See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/328726373

An efficient zigzag theory based finite element modeling of composite and

sandwich plates with multiple delaminations using a hybrid continuity
method

Article  in  Computer Methods in Applied Mechanics and Engineering · November 2018

DOI: 10.1016/j.cma.2018.10.035

CITATION READS

1 54

2 authors:

Santosh Kapuria Adnan Ahmed

Indian Institute of Technology Delhi CSIR Structural Engineering Research Centre
160 PUBLICATIONS   2,683 CITATIONS    5 PUBLICATIONS   12 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

A facet shell element for laminated shells of general shapes View project

Active vibration control of FGM structures View project

All content following this page was uploaded by Adnan Ahmed on 28 August 2019.

The user has requested enhancement of the downloaded file.

 Available online at www.sciencedirect.com
ScienceDirect

Comput. Methods Appl. Mech. Engrg. 345 (2019) 212–232

www.elsevier.com/locate/cma

An efficient zigzag theory based finite element modeling of

composite and sandwich plates with multiple delaminations using a
hybrid continuity method
Santosh Kapuria a , b , ∗, Adnan Ahmed a
a Department of Applied Mechanics, Indian Institute of Technology Delhi, Hauz Khas, New Delhi 110016, India
b CSIR-Structural Engineering Research Centre, CSIR Campus, Taramani, Chennai 600113, India

Received 27 September 2017; received in revised form 15 October 2018; accepted 19 October 2018
Available online 3 November 2018

Highlights

• Zigzag theory based FE formulation for laminated plates with multiple delaminations.
• First efficient layerwise theory based formulation using region approach.
• A hybrid approach used for tying adjacent plate elements at the delamination fronts.
• Assessed for stress and free vibration analysis in comparison with 3D FE solutions.
• Yields accurate results even for highly inhomogeneous delaminated sandwich plates.

Abstract

We present a finite element (FE) formulation based on an efficient layerwise (zigzag) theory for stress and vibration analysis of
highly inhomogeneous composite and sandwich plates with multiple delaminations using the region method. The delaminations are
assumed to be present at multiple interfacial and/or planar locations, and are not allowed to change in size during the deformations.
Following the free mode model, the delaminated faces are assumed to have no mutual interaction during deformations. Using
a hybrid method, the continuity of inplane displacements at the delamination front is satisfied exactly at the midplanes of the
sublaminates separated by delaminations, while the deviations of their through-thickness variations in the intact and delaminated
segments are minimized with respect to the rotation variables, using the least squares method. The formulation is shown to yield
accurate results with reference to the full-field three dimensional FE solutions, for the deflection, stresses, natural frequencies
and mode shapes for delaminated composite as well as highly inhomogeneous single- and double-core sandwich plates. The
conventional point and least squares continuity methods, however, show large error for moderately thick plates and for higher
than fundamental vibration modes. The smeared third order theory, which has the same number of degrees of freedom as the
zigzag theory, is shown to yield grossly inaccurate results for delaminated sandwich plates. The present formulation is more

∗ Corresponding author at: Department of Applied Mechanics, Indian Institute of Technology Delhi, Hauz Khas, New Delhi 110016, India.
E-mail addresses: [email protected], [email protected] (S. Kapuria).

https://doi.org/10.1016/j.cma.2018.10.035
0045-7825/⃝ c 2018 Elsevier B.V. All rights reserved.
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 213

computationally efficient than the layerwise theories that are usually used for such analysis, but is at the same time accurate, simple
and robust.
⃝c 2018 Elsevier B.V. All rights reserved.

Keywords: Finite element; Delamination; Sandwich plate; Composite plate; Zigzag theory

1. Introduction
Delaminations can appear in laminated structures during manufacturing due to incomplete wetting, air entrapment
etc. or during service due to impact loading, highly localized out-of-plane free edge stresses etc. and is considered to be
a significant and severe mode of damage [1]. Early experimental [2] and analytical [3,4] studies on delaminated beams
established that delaminations can significantly affect the natural frequencies and mode shapes in laminated structures.
Owing to its severity, there have been several studies on the modeling of delaminations in laminated structures as
summarized in the review paper [5]. The change in the vibration signatures due to the delaminations can also be used
to detect them for structural health monitoring [6,7].
Most of the studies on delaminated beams and plates employ the region approach to model the delaminations,
wherein the sublaminates created by the delaminations are modeled as individual segments and connected to the intact
region at the delamination fronts. The sublaminates and the intact laminate are often modeled using the equivalent
single layer (ESL) theories like the first order shear deformation theory (FSDT) and the refined third order theory
(TOT). These theories assume the displacements to follow a single global variation along the thickness direction
across all layers. Finite element (FE) analyses based on the FSDT have been used for the free vibration analysis
of delaminated composite plates using the region approach [8–11]. Here, the continuity at the delamination front is
enforced by equating the displacement and rotation variables of the elements in the sublaminates of the delaminated
region with those of the adjacent elements in the integral region, at z-locations of the reference surfaces (usually
midsurface) of the sublaminates. This method will be called hereafter as point continuity method. As the displacement
field is assumed to be linear across the thickness in the FSDT, this method automatically ensures the continuity
of the displacements at all points across the sublaminates. The continuity conditions have been satisfied either by the
penalty parameter approach [8], which is analogous to connecting the coincident nodes with very stiff springs, or more
exactly by applying a transformation to the element stiffness and mass matrices of the elements in the delaminated
region connected to the delamination front [9].
In the aforementioned models, the delaminated faces are assumed to have no mutual interaction during the
deformation and are, therefore, called the free mode model. It may, however, sometimes predict physically
inadmissible mode shapes with interpenetrating sublaminates, for off-midplane delaminations. To avoid this, the
constrained mode model [4] was proposed, in which the delaminated layers are assumed to have identical transverse
deformation but are free to slide over each other. However, the results of this and subsequent studies [12] have shown
that while the constrained mode models cannot predict the opening mode shapes found in experiments [13], when
opening does not occur, their predictions do not show any significant difference from the free mode models for
the natural frequency. Methods like incorporation of normal contact forces between the delaminated layers [14]
and insertion of piecewise linear springs between the delaminated sublaminates [15] have been adopted to prevent
interlayer penetration in case of forced vibration response.
Since the FSDT relies on shear correction factors whose accurate estimation particularly for dynamic problems is
not trivial [16], higher order ESL theories have been developed for the analysis of composite structures, which do
not require any correction factor. Several studies have employed FEs based on the TOT [17] for free vibration and
dynamic instability analysis of composite plates with delaminations, using the point continuity conditions [18–21].
But, since the displacement field in the higher order theories is nonlinear across the thickness, the point continuity
method does not ensure the continuity of the displacements at all points across the thickness at the delamination
front. Hu et al. [12] developed an FE model based on the TOT, employing the least squares method for satisfying
the continuity of the inplane displacements across the sublaminates. The authors [22] have recently presented a
quadrilateral element based on the TOT for analysis of delaminated composite plates, in which a hybrid method
of satisfying the continuity conditions at the delamination fronts was proposed and evaluated. In this method,
214 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

