-based optimization scheme for obtaining the maximum
coupling in thin-walled composite beams subject to hygrothermal
and frequency constraints. Roque and Martins [28] used differential evolution optimization to find the volume fraction that can
maximize the first natural frequency for a functionally graded
beam.
The optimization methods for the composite beam structures,
as mentioned above, can be classified into gradient-based and
non gradient-based algorithms. The non gradient-based algorithms
are also called random search algorithms. The random search algorithms can implement the optimization design without the gradient information. However, the gradient-based algorithms require
the gradient to construct the searching algorithm. Therefore, the
non gradient-based algorithms are easier to be carried out than
the gradient-based algorithms. However, the random search algorithms have to implement a large number of structural reanalysis
and may not find the optimum design if the sampling number is
not big enough. Compared with the random search algorithms,
the gradient-based algorithms are more efficient and can find the

optimum design (at least the local optimum) if the gradient can
be computed efficiently and accurately [2931]. Therefore, the sensitivity analysis is the key technique for the gradient-based algorithms. Today the works on the sensitivity analysis for composite
laminated beams are almost limited to the sensitivity analysis of
the frequency and deflection. A few work is reported on the sensitivity analysis of the stresses for composite laminated beams.
The aim of this paper is to develop the novel optimization techniques for the lightweight designs of Timoshenko composite laminated beams (continuous model) subjected to the static
loadings. The optimization model for the lightweight design is to
find the width and depth (or layer thickness) of the solid composite
beams to minimize the mass of the beams under the stiffness failure, strength failure and delamination failure constraints. The
paper is arranged as follows. In Section 2, the exact analytical
expressions of the displacements and stresses are derived using
the Timoshenko composite continuous beam model. In Section 3,
the exact analytical sensitivity formulae of deflections and stresses
are achieved by direct differentiation. In Section 4, the main failure
criteria for the composite laminated beams are analysed and
employed. In Section 5, the lightweight design optimization model
is formulated. In Section 6, the analytical gradients of objective
function (mass) and constraint functions are achieved by using
the sensitivity information of the deflections and stresses. In Section 7, the standard gradient-based nonlinear programming algorithms coded in Matlab, i.e., Sequential Quadratic Programming
(SQP), interior-point penalty algorithm and active-set algorithm,
are proposed to find the lightweight designs of Timoshenko composite laminated beams. In Section 8, the lightweight designs of
the composite laminated beams with different boundary conditions (pinned-pinned, fixed-fixed, fixed-free and fixed-pinned)
are performed using different optimization algorithms with the
exact sensitivities proposed in this paper. The finite difference sensitivities are also compared with the exact sensitivities proposed in
this paper. The lightweight designs of the composite beams
obtained by active set method using both exact sensitivities and
finite difference sensitivities are discussed.
2. Exact analytical displacements and stresses
2.1. Exact analytical displacements
Fig. 1 is a segment dx of composite laminated beam with rectangular cross section. The number of layers are denoted by
1; 2; . . . ; N and the fiber orientations of layers are denoted by
h1 , h2 ; . . . ; hN . The layered positions are denoted by z1 ; z2 ; . . . ; zN

z N +1

zN

z2
z1

h
2

(N )

zk +1
zI

( N 1)

(N)

( N 1)

(k )

(k )
(I )
( 2)

(I )

( 2)

(1)

(1)
dx

Fig. 1. Composite laminated beam (segment dx).

x
h
2

274

Q. Liu / Composite Structures 143 (2016) 272286

q ( x)

and zN1 . The width and thickness of the rectangular cross section
of the beams are denoted by b and h, respectively. Based on the
first-order shear deformation theory (also called Timoshenko beam
theory), the assumed displacement fields for the composite laminated beam are

ux; z u0 x z/x

wx; z w0 x

where u0 x and w0 x are the displacements of the points on the


middle plane along the x axis and z axis, respectively. /x is the
rotation of the normal to the middle plane about the y axis. ux; z
and wx; z are the in-plane displacement and deflection of the composite beam.
The strain and displacement relationships are given by the following equations.

ex

du0 x
d/x
z
dx
dx

cxz /x

dw0 x
dx

du0 x
d/x
bB11
Nx bA11
dx
dx
du0 x
d/x
bD11
dx
dx


dw0 x
Q x bKA55 /x
dx

M x bB11

N
X
 k zk1  zk
Q
11

dx

dNx
px 0
dx

12

dQ x
qx 0
dx

13

dMx
 Q x mx 0
dx

14

By substituting Eqs. (5)(7) into Eqs. (12)(14), we have


2

d u0 x

d u0 x
2

dx

A55

dx



dw0 x
mx 0
 bKA55 /x
dx

d u0 x
2

bB11

d /x

0
2
dx
dx
"
#
2
d/x d w0 x
bKA55

qx 0
2
dx
dx

bA11

d u0 x

bB11

dx

bD11

d /x
2

dx



dw0 x
 bKA55 /x
0
dx

18

19

20

Eq. (18) can be rewritten as

N
1X
 k z2  z2
Q
k
2 k1 11 k1

N
1X
 k z3  z3

Q
k
3 k1 11 k1

10

55

d /x

16

When only the transversal loading qx is applied to the composite laminated beam, Eqs. (15)(17) can be simplified as

d u0 x
2

N
X
 k zk1  zk

bD11

15

17

dx

D11

bB11

d /x

px 0
2
2
dx
dx
"
#
2
d/x d w0 x
qx 0
bKA55

2
dx
dx

bA11

k1

B11

Qx + dQx

Fig. 2. Free-body diagram.

bB11

N x + dN x

m ( x)

where Nx ,Mx and Q x are the normal force along the x axis, bending
moment about the y axis and shear force along the z axis on the
cross section of the composite laminated beam, respectively. The
subscript x indicates the plane of action of the internal forces. b is
the width of the beam. K is the shear correction factor, taken as
5/6 to account for the parabolic variation of transverse shear stress.
The extensional stiffness A11 , bending stiffness D11 , bendingextensional coupling stiffness B11 and extensional stiffness A55 of
the composite laminate are given by

A11

M x + dM x
p ( x)

Nx

where ex is normal strain and cxz is shear strain. All the other strain
components (i.e., ey , ez , cxy and cyz ) are zero.
The integrated constitutive equations can be written as

Qx

Mx

B11 d /x
A11 dx2

21

By substituting Eq. (21) into Eq. (20), we have

D11 

11

! 2


B211 d /x
dw0 x
0
 KA55 /x
2
dx
A11
dx

22

Differentiating Eq. (22) with respect to x leads to

!
"
#
3
2
B211 d /x
d/x d w0 x

b D11 
 bKA55
0
3
2
dx
A11
dx
dx

23

k1

where zk and zk1 are the position coordinates of the bottom and top
surfaces of the kth lamina, as shown in Fig. 1. The stiffness coeffi k of the kth lamina in the laminate coordinate system
cients Q
ij

can be computed by using Eq. (60) in Section 2.2.


