Composite Structures 307 (2023) 116627

Contents lists available at ScienceDirect

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

Nonlinear dynamic analysis of unsymmetric layered composite shells

Alireza Beheshti *, Reza Ansari
Faculty of Mechanical Engineering, University of Guilan, Rasht, Iran

A R T I C L E I N F O A B S T R A C T

Keywords: The concentration of the present study is on the transient dynamic analysis of composite shells given geometric
Dynamic analysis nonlinearity. For this purpose, use is made of a 12-parameter shell model suited for thick composite structures
Shell and then strains and stresses are extracted. The so-called weak form is constructed based on the Hamilton’s
Finite deformation
principle and for gaining the solution to the dynamic problem, the nonlinear finite element method, the total
Layered composites
Lagrangian scheme, as well as the composite time integration approach are employed. To overcome the nu­
merical anomalies pertaining to the four-node shell element, transverse shear, membrane and curvatures-
thickness locking, appropriate approaches are taken into account. Finally, some problems are solved to eval­
uate the newly developed shell element and a comparison with the data available in the literature is made.

1. Introduction Kommineni [7]. Vu-Quoc and Tan [8] devised a solid-shell element for
multi-layer composite structures and a comparison among various im­
The application of composite materials in industries is fast-growing plicit time integration schemes was made. Naboulsi and Palazotto [9]
on the account of unique characteristics of this class of materials, say presented a nonlinear finite element formulation for the dynamic anal­
lightness, cheapness and strength. Also, structures made of the artificial ysis of composite shells based on various shell models including
material have many commercial applications and they are subjected to simplified large rotation, large rotation and displacement and Jaumann
dynamic loads frequently. Additionally, in many practical deformations, analysis of general shells. In order to study the hygrothermo-mechanical
displacements are large, but strains are small and it implies a geometric dynamic analysis of composite shells considering geometric nonline­
nonlinearity. Consequently, a rigorous investigation of nonlinear dy­ arity, use was made of Newmark method in conjunction with the iter­
namic behavior of composite shells is of central importance [1,2]. ative Newton-Raphson scheme by Naidu and Sinha [10]. Next,
In the present paragraph, an effort is made to introduce all the works geometrically nonlinear transient analysis of doubly curved shells was
done in the area of nonlinear dynamic analysis of multi-layer composite carried out by Kundu and Sinha [11] via development of a 9-node shell
shells. Chao and Reddy [3] took advantage of a 3D degenerated element element and utilization of the classical Newmark time integration
to analyze the transient dynamic analysis of plates and shells. Nonlinear method. Rahman et al. [12] developed a reduced-order finite element
dynamic buckling of composite shells subjected to internal hydrostatic formulation for dynamic buckling of composite shells using perturbation
pressure was addressed by Schokker et al. [4]. In this work, the so-called method. By employing the corotational kinematic model and an implicit
Newmark time integration scheme and the Newton-Raphson approach time integration scheme conserving energy approximately, Almeida and
were employed to solve the problem. Dynamic transient analysis of Awruch [13] studied dynamics of composite structures via a flat trian­
shells was studied by To and Wang [5] via introduction of a flat three- gular shell element.
node triangular element and use of trapezoidal rule associated with There are well-documented evidences on accuracy and robustness of
the time integration based on the first-order shear deformation model. the standard Newmark scheme in solving time-dependent linear prob­
Ganapathi and Varadan [6] relied on the von-Karman strains for lems. However, in nonlinear problems the scheme generally does not
considering geometric nonlinearity and the Newmark time integration satisfy the equations of conservation of angular momentum and energy
scheme in order to study the deformation of layered composite shells and it is the reason why some new approaches have been developed over
employing the finite element method. A finite element investigation into years. In this work, we rely on the composite time integration scheme of
nonlinear dynamic behavior of composite shells allowing for von Kar­ Bathe and Baig [14]. The scheme is computationally more expensive
man strains and a higher-order shell model was conducted by Kant and than the Newmark approach, however the converged solution is

* Corresponding author.
E-mail addresses: [email protected] (A. Beheshti), [email protected] (R. Ansari).

https://doi.org/10.1016/j.compstruct.2022.116627
Received 30 July 2022; Received in revised form 3 December 2022; Accepted 25 December 2022
Available online 29 December 2022
0263-8223/© 2022 Elsevier Ltd. All rights reserved.
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