the continuity of the inplane displacements is satisfied exactly at a reference plane (midplane) of the sublaminate
as in the point continuity method, while the deviations of their through-thickness variations are minimized with
respect to the rotation variables, using the least squares method. The numerical study revealed that the least squares
method does not always yield higher accuracy for the displacements, stresses and natural frequencies than the point
continuity method, but the hybrid method is superior to both these methods for modeling delaminations in composite
plates.
The ESL theories are computationally efficient because the number of displacement variables in these theories
are independent of the number of layers in the laminate. But, the results obtained using these theories may be
inaccurate when the adjacent layers of the laminate have widely different material properties as in the case of soft
core sandwich laminates and hybrid laminates made of multiple materials. Accurate results can be obtained using the
three-dimensional solid elements [23–25], in which the delaminations are modeled directly as unconnected nodes for
the free mode models, but such analyses are often too computationally involved and time consuming to be used for
practical design. Lekhnitskii [26] first proposed to use a layerwise theory (LWT) for analysis of composite beams. The
LWTs [27–29] have also been adapted to the analysis of delaminated composite and sandwich plates, using Heaviside
functions to include the discontinuities in the displacements caused by delaminations [30–33]. But they too suffer
from the same computational disadvantage as the full 3D FE analysis, as the number of degrees of freedom (DOFs)
increases proportionately with the number of layers in the laminate.
In efficient layerwise theories (ELTs), the number of primary displacement unknowns of originally an LWT is
reduced by imposing the conditions of transverse shear stress continuity at the layer interfaces and zero transverse
shear stresses at the top and bottom surfaces. Several such ELTs, also called zigzag theories (ZIGTs), have been
presented for analysis of laminated plates [34–37]. Kapuria and Kulkarni [38] developed a four-node quadrilateral
element for multilayered anisotropic plates, based on the third order ZIGT [36,39], using the improved discrete
Kirchhoff constraint (IDKQ) technique for satisfying the required C1 -continuity of the deflection variable. The
element has only seven DOFs per node, which is the same as in TOT based elements. It was shown to yield accurate
results for the static and dynamic response of laminated composite and sandwich plates of different lay-ups, shapes
and boundary conditions, and to be superior to other existing plate elements in terms of accuracy, computational
efficiency and robustness.
Librescu and coworkers [40–42] have extended the ZIGTs to incorporate the weak interfacial bonding allowing
interlayer slips, and obtained solutions for problems with weak interfaces spread over the entire plate area using
global trial functions. Cho and Kim [43] extended the third order ZIGT of Cho and Parmerter [35] for the analysis
of composite plates with multiple delaminations by including the displacement jumps at the delaminated interfaces
using the Heaviside step function. A two-node C1 -continuous delaminated beam finite element was used to study
the buckling and free vibration of composite beams with multiple through-width delaminations. Kim et al. [44] and
Oh et al. [45] presented four-node elements based on ELTs with displacement jumps at delaminated interfaces, for
delaminated composite plates. Such treatments of delaminations using displacement jumps, however, do not exactly
satisfy the continuity of the displacement field at the delamination front, and also the shear traction free conditions at
the delaminated interfaces are not explicitly satisfied. No FE models based on the ELTs have been presented so far for
laminated plates featuring delaminations, using the region approach. Further, no study has been reported on modeling
delaminated sandwich plates except for the ones using computationally involved LWTs [31].
In this work, we present an FE formulation based on the efficient ZIGT for the static and free vibration analysis
of laminated composite and sandwich plates having multiple delaminations at different interfacial and/or planar
locations. The four-node IDKQ quadrilateral element of Kapuria and Kulkarni [38] with seven DOFs per node is
employed, and the delaminations are modeled using the region approach. Following the free mode model, the present
work neglects ensuring that there is no interpenetration across the delaminated faces. The continuity conditions for
the inplane displacements at the delamination fronts are derived for the ZIGT using the point continuity method,
least squares method and the recently proposed hybrid point–least squares method [22]. The continuity conditions
are satisfied by expressing them in terms of a transformation matrix for the nodal DOFs of adjacent elements
across the delamination front, and applying the transformation to the element matrices for the delaminated region.
While in [22] the element is based on the smeared TOT and the continuity conditions were derived for a global
cubic variation of the inplane displacements and implemented for a single delamination, in the present work based
on the ZIGT, the continuity conditions at the delamination front have to deal with a layerwise variation for the
inplane displacements and are implemented for multiple delaminations. The accuracy of the element and the relative
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 215

Fig. 1. (a) Laminated plate with multiple delaminations at different interfacial and planar locations (b) Geometry of a sublaminate.

performance of the three continuity methods are assessed in comparison with the converged 3D FE solutions
for composite and sandwich plates with multiple delaminations. The performance of the TOT [22] having the
same number of DOFs as the present element is also assessed simultaneously, for both static and free vibration
responses.

2. Zigzag theory approximations

We consider a multilayered laminated plate with multiple delaminations which can have different locations both
across the thickness and over the plane as shown in Fig. 1(a). Delaminations are assumed to pre-exist in laminates
and they do not change in size during deformations. Following the region approach with free mode model, each of
the sublaminates separated by the delaminated interfaces in the delaminated regions is modeled separately to allow
the transverse as well as the inplane displacements to be discontinuous (opening and slipping) across the delaminated
interface, but it is connected with the intact region(s) at the delamination front(s) by ensuring the continuity of the
displacement fields in the intact laminate and the sublaminate across the thickness direction. For any sublaminate
(Fig. 1(b)), the middle surface is chosen to be the reference plane (z = 0) and the z-coordinate of the bottom
surface of the kth layer from the bottom of the sublaminate is denoted as z k−1 . The interface between the kth and
(k +1)th layers is denoted as the kth interface. The layers are made of generally orthotropic materials, and the material
symmetry direction 1 of the kth layer makes an angle of θk (anti-clockwise) with the x-axis. The total thickness of the
sublaminate is taken as h.
In the ZIGT, the transverse normal stress σz is assumed to be negligible in comparison with other stress components.
Using this assumption, the linear 3D constitutive equations relating the inplane stresses σ and transverse shear stresses
τ to the inplane strains ε and transverse shear strains γ of a lamina reduce to

σ = Q̄ε, τ = Q̂γ , (1)

216 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

where
⎡ ⎤ ⎡ ⎤ ⎡ ⎤
σx εx Q̄ 11 Q̄ 12 Q̄ 16
σ = ⎢
⎣ σy ⎦ , ε=⎢
⎣ εy ⎦ , ⎣ Q̄ 12 Q̄ 22 Q̄ 26 ⎦ ,
⎢ ⎥ ⎢ ⎥ ⎢ ⎥
⎥ ⎥ Q̄ = ⎢ ⎥

τ γ Q̄ 16 Q̄ 26 Q̄ 66 (2)
[ x y] [ x y] [ ]
τ γ Q̄ 55 Q̄ 45
τ = zx , γ = zx , Q̂ =
τ yz γ yz Q̄ 45 Q̄ 44
and Q i j are the reduced stiffness coefficients.
The transverse displacement w of the sublaminate is approximated to be constant over the thickness, and the
inplane displacements u x and u y are assumed to follow a global third order variation in z, combined with a layerwise
linear variation. Accordingly, the displacement field for the kth layer is expressed as [38]:

u(x, y, z, t) = uk (x, y, t) − zw0d (x, y, t) + zψ k (x, y, t) + z 2 ξ (x, y, t) + z 3 η(x, y, t) (3)

w(x, y, z, t) = w0 (x, y, t) (4)
where
w0,x ψk x ξ η
[ ] [ ] [ ] [
] [ ] [ ]
u u kx
u= x w0 d = , uk = , ψk = , ξ= x , η= x (5)
uy w0,y uky ψk y ξy ηy
uk denotes the translation components of the kth layer and ψ k is related to its shear rotations. ξ and η are the
global quadratic and cubic terms in z. Thus, uk and ψ k in Eq. (3) represent the layerwise linear component of the
assumed displacement field, while the other remaining terms represent the] global third order variation. By imposing
T
the conditions of continuity of the transverse shear stresses τ = τzx τzy and inplane displacements u at the layer
[

interfaces, and the zero shear tractions (τ = 0) at the top and bottom surfaces of the sublaminate, the unknowns
uk , ψ k , ξ and η for the L-layered sublaminate are expressed in terms of only five variables u0 , ψ 0 and w0 to yield

u(x, y, z, t) = u0 (x, y, t) − zw0d (x, y, t) + Rk (z)ψ 0 (x, y, t) (6)

where u0 and ψ 0 denote the displacements and shear rotations of the reference surface of the sublaminate. Rk (z) is a
2 × 2 matrix of layerwise functions of z of the form