When a segment dx of composite laminated beam is subjected
to the distributed normal force px, transversal force qx and
moment mx, as shown in Fig. 2. The equilibrium equations of a
segment dx are as follows.

By substituting Eq. (19) into Eq. (23), we obtain


3

d /x
3

dx

A11
bB211  A11 D11

qx

24

Integrating Eq. (24) leads to


2

d /x
2

dx

A11
bB211  A11 D11

Z


qxdx

25

275

Q. Liu / Composite Structures 143 (2016) 272286

where qxdx produces one indefinite integration constant C 1 .


Thereafter the other formulae have the same integration constant
R
(i.e., C 1 ) when qxdx is in a square bracket.


u0 x B

We integrate Eq. (25) and have

d/x
A11

dx
bB211  A11 D11

ZZ


qxdxdx

26

RR


where
qxdxdx produces two indefinite integration constants
C 1 and C 2 . Thereafter the other formulae have the same integration
RR
constants (i.e., C 1 and C 2 ) when qxdxdx is in a square bracket.
Further integrating Eq. (26) leads to

/x

ZZZ

A11
bB211  A11 D11


qxdxdxdx

27

RRR
where qxdxdxdx produces three indefinite integration constant C 1 , C 2 and C 3 . Thereafter the other formulae have the same
RRR
integration constants (i.e., C 1 , C 2 and C 3 ) when qxdxdxdxis in
a square bracket.
By substituting Eq. (25) into Eq. (21) and integrating, we have
2

d u0 x
2

dx

Z

B11
bB211  A11 D11

du0 x
B11

dx
bB211  A11 D11
u0 x 

B11
bB211

 A11 D11


qxdx

Z Z

ZZ Z

28


qxdx dx

29


qxdx dxdx

30

The integral term in Eq. (30) produces three indefinite integration constants, i.e., C 1 , C 4 and C 5 , where C 1 is the same indefinite
integration constant as that of Eq. (25), and produced by
R
qxdx .
By substituting Eq. (26) into Eq. (19) and integrating, we obtain
2

d w0 x
2

dx

1
A11

qx 
2
bKA55
bB11  A11 D11


/x A

ZZZ

ZZ


qxdxdx

31

ZZ
)
Z (
dw0 x
1
A11


qx 
qxdxdx dx
dx
bKA55
bB211  A11 D11


qxdxdxdx

ZZ Z

ZZ 
w0 x

37


qxdx dxdx



Cqx
A

ZZ


qxdxdx dxdx

38

39

The integral expressions of displacements and rotation in Eqs.


(37)(39) can be applied to any integrable loading forms. In this
paper, we take the uniform distributed loading as an example to
find the analytical expressions of the displacements and rotation.
When qx q0 , Eqs. (37)(39) lead to the following explicit
formulae.



 q 0 x3 1 C 1 x 2 C 2 x C 3
/x A
2
6

40



 q 0 x3 1 C 1 x 2 C 4 x C 5
u0 x B
2
6

41



 1 x3  C
 2 x2 C 6 x C 7
 q0 x4  1 AC
 q0 1 AC
w0 x A
6
24
2 2

42

The indefinite integration constants, i.e., C 1 , C 2 , C 3 , C 4 , C 5 , C 6 and


C 7 , can be determined using the boundary conditions of the composite beams. The internal forces also need to be used to express
the boundary conditions. The internal forces can be computed as
follows.
By substituting Eqs. (41) and (40) into Eq. (5), we have




 q0  bA11 B
  bA11 B
 q0 x2 bB11 A
 C1x
Nx bB11 A
2
3
 2  bA11 BC
 4
bB11 AC

43

By substituting Eqs. (41) and (40) into Eq. (6), we obtain




 q0  bB11 B
  bB11 B
 q0 x2 bD11 A
 C1x
Mx bD11 A
2
3
 2  bB11 BC
 4
bD11 AC

44

By substituting Eqs. (42) and (40) into Eq. (7), we have

 3 bKA55 C 6
 bKA55 x bKA55 AC
Q x Cq
0

45

32
w0 x

ZZ (


1
A11
qx 
bKA55
bB211  A11 D11

ZZ

)
qxdxdx dxdx
33

The four indefinite integration constants of Eq. (33) are C 1 , C 2 ,


C 6 and C 7 , where C 1 and C 2 are the same indefinite integration conRR

stants as those of Eq. (26), and produced by
qxdxdx .
We define the following three parameters so that the formulae
become more concise and convenient.


A

B

C

A11
bB211  A11 D11
B11
bB211  A11 D11
1
bKA55

34

35

36

The displacements, i.e., Eqs. (30) and (33), and rotation, i.e., Eq.
(27) can be written as

2.2. Exact analytical stresses


The nonzero strain components can also be achieved. By substituting Eqs. (41) and (40) into Eq. (3), we have


0 2
 q 0 x2 C 1 x C 2
x C 1 x C 4 zA
3
2

ex B

46

By substituting Eqs. (42) and (40) into Eq. (4), we obtain



q0 3 1
 1 x2
 q0 x3  1 AC
x C 1 x2 C 2 x C 3  A
2
2
6
6

 2 x C 6
 AC
 Cq

cxz A

47

The material coordinate system (123) and beam coordinate system/laminate coordinate system (xyz) are shown in Fig. 3. The
angle between the x axis and 1 axis is hk (counter clockwise).
The 1 axis is along the direction of the fiber. In this paper, the
notations of stress and strain are the same as those in Jones text
book [1].
Therefore, the plane stress components in the material coordinate system (123) are

276

Q. Liu / Composite Structures 143 (2016) 272286

 k Q cos4 hk 2Q 2Q sin2 hk cos2 hk Q sin4 hk


Q
11
12
66
22
11

y
1

z=3

 k Q Q  4Q sin2 hk cos2 hk Q sin4 hk cos4 hk


Q
11
22
66
12
12
x

(k )

 k Q sin4 hk 2Q 2Q sin2 hk cos2 hk Q cos4 hk


Q
11
12
66
22
22
 k Q  Q  2Q sinhk cos3 hk
Q
11
12
66
16
Q 12  Q 22 2Q 66 sin hk cos hk
3

Fig. 3. The material coordinate system (123) and laminate coordinate system (xyz).