obtained using larger time increment and the point balances the In what follows, Greek letters take the values of 1 and 2, whereas the
computational time and makes the scheme more robust than the clas­ Latin letters accept 1, 2 and 3. In the formulation of the shell structure,
sical Newmark method. Also, it is worth noting that the superiority of there exists a need for covariant base vectors in the curvilinear coordi­
the approach compared to the Newmark scheme has been checked in nate system. They are defined in the spatial and material configurations,
[14] by solving some typical examples. The alternatives to the classical respectively, by
trapezoidal approach (Newmark) are energy conserving scheme and the
∂x ∂X
generalized α-method proposed by Simo and Tarnow [15] and Chung gi = and Gi = i (4)
∂ξi ∂ξ
and Hulbert [16], respectively.
The novelty of the present article is to employ the 12-parameter shell By bearing Eqs. (1) and (2) in mind, the above vectors can be
model devised originally for unsymmetric multilayer composite plates rewritten as follows:
by Kant and Manjunatha [17] and the composite time integration ⎧ (0) (1) (2) (3)
scheme developed by Bathe and Baig [14] which has the superiority to ⎨ g = ∂x + ξ3 ∂d + ξ32 ∂d + ξ33 ∂d
⎪
(5)
α
the conventional Newmark time integration scheme in nonlinear ana­ ∂ξα ∂ξα ∂ξα ∂ξα
⎪
⎩
lyses. To authors’ knowledge it is the first work in which the robust time 2
g3 = d(1) + 2ξ3 d(2) + 3ξ3 d(3)
integration scheme is combined with the higher-order shell model in
⎧
order to study the transient dynamic analysis of layered composite (0)
⎨ G = ∂ X + ξ3 ∂ D
⎪
shells. It is also worth mentioning that the classical shell models α
∂ξα ∂ξα (6)
(Kirchhoff-Love type and Reissner–Mindlin type theories) can be used to ⎪
⎩
G3 = D
estimate the behavior of anisotropic shells, but the classical supposition
that fibers initially normal to shell midsurface remain straight during the In the present work, base vectors of surfaces for which ξ3 = 0 are
deformation is loose in modeling the mechanical behavior of shells made employed in the development of equations. They are
of layered composite materials. Comparison of simulations based on the ⎧
structural mechanics with the 3D-elasticity results indicated that for
(0)
⎨ aα = gα|ξ3 =0 = ∂x α
⎪
unsymmetric thick shell the twelve parameter shell/plate model pro­ ∂ξ (7)
⎪
vides us with accurate results and it is the reason why the higher-order ⎩
a3 = g3|ξ3 =0 = d(1)
kinematic model is employed in the present work. The main aim of the
current work is to study the geometric nonlinearity of the composite ⎧
(0)
⎨ A = G 3 = ∂X
⎪
structures based on a transient dynamic analysis. However, features like
(8)
α α|ξ =0
∂ξα
the material nonlinearity and the viscoelastic effects are not considered. ⎪
⎩
G3 = D
The arrangement of the contribution is as follows. Upon an intro­
duction of dynamic analysis of shells, the kinematics of the 12-param­ The deformation gradient tensor in the convective curvilinear coor­
eter shell model is presented in Sect. 2. Then, the constitutive dinate system is expressed by
equation of the composite shell is introduced in Sect. 3. After presenting
the finite element formulation of the shell structure considering time- F = gi ⊗ Gi (9)
dependent parameters in Sect. 4, the focus of Sect. 5 is on the solution where Gi are contravariant base vectors in the Lagrangian configu­
of some standard numerical examples. Finally, Sect. 6 is concerned with
ration and are defined through Gi .Gj = δji where δji is the so-called Kro­
the conclusion of the current study.
necker delta. It is worth noting that the spatial contravariant base
vectors are accomplished using the same procedure. Next, with the
2. Kinematics
knowledge of definition of the right Cauchy-Green deformation tensor as
well as Eq. (9), we have
For description of the shell deformation based on the single-layer
shell scheme herein, use is made of a curvilinear coordinate system C = FT F = gij Gi ⊗ Gj (10)
with in-plane components of ξ1 and ξ2 and the out-of-plane parameter of
ξ3 ranging from –h/2 to h/2 which h is the uniform thickness of the shell where gij = gi .gj are components of the spatial metric tensor g.
in material configuration. In this work, the Eulerian position vector is Then, with the aid of the shifter Z = Gi ⊗ Ai = A − ξ3 B where A =
written on the basis of twelve parameters, while the Lagrangian one is Ai ⊗ Ai is middle-surface identity tensor and B = − Aα,3 ⊗ Aα is referred
similar to the classics. Accordingly, the spatial and material position to as the curvature tensor, one is able to extract the relation Gi = Z− T Ai
vectors can be expressed by and subsequently C = Z− T CẐ − 1 where C ̂ = gij Ai ⊗ Aj . Given the geo­
2
x = x(0) + ξ3 d(1) + ξ3 d(2) + ξ3 d(3)
3
(1) metric nonlinearity addressed in this work, use is made of the Green-
Lagrange strain tensor E = (C − G)/2 where G is the material metric
X = X(0) + ξ3 D (2) tensor. Similar to the right Cauchy-Green deformation tensor, the strain
tensor can be written based on material middle-surface base vectors as
where x and X describe the working planes in Lagrangian and
(0) (0)
̂ = Eij Ai ⊗ Aj with Eij = (gij − Gij )/2.
E
Eulerian configurations and are obtained by setting ξ3 = 0. Notice that By bearing in mind the deformation field for the 12-parameter shell
the spatial plane, described by x(0) , is not essentially the middle surface model, the adjusted right Cauchy-Green deformation tensor can be
of the shell in the deformed configuration and fibers normal to the expanded with respect to the thickness parameter in the following form:
material mid-surface do not remain basically straight over the defor­ ∑6
̂ = p
̂ (p) (11)
mation. Furthermore, d(i) i ∈ {1, 2, 3} are director vectors of the C p=0
ξ3 C
employed shell model and D is the normal vector to the middle surface in
the undeformed configuration. In addition, the displacement field based ̂ = g(p) Ai ⊗ Aj p ∈ {0, 1, 2, 3, 4, 5, 6} and the associated
where C
(p)
ij
on the above equations can be expressed by components are
( )
(3)
2 3
x(0) − X(0) + ξ3 d(1) − D + ξ3 d(2) + ξ3 d(3)
u = ⏟̅̅̅̅̅̅⏞⏞̅̅̅̅̅̅⏟ g(0) (0) (0)
αβ = x,α .x,β
⏟̅̅̅̅̅̅̅⏞⏞̅̅̅̅̅̅̅⏟
u(0) w
(1) (1)
(1) (0)
gαβ = x(0)
,α .d,β + d,α .x,β

2
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

g(2) (0) (2) (1) (1) (2) (0)

αβ = x,α .d,β + d,α .d,β + d,α .x,β ∂Ws ∂Ws
S= and ℂ = (21)
∂E ∂E∂E
(3) (3) (2) (1) (0)
where Ws is the energy density function. The focus in the present
(1) (2) (3)
gαβ = x(0)
,α .d,β + d,α .d,β + d,α .d,β + d,α .x,β
study is on the multi-layer single-family fiber reinforced composites and
g(4) (1) (3) (2) (2) (3) (1)
αβ = d,α .d,β + d,α .d,β + d,α .d,β it is supposed that there exists a curvilinear coordinate system defined in
1 2 3
the undeformed configuration with the parameters ξ , ξ and ξ in which
g(5) axes are normal to each other and the last parameter coincides with the
(2) (3) (3) (2)
αβ = d,α .d,β + d,α .d,β
thickness parameter. The constants of the local material elasticity tensor
g(6) (3) (3)
αβ = d,α .d,β (12) are described by [18].
1111 1 − ν23 ν32 2222 1 − ν13 ν31 3333 1 − ν12 ν21
(0)
g3α = d(1) .x,β
(0) C = , C = , C = ,
ΔE2 E3 ΔE1 E3 ΔE1 E2

g(1) (1) (1) (2) (0)