Rk (z) = R̂k1 + z R̂k2 + z 2 R̂3 + z 3 R̂4 (7)

where R̂k1 , R̂k2 , R̂3 and R̂4 are 2 × 2 coefficient matrices which depend on the material properties and the lay-ups of
the sublaminate. Their expressions are available in Ref. [38]. For the smeared third order theory (TOT) [17,22], Rk (z)
is not a layerwise function and is given by

Rk (z) = R(z) = [z − 4z 3 /3h 2 ]I2 , (8)

where In is the n × n identity matrix.
Eq. (6) can be expressed in compact form as

u = f0 (z)ū1 (9)
where
ψ0 x ψ0 y ,
[ ]
ū1 = u 0x u 0 y −w0,x −w0,y (10)
f0 (z) = I2 zI2 Rk (z)
[ ]
(11)
Using the strain–displacement relations for small displacement gradients, the strains ε and γ can be expressed in
terms of the displacement variables as

ε = f1 (z)ε̄ 1 , γ = Rk,z ψ 0 (12)

 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 217

where
ε0
⎡ ⎤

ε̄1 = ⎣ κ 0 ⎦ ,
f1 (z) = I3 zI3 Φ̄k (z) ,
[ ]
ψ
⎡ 0d
⎡ k k ⎤
R11 0 R12 0 ⎤
u 0x,x
Φ̄k (z) = ⎣ 0 R21 k
0 R22 k ⎦
,
k k k k ε0 = ⎣ u 0 y,y ⎦,
R21 R11 R22 R12 (13)
u +u
w0,x x ⎡ 0 y,x ⎤ 0x,y
⎡ ⎤
ψ0x,x
κ 0 = − ⎣ w0, yy ⎦ , ⎢ψ0x,y ⎥
2w0,x y ψ 0d = ⎢
⎣ψ0 ⎦ .
⎥
y,x
ψ0 y,y

3. Finite element formulation

The weak variational form of the equations of motion of the plate (Fig. 1) as obtained using the Hamilton’s principle
∑ L e ∫ zk−
can be expressed, using the notation ⟨. . .⟩ = k=1 z+
(. . .) dz for integration across the thickness, as
k−1
ne ∫
∑
[⟨ρ k (δuT ü + ẅδw) + δεT σ + δγ T τ ⟩ − pze δw(x, y, t)] d Ae
e=1 Ae
∫
− ⟨σn δu n + τns δu s + τnz δw⟩ ds = 0 (14)
ΓL

∀ admissible δu 0 , δw0 and δψ0 , where n e denotes the number of elements used to discretize the plate mid-surface, L e
is the number of layers in eth element, Ae denotes the midplane surface area of the element and Γ L is the boundary
curve with normal n and tangent s. pze is the normal traction on the element surface and ρ k denotes the material mass
density of the kth layer.
Using Eqs. (4) and (6), the inertial terms in Eq. (14) can be expressed as

⟨ρ k (δuT ü + ẅδw)⟩ = δ ūT1 Iū¨ 1 + δw0 I¯ẅ0 (15)

where I and I¯ are inertial constants defined as
I = ⟨ρ k fT0 (z)f0 (z)⟩, I¯ = ⟨ρ k ⟩ = I11 (16)
Substituting Eq. (11) into Eq. (15), the submatrices of I are obtained as
[[ ] [ ] [ ]]
I11 0 I 0 I I
, 13 , 15 16 = ⟨ρ k I2 zI2 Rk (z) ⟩
[ ]
0 I22 0 I24 I25 I26
[[ ] [ ] [ ]]
I33 0 I35 I36 I55 I56
, , = ⟨ρ k z 2 I2 Rk (z)I2 {Rk (z)}T Rk (z) ⟩
[ ]
(17)
0 I33 I45 I46 I65 I66
Using Eq. (12), the strain energy terms in Eq. (14) can be expressed as

⟨(δεT σ + δγ T τ )⟩ = δ ε̄T1 F1 + δψ T0 Q (18)

where
⎡ ⎤ ⎡ ⎤
N ⟨σ ⟩
F1 = ⎣M⎦ = ⟨fT1 (z)σ ⟩ = ⎣ ⟨zσ ⟩ ⎦ , Q = ⟨{Rk (z),z }T τ ⟩
T
P ⟨Φk (z)σ ⟩ ⎡ ⎤
⎡ ⎤ ⎡ ⎤ Px (19)
Nx Mx ⎢ Py ⎥
N = ⎣ Ny ⎦ , M = ⎣ My ⎦ , P=⎢ ⎥
⎣ Px y ⎦
Nx y Mx y
Pyx
218 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

N, M, P and Q denote, the inplane stress resultants, bending moments, higher order bending moments and shear
resultants, respectively.
Substituting Eqs. (15) and (18) into Eq. (14), the variational equation corresponding to the zigzag theory is
obtained as
ne ∫ ∫
∑ T
[δuT Iū¨ + δw0 I¯ẅ0 + δ ε̄T1 F1 + δ ψ̄ 0 Q − F3 δw0 ]d Ae − [Nn δu 0n + Nns δu 0s − Mn δw0,n
n=1 Ae ΓL
∑
+(Vn + Mns,s )δw0 + Pn δψ0n + Pns δψ0s ] ds − ∆Mns (si )δw0 (si ) = 0 (20)
i

where F3 = pze , Vn = ⟨τ ⟩, si denotes the corner locations of the lateral surfaces, Nn , Nns , Mn , Mns , Vn , Pn and Pns
denote the various stress resultants on the boundary with normal n and tangent s.
A four-node quadrilateral element with seven DOFs, u 0x , u 0 y , w0 , w0,x , w0,y , ψ0x and ψ0 y per node is used to
model the plate. For inplane displacements u 0x and u 0 y , and shear rotations ψ0x and ψ0 y , the highest order of
derivatives appearing in the variational equation (Eq. (14)) are of first order. Therefore, these are interpolated using
C0 -continuous bilinear Lagrange interpolation functions. For deflection w0 , however, second order derivatives occur
in the variational equation. Hence, the interpolation functions used should satisfy continuity of w0 as well as
its derivatives w0x and w0 y (C1 -continuity) at the element boundaries. Since this is difficult to be achieved in a
quadrilateral element, we circumvent this difficulty by using the concept of improved discrete Kirchhoff constraint
(IDKQ) technique [46]. In this approach, the deflection gradients w0,x and w0,y are first replaced by rotation
variables θ0x and θ0 y , which are interpolated using C0 -continuous quadratic Lagrange functions, and w0 is interpolated
independently. Subsequently, the two are related by equating θ0i = w0,i at the corner nodes, midside points and at
the element center. The technique ensures complete compatibility of the deflection field along the element sides [47].
The deflection w0 appearing in the inertial and loading terms in Eq. (14) is interpolated using a bi-cubic function
involving 12 constants, and the same function is used for computing w0,i at the element center for equating it
with θ0i .
Substituting the shape functions in the inertial, strain energy and loading terms in the area integral of Eq. (20)
e e
and performing the integral over the element area yields the element mass matrix M , element stiffness matrix K
e
and element load vector P . Summing up the contributions of all elements in Eq. (20) yields the global equations of
motion for the global generalized displacement vector U of the plate

MÜ + KU = P (21)
where M, K and P are the assembled mass matrix, stiffness matrix and load vectors, respectively. For free vibration,
P is set to a null vector, resulting in an eigenvalue problem, It is solved using the subspace iteration technique to
determine the natural frequencies ωn and mode shapes (normalized with respect to the mass matrix). The boundary
conditions for the various support conditions are [48]

simply-supported: Nn = 0, u 0s = 0, w0 = 0, Mn = 0, Pn = 0, ψ0s = 0
clamped: u 0n = 0, u 0s = 0, w0 = 0, w0,n (= θn ) = 0, ψ0n = 0, ψ0s = 0 (22)
free: Nn = 0, Nns = 0, Vn + Mns,s = 0, Mn = 0, Pn = 0, Pns = 0

4. Continuity conditions at delamination front

Multiple delaminations in the laminated plate are modeled using the region approach, wherein each of the
sublaminates in the delaminated region are modeled separately. Fig. 2 shows two adjacent elements at delamination
front, one modeling the intact laminate in the undelaminated region and the other modeling a sublaminate in the
delaminated region. The equilibrium of stress resultants at the delamination front is accomplished by having common
nodes (2 and 3) at the delamination front. Further, the displacement field at all points along the thickness of the
sublaminate should be continuous at the delamination front. Since the variation of the inplane displacements along
the thickness direction is nonlinear, ensuring the continuity at all points across the thickness at the delamination
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 219

Fig. 2. a) Elements at a delamination front (b) Inplane displacement variations at intact laminate and sublaminate at the delamination front.

front is not trivial. In this segment, three approaches are described: two of them have been widely used in previous
studies on delaminated plates based on the other theories and the third one is a new approach recently developed by
the authors for the TOT [22].