r1

B
C
k  k B
@ r2 A T Q
@

s12

 k Q  Q  2Q sin3 hk coshk
Q
11
12
66
26

ex
ey C
A; zk1 6 z 6 zk
cxy

48

Q 12  Q 22 2Q 66 sinhk cos3 hk
 k Q Q  2Q  2Q sin2 hk cos2 hk
Q
11
22
12
66
66
Q 66 sin hk cos4 hk
4

ey = 0, cxy = 0, and

where

cos2 hk

6
T k 4

sin hk

sin hk

cos2 hk

 sin h

cos h

sin h

cos h

 k Q cos2 hk Q sin2 hk
Q
44
55
44

7
2 sin hk cos hk 5

 k Q  Q coshk sinhk
Q
55
44
45

2 sin hk cos hk

cos2 h

 sin h

49
2

 k
Q
12
 k
Q
22
 k
Q
26

 k
Q
6 11
6Q
 k
4 12
 k
Q
16

 k
Q

 k
Q
16
 k
Q
26
 k
Q
66

3
7
7
5

50

r1
C
rp1 B
@ r2 A
s12
0

ex

51

C
epx B
@0A

52

0
We rewrite Eq. (48) as

rp1 T k Q k epx ; zk1 6 z 6 zk

53

The shear stress components in the material coordinate systems

s23
 k cyz ; zk1 6 z 6 zk
T sk Q
s
cxz
s13

where cyz = 0 and

"

T k
s

sin hk
cos h

"
 k
Q
s

 k
Q
44
 k
Q
45

cos hk

54

cs

s23
s13
0

where the stiffness coefficients Q ij of the lamina in material coordinate system are

E1
m12 E2
; Q 12
;
1  m12 m21
1  m12 m21
G12 ; Q 44 G23 ; Q 55 G13

 sin hk

where E1 and E2 are the modulus of elasticity, m12 and m21 are the
Poissons ratio. G12 , G23 and G13 are the shear modulus, of the composite lamina.
The indefinite integration constants are derived by using the
different boundary conditions (pinned-pinned, fixed-fixed, fixedfree and fixed-pined), i.e.,
(1) Pinned-pinned
The boundary conditions, i.e., u0 0 0, w0 0 0, N x 0 0,
Mx 0 0, w0 L 0, M x L 0, Q x 2L 0, are used to determine
the indefinite integration constants. The seven indefinite integration constants are

C1

 L
  3D11 Aq
2B11 B
0
;


6D11 A  B11 B
C 5 0;

C6

C 2 0;

C3

  B11 B

3D11 A
3

 q0 L ;
72D11 A  B11 B


  3D11 A
B11 B
 3 1

 Aq0 L 2 Cq0 L;
72D11 A  B11 B

C7 0
62

The boundary conditions, i.e., u0 0 0, w0 0 0, /0 0,



u0 L 0, w0 L 0, /L 0, u0 2L 0, are employed to determine the indefinite integration constants. The seven indefinite
integration constants are

57

1
1
q L2 ;
C 1  q0 L; C 2
2
12 0
1
C 6 Cq
L; C 7 0
2 0


58

C 3 0;

C4

1
q L2 ;
12 0

C 5 0;
63

(3) Fixed-pinned

So that we can rewrite Eq. (54) as

59

 k of the kth lamina in the laminate coorThe stiffness coefficients Q


ij
dinate system in Eqs. (50) and (56) are

61

56

ss T sk Q sk cs ; zk1 6 z 6 zk

E2
1  m12 m21

(2) Fixed-fixed

55

cxz

Q 22

55

#
k


Q
45
 k
Q

Q 66

C 4 0;

We also denote the shear stress and shear strain as

ss

60

Q 11

We use the following notations to depict the plane stress and


strain in the material coordinate systems.

 k Q sin2 hk Q cos2 hk
Q
44
55
55

The boundary conditions, i.e., u0 0 0, w0 0 0, /0 0,


0, are used to determine
w0 L 0, N x L 0, M x L 0, Q x 5L
8
the indefinite integration constants. The seven indefinite integration constants are

277

Q. Liu / Composite Structures 143 (2016) 272286

 L2  3Cq
 L2 3Cq


5Aq
Aq
0
0
0
0

;
C
; C 3 0;
2


8AL
8A
2


7Aq0 L 9Cq0
5
; C 5 0; C 6 Cq
L; C 7 0
C4

8 0
24A

By substituting Eqs. (67) and (68) into Eqs. (8)(11), we have

C1

64

(4) Cantilever/Fixed-free

A11

 11 Lbq  3Lq  2B11 BLbq



 2 bq
3AD
3L2 q0  B11 BL
0
0
0
0
; C2

 11 ;


6bB11 B  AD11
6bB11 B  AD
 11 L2 bq
3L2 q0  AD
0

C 3 0; C 4
 11 ; C 5 0; C 6 Cq0 L; C 7 0
  AD
6bB11 B

C1

B11

D11

N
1X
 k
Q

3 k1 11

65

3.1. Option of design variables

66

where d1 b and d2 h.
3.2. First derivatives of A11 , B11 , D11 , and A55
The deflection of the beam, i.e., Eq. (42), the strain components,
i.e., Eqs. (46) and (47), and the indefinite integration constants, i.e.,
Eqs. (62)(65), show it is necessary to obtain the first derivatives of
A11 , B11 , D11 , and A55 so that the first derivatives of the deflection
and stresses can be achieved. In this paper, the thickness of each
lamina are supposed to be the same. The position coordinates of
the bottom and top surfaces of the kth lamina can be expressed
used the depth of the beam, i.e.,

h h
2k  N  2
h
zk  k  1
2 N
2N
h h
2k  N
zk1  k 1  1
h
2 N
2N

"
3 
3 #
2k  N
2k  N  2
h 
h
2N
2N

71



N
N
X
X
 k 2k  N h  2k  N  2 h h
 k
Q
Q
55
55
2N
2N
N
k1
k1

72

The first derivatives of the stiffness coefficients A11 , B11 , D11 , and
A55 with respect to design variables are obtained by direct
differentiation.