3α = d .d,α + 2d .x,α C
1212
= G12 , C
1313
= G31 , C
2323
= G23 ,

(2) ν21 + ν31 ν23 ν31 + ν21 ν32 ν32 + ν12 ν31
g3α = d(1) .d(2) (22)
(2) (1) (3) (0) 1122 1133 2233
,α + 2d .d,α + 3d .x,α C = , C = , C = ,
ΔE2 E3 ΔE2 E3 ΔE1 E3
g(3) (1) (3) (2) (2) (3) (1)
3α = d .d,α + 2d .d,α + 3d .d,α where
/
1 − ν12 ν21 − ν23 ν23 − ν31 ν13 − 2ν21 ν32 ν13
g(4) (2) (3) (3) (2)
3α = 2d .d,α + 3d .d,α
Δ= with νij = νji Ei Ej for i ∕
=j
E1 E2 E3
(23)
g(5) (3) (3)
3α = 3d .d,α (13)
-
It is worth noting that the overscript refers to the local material
(0) (1) (1) coordinate system. Next, by considering the orientation of the base
g33 = d .d
vectors with respect to the local ones, one is able to accomplish the
components of the elasticity tensor C.
g(1) (1) (2)
33 = 2d .d + 2d(2) .d(1)

4. Finite element procedure

g(2) (1) (3)
33 = 3d .d + 4d(2) .d(2) + 3d(3) .d(1)
In this section, the finite element formulation of the 12-parameter
g(3) (2) (3)
33 = 6d .d + 6d(3) .d(2)
shell model is addressed based on the dynamic analysis.

g(4) (3) (3)

33 = 9d .d (14)
4.1. Hamilton’s principle
Furthermore, the non-zero components of the material metric tensor,
∑ To construct the integral form of dynamic equilibrium equations of a
G = 6p=0 ξ3 G
p
̂ ̂ (p) , are expressed by
deformable body, one is able to employ the Hamilton’s principle which
Gαβ = G(0) 3 (1) 3 (2) 2
(15) can be written as follows:
αβ + ξ Gαβ + ξ Gαβ
∫ t2
δ [K − Π]dt = 0 (24)
G3α = Gα3 = G(0) 3 (1)
3α + ξ G3α (16) t1

where K and Π stand for, respectively, the kinetic and total potential
G33 = G(0) (17)
33
energies. The variation of the parameters with respect to the material
where configuration can be expressed by
∫
G(0)
αβ = X(0) (0) (1) (0) (0) (2)
,α .X,β , Gαβ = X,α .D,β + X,β .D,α , Gαβ = D,α .D,β (18) δK =
∂u ∂u
ρ0 .δ dV (25)
V ∂t ∂t
(19)
(0) (1)
G3α = X(0)
,α .D, G3α = D,α .D ∫ ∫ ∫
δΠ = S : δEdV − f 0 .δudV − T.δudA (26)
(20)
V V A
G(0)
33 = D.D
where ρ0 and V are the mass density and the volume of the body in
̂ = ∑6 ξ3 p E
Finally, the adjusted strain tensor can be written as E ̂ (p) the undeformed state. Also, f 0 is the body force per unit referential
p=0

̂ (p) − G
̂ (p) )/2. volume and T is the material traction vector applied to surface A.
where E
̂ (p) = ( C

4.2. Space discretization

3. Constitutive equation

Based on the tradition of the finite element method, the geometry

Now, let us focus on the constitutive equation for the stress tensor.
under study, a surface herein, is chopped into smaller regions referred to
The fourth-order elasticity tensor, C, is used to extract the stress field
with the knowledge of the strain field through ΔS = C : ΔE. In this as elements such that S0 ≈ S0h = Ane=1
e
S0e where ne stands for the total
work, similar to kinematic tensors, the adjusted form of the elasticity number of elements. Using bilinear interpolation functions consistent
̂ = Cijkl Ai ⊗ Aj ⊗ Ak ⊗ Al , is applied which is obtained through with the four-node element developed in the present work, the geo­
tensor, ℂ
metric parameters and variables are approximated employing
Δ̂S =C ̂ From the nonlinear elasticity, the second Piola-Kirchhoff
̂ : Δ E.
〈{ (0) } 〉 〈{ (0) } 〉 ∑4 〈{ } }
stress and the elasticity tensors are obtained from the following re­ X , {D} ≈ Xh , {Dh } = (0)
[NA ] XA , {DA 〉 (27)
lations:
A=1

3
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

where ̂ S = Sij Ai ⊗ Aj . Now, by bearing in mind dV = det[Z]dSdξ3 the first

term on the right hand side of Eq. (26) can be written as
∫ ∫ ∫ (∑ )
6 (I)
S : δE dV = ̂ ̂
S : δ E dV = ̂ ̂
S (I) : δ E dS
I=0
V V S

where S is the area of undeformed reference surface and the stress

resultants are introduced by
∫
(33)
p
̂ (I) =
S ̂
Sξ3 det[Z]dξ3 I ∈ {0, 1, 2, 3, 4, 5, 6}
h