4.1. Point continuity method

This is the most convenient and widely used method, in which the inplane displacements (u xD and u yD ) at the
midplane of the delaminated sublaminate are equated to those (u xI and u yI ) of the integral laminate at the same
z-location at the delamination front. Also, the rotational variables of the intact and delaminated segments at the
delamination front are assumed to be equal.
Consider that the bottom of the delaminated segment occurs at the bth interface of the intact laminate and there
are p layers in this sublaminate. For clarity, all the displacement variables and coefficient matrix Rk for the intact and
delaminated segments are denoted with superscripts/subscripts I and D, respectively. Thus, the inplane displacements
of the adjacent segments can be expressed as
u I (x, y, z, t) = u0I (x, y, t) − zw0I d (x, y, t) + RkI (z)ψ 0I (x, y, t)
u D (x, y, z, t) = u0D (x, y, t) − z ′ w0Dd (x, y, t) + RkD (z ′ )ψ 0D (x, y, t) (23)
where
[ ] ⎡ ⎤ [ ] [ ]
u xI,D u I,D w0,x
I,D
ψ0x
I,D
u I,D
= , u0I,D = ⎣ 0x ⎦ , w0I,D = , ψ 0I,D = (24)
u yI,D u 0I,D
y
d
w0,y
I,D
ψ0y
I,D
220 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

z and z ′ denote the z-coordinates in the intact laminate and delaminated sublaminate from their respective midplanes
such that z ′ = z − (z b + z b+ p )/2, Using Eq. (23), the aforementioned continuity conditions can be stated as
1 1
u0D = u0I − (z b + z b+ p )I2 w0,d
I
+ RkI ( (z b + z b+ p ))ψ 0I
2 2
= u0I + λ̄w0,d
I
+ λ̃ψ 0I (25)
w0,x
D
= w0,x
I
, w0,y
D
= w0,y
I
, ψ0Dx = ψ0Ix , ψ0Dy = ψ0Iy (26)

where λ̄ = − 12 (z b + z b+ p )I2 and λ̃ = RkI ( 12 (z b + z b+ p )). Using Eqs. (25) and (26), the nodal DOFs at any common
node of the elements in the delaminated and intact segments can be related through a transformation matrix T as

U D = TU I (27)
where
U D = [u 0Dx u 0Dy w0D w0,x
D
w0,y
D
ψ0Dx ψ0Dy ]T ,
(28)
U I = [u 0I x u 0I y w0I w0,x
I
w0,y
I
ψ0Ix ψ0Iy ]T ,
⎡ ⎤
1 0 0 λ̄11 0 λ̃11 λ̃12
λ̄22 λ̃21 λ̃22 ⎥
⎢ ⎥
⎢0 1 0 0
⎢ ⎥
⎢ ⎥
⎢0 0 1 0 0 0 0⎥
⎢ ⎥
T=⎢ ⎢0 0 0 1 0 0 0⎥ ⎥ (29)
⎢ ⎥
⎢0 0 0 0 1 0 0⎥
⎢ ⎥
⎢ ⎥
⎣0 0 0 0 0 1 0⎦
0 0 0 0 0 0 1

4.2. Least squares method

In this method [12], the continuity of the inplane displacements is sought to be satisfied in the least squares sense
across the thickness of the sublaminate. It is achieved by minimizing the mean square deviation of the inplane
displacements of the delaminated and intact segments over the sublaminate thickness. The error functions for the
sublaminate are defined as
b+ p ∫ z b+ p ∫ z
∑ k ∑ k
F1 = (u xI − u xD )2 dz F2 = (u yI − u yD )2 dz (30)
k=b+1 z k−1 k=b+1 z k−1

F1 and F2 will be minimum if

∂ F1 ∂ F1 ∂ F1
= = =0
∂u 0x
D
∂w0D,x ∂ψ0x
D

∂ F2 ∂ F2 ∂ F2
= = =0 (31)
∂u 0y
D
∂w0D,y ∂ψ0y
D

Substituting the expressions of u iI and u iD (i = x, y) from Eq. (23) into Eq. (30), and using the resulting expressions
of F1 and F2 in Eq. (31), we obtain

u0D − u0I + a1 w0,d

I
− a2 ψ 0I + a3 ψ 0D = 0 (32)

b1 w0,d
D
− b1 w0,d
I
+ b2 ψ 0I − b3 ψ 0D = 0 (33)

c1 u0I − c1 u0D − c2 w0,d

I
+ c3 w0,d
D
+ c4 ψ 0I − c5 ψ 0D = 0 (34)
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 221

where
b+ p ∫ z
1 1 ∑ k
a1 = (z b+ p + z b )I2 , a2 = Rk (z) dz
2 h d k=b+1 zk−1 I
b+ p ∫ z
1 ∑ k h 3d
a3 = Rk (z ′ ) dz, b1 = I2
h d k=b+1 zk−1 D 12
b+ p ∫ z b+ p ∫ zk
∑ k ∑
b2 = z ′ RkI (z) dz, b3 = z ′ RkD (z ′ ) dz
k=b+1 z k−1 k=b+1 z k−1
b+ p ∫ z b+ p ∫ z (35)
∑ k ∑ k
1 2
c = R̃kD (z ′ ) dz, c = z R̃kD (z ′ ) dz
k=b+1 z k−1 k=b+1 z k−1
b+ p ∫ z b+ p ∫ z
∑ k ∑ k
c3 = z ′ R̃kD (z ′ ) dz, c4 = R̃kD (z ′ )RkI (z) dz
k=b+1 z k−1 k=b+1 z k−1
b+ p ∫ z
∑ k
c5 = R̃kD (z ′ )RkD (z ′ ) dz, R̃kD (z ′ ) = diag(RkD (z ′ ))
k=b+1 z k−1

Eqs. (32), (33) and (34) can be written in the matrix form as
⎡ ⎤ ⎡ ⎤
u0D u0I
⎢ ⎥ ⎢ ⎥
C̄ ⎢ D ⎥ ⎢ I ⎥
⎣w0,d ⎦ = C̃ ⎣w0,d ⎦ (36)
ψ 0D ψ 0I
⎡ ⎤ ⎡ ⎤
u0D u0I
⎣w0,d ⎦ = λ ⎣w0,d ⎦
⎢ ⎥ ⎢ ⎥
⇒⎢ D ⎥ ⎢ I ⎥ (37)
ψ 0D ψ 0I
where
a3 −a1 a2
⎡ ⎤ ⎡ ⎤
I2 0 I2
λ = C̄−1 C̃, C̄ = ⎣ 0 b1 −b3 ⎦ , C̃ = ⎣ 0 b1 −b2 ⎦ (38)
c1 −c3 c5 c1 −c2 c4
Using Eq. (37), U D and U I of Eq. (27) can be related with the transformation matrix defined as
λ11 λ12 0 λ13 λ14 λ15 λ16
⎡ ⎤
⎢λ21 λ22 0 λ23 λ24 λ25 λ26 ⎥
⎢ ⎥
⎢0 0 1 0 0 0 0⎥
λ λ λ33 λ34 λ35 λ36 ⎥
⎢ ⎥
T=⎢ 31 32 0 (39)
⎢λ41 λ42 0 λ43 λ44 λ45 λ46 ⎥
⎢ ⎥
⎣λ51 λ52 0 λ53 λ54 λ55 λ56 ⎦
⎢ ⎥

λ61 λ62 0 λ63 λ64 λ65 λ66

where λi j denotes the elements of matrix λ.