73

k1

The design variables of the composite laminated beam can be


the fiber orientations, width and depth (thickness of layer) of the
beam, or even the fiber volume fractions. Since the Timoshenko
beam model in this paper is one dimensional beam, it is not necessary to optimize the fiber orientations because all the fiber orientations will go to zero to obtain the maximum bending stiffness.
In this paper, the stiffness failure, strength failure and delamination failure are included in the optimization mathematical model
to achieve the lightweight design of the composite laminated
beams. As pointed out in the authors previous work [5], the
strength of the lamina will change when the fiber volume fraction
of the lamina (layer) varies in the optimization process. Today the
strength of the lamina is often determined by using experiment
methods. It is still difficult to accurately predict the strength of
the lamina using any mechanics theories [1], although the moduli
of elasticity can be accurately calculated using micromechanics
theory. Therefore, when the strength failure and delamination failure are included in the optimization mathematical model, it is
unpractical to choose the fiber volume fractions as the design variables. Therefore, the width and depth of the beam are chosen as the
design variables to achieve the lightweight designs of the composite laminated beams, i.e.,
T

70

8
when di b
> 0;
@A11 < X
1 N  k
>
@di
Q 11 ; when di h
:N

3. Exact analytical sensitivity of deflections and stresses

d b h

A55

69

"
2 
2 #
2k  N
2k  N  2
h 
h
2N
2N

N
1X
 k

Q
2 k1 11

The boundary conditions, i.e., u0 0 0, w0 0 0, /0 0,


Mx 0  12 q0 L2 , N x L 0, M x L 0, Q x L 0, are employed to
determine the indefinite integration constants. The seven indefinite integration constants are



N
N
X
X
 k 2k  N  2k  N  2 h h
 k
Q
Q
11
11
2N
2N
N
k1
k1

67

8
when di b
> 0;
@B11 < X
N
h
i


 k 2kN 2  2kN2 2 h; when di h


>
@di
Q
:
11
2N
2N
k1

8
when di b
> 0;
@D11 < X
N
h
i


 k 2kN 3  2kN2 3 h2 ; when di h


>
Q
@di
:
11
2N
2N

75

k1

8
when di b
> 0;
@A55 < X
N
1
 k ; when di h
>N
Q
@di
:
55

76

k1

3.3. First derivatives of deflections and stresses


Since the first derivatives of A11 , B11 , D11 , and A55 with respect to
design variables have been computed in Section 3.2, the first
derivatives of deflections and stresses with respect to design variables can be obtained by direct differentiating Eqs. (42), (53) and
(59). We have

 

 q
@w0 x
@A
1 @A
0 4
 @C 1 x3

C1 A
x 
@di
@di 24
6 @di
@di
 


@ C q0 1 @ A
1  @C 2 2 @C 6
@C 7
x

C2 A
x

@di 2 2 @di
2 @di
@di
@di
 k
@ rp1 @T k  k
@Q
@e

epx T k Q k px ;
Q epx T k
@di
@di
@di
@di

77

zk1 6 z 6 zk
78

where the element

@ ex
@di

in

@ epx
@di

is





q
@ ex
@B
0 2
 @C 1 x @C 4

x C1x C4  B
@di 3
@di
@di
@di




q
@A
@C
@C 2
1
0 2

x
z
x C 1 x C 2 zA
@di 2
@di
@di
@e

The elements @dyi and

68

74

@ cxy
@di

in

@ epx
@di

79
k

 k

Q
are equal to zero. @T@di and @@d
can
i

be obtained by direct differentiating Eqs. (49) and (50) with the


design variable di .

278

Q. Liu / Composite Structures 143 (2016) 272286

s
@ ss @T sk  k
@Q
@c

c T sk Q sk s ;
Q c T sk
@di
@di s s
@di s
@di
k

r1 r2
r2 s2 X c  X t
Y  Yt
 p 2 12

r1 c
r2 1
Xt Xc
Xt Xc
YtYc
X t X c Y t Y c Y t Y c S2
r21

zk1 6 z 6 zk
80

where the element

@ cxz
@di

in

@ cs
@di

is


 q
@ cxz @ A
1
0 3

x C 1 x2 C 2 x C 3
@di 6
2
@di


 q

1
@C
@C 2
@C 3
@A
1 @A
1 2
0 3

A

x
x
C 1 x2
x 
2 @di
@di 6
2 @di
@di
@di
 


1  @C 1 2
@C
@A
 @C 2 x @C 6
 A
x 
q0
C2 A
2 @di
@di
@di
@di
@di
The element

@ cyz
@di

in

@ cs
@di

is equal to zero.

@T s
@di

and


@Q
s
@di

81
can be

obtained by direct differentiating Eqs. (55) and (56) with the


design variable di .
The analytical expressions of

 @B


@A
, , @C
@di @di @di

and

@C j
@di

(j 1; 2;    7) in

Eqs. (77)(81) can be derived using Eqs. (73)(76).

where X t and X c are the tensile strength and compressive strength


along the 1 axis of the material coordinate system, respectively. Y t
and Y c are the tensile strength and compressive strength along the 2
axis of the material coordinate system, respectively. S is the shear
strength on the plane 102 of the material coordinate system. The
stress state (r1 , r2 and s12 ) are depicted in the material coordinate
system 123, as shown in Fig. 3.
4.3. Stiffness failure criterion
In the serviceability, if the deflection of the composite laminated beam is greater than the serviceability limit. The stiffness
failure is supposed to happen. Therefore, the deflection of the
composite laminated beam is not allowed to be greater than
the serviceability limit. We have the following stiffness failure
criterion.


w0 6 w
4. Failure criteria of a composite laminated beam

Delamination, the separation of two adjacent layers in the


composite laminates, is one of the most critical failure modes
in composite laminated structures. Debonding between the
adjacent layers depends on the stresses acting on that interface, i.e., the normal stress component r3 and the two shear
stresses s12 and s23 . The researchers have developed many
delamination failure criteria to predict the delamination failure
of the composite laminated structures. In this paper, the
delamination failure criterion by Yeh and Kim [32] is adopted
to predict when the delamination will occur. The tensile
delamination occurs if

r3

ZT

s13

S13

s23
S23

2
P 1;

when

r3 > 0

82

where S12 and S23 are the shear strength on the planes 103 and 203
of the material coordinate system 123. Z T is the tensile strength
along the 3 axis of the material coordinate system.The shear delamination occurs if

s13
S13

s23
S23

2
P 1;

when

r3 < 0

85

 is the serviceability limit on the deflection of the composwhere w


ite laminated beams.

4.1. Delamination failure criterion

84

83

In this paper, we let S13 S23 . For a composite laminated beam


subjected to the transversal loading, the normal stress component
on the plane 102 of the material coordinate system is less than
zero, i.e., r3 < 0, therefore, the failure is shear delamination, not
tensile delamination.
4.2. Tsai-Wu strength failure criterion
Strength failure is also one of the most critical failure modes in
composite laminated structures. Over the last five decades, there
have been continuous efforts in developing strength failure criteria
for composite laminated structures. Today, a large number of lamina failure criteria and laminate failure analysis methods have been
developed, such as Tsai-Hill, Hoffman and Tsai-Wu failure criteria.
In this paper, Tsai-Wu failure criterion is used to keep strength failure away from the design since Tsai-Wu strength failure criterion
gives prediction that range from acceptable to excellent when it
is compared with test results.
The Tsai-Wu strength failure criterion has the following
form [33].