Fig. 1. Node numbers and tying points for a typical shell element. Notice that the determinant of the shifter is obtained from
2
〈{ } { (1) } { (2) } { (3) } { (0) } 〉 det[Z] = 1 − 2ξ3 H +ξ3 K with H = Bαα /2 and K = det[Bαβ ].
As for the last term on right hand side of Eq. (26) for the case in
(0)
x , d , d , d , u , {w}
〈{ } { } { } { } { } 〉
≈
(0) (1) (2)
xh , dh , dh , dh , uh , {wh }
(3) (0) which the top surface of shell is under the traction of T+ , it can expanded
∑4 〈{ } { } { } { } { } } as follows:
(28)
(0) (1) (2) (3) (0)
= [NA ] xA , dA , dA , dA , uA , {wA 〉 ∫ ∫
A=1 ( )
T.δxdS+ = T(0) .δx(0) + T(1) .δd(1) + T(2) .δd(2) +T(3) .δd(3) dS (34)
[NA ] is a diagonal matrix including standard bilinear shape functions S+ S

and the subscript A refers to the node number A. √̅̅̅̅̅̅̅̅̅̅̅̅ √̅̅̅̅̅̅̅ ( +

with T(I) = T+ (h/2)I G+ /A I ∈ {0, 1, 2, 3} where G+ = G1 ×
The displacement-based low-order shell element addressed in this ) √̅̅̅̅
G+ .D and A is obtained in a similar way given the reference base
work suffers from the transverse shear and membrane locking and ANS 2
vectors. Also, it is supposed that there is no body force.
(assumed natural strain) method is adopted to circumvent the numerical
In the finite element formulation, it is customary to convert all
anomalies. Thus, the strains are extracted from the following relations
tensor-based relations to the matrix-based counterparts. For this pur­
[19,20].
pose, the stress resultant and strain vectors are introduced as below:
⎧
⎪ (0) ANS 1 [( ) (0) A ( ) (0) C ]
{ } { 11 22 33 12 13 23 }T
⎪
⎨ E13 = 1 − ξ2 E13 + 1 + ξ2 E13
2 ̂
S (I) = S(I) , S(I) , S(I) , S(I) , S(I) , S(I) (35)
(29)
⎪
⎪
⎩ E23(0) ANS 1 [( ) (0) D ( ) (0) B ]
= 1 − ξ1 E23 + 1 + ξ1 E23 { } { }T
2 ̂ (I) (36)
(I) (I) (I) (I) (I) (I)
E = E11 , E22 , E33 , 2E12 , 2E13 , 2E23

Now, by bearing in mind the strains introduced previously in Sect. 2,

⎧
⎪ (0) ANS 1( 2
)
(0) A
( 2
)
(0) B the variational strain vector can be approximated in the following form:
⎪
⎪
⎪ E11 = 1− 2aB − ξ2 +2aB ξ2 E11 +aC − 1+ξ2 E22
⎪
⎪
⎪
2 { (I) } ∑4 [ (I) ]
⎪ ̂
⎪
⎪
⎪ 1( 2 22
)
(0) C
(
22
)
(0) D
(
22
)
(0) E
δE = A=1
BA {δuA } with {δuA }
⎪ +2 1− 2aA +ξ +2aA ξ E11 +aD − 1+ξ E22 +aE − 1+ξ E12
⎪
⎪ {{ }T { }T { }T }T
⎪
⎪
, {δwA }T , δdA (37)
⎪ (0) (2) (3)
⎪
⎪ ( ) 1( ) = δuA , δdA
⎪ (0) ANS 2 (0) A 2 (0) B
⎪
⎨ E22 = aB − 1+ξ1 E22 + 1− 2aC +ξ1 +2aC ξ1 E22
2 [ ]
⎪
⎪
(
12
)
(0) C 1( 12
)
(0) D
(
12
)
(0) E where B(I)
A is strain–displacement matrix and {δuA } is the variation
⎪ 1
⎪ +aA − 1+ξ E11 + 2 1− 2aD − ξ +2aD ξ E22 +aE − 1+ξ E12
⎪
⎪
⎪
⎪ of the generalized displacement vector. With the aid of the above
⎪
⎪
⎪
⎪
⎪ (0) ANS
E12
1(
= − ξ1 +4aB ξ1 ξ2 E11
) (0) A 1 ( 2 ) (0) B
+ ξ +4aC ξ1 ξ2 E22 approximation, the variation of the nodal internal force can be expressed
⎪
⎪
⎪
⎪ 4 4 by
⎪
⎪
⎪ ( ) (0) C 1 ( ) (0) D ( ) (0) E
⎩ +1 ξ1 +4aA ξ1 ξ2 E11
⎪ + − ξ2 +4aD ξ1 ξ2 E22 + 1+4aE ξ1 ξ2 E12 { int } ∑6
∫ ([ ]T { }
)
4 4 fA = I=0
B (I)
A
̂
S (I) dS e (38)
(30) Se

Then, by direct application of approximated fields in Eq. (34), the

where in fact the strain fields are interpolated using new functions
nodal external force can be accomplished as
based on the related values at the tying points, illustrated in Fig. 1. In the
∫ ({
studies of Choi and Paik [21,22], Kulikov and Plotnikova [23,24] and { ext } { }T { }T { }T { }T } )
fA = T(0) [NA ], T(1) [NA ], T(2) [NA ], T(3) [NA ]T dSe
Roh and Cho [25] only linear interpolations for in-plane strains are Se
presented to cure membrane locking in 4-node shell elements. The (39)
nonlinear interpolations of the strains proposed in Kulikov and Plotni­
kova [26] are more suitable for finite element implementations because Next, the variation of the total potential energy for a typical element
they allow the element to pass the bending test. Due to the direct can be expressed by
interpolation of the main director in this work, the element is prone to ∑4 ( { } { ext }) ( { } { ext })
δΠe = {δuA }T f int − fA = {δUe }T f int − fe (40)
the curvature-thickness locking and the remedy of the assumed strain A=1 A e

scheme is used for the thickness strain as follows [27]: { } { }

where f inte and f ext e are element internal and external force vec­
(0) AS
∑4 ( ) (0) ( 1 2 ) { }T
E33 = A=1
NA ξ1 , ξ2 E33 ξA , ξA (31) tors and {δUe } = {δu1 }T , {δu2 }T , {δu3 }T , {δu4 }T .
Finally, the variation of the total potential energy can be written as
( { } { ext } )
4.3. Internal and external load vectors δΠ = Ane=1
e
{δUe }T f inte − fe = {δU}T ({F(U) } − {P}) (41)

where {δU} = {{δU1 }, {δU2 }, ⋯, {δUne } }T is variation of global

In the definition of the strain tensor, it was shown that it can be
generalized displacement vector. Also, {F} and {P} are total internal
described with respect to the base vectors of the material mid-surface by
and external force vectors.
taking advantage of the shifter Z. Thus, one is able to write S = Z ̂SZT