4.3. Hybrid point–least squares method

This method was proposed recently by the authors [22] in the context of the TOT and was compared to
point continuity and least squares methods for composite plates with single delamination. In the point continuity
method, the inplane displacements in the intact and delaminated segments are equated at the midplane of the
delaminated sublaminate, but there is no attempt to minimize the deviation between their through-thickness variations
in the adjacent segments. On the other hand, while the deviations in through-thickness variations of the inplane
displacements in adjacent segments at the delamination front are sought to be minimized, at no given point across the
222 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

Table 1
Material properties.
Mat. No. Y1 Y2 Y3 G 23 G 13 G 12 ν23 ν13 ν12 ρ (kg m−3 )
(GPa)
1 134.0 10.3 10.3 3.28 5.0 5.0 0.33 0.33 0.33 1480
2 6.89×10−3 6.89×10−3 6.89×10−3 3.45×10−3 3.45×10−3 3.45×10−3 0 0 0 97
3 131.0 10.34 10.34 6.895 6.205 6.895 0.49 0.22 0.22 1627

thickness of the sublaminate, the inplane displacements are made to be equal. In the hybrid approach, the continuity
of the inplane displacements is satisfied exactly at a reference plane (midplane) of the sublaminate as in the point-
continuity method and at the same time, their deviations F1 and F2 are minimized with respect to the rotation variables
w0D,x , w0D,y , ψ0Dx and ψ0Dy , using the least squares method. Thus, the continuity relations in this method are given by Eqs.
(25), (33) and (34), which finally yield
⎡ ⎤ ⎡ ⎤
u0D u0I
⎢w D ⎥ = λ ⎢w I ⎥
⎢ ⎥ ⎢ ⎥
⎣ 0,d ⎦ ⎣ 0,d ⎦ (40)
ψ 0D ψ 0I
where
λ̄ λ̃
⎡ ⎤ ⎡ ⎤
I2 0 0 I2
λ = C̄−1 C̃, C̄ = ⎣ 0 b1 −b3 ⎦ , C̃ = ⎣ 0 b1 −b2 ⎦ (41)
c1 −c3 c5 c1 −c2 c4
The transformation matrix T remains the same as that given by Eq. (39) with λi j to be taken from Eq. (41).
Eqs. (29), (39) and (41) are valid for any general sublaminate in the delaminated region, and hence all the above
three methods are applicable to model multiple delaminations across the laminate thickness. At the delamination front,
the nodal DOFs of the elements in the intact segment are taken as the primary variables and the nodal DOFs of the
elements of delaminated segments are related to these primary variables using the transformation matrix Ti (for the ith
node). Accordingly, the stiffness and mass matrices of the elements in the delaminated segments at the delamination
e e e
front are transformed as Ke = TTe K Te , Me = TTe M Te and Pe = TTe P where
⎡ ⎤
T1
⎢ T2 ⎥
Te = ⎢ ⎥ (42)
⎣ T3 ⎦
T4
where Ti is obtained using Eqs. (29), (39) and (41) depending upon the method, for those nodes which are on the
delamination front. For the other nodes of the element Ti will be an identity matrix of size 7 × 7.

5. Numerical results and discussion

In this section, problems on static and free vibration responses of composite beams and sandwich plates containing
single and multiple delaminations are considered to assess the performance of the developed FE based on the ZIGT
obtained with point continuity (PC), least squares (LS) and hybrid point–least squares (HPLS) methods of satisfying
continuity of displacements at delamination fronts. The accuracy of the present results for delaminated plates is
assessed in comparison with existing LWT based FE results available for single delamination and 3D FE results
obtained using the commercial FE software ABAQUS. The layer stacking order is mentioned from bottom to top, and
the material properties of the laminas are listed in Table 1. The span to thickness ratio is defined as S = a/ h, where a
is the span of the beam/plate and h is the total thickness. The results presented here are the converged values and the
mesh sizes used are given in the respective problem descriptions. For the 3D FE models in ABAQUS, the following
boundary conditions are imposed for various support conditions (stated for x-normal edges):
simply-supported: σx = 0, u y = 0, w = 0
clamped: u x = 0, u y = 0, w = 0 (43)
free: σx = 0, τx y = 0, τx z = 0
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 223

Table 2
Deflection, stresses and fundamental frequency of a three-layer simply supported plate.
w̄ = wY1a / hq0 σ̄x = σx /q0 σ̄ y = σ y /q0 τ̄zx = τzx /q0 ω¯1 = ω1 h ρ/Y1a
√
Solution
Exact 3D [49] −121.72 66.787 46.424 −3.5768 0.11203
Present (12 × 12) −121.99 66.901 46.689 −3.5568 0.11197
Present (16 × 16) −121.95 66.829 46.642 −3.6115 0.11199
(x, y, z) locations: w̄(a/2, a/2, 0), σ̄x (a/2, a/2, h/2), σ̄ y (a/2, a/2, h/2), τ̄x z (0, a/2, −0.4h) a for middle layer.

Fig. 3. Locations of delaminations in a composite beam.

Before analyzing the delaminated plates, the accuracy of the FE formulation is established first for the stress and
free vibration analysis of laminated plates without any delamination, by comparing the results with the exact analytical
solutions available in the literature. The analysis of the delaminated plates is presented in the subsequent subsections.

5.1. Stress and free vibration analysis of simply supported laminated plates without delamination

A simply supported square three-layer sandwich plate is analyzed for the static and free vibration response,
for which the exact 3D elasticity solution has been presented by Srinivas and Rao [49]. The plate is moderately
thick with S = 10 and layer thicknesses (0.1h/0.8h/0.1h). The material properties of the orthotropic layers are:
Y2 /Y1 = 0.5250, Y3 /Y1 = 0.5694, G 23 /E 1 = 0.2971, G 13 /E 1 = 0.1781, G 12 /E 1 = 0.2928, ν12 = 0.4404,
ν13 = −0.0613 and ν23 = 0.1806. The middle layer is relatively softer with its Y1 being 15 1
th of that of the top and
bottom layers, but all layers have the same mass density. All the layers have the material direction 1 along x-direction.
The plate is subjected to a uniformly distributed pressure q0 . A quarter of the plate is modeled using symmetry. The
present results for the deflection, inplane normal stresses, transverse shear stress τzx and the fundamental frequency
obtained with mesh size of 12 × 12 and 16 × 16 are compared in Table 2 with the exact analytical solution [49]. It
is seen that the present ZIGT FE yields highly accurate results with an error of just 0.2%, 0.6% and 0.06% for the
deflection, stresses and frequency respectively, with respect to the exact 3D elasticity solution, for the 12 × 12 mesh.