5. Lightweight design optimization model


Considering the stiffness failure criterion, strength failure criterion and delamination failure criterion, we have the lightweight
design optimization model of composite beams as follows.
T

Find d b h
Minimize Wd


r2
r2
s2
r1 r2
Y c Y t
t
Subject to g j X t X1 c  p
r

r
1
<0
Y t Y2 c S122 XXctX
1
2
Xc
Yt Yc
fj

s213
S213

s223

Xt Xc Y t Y c

S2 1 < 0
23

0 6 0
r w0 aL  w

b6b6b

h6h6h
86
where Wd is the mass of a composite laminated beam. aL is the
location (x coordinate) where the deflection of the beam is monitored. a is different for different boundary conditions, i.e., PP:
a 1=2, FF: a 1=2, FP: a 505=873, CL: a 1. b and b are the
 are the
lower and upper limits on the width of the beam. h and h
lower and upper limits on the depth of the beam, respectively. In
this paper, g, f and r are called strength failure function, delamination failure function and stiffness failure function, respectively. The
subscript (j 1; 2; . . . ; N m :) means the jth monitored point of the
strength and delamination, and Nm is the total number of the monitored points. It is noted that the analytical expressions of the
strength failure function g, delamination failure function f and stiffness failure function r are naturally obtained by just substituting
the analytical stresses (r1 , r2 , s12 , s23 and s13 ) and deflection
w0 aL in their right hand sides.
When the composite laminated beams are subjected to the uniform distributed loading, the maximum of the shear force Q x is at
x 0 or x L for PP, FF, FP and CL. The maximum of the bending
moment M x is at x L=2 for PP and FF, at x 5L=8 for FP, at
x 0 for CL. Therefore, we have the monitored points for strength
failure and delamination failure, as shown in Fig. 4 for PP and FF,
Fig. 5 for FP and Fig. 6 for CL. Double circles means two monitored
points for the adjacent layers since their fiber orientations are not
the same.

279

Q. Liu / Composite Structures 143 (2016) 272286


Table 2
Material properties of lamina.

L
4

L
8

L
8

L
8

L
8

L
4

Fig. 4. Monitored points (Nm = 36) for strength failure and delamination failure (PP
and FF).

Property

T300/5208

E1 (GPa)
E2 E3 (GPa)
G12 G13 (GPa)
G23 (GPa)
m12 m13

m23
q (kg/m3)

136.00
9.80
4.70
5.20
0.28
0.15
1540

X t (MPa)
X c (MPa)
Y t (MPa)
Y c (MPa)
S (MPa)
S13 S23 (MPa)

1550
1090
59
59
75
75

Table 3
Deflection and sensitivity of deflection at monitored points (initial design).

L
4

L
8

L
8

L
8

L
8

L
4

Fig. 5. Monitored points (Nm = 36) for strength failure and delamination failure
(FP).

Beam type
(location)

PP
(x L=2)

FF
(x L=2)

FP
(x 505L=873)

CL
(x L)

w0 (m)
@w0 =@b
@w0 =@h

0.0145
0.0484
0.0863

0.0038
0.0126
0.0191

0.0068
0.0228
0.0375

0.1334
0.4445
0.8153

The first derivatives of the mass with respect to the design variables are

L
4

L
8

L
8

L
8

L
8

L
4

@W
hLq
@b

88

@W
bLq
@h

89

We have the analytical gradient of the objective functions as

rWT
Fig. 6. Monitored points (Nm = 12) for strength failure and delamination failure
(CL).

 @W


@W T
@h

@b

90

6.2. Analytical gradients of constraint functions


The first derivatives of the strength failure function with respect
to design variables can be achieved by direct differentiation.

Table 1
Main frame for nonlinear programming algorithms in Matlab.
%% Choose optimization algorithm using gradient
% SQP or interior-point or active-set
options = optimset(GradObj, on, GradConstr, on, Algorithm, sqp);
%% Implement optimization algorithms
 @mycon, options);
d = fmincon(@myfun, d , [], [], [], [], d, d,
0

% d0 is the initial design.



@g
2r1 @ r1
1
@ r1
@ r2
2r2 @ r2

 p r2
r1
@di X t X c @di
@di
@di
Y t Y c @di
Xt Xc Y t Y c
2s12 @ s12 X c  X t @ r1 Y c  Y t @ r2
2

@di
X t X c @di
Y t Y c @di
S

91

We have the analytical gradient of strength failure function as

 are the lower and upper limits on the design variables.


% d and d
%% Computing objective and its gradient
[mass, gradient_mass] = myfun(d);
%% Computing constraints and their gradients
[constraint, gradient_constraints] = mycon(d);

rgT

 @g

@b


@g T
@h

92

The first derivatives of the delamination failure function with


respect to design variables can also be achieved by direct
differentiation.

6. Analytical gradients of objective function and constraint


functions
6.1. Analytical gradient of objective function

@f
2s13 @ s13 2s23 @ s23
 2
 2
@di
S12 @di
S23 @di

93

The analytical gradient of delamination failure function is


T

rf 

@f
@b

@f
@h

iT

94

The mass of the composite laminated beam can be calculated as

W bhLq
where q is the density of the composite material.

87

7. Nonlinear programming algorithms


Since the analytical expressions and analytical gradients of the
objective function and constrained functions have been achieved,

280

Q. Liu / Composite Structures 143 (2016) 272286

Table 4
Normal stresses and sensitivities of normal stresses (initial design).
Beam type (location)

PP (x L=2, z h=2)

FF (x L=2, z h=2)

FP (x 5L=8, z h=2)

CL (x 0, z h=2)

r1 (MPa)
@ r1 =@b (MPa.m1)
@ r1 =@h (MPa.m1)
r2 (MPa)
@ r2 =@b (MPa.m1)
@ r2 =@h (MPa.m1)

88.13
293.77
550.81
1.78
5.93
11.11

29.38
97.92
183.60
0.59
1.98
3.70

45.18
150.61
277.74
0.91
3.04
5.60

352.52
1175.10
2203.30
7.11
23.71
44.45

Table 5
Shear stresses and sensitivities of shear stresses (initial design) (at the bottom of the 4th layer).
Beam type (location)

PP (x = 0, z = 0)

FF (x = 0, z = 0)

FP (x = 0, z = 0)

CL (x = 0, z = 0)

s12 (MPa)
@ s12 =@b (MPa.m1)
@ s12 =@h (MPa.m1)
s23 (MPa)
@ s23 =@b (MPa.m1)
@ s23 =@h (MPa.m1)
s13 (MPa)
@ s13 =@b (m1)
@ s13 =@h (m1)

0
0
0
2.01
6.71
4.20
2.23
7.43
4.64

0
0
0
2.01
6.71
4.20
2.23
7.43
4.65

0
0
0
2.52
8.39
5.25
2.79
9.29
5.80

0
0
0
4.03
13.43
8.39
4.46
14.86
9.29

Fig. 7. Pinned-pinned beam (initial design).