4
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

4.4. Mass matrix { } { } [{ } { }]

γΔt
U̇n+γ = U̇n + Ün + Ün+γ (48)
The first step in the extraction of the consistent mass matrix based on 2
the finite element method is to rewrite the variation of the kinetic energy { } ( )2 [{ } { }]
{ } γΔt
in the following form: Un+γ = {Un } + γΔt U̇n + Ün + Ün+γ (49)
2
∫
∂2 u
δK = − ρ0 2 .δudV (42) where the overscript ⋅ stands for the derivative with respect to time
V ∂t and
By inserting the displacement field presented in Eq. (3) and then { } ( )2 [ { }] { }
2 { }
taking integration along the thickness parameter and converting the Ün+γ = Un+γ − {Un } − γΔt U̇n − Ün (50)
γΔt
above relation into the matrix form we have
∫ { } Now, by bearing Eq. (47) in mind, the equilibrium equation at tn+γ
δK = − {δU}T [m] U ¨ dS (43) can be written as
S
{ }
{ { }T { }T }T { ( )} { }
{ }T [M] Ün+γ + F Un+γ = Pn+γ (51)
where {U} = u(0) , {w}T , d(2) , d(3) and

By taking advantage of Eqs. (49) and (50) and a systematic lineari­

⎡ [I0 ] [I1 ] [I2 ] [I3 ] ⎤ ⎡ ⎤ zation process based on the Newton-Raphson scheme, the incremental
∫
⎢ [I1 ] [I2 ] [I3 ] [I4 ] ⎥ displacement field associated with the i-th iteration can be obtained
ρ0 ξ3 [I]det⎢ ⎥
k
[m] = ⎣ ⎦ with [Ik ] = ⎣Z⎦dξ3
[I2 ] [I3 ] [I4 ] [I5 ] h through an iterative procedure as follows:
[I3 ] [I4 ] [I5 ] [I6 ]
(44)

( )
[ ] ( 2 )2 { i} { } { ( )}
(( )2 ({
2 } ) 4
{ } { }
Ki−n+γ1 + [M] ΔU = Pn+γ − Fi− 1 Un+γ − [M] Ui−n+γ1 − {Un } − U̇n − Ün ) (52)
γΔt γΔt γΔt

[ ]
which [I] is the third-order identity matrix. which Ki−n+γ1 is the tangent stiffness matrix and the displacement
Then using the standard finite element approximation of the { } { } { }
displacement field, the kinetic energy for an element is field is updated through Uin+γ = Ui−n+γ1 + ΔUi and the iterations
∑4 ∑4 (∫ ){ } continue up until the convergence requirement is fulfilled. Next, the
velocity and acceleration fields are accomplished. With the knowledge
T
δKe = − A=1 B=1
{δuA } N N
A B [m]dS üB
of the kinematic fields inside the time domain, the concentration is on
Se
{ }
T
= − {δUe } [Me ] Üe (45) the fields at the end of the time interval, tn+1 . For this purpose, use is
made of the following relation:
Through an assembling process the total variational kinetic energy is
f˙n+1 = c1 fn + c2 fn+γ + c3 fn+1 (53)
{ } { }
δK = − Ane=1
e
{δUe }T [Me ] Üe = − {δU}T [M] Ü (46) where f is typical function and the constants are obtained from
(1 − γ) 1 (2 − γ)
where [M] is the global mass matrix. c1 = , c2 = − , c3 = (54)
γΔt (1 − γ)γΔt (1 − γ)Δt
By combining Eqs. (24), (41) and (46), the finite element formulation
of the present shell model leads to the following nonlinear algebraic With the aid of Eq. (53), the velocity and acceleration at the end of
equation: the time domain can be expressed by
{ } { }
[M] Ü + {F(U)} = {P} (47) { }
U̇n+1 = c1 {Un } + c2 Un+γ + c3 {Un+1 } (55)

{ } { } { } { }
4.5. Time discretization Ün+1 = c1 U̇n + c2 U̇n+γ + c3 U̇n+1 (56)

After the derivation of the governing algebraic equation in the pre­ Now, the equilibrium equation is evaluated at tn+1 as below:
vious section, the focus is on the discretization in time based on the { }
composite time integration scheme devised by Bathe and Baig [14]. [M] Ün+1 + {F(Un+1 )} = {Pn+1 } (57)
Suppose that t stands for the time and it ranges from 0 to the total time
period, T which is chopped into smaller uniform time increment, Δt = Then, by performing the standard linearization scheme on the above
tn+1 − tn , where tn = nΔt. In this scheme, the parameter γ ∈ [0, 1] is used equation, we have
to divide the every time interval [tn , tn+1 ] into two parts and the associ­
ated inside time parameter is obtained from tn+γ = tn + γΔt. Notice that
in the composite time integration scheme the standard Newmark
approach is employed to extract data at tn+γ and subsequently an
expansion is used to accomplish data at the end of the time interval.
Accordingly, we have

5
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

5.1. Simply-supported plate

The example is concerned with dynamic analysis of a plate which is

simply-supported in all sides and the top surface of the structure is under
a transverse step loading with the intensity of q. The length of the square
plate is a = 25 cm and the thickness h = 5 cm. Both cross-ply and angle-
ply laminated composite plates with the lay-up of [0◦ /90◦ ] and [45◦ /-
45◦ ] are considered in the present example. Only one-fourth of the
whole geometry is used for the finite element simulation for the former
allowing for symmetries, while the full model is used in the numerical
investigation for the latter. The total number of 100 and 400 elements
distributed uniformly and equally along square sides are used to mesh
the geometry.
Each layer of the laminated composite herein possesses the following
material constants [28]:
/
E2 = E3 = 2.1 × 106 N cm2 , E1 = 25E2 , G12 = G13 = G23 = 0.5E2 , ν12
/
= ν13 = ν23 = 0.25, ρ = 8 × 10− 6 N.s2 cm4

Fig. 2. Central displacement versus time for cross-ply and angle-ply layered
composite plates.