5.2. Free vibration of composite beams with multiple delamination

Free vibration response of cantilever composite beams with delaminations at multiple interfaces is studied first. The
beams are made of an eight-layer cross-ply (0◦ /90◦ )2s graphite/epoxy laminate of material 1 and have dimensions of
127 mm (length) × 12.7 mm (width) × 1.02 mm (thickness). All layers are of equal thickness. The beams have
through-width delaminations of varying length located symmetrically about the mid-section of the beam and at one
to three interfaces simultaneously, as shown in Fig. 3. The same beams were studied by Shen and Grady [13] but with
a single delamination. The beams are modeled with a 20 × 1 mesh size using the present element. The results for
the natural frequencies of the first two modes obtained using the HPLS continuity method are presented in Table 3.
The results are compared with the available analytical solution based on the FSDT presented by Shen and Grady [13]
for the fundamental frequency for the single delamination case. The 3D FE solutions obtained using ABAQUS are
listed in the table as reference for assessing accuracy. In the 3D FE model, the layers are modeled using twenty-node
224 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

Table 3
Natural frequencies (Hz) of cantilever cross-ply (0◦ /90◦ )2s beams with multiple delaminations.
Del. length (mm) No. of Present FSDT [13] 3D FE
delaminations Mode 1 Mode 2 Mode 1 Mode 1 Mode 2
Intact 0 82.212 514.35 82.042 82.215 513.38
50.8 1 80.945 446.52 80.383 80.617 437.07
2 80.394 135.95 – 80.567 136.55
3 70.904 135.54 – 74.245 136.47
101.6 1 73.303 230.60 73.147 72.120 224.70
2 33.361 73.340 – 33.351 73.028
3 33.022 51.669 – 33.019 51.588

quadratic solid element with reduced integration (C3D20R) and considering consistent mass matrix. The laminate is
modeled with 80 elements along the length, 8 elements along the width direction and each layer is modeled with two
elements in the thickness direction. The delamination is modeled by dividing the beam into different parts separated
by the delamination interfaces and imposing TIE constraints on the nodes on the adjacent surfaces at the interfaces
over the intact areas only. The convergence of the 3D FE solution has been verified by doubling the mesh in length
and width directions, which gives a maximum difference of 0.03% in the frequencies of the first 2 modes of the
delaminated beams.
It can be seen from Table 3 that the present results are in excellent agreement with the 3D FE solution for both
delamination lengths and for single as well as multiple delaminations. The results for the fundamental frequency for
the single delamination case are found to match well with the FSDT based solution as well, since the beams are thin.
As expected, an increase in the delamination length as well as the number of delaminations results in a decrease in the
natural frequencies. By increasing the number of delaminated interfaces, the natural frequencies of all modes may not
be affected equally. For example, the natural frequency of the second mode corresponding to 2 and 3 delaminations
do not differ much, as in both cases this mode pertains to a local mode corresponding to bending of the 2nd layer from
the bottom.
As an example of the computational efficiency of the present element, the run time for this problem is compared.
In a workstation of 2.4 GHz frequency and 256 GB RAM, while the 3D FE solution takes 4 h 40 min, the present
model takes only 3 min 10 s, which is a ninety-fold reduction.

5.3. Free vibration of sandwich plates

To assess the accuracy of the present ZIGT based FE formulation and the relative performance of the continuity
schemes, PC, LS and HPLS, for free vibration of delaminated sandwich plates, a simply supported square sandwich
plate with a central delamination of size 0.5a × 0.5a as shown in Fig. 4 is considered. For this problem, FE
results based on the LWT have been presented by Marjanović and Vuksanović [31]. The lay-up of the laminate is
(0◦ /90◦ /Core/90◦ /0◦ ), in which the ratio of thickness of the core (h c ) to that of each face sheet (h f ) is 10. Each
layer in the face sheets has equal thickness. The core is made of material 2 and the face sheets are made of material
3. The delamination is located at the top interface in the face sheet. The present results are obtained using a mesh
size of 20 × 20. For comparison with 3D FE solution, the delaminated plate is modeled in ABAQUS by dividing the
plate into a lower part consisting of four layers and an upper part consisting of the topmost layer and then imposing
the TIE constraint on the nodes which are on the intact areas. A mesh size of 100 × 100 × 16 of twenty-node solid
elements (C3D20R) is used, which gives the converged results with a maximum difference of 0.002% among the first
10 frequencies from the results obtained with 160 × 160 × 16 mesh.
The present ZIGT results obtained using the PC, LS and HPLS methods are compared with the LWT FE solution
of [31] and the 3D FE solution in Table 4. The results based on the TOT FE [22] which have the same number
of DOFs per element as the present ZIGT FE are also presented in the table for comparison. The TOT results are
obtained using the HPLS method for the delamination front continuity. The results are compared for both intact and
delaminated plates for a moderately thick plate with S = 10 for seven modes as given in the reference solution [31],
and also for a thinner plate with S = 40 for first ten modes. The mode shapes (normalized with respect to the mass
matrix) for the delaminated plate with S = 10 are shown in Fig. 5. The mode shapes for the delaminated plates consist
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 225

Fig. 4. Single-core sandwich plate with a central delamination.

Table 4
Nondimensional frequencies (ω̄ = ωa 2 / h(ρ/Y2 )1/2 ) of square sandwich plate with a central square delamination.
S Intact plate Delaminated plate
Modea Present 3D FE LWT [31] TOT Modea Present 3D FE LWT [31] TOT
PC LS HPLS
10 (1,1) 1.8557 1.8505 1.85 7.1876 (1,1) 2.1536 1.8934 1.8367 1.8384 1.83 –
(1,2) 3.2325 3.2772 3.25 12.279 (1,2) 3.8852 3.5451 3.2429 3.2042 3.18 –
(2,2) 4.4090 4.2954 4.32 15.646 (2,2) 5.6622 5.1644 4.3689 4.2567 4.29 –
(1,3) 5.4012 5.3160 5.36 17.926 (1,3) 6.4815 5.6455 5.2747 5.1951 5.53 –
(2,3) 6.4379 6.5414 6.21 20.414 (2,3) 7.8902 6.9668 6.2541 6.0582 6.18 –
Local-1 4.1531 4.0694 4.0133 4.6794 3.95 3.9802
Local-2 5.2932 5.2078 5.1813 5.4072 5.28 5.1680
40 (1,1) 6.4703 6.4724 – 14.516 Local-1 4.0495 3.9617 3.9568 3.7956 – 4.0869
(2,1) 10.599 10.593 – 32.673 Local-2 5.2056 5.1642 5.1581 5.0099 – 5.2149
(1,2) 11.105 11.106 – 33.931 Mixed-1 6.9884 6.7334 6.7268 6.6942 – 7.5088
(2,2) 13.945 13.936 – 45.569 Local-3 7.7354 7.6883 7.6867 7.5901 – 10.712
(3,1) 15.367 15.340 – 55.799 Mixed-2 10.373 9.7517 9.6906 9.3264 – 11.007
(1,3) 16.302 16.288 – 57.878 Mixed-3 10.845 10.438 10.377 10.321 – 13.151
(3,2) 17.923 17.848 – 64.535 Local-4 11.484 11.275 11.218 10.744 – 14.742
(2,3) 18.426 18.371 – 65.681 Mixed-4 12.380 11.493 11.458 11.402 – 15.890
(4,1) 20.541 20.458 – 79.728 Mixed-5 12.719 11.760 11.721 11.536 – 19.840
(3,3) 21.749 21.549 – 80.039 Local-5 13.384 13.156 13.139 12.779 – 21.066
a (m, n) denotes the global double Fourier type mode in a simply supported plate with m and n denoting the mode numbers in x and y directions,
respectively.

of (i) global modes (m, n) as in an undelaminated plate, (ii) local modes where only the delaminated areas vibrate,
and (iii) mixed modes where the local vibration of the delaminated portion occurs along with the global vibration.
The results in Table 4 show that the present ZIGT FE predicts the natural frequencies of the undelaminated
sandwich plates with excellent accuracy in comparison with the converged 3D FE solution, with a maximum error of
2.6% and 0.17% for S = 10 and 40, respectively, for the first five modes and 0.9% for S = 40 for up to ten modes.
The present results are as accurate as those based on the LWT [31] for the moderately thick plate, in which the number
of DOFs is at least 18 per node for this problem (as against 7 in the ZIGT FE) and increases proportionally with the
number of layers. The TOT, on the other hand, yields grossly inaccurate results with as large as 288% and 124% error
for S = 10 and 40, respectively, even for the fundamental frequency.
For the delaminated plates, the HPLS method gives by far the best and accurate results out of the three methods
of satisfying the displacement continuity conditions at the delamination front. The ZIGT with HPLS predicts the
frequencies of the global modes of the moderately thick (S = 10) plate within an error of 3% with respect to the
226 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