Q. Liu / Composite Structures 143 (2016) 272286

281

Fig. 8. Fixed-fixed beam (initial design).

any standard gradient-based nonlinear programming algorithms


can be used to solve the lightweight design optimization model,
i.e., Eq. (86). In this paper, the standard gradient-based nonlinear
programming algorithms coded in Matlab, i.e., Sequential Quadratic Programming (SQP), interior-point penalty algorithm and
active-set algorithm, are used to find the lightweight designs of
Timoshenko composite laminated beams. The main frame to
implement the nonlinear programming algorithms in Matlab is
shown in Table 1.
8. Examples
The span of the composite laminated beams (PP, FF, FP and CL)
are L = 7.2 m. The composite beams are subjected to the uniform
distributed loading with intensity q0 = 105 N/m. The beams have
the layered number N = 8. The material properties of the composite
lamina is shown in Table 2. The given fiber orientations are
[0/90/45/45]s in the examples. The initial design of the composite
laminated beams are b = 0.3 m and h = 0.48 m (the thickness of
each layer is 0.06 m), mass W = 1597 kg. The geometric design
 = 2 m, h = 0.2 m and h
 = 2 m. The serviceability
space is b = 0.1 m, b

 = 0.01 m in Eq. (86). All the computations


limit of the deflection w
are implemented using the same computer (Processor: Intel(R)
Core(TM) i5-2320 CPU @ 3.00 GHz, RAM: 8 GB).
8.1. Sensitivity of deflection and stresses
The deflection and its first derivatives with respect to the width
and depth of the composite laminated beam can be calculated
using the exact analytical formulae Eqs. (42) and (77), respectively.
The maximum deflections and their first derivatives with respect
to design variables are shown in Table 3. The stresses in the material coordinate system can be computed using the exact analytical
formulae Eqs. (48) and (54). The first derivatives of the stresses in
the material coordinate system can be achieved using the exact
analytical formulae Eqs. (78) and (80). The normal and shear stresses, their first derivatives with respect to design variables at the
certain monitored points are listed in Tables 4 and 5.
Compared with the stress, the deflection has more intuitively
for the readers to see the deformation behavior of the composite
laminated beams. All the deflections and their first derivatives with
respect to the width and depth of the initial design with different

282

Q. Liu / Composite Structures 143 (2016) 272286

Fig. 9. Fixed-pinned beam (initial design).

boundary conditions are plotted in Figs. 710. Since the allowable


 = 0.01 m, we can see in Table 1 and Figs. 7(a) and 10
deflection is w
(a) that the initial beams with pinned-pinned and fixed-free
boundary conditions are infeasible design. It is noted that the initial design is not necessary to be feasible design for the three optimization methods, i.e., sequential quadratic programming,
interior-point penalty algorithm and active-set algorithm, in this
paper. The sensitivity information in Figs. 7(b), 8(b), 9(b) and 10
(b) indicates that, for the initial design, the depth is more sensitive
to the deflection than the width of the beam.
8.2. Lightweight designs of composite beams
In the section, the lightweight designs of Timoshenko composite laminated beams with pinned-pinned, fixed-fixed, fixed-pinned
and fixed-free boundary conditions are achieved by using the optimization techniques proposed in this paper. The three standard
gradient-based optimization methods coded in Matlab, i.e.,
Sequential Quadratic Programming method (SQP), interior-point
penalty method (interior-point) and active set method (activeset), are employed to achieve the lightweight designs (optimal
designs) of the composite laminated beams. The lightweight

designs are shown in Table 6 by SQP, Table 7 by interior-point


and Table 8 by active set. The computational effort of the three
optimization methods are given in Tables 68.
The computational time to calculate the objective function, constrained functions and their gradients (one time) is shown in
Table 9.
The lightweight designs of the composite laminated beams with
PP, FF, FP and CL boundary conditions in Tables 68 indicate that
the three gradient-based optimization methods converge to the
same lightweight design. Therefore, any one of the three
gradient-based optimization methods can be employed to find
the lightweight designs of the composite laminated beams. However, the iterations and function counts of the three optimization
methods are different. The SQP use the least iteration and function
count to achieve the lightweight designs. However, the CPU time
depends not only on the iteration and function count, but also on
the computational efficiency of objective function, constrained
functions and their gradients. For example, the computational time
of objective function, constrained functions and their gradients
(one time) for FF beam, as shown in Table 9, is much less than
those of PP, FP and CL beams. Even though the iteration and function count to achieve the lightweight designs of FF beam are

283

Q. Liu / Composite Structures 143 (2016) 272286

Fig. 10. Fixed-free beam (initial design).

Table 6
Lightweight designs of Timoshenko composite laminated beams (SQP).
Beam type

PP

FF

FP

CL

Optimal design

Mass W (kg)
Width b (m)
Depth h (m)
Layer thickness h=N (m)

909.2634
0.1000
0.8200
0.1025

560.7427
0.1000
0.5057
0.0632

706.5145
0.1000
0.6372
0.0796

2065
0.1000
1.8623
0.2328

Computational effort

Iteration and
function count
CPU time (s)

7
7
45

8
9
6

6
7
30

9
11
32

Table 7
Lightweight designs of Timoshenko composite laminated beams (interior-point).
Beam type

PP

FF

FP

CL

Optimal design

Mass W (kg)
Width b (m)
Depth h (m)
Layer thickness h=N (m)

909.2634
0.1000
0.8200
0.1025

560.7427
0.1000
0.5057
0.0632

706.5145
0.1000
0.6372
0.0796

2065
0.1000
1.8623
0.2328

Computational effort

Iteration and
function count
CPU time (s)

11
17
60

13
32
11

12
16
38

11
13
21

284

Q. Liu / Composite Structures 143 (2016) 272286

Table 8
Lightweight designs of Timoshenko composite laminated beams (active-set).
Beam type