( { }
( [ i− 1 ] ){ } { }
Kn+1 + c3 c3 [M] ΔUi = {Pn+1 } − Fi− 1 (Un+1 ) − [M] c1 U̇n
{ }
{ }
+ c2 U̇n+γ + c1 c3 {Un } + c2 c3 Un+γ
)
{ }
+ c3 c3 Ui−n+11 Fig. 3. Schematic of a spherical panel.

(58)
Similar to Eq. (52), based on an iterative algorithm the above
equation should be solved and after satisfaction of the convergence
requirement the solution is available to us at the end of the time interval.
As can be seen, the solution procedure in every time interval is twofold:
first the kinematic unknowns are obtained inside the time domain and
subsequently the terminal unknowns are accomplished through
considering the linearized form of the associated equations.

5. Numerical examples

In the preceding section, the finite element formulation for dynamic

analysis of shells was addressed and in the present section the accuracy
and validity of the developed relations are investigated through solving
some examples involving dynamic analysis of a plate, a spherical panel,
a short cylinder and a hyperboloidal shell. It is worth noting that for
carrying out the numerical integration associated with the tangent
stiffness matrix a 2 × 2 scheme is adopted and over the thickness
parameter 8 Gaussian points are taken into account. Additionally, the
initial values of velocity and acceleration vectors are supposed to be
zero.

Fig. 4. Displacement-time curves for cross-ply and angle-ply spherical panels

with a/h = 100.

Table 1
Displacement of the center of layered composite plates (cm) for cross-ply and angle-ply lay-ups.
Layup Time (µs)

40 80 120 160 200 240 280 320 360 400

[0 /90 ]
◦ ◦
1.0444 3.3231 3.5467 1.1593 − 0.0895 1.3571 3.3154 3.3995 1.2691 − 0.1076
[45◦ /-45◦ ] 1.0120 2.1015 0.8988 0.0224 1.5487 1.9793 0.2839 0.3015 2.0828 1.6477

6
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

Table 2
Non-dimensional displacements of the center of multilayer thin spherical panels.
Layup Time (ms)

1 2 3 4 5 6 7 8 9 10

[0◦ /90◦ ] 0.1188 0.5631 0.6523 0.4689 0.0451 0.0206 0.2337 0.6458 0.6185 0.3508
[45◦ /-45◦ ] 0.0874 0.2545 0.1310 − 1.0925 × 10-4 0.0813 0.2334 0.1613 − 0.0139 0.0785 0.2178

Also, a uniform 10 × 10 discretization scheme is used for the cross-

ply case with the layup of [0◦ /90◦ ]. As expected, for angle-ply case,
[45◦ /-45◦ ], the entire geometry is meshed using 400 elements. As for the
discretization in time, the stable results are obtained with Δt = 0.04 ms.
First, the present result is compared with the report of Kundu and Sinha
[11] for the thin cross-ply composite shell and subsequently the thick
one is addressed numerically. A step loading with the value of q = 1 ×
10-4 N/m2 is applied to the body for a/h = 100. For the case in which a/
h = 10, the thick shell, the magnitude of the internal pressure is q = 5 ×

Fig. 6. A short cylinder under point forces.

Fig. 5. Non-dimensional displacement versus time for thick spherical panels.

It is also worth noting that the intensity of the applied load for both
types of layup is q = 5 × 104 N/cm2. The behavior of the plate is studied
for total time of T = 400 μs and the increment of Δt = 5 μs is used for
the time discretization. It is observed from Fig. 2 that the angle-ply
layered composite plate shows a stiffer behavior than the cross-ply
one under the same magnitude of loading. Notice that the figure has
been constructed for the displacement of the plate center. Additionally,
the information regarding the composite plate displacement for certain
times has also been reported in Table 1.

5.2. Spherical panel

The focus of the current example is on spherical panels with simply

supported boundary conditions under the application of an internal
pressure with intensity of q, see Fig. 3. Similar to the previous case, not
only the cross-ply but also the angle-ply laminated composite shells are
dealt with herein. The geometric properties of the structure shown in
Fig. 3 are described as follows: R1 = R2 = 5 m, a = 0.5 m, b = 0.5 m, h =
0.005 m and h = 0.05 m. Every lamina is made of the T300/5208 CFRP
and has the following material properties [11]:

E1 = 181GPa, E2 = E3 = 10.3GPa, G12 = G13 = 7.17GPa, G23

/
= 3.58GPa, ν12 = ν13 = ν23 = 0.28, ρ = 1600kg m3 Fig. 7. Displacements of point A versus time.

Table 3
Central non-dimensional displacements of layered composite panels with a/h = 10.
Layup Time (ms)

0.4 0.8 1.2 1.6 2 2.4 2.8 3.2 3.6 4

[0 /90 ]
◦ ◦
1.0149 0.5611 0.1894 1.2242 0.1027 0.6863 0.9132 0.0018 1.1617 0.3635
[45◦ /-45◦ ] 0.7945 0.1505 0.5570 0.4490 0.2371 0.7322 0.0217 0.8398 0.0422 0.7049

7
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

of D = 15 cm, the length of L = 3 cm and the thickness of h = 0.02 cm.