Fig. 5. Mode shapes of a moderately thick (S = 10) square sandwich plate with a central delamination.

converged 3D FE solution, whereas the corresponding maximum error in the PC and LS methods is 33% and 21%,
respectively. The present results using the HPLS method are also close to the LWT results for both global and local
modes, but differ from the 3D FE solution by 14% for local mode 1. The TOT overpredicts the natural frequencies so
much that the global modes of the delaminated plate could not be found in the first 15 modes predicted by it. The local
modes, which correspond to vibration of the top layer of the plate in the delamination area, are, however predicted by
the TOT with good accuracy. For the thinner plate with S = 40, the first ten modes presented are either local or mixed.
For these modes too, the present ZIGT with the HPLS method yields the most accurate results with a maximum error
of 4.4%, whereas the PC method shows an error of up to 11.2% for the same. The TOT shows an error of up to 72%
for these modes.

5.4. Stress analysis of sandwich plates

For assessing the performance of the present FE formulation for stress analysis of delaminated sandwich plates,
a simply supported double-core sandwich plate with full width-wise delamination located at the middle interface as
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 227

Fig. 6. Double-core sandwich plate with through-width delamination.

shown in Fig. 6 is considered. The laminate lay-up is (0◦ /90◦ /Core/90◦ /0◦ )s , in which the ratio of thickness of each
core (h c ) to that of each face sheet (h f ) is 6. Each layer in the face sheets has equal thickness. The cores are made
of material 2 and the face sheets are made of material 3. The various dimensions of the plate and the delamination
are: length of delamination, d = 105 mm, length of the intact segment, c = 45 mm, plate width, b = 100 mm and
thickness of upper and lower sublaminates, h/2 = 2 mm. The plate is subjected to a uniformly distributed load of
q = 1000 N/m2 . It is modeled with a mesh size of 30 × 30. The longitudinal variations of the deflection and the
inplane stresses σx and σ y at the top and bottom surfaces of the delaminated plate are plotted in Fig. 7 and compared
with the TOT FE and 3D FE solutions. The stresses are computed using Eqs. (1), (4), (6) and (12) at the Gauss points
and then extrapolated to the nodes. In case multiple elements are connected to a node, which have the same material,
their stresses are averaged. The present ZIGT FE results are obtained using the HPLS method, and the 3D FE results
are obtained using a mesh size of 120 × 80 × 24 of C3D20R element. The convergence of the results is verified by
obtaining the results with a denser mesh of 180 × 120 × 24, in which the difference in peak deflection and stresses
is found to be within 0.004%. The present results match very well with the 3D FE solution for the profiles and peak
magnitudes of both deflection and stresses, whereas the TOT shows large error. There is a sharp rise in the inplane
stress σx of the bottom surface at the delamination front, which too is well captured by the ZIGT FE.
Through-the-thickness distributions of the five stress components in the upper delaminated sublaminate, obtained
from the present ZIGT FE, 3D FE and TOT FE, are plotted in Fig. 8 at locations where they are large. The inplane
stresses σx and σ y are plotted near the mid-section of the upper sublaminate and the inplane shear stress τx y is plotted
at its corner. The out-of-plane shear stresses τzx and τ yz are plotted near the edges, x = a and y = 0, respectively. It is
seen that the present FE predicts the distributions of the inplane stresses and the predominant transverse shear stress
τzx across the delaminated sublaminate very accurately, whereas the TOT underestimates the inplane stresses in face
sheets by a large margin.

5.5. Free vibration of sandwich plates with two delaminations

To examine the accuracy of the FE formulation for multiple delaminations, we study the free vibration response
of a simply supported square sandwich plate with two delamination areas located at different planar and interfacial
locations. The geometry and laminate lay-up of the plate are shown in Fig. 9. The dimensions and material properties
of the plate are the same as mentioned in Section 5.2. The present ZIGT FE results for the first ten natural frequencies
are obtained using the three continuity schemes and compared with 3D FE and TOT FE results in Table 5 for S = 10
and 20. The ZIGT and TOT FE results are obtained using a mesh size of 40 × 40, and the 3D FE results are obtained
228 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

Fig. 7. Longitudinal variations of deflection and inplane stresses at the top and bottom surfaces of a delaminated sandwich plate under uniformly
distributed load.

using ABAQUS, employing the C3D20R element with a mesh size of 100 × 100 × 16, which gives the same order of
convergence as reported in Section 5.3. The first ten frequencies are given and all the modes corresponding to these
frequencies have been found to be of the mixed type with both global and local vibrations. It can be seen that, for this
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 229

Fig. 8. Through-thickness distribution of stresses at different locations for the upper delaminated sublaminate.

problem with two delaminations also, the present ZIGT FE formulation, when the continuity conditions are satisfied
using the hybrid method, yields excellent results with a maximum error of 3.9% with respect to the 3D FE solution
for the first ten modes. In comparison, both the conventional methods, PC and LS, yield much worse results with an
error of up to 26% and 20.2%, respectively for S = 10, and 25.5% and 8.3% for S = 20. The TOT, on the other
230 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

Fig. 9. Single-core sandwich plate with multiple delaminations.

hand even failed to predict any mode shape correctly and yields extremely poor results for the fundamental mode with
unacceptably large error of 296% and 204%, for S = 10 and 20, respectively (see Table 5).

6. Conclusions
An FE formulation with a four-node quadrilateral element based on an efficient ZIGT has been developed for
stress and free vibration analysis of laminated plates containing multiple delaminations, using the region approach.
The multiple delaminations can occur at different interfacial and/or planar locations. Three different methods have
been used to satisfy the continuity conditions at the delamination front, out of which two (PC and LS) have been
generally used in previous studies using other higher order theories and the third one (HPLS) was recently proposed
by the authors in [22]. These conditions are satisfied through a transformation matrix, instead of the widely used
approximate penalty approach.
The comparisons with the full-field 3D FE solutions reveal the following:
1. The present formulation with the HPLS method of satisfying continuity yields accurate results with reference
to the full-field 3D FE solutions, for the static and free vibration response of composite as well as highly
inhomogeneous single- and double-core sandwich plates having multiple delaminations that are at different
interfacial and/or planar locations.
2. The formulation accurately predicts the deflection and stresses including the sharp rise in the inplane stresses
at the delamination front, and the natural frequencies and mode shapes of the global modes, most local modes
and hybrid modes of the delaminated sandwich plates with span-to-thickness ratio of 10 and higher.
3. The most widely used PC method, however, can yield very large error for moderately thick plates, and
fundamental and higher modes of vibration. The LS method shows improvement over the PC method, but
the HPLS method gives consistently by far the best accuracy out of these 3 methods.
 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232 231

Table 5
Natural frequencies (Hz) of square sandwich plate with multiple delaminations.
S Mode Present 3D FE TOT (HPLS)
PC LS HPLS
10 1 93.353 86.566 74.424 74.092 293.53
2 155.44 142.22 127.77 126.79 416.54
3 163.54 157.59 131.91 131.15 452.43
4 202.75 196.61 173.51 171.36 518.16
5 242.17 222.98 206.83 204.24 538.96
6 263.99 249.68 213.21 210.92 566.06
7 286.12 256.49 245.09 238.62 576.11
8 290.27 282.55 250.43 245.55 659.31
9 338.34 328.98 311.89 302.09 728.83
10 356.59 343.39 314.65 307.05 753.74
20 1 82.779 73.592 70.140 69.571 211.88
2 136.68 119.59 115.03 112.49 239.17
3 137.01 125.06 116.45 115.91 288.45
4 175.05 160.05 153.64 147.84 386.37
5 190.98 172.35 166.51 161.14 417.71
6 215.68 184.61 172.55 171.92 435.23
7 230.25 197.40 193.35 198.54 469.86
8 240.97 215.02 204.62 203.68 509.63
9 256.57 242.48 240.97 237.79 606.47
10 275.95 253.34 249.25 241.63 614.89