PP

FF

FP

CL

Optimal design

Mass W (kg)
Width b (m)
Depth h (m)
Layer thickness h=N (m)

909.2634
0.1000
0.8200
0.1025

560.7427
0.1000
0.5057
0.0632

706.5145
0.1000
0.6372
0.0796

2065
0.1000
1.8623
0.2328

Computational effort

Iteration and
function count
CPU time (s)

7
15
52

24
163
56

18
74
174

9
17
27

Table 9
Computational time for objective function, constrained functions and their gradients.
Beam type

PP

FF

FP

CL

Computational time (s)

3.5

0.3

2.4

1.6

lightweight designs, the stiffness failure is completely active


constraints (equal to zero), the strength failure and delamination
failure are not really active (not equal to zero). This numerical
 increases, it is still
results mean if the allowable deflection w
possible to obtain the lighter weight designs of composite beams.
8.3. Comparison with central difference method

Table 10
Deflection and sensitivity of deflection (lightweight design).
Beam type
(location)

PP
(x L=2)

FF
(x L=2)

FP
(x 505L=873)

CL
(x L)

w0 (m)
@w0 =@b
@w0 =@h

0.0100
0.1000
0.0319

0.0100
0.1000
0.0470

0.0100
0.1000
0.0382

0.0100
0.1000
0.0125

greater than those of PP, FP and CL beams (as shown in Table 7


more obviously), the CPU time to achieve the lightweight design
of FF beam is far less than those of PP, FP and CL beams. This indicate that the high efficient sensitivity and gradient computation
methods can greatly improve the efficiency of the optimization
methods. It is very important to the lightweight design of the large
scale composite beam structure system. Therefore, the computational efficiency of the sensitivity and gradient is often called the
bottle neck for the gradient-based optimization methods in structural optimization. The analytical sensitivity and gradient methods
are often more efficient than the numerical sensitivity and gradient
analysis methods. However, it is often tedious and of heavy work
to derive the exact analytical expression formulae of sensitivity
and gradient. It should be noted that all the exact analytical
expression formulae of sensitivity and gradient for the composite
laminated beams in this paper are derived by using the Symbolic
Math Toolbox of Matlab R2015a.
The maximum deflection and its first derivatives with respect to
the width and depth of the lightweight design are listed in Table 10.
The deflections, as shown in Table 10, indicate that the maximum
deflection of the lightweight designs is equal to the allowable
deflection. The sensitivity of the deflection in Table 10 indicates
that, for the lightweight designs, the width is more sensitive to
the deflection than the depth of the composite beams. Thats
because the width of the lightweight design beams has reached
the lower limit of the width.
The numerical value/scope of the constrained functions of lightweight designs at all the monitored points are listed in Table 11.
The numerical value/scope in Table 11 shows that, for the

In this section, the Central Difference Method (CDM) is


employed to compute the first derivatives of the responses of
Timoshenko composite laminated beams with respect to the width
and depth. The central difference method has higher accuracy to
obtain the sensitivity than both forward and backward difference
methods. The sensitivity obtained by the central difference method
is used in the gradient-based optimization methods in this paper.
Due to space limitation, the sensitivity analysis and lightweight
design of the composite beam with fixed-fixed boundary condition
using CDM and active set method are implemented in this paper.
The sensitivities of deflection with respect to width and depth
are shown in Figs. 11 and 12, respectively. The lightweight designs
of Timoshenko composite laminated beam are listed in Table 12.
The results in Figs. 11 and 12 indicate that the sensitivities of
deflection with respect to width and depth obtained by the central
difference method are going to the exact solution by the proposed
analytical method in this paper while the difference step size is
getting smaller and smaller. The lightweight designs using exact
derivatives and central difference derivatives (Table 12) show that
the active set method using difference derivatives may not achieve
the optimal design (the mass the optimum is 675.1210 kg using
difference derivatives, 560.7427 kg using exact derivatives),
although the computational efficiency of active set method using
difference derivatives is equivalent to that of active set method
using exact derivatives (CPU times are about 56 s). Therefore, it
is necessary to develop the exact sensitivity analysis method
instead of the difference methods for the gradient-based optimization algorithms.
9. Conclusion
This paper developed the lightweight design methods for the
composite laminated beams subjected to static loading. It should
be noted that the proposed lightweight design methods are only
applied to the continuous design variables. When the design
variables are discrete, the evolutionary algorithms such as

Table 11
Numerical value/scope of constrained functions of lightweight designs at monitored points.
Beam type

PP

FF

FP

CL

Stiffness failure function r


Strength failure function g j

0.00000
0.97168
1.02101
0.99455
1.00000

0.00647
0.99137
1.01482
0.98193
1.00361

0.00000
0.97297
1.02987
0.98591
1.00000

0.00000
0.99433
1.01693
0.97867
0.99799

Delamination failure function f j

285

Q. Liu / Composite Structures 143 (2016) 272286

Fig. 11. Sensitivity of deflection with respect to width.

Fig. 12. Sensitivity of deflection with respect to depth.

Table 12
Lightweight designs using exact derivatives and difference derivatives (active-set).
Exact derivatives

Difference derivatives (db b=5; dh h=5)

Optimal design of composite beam (FF)

Mass W (kg)
Width b (m)
Depth h (m)
Layer thickness (m)

560.7427
0.1000
0.5057
0.0632

675.1210
0.1371
0.4442
0.0555

Computational
effort

Iteration and
function count
CPU time (s)

24
163
56

17
201
56

Sensitivity analysis using

Genetic Algorithms and Swarm Optimization Algorithms can be


employed to solve optimization problems with discrete design
variables. The main contributions and conclusions of this work
are as follows.
(1) The exact analytical formulae of displacements and stresses
of the composite laminated beams with pinned-pinned,
fixed-fixed, fixed-pinned and fixed-free boundary conditions
are derived using Timoshenko beam theory (continuous
beam model).