Furthermore, the function f(t) varies with respect to time according to
⎧
⎨ 0.3t, 0 ≤ t ≤ 0.05
f (t) = − 0.3t + 0.03, 0.05 < t ≤ 0.1
⎩
0, t > 0.1

implies that f(t) linearly increases to the value of 0.015 then there
exists a linear decrease to zero and subsequently it is kept constant.
The lamina has the following material properties [29]:
/ /
E1 = 40 × 105 N cm2 , E2 = E3 = 8.28 × 105 N cm2 , G12 = G13
/ /
= 4.13 × 105 N cm2 , G23 = 0.03 × 105 N cm2 , ν12 = ν13 = ν23
/
= 0.25, ρ = 19.015 × 10− 6 N.s2 cm4

For carrying out the finite element analysis, a cross-ply composite

shell with the lamination scheme of [0◦ /90◦ ] measured with respect to
the cylinder axis is chosen and the analysis is based on the 3 × 4 × 8
mesh discretization, as illustrated in Fig. 6. The point forces are applied
to the nodes located at the angles of 0◦ , 90◦ , 180◦ and 270◦ , as shown in
Fig. 6. It should be pointed out that the time increment of 0.01 s provides
us with stable and reliable results based on the composite time inte­
gration approach. The displacements and velocities of the point A along
the axes of the Cartesian coordinate system, {X1, X2, X3}, such that U =
Fig. 8. Velocities of point A versus time. Ui Ei and V = Vi Ei where Ei are base vectors of the coordinate system, for
the time period of 0.6 s have been depicted in Fig. 7 and Fig. 8,
107 N/m2. The non-dimensional displacement of the center of the thin
shell is reported in Fig. 4 for two types of layup. As can be seen there is a
decent consistency with the work of Kundu and Sinha [11] for cross-ply
composite shell and the displacement amplitude and time period of the
center motion of the two-layer angle-ply composite is lower than the
cross-ply composite spherical panel. As further information, the dis­
placement–time data of the thin panels are tabulated in Table 2.
As for the thick panel with a/h = 10, the variation of the non-
dimensional displacement of the center with respect to time is depic­
ted in Fig. 5. It is transparent that the two-layer angle-ply shell indicates
stiffer behavior than the cross-ply one under the same step loading, but
the discrepancy is less serious than the thin spherical panel. Addition­
ally, the results are also provided in the tabular form in Table 3.

5.3. Motion of a short cylinder

In the present example, the motion of a short cylinder under the

application of point forces is taken into account as shown in Fig. 6. It is
worth noting that the problem was first solved for an elastic homoge­
Fig. 10. Schematic of a hyperboloidal shell under point forces.
neous material by Simo and Tarnow [15]. The cylinder has the diameter

Fig. 9. Undeformed and deformed configurations of the short cylinder.

8
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

Fig. 13. Deformed shape of hyperboloidal shell under dynamic loading.

increases linearly with respect to time from zero at t = 0 to P = 18 × 104

N at t = t* where t* takes the values of 0.01 s and 0.001 s, respectively.
Every layer is made of a material similar to the previous example and the
analyses are carried out for a two-layer composite with the lay-up of
Fig. 11. Load-deflection data of a hyperboloidal shell under quasi-static and [90◦ /0◦ ]. With respect to geometric data, we have R0 = 7.5 cm, L = 20
dynamic loads. cm and h = 1 cm, see Fig. 10, and radius varies according to R =
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
2
R0 1 + 3 XL32 . Also, as raised before, the maximum value of the point
force is P = 18 × 104 N. It should be remarked that for the quasi-static
deformation, the solution to the problem is carried out using 30 uniform
load steps, while for the dynamic analyses the time increment is chosen
to be Δt = 0.001 s and Δt = 0.0001 s, respectively, implying use of 100
time increments. The reported values provide us with accurate results.
Also, due to the symmetry, just one eighth of the entire geometry is
considered for the finite element simulations and the region is dis­
cretized using 20 × 20 elements as shown in Fig. 10.
Force-displacement curves associated with points A and B for quasi-
static and dynamic analyses have been plotted in Fig. 11. As can be seen,
for the loading time of t*=0.01 s, the force–displacement curves of
points A and B fluctuate around the respective curves related to the
quasi-static deformation. However, for t*=0.001 s, a stiffer behavior can
be clearly observed compared to the quasi-static deformation. Addi­
tionally, the displacements of points A and B along X1 and X2 axes for
both cases versus time are illustrated in Fig. 12. For more information,
the deformed shape of the shell for t*=0.001 s has been illustrated in
Fig. 13.

6. Conclusion

The focus of the current contribution was on the dynamic transient

analysis of multilayer composite shells. For this purpose, the higher-
Fig. 12. Displacement-time curves of multilayer composite hyperboloidal shell order kinematic shell model with 12 parameters suited for unsym­
under two loading rates. metric layered composites was used. Also, for solving the so-called weak
form associated with dynamics of the shell, the composite time inte­
respectively. Notice that the forces are applied to the cylinder up to 0.1 s gration scheme as well as the nonlinear finite element method were
and after that the shell freely flies in the air as shown in Fig. 9. employed. For space discretization, a four-node shell element consid­
ering remedies for transverse shear, membrane and curvature-thickness
locking problems was designed. In order to verify the validity of the
5.4. Hyperboloidal shell finite element formulation and the time integration scheme some typical
examples were taken into account and an analogy was made with the
In the last example, the deformation of a hyperboloidal shell under solved problems in the literature. It was shown that the proposed shell
point forces, as depicted in Fig. 10, is studied. A loading with three element and the composite time integration scheme are able to solve
various rates, one quasi-static and two dynamic deformations, is applied complex problems in the area of nonlinear dynamics allowing for geo­
to the body while the maximum magnitude of the force is constant. In metric nonlinearity.
such a way, we are able to witness the effect of rate of an applied loading It was observed that the higher-order shell model has a decent
to the structure. For dynamic loadings, the magnitude of the load agreement with the classical one when the structure is thin. Also, stable