(2) The exact analytical sensitivity formulae of deflection and


stresses, with respect to the width and depth (thickness of
layer), of the composite laminated beams are derived using
direct differentiation.
(3) The lightweight design optimization model considering stiffness failure criterion, strength failure criterion and delamination failure criterion is proposed in this paper.
(4) The analytical expression formulae for the mass, stiffness
failure function, strength failure functions, delamination
failure functions and their analytical gradients are derived

286

Q. Liu / Composite Structures 143 (2016) 272286

using the exact analytical deflection, stresses and their analytical sensitivities with respect to the width and depth
(thickness of layer) of the beams.
(5) Three standard gradient-based optimization methods, i.e.,
sequential quadratic programming algorithm, interiorpoint penalty algorithm and active-set algorithm, coded in
Matlab, are proposed to the lightweight designs of the composite laminated beams. The lightweight designs of the composite laminated beams with pinned-pinned, fixed-fixed,
fixed-pinned and fixed-free boundary conditions are
performed.
(6) The three standard gradient-based optimization methods
(SQP, interior-point and active-set) converge to the same
lightweight design of the composite laminated beams. Any
one of the three gradient-based optimization methods can
be employed to find the lightweight designs of the composite laminated beams. The efficiency of the three optimization
methods depends not only on the iteration and function
count, but also on the computational efficiency of objective
function, constrained functions and their gradients. The high
efficient sensitivity and gradient computation methods can
greatly improve the efficiency of the gradient-based optimization methods.
(7) The optimization algorithms using the difference sensitivities may not lead to the optimal lightweight designs of the
composite beams. It is necessary to develop the exact sensitivity analysis method instead of the difference methods for
the gradient-based optimization algorithms.

Acknowledgements
The author would like to thank the reviewers for their constructive comments.
References
[1] Jones RM. Mechanics of composite materials. 2nd ed. Philadelphia (PA): Taylor
& Francis; 1999.
[2] Reddy JN. Mechanics of laminated composite plates and shells: theory and
analysis. 2nd ed. Washington (DC): CRC Press; 2003.
[3] Hale J. Boeing 787 from the ground up. boeing.com/commercial/aeromagazine
2006; 24(4): 17-23. <http://www.boeing.com/commercial/aeromagazine/
articles/qtr_4_06/article_04_2.html>.
[4] Vinson JR, Sierakowski RL. Behavior of structures composed of composite
materials. 2nd ed. Secaucus, NJ, USA: Kluwer Academic Publishers; 2002.
[5] Liu Q. Analytical sensitivity analysis of eigenvalues and lightweight design of
composite laminated beams. Compos Struct 2015;134:91826.
[6] Dancila DS, Armanios EA. The influence of coupling on the free vibration of
anisotropic thin-walled closed-section beams. Int J Solids Struct
1998;35:310519.
[7] Banerjee JR. Explicit analytical expressions for frequency equation and mode
shape of composite beams. Int J Solids Struct 2001;38:241526.

[8] Chandrashekhara K, Krishnamurthy K, Roy S. Free vibration of composite


beams including rotary inertia and shear deformation. Compos Struct
1990;14:26979.
[9] Chandrashekhara K, Krishnamurthy K, Wu WZB. Analytical solution to
vibration of generally layered composite beams. J Sound Vib 1992;159
(1):8599.
[10] Banerjee JR. Frequency equation and mode shape formulae for composite
Timoshenko beams. Compos Struct 2001;51:3818.
[11] Blasques JP, Stolpe M. Maximum stiffness and minimum weight optimization
of laminated composite beams using continuous fibre angle. Struct Multidiscip
Optim 2011;43:57388.
[12] Cardoso JB, Sousa LG, Castro JA, Valido AJ. Optimal design of laminated
composite beam structures. Struct Multidiscip Optim 2001;24:20511.
[13] Cardoso JB, Valido AJ. Cross-section optimal design of composite laminated
thin-walled beams. Comput Struct 2011;89:106976.
[14] Liu Q. Analytical sensitivity analysis of frequencies and modes for composite
laminated structures. Int J Mech Sci 2015;90:25877.
[15] Liu Q, Paavola J. Lightweight design of composite laminated structures with
frequency constraint. Compos Struct 2015. http://dx.doi.org/10.1016/
j.compstruct.2015.08.116.
[16] Netoa MA, Yu W, Leal RP. Generalized Timoshenko modelling of composite
beam structures: sensitivity analysis and optimal design. Eng Optim 2008;40
(10):891906.
[17] Sedaghati R, Zabihollah A, Ahari M. Sensitivity analysis and optimal design of
smart piezolaminated composite beams. AIAA J 2006;44(12):298796.
[18] Hamdaoui M, Jrad M, Daya EM. Optimal design of frequency dependent threelayered rectangular composite beams for low mass and high damping. Compos
Struct 2015;120:17482.
[19] Valido AJ, Cardoso JB. Geometrically nonlinear composite beam structures:
optimization design. Eng Optim 2003;35(5):55360.
[20] Kim DH, Kim HG, Kim HS. Design optimization and manufacture of hybrid
glass/carbon fiber reinforced composite bumper beam for automobile vehicle.
Compos Struct 2015;131:74252.
[21] Belingardi G, Beyene AT, Koricho EG. Geometrical optimization of bumper
beam profile made of pultruded composite by numerical simulation. Compos
Struct 2013;102:21725.
[22] Blasques JP. Multi-material topology optimization of laminated composite
beams with eigenfrequency constraints. Compos Struct 2014;111:4555.
[23] Blasques JP, Stolpe M. Multi-material topology optimization of laminated
composite beam cross sections. Compos Struct 2012;94:327889.
[24] Suresh S, Sujit PB, Rao AK. Particle swarm optimization approach for multiobjective composite box-beam design. Compos Struct 2007;81:598605.
[25] Murugan MS, Suresh S, Ganguli R, Mani V. Target vector optimization of
composite box beam using real-coded genetic algorithm: a decomposition
approach. Struct Multidiscip Optim 2007;33:13146.
[26] Kathiravan R, Ganguli R. Strength design of composite beam using gradient
and particle swarm optimization. Compos Struct 2007;81:4719.
[27] Lentz WK, Armanios E. Optimum coupling in thin-walled, closed-section
composite beams. J Aerosp Eng 1998;11:819.
[28] Roque CMC, Martins PALS. Differential evolution for optimization of
functionally graded beams. Compos Struct 2015;133:11917.
[29] Liu Q, Paavola J, Zhang Jing. Shape and cross-section optimization of plane
truss structures subjected to earthquake excitation using gradient and Hessian
matrix calculations. Mech Adv Mater Struct 2016;23(2):15669.
[30] Ganguli R, Chopra I. Aeroelastic optimization of a helicopter rotor with
composite coupling. J Aircr 1995;32(6):132634.
[31] Ganguli R, Chopra I. Aeroelastic optimization of a helicopter rotor with twocell composite blades. AIAA J 1996;34(4):83541.
[32] Yeh HY, Kim CH. The Yeh-Stratton criterion for composite materials. J Compos
Mater 1994;28(10):92639.
[33] Tsai SW, Wu EM. A general theory of strength for anisotropic materials. J
Compos Mater 1971;5(1):5880.

Você também pode gostar