9
A. Beheshti and R. Ansari Composite Structures 307 (2023) 116627

results were achieved employing the composite time integration [8] Vu-Quoc L, Tan XG. Optimal solid shells for non-linear analyses of multilayer
composites. II. Dynamics. Comput Methods Appl Mech Eng 2003;192(9–10):
approach with a reasonable time increment under nonlinear de­
1017–59.
formations. It was also illustrated that the rate of loadings has an impact [9] Naboulsi SK, Palazotto AN. Non-linear static–dynamic finite element formulation
on the mechanical behavior of structures. In addition, the present results for composite shells. Int J Non Linear Mech 2003;38(1):87–110.
can be regarded as references for future studies. Moreover, the ANS [10] Naidu NVS, Sinha PK. Nonlinear transient analysis of laminated composite shells in
hygrothermal environments. Compos Struct 2006;72(3):280–8.
scheme for the transverse shear and the membrane locking and the [11] Kundu CK, Sinha PK. Nonlinear transient analysis of laminated composite shells.
assumed strain approach for the thickness-curvature locking are robust J Reinf Plast Compos 2006;25(11):1129–47.
in the nonlinear dynamics of composite shells and provide us with [12] Rahman T, Jansen EL, Gürdal Z. Dynamic buckling analysis of composite
cylindrical shells using a finite element based perturbation method. Nonlinear Dyn
convergent results. 2011;66(3):389–401.
[13] Almeida FS, Awruch AM. Corotational nonlinear dynamic analysis of laminated
CRediT authorship contribution statement composite shells. Finite Elem Anal Des 2011;47(10):1131–45.
[14] Bathe K-J, Baig MMI. On a composite implicit time integration procedure for
nonlinear dynamics. Comput Struct 2005;83(31–32):2513–24.
Alireza Beheshti: Conceptualization, Methodology, Software, [15] Simo JC, Tarnow N. A new energy and momentum conserving algorithm for the
Writing – original draft, Visualization. Reza Ansari: Conceptualization, non-linear dynamics of shells. Int J Numer Meth Eng 1994;37(15):2527–49.
[16] Chung J, Hulbert GM. A time integration algorithm for structural dynamics with
Validation, Writing – review & editing, Supervision. improved numerical dissipation: the generalized-α method. J Appl Mech 1993;60
(2):371–5.
Declaration of Competing Interest [17] Kant T, Manjunatha BS. An unsymmetric FRC laminate C◦ finite element model
with 12 degrees of freedom per node. Eng Comput 1988;5(4):300–8.
[18] Başar Y. Finite-rotation theories for composite laminates. Acta Mech 1993;98(1–4):
The authors declare that they have no known competing financial 159–76.
interests or personal relationships that could have appeared to influence [19] Dvorkin EN, Bathe KJ. A continuum mechanics based four-node shell element for
the work reported in this paper. general non-linear analysis. Eng Comput 1984;1(1):77–88.
[20] Ko Y, Lee P-S, Bathe K-J. The MITC4+ shell element in geometric nonlinear
analysis. Comput Struct 2017;185:1–14.
Data availability [21] Choi CK, Paik JG. An efficient four node degenerated shell element based on the
assumed covariant strain. Struct Eng Mech 1994;2(1):17–34.
[22] Choi CK, Paik JG. An effective four node degenerated shell element for
The data that has been used is confidential. geometrically nonlinear analysis. Thin-Walled Struct 1996;24(3):261–83.
[23] Kulikov GM, Plotnikova SV. Simple and effective elements based upon
References Timoshenko-Mindlin shell theory. Comput Methods Appl Mech Eng 2002;191
(11–12):1173–87.
[24] Kulikov GM, Plotnikova SV. Non-linear strain–displacement equations exactly
[1] Chen J, Li Q-S. Nonlinear aeroelastic flutter and dynamic response of composite
representing large rigid-body motions. Part I Timoshenko-Mindlin shell theory.
laminated cylindrical shell in supersonic air flow. Compos Struct 2017;168:
Comput Methods Appl Mech Eng 2003;192(7–8):851–75.
474–84.
[25] Roh HY, Cho M. The application of geometrically exact shell elements to B-spline
[2] Amabili M, Balasubramanian P. Nonlinear forced vibrations of laminated
surfaces. Comput Methods Appl Mech Eng 2004;193(23–26):2261–99.
composite conical shells by using a refined shear deformation theory. Compos
[26] Kulikov GM, Plotnikova SV. A family of ANS four-node exact geometry shell
Struct 2020;249:112522.
elements in general convected curvilinear coordinates. Int J Numer Meth Eng
[3] Chao WC, Reddy JN. Analysis of laminated composite shells using a degenerated 3-
2010;83(10):1376–406.
D element. Int J Numer Meth Eng 1984;20(11):1991–2007.
[27] Betsch P, Stein E. An assumed strain approach avoiding artificial thickness
[4] Schokker A, Sridharan S, Kasagi A. Dynamic buckling of composite shells. Comput
straining for a non-linear 4-node shell element. Commun Numer Methods Eng
Struct 1996;59(1):43–53.
1995;11(11):899–909.
[5] To CWS, Wang B. Transient responses of geometrically nonlinear laminated
[28] Kant T, Ravichandran RV, Pandya BN, Mallikarjuna BN. Finite element transient
composite shell structures. Finite Elem Anal Des 1998;31(2):117–34.
dynamic analysis of isotropic and fibre reinforced composite plates using a higher-
[6] Ganapathi M, Varadan TK. Application of a field-consistent shear flexible element
order theory. Compos Struct 1988;9(4):319–42.
for nonlinear dynamic analysis of laminated shells. Finite Elem Anal Des 1992;12
[29] Kant T, Varaiya JH, Arora CP. Finite element transient analysis of composite and
(2):105–16.
sandwich plates based on a refined theory and implicit time integration schemes.
[7] Kant T, Kommineni JR. Geometrically non-linear analysis of doubly curved
Comput Struct 1990;36(3):401–20.
laminated and sandwich fibre reinforced composite shells with a higher order
theory and C◦ finite elements. J Reinf Plast Compos 1992;11(9):1048–76.

10