4. The TOT, which has the same number of DOFs as the present theory, even when the HPLS method is used for
continuity, yields grossly inaccurate results for the deflection, stresses, natural frequencies and mode shapes
of delaminated sandwich plates. It overpredicts the natural frequencies so much that the global modes of the
delaminated plates could not be found even in the first 15 modes predicted by it, where the fundamental mode
itself is a global mode.
Finally, the present ZIGT with the hybrid method of satisfying the displacement continuity provides an accurate,
computationally efficient, robust and simple tool for analysis of thin to moderately thick delaminated inhomogeneous
laminated plates including soft-core sandwich plates, which was hitherto not available in literature.
The future work on the formulation will include methods to prevent interpenetration of layers at delaminated
interfaces.

References
[1] A.C. Garg, Delamination–a damage mode in composite structures, Eng. Fract. Mech. 29 (1988) 557–584.
[2] J. Wang, Y. Liu, J. Gibby, Vibrations of split beams, J. Sound Vib. 84 (1982) 491–502.
[3] R. Ramkumar, S. Kulkarni, R. Pipes, Free vibration frequencies of a delaminated beam, in: 34th Annual Technical Conference Proceedings,
Reinforced/Composite Institute, Society of Plastics Industry, 1979, pp. 1–5. Section 22-E.
[4] P. Mujumdar, S. Suryanarayan, Flexural vibrations of beams with delaminations, J. Sound Vib. 125 (1988) 441–461.
[5] C.N. Della, D. Shu, Vibration of delaminated composite laminates: A review, Appl. Mech. Rev. 60 (2007) 1–20.
[6] Y. Zou, L. Tong, G. Steven, Vibration-based model-dependent damage (delamination) identification and health monitoring for composite
structures-a review, J. Sound Vib. 230 (2000) 357–378.
[7] B. Bombale, M. Singha, S. Kapuria, Detection of delamination damage in composite beams and plates using wavelet analysis, Struct. Eng.
Mech. 30 (2008) 699–712.
[8] R.W. Campanelli, J.J. Engblom, The effect of aminations in graphite/peek composite plates on modal dynamic characteristics, Compos. Struct.
31 (1995) 195–202.
[9] F. Ju, H. Lee, K. Lee, Finite element analysis of free vibration of delaminated composite plates, Compos. Eng. 5 (1995) 195–209.
[10] M. Krawczuk, W. Ostachowicz, A. Zak, Dynamics of cracked composite material structures, Comput. Mech. 20 (1997) 79–83.
[11] A. Żak, M. Krawczuk, W. Ostachowicz, Numerical and experimental investigation of free vibration of multilayer delaminated composite
beams and plates, Comput. Mech. 26 (2000) 309–315.
[12] N. Hu, H. Fukunaga, M. Kameyama, Y. Aramaki, F. Chang, Vibration analysis of delaminated composite beams and plates using a higher-order
finite element, Int. J. Mech. Sci. 44 (2002) 1479–1503.
[13] M.-H. Shen, J. Grady, Free vibrations of delaminated beams, AIAA J. 30 (1992) 1361–1370.
232 S. Kapuria and A. Ahmed / Computer Methods in Applied Mechanics and Engineering 345 (2019) 212–232

[14] A. Żak, M. Krawczuk, W. Ostachowicz, Vibration of a laminated composite plate with closing delamination, J. Intell. Mater. Syst. Struct. 12
(2001) 545–551.
[15] H. Chen, M. Hong, Y. Liu, Dynamic behavior of delaminated plates considering progressive failure process, Compos. Struct. 66 (2004)
459–466.
[16] H. Altenbach, An alternative determination of transverse shear stiffnesses for sandwich and laminated plates, Int. J. Solids Struct. 37 (2000)
3503–3520.
[17] J.N. Reddy, Exact solutions of moderately thick laminated shells, J. Eng. Mech. 110 (1984) 794–809.
[18] A.G. Radu, A. Chattopadhyay, Dynamic stability analysis of composite plates including delaminations using a higher order theory and
transformation matrix approach, Int. J. Solids Struct. 39 (2002) 1949–1965.
[19] A. Kumar, R. Shrivastava, Free vibration of square laminates with delamination around a central cutout using HSDT, Compos. Struct. 70
(2005) 317–333.
[20] T. Park, S.Y. Lee, G. Voyiadjis, Finite element vibration analysis of composite skew laminates containing delaminations around quadrilateral
cutouts, Composites B 40 (2009) 225–236.
[21] M.-H. Noh, S.-Y. Lee, Dynamic instability of delaminated composite skew plates subjected to combined static and dynamic loads based on
HSDT, Composites B 58 (2014) 113–121.
[22] A. Ahmed, S. Kapuria, Third order theory based quadrilateral element for delaminated composite plates with a hybrid method for satisfying
continuity at delamination fronts, Compos. Struct. 181 (2017) 84–95.
[23] L.H. Tenek, E.G. Henneke, M.D. Gunzburger, Vibration of delaminated composite plates and some applications to non-destructive testing,
Compos. Struct. 23 (1993) 253–262.
[24] L. Yam, Z. Wei, L. Cheng, W. Wong, Numerical analysis of multi-layer composite plates with internal delamination, Comput. Struct. 82
(2004) 627–637.
[25] K. Alnefaie, Finite element modeling of composite plates with internal delamination, Compos. Struct. 90 (2009) 21–27.
[26] S.G. Lekhnitskii, Strength calculation of composite beams, Vest. Inzh. Tek. 9 (1935) 137–148.
[27] D. Robbins, J. Reddy, Modelling of thick composites using a layerwise laminate theory, Internat. J. Numer. Methods Engrg. 36 (1993)
655–677.
[28] A. Toledano, H. Murakami, A composite plate theory for arbitrary laminate configurations, ASME J. Appl. Mech. 54 (1) (1987) 181–189.
[29] E. Carrera, Evaluation of layerwise mixed theories for laminated plates analysis, AIAA J. 36 (1998) 830–839.
[30] E. Barbero, J. Reddy, Modeling of delamination in composite laminates using a layer-wise plate theory, Int. J. Solids Struct. 28 (1991)
373–388.
[31] M. Marjanović, D. Vuksanović, Layerwise solution of free vibrations and buckling of laminated composite and sandwich plates with
embedded delaminations, Compos. Struct. 108 (2014) 9–20.
[32] S.K. Kumar, M. Cinefra, E. Carrera, R. Ganguli, D. Harursampath, Finite element analysis of free vibration of the delaminated composite
plate with variable kinematic multilayered plate elements, Composites B 66 (2014) 453–465.
[33] N. Kharghani, C.G. Soares, Behavior of composite laminates with embedded delaminations, Compos. Struct. 150 (2016) 226–239.
[34] E. Carrera, C0 Reissner-Mindlin multilayered plate elements including zig-zag and interlaminar stress continuity, Internat. J. Numer. Methods
Engrg. 39 (1996) 1797–1820.
[35] M. Cho, R. Parmerter, Efficient higher order composite plate theory for general lamination configurations, AIAA J. 31 (1993) 1299–1306.
[36] S. Kapuria, An efficient coupled theory for multilayered beams with embedded piezoelectric sensory and active layers, Int. J. Solids Struct.
38 (2001) 9179–9199.
[37] E. Carrera, Historical review of zig-zag theories for multilayered plates and shells, Appl. Mech. Rev. 56 (3) (2003) 287–308.
[38] S. Kapuria, S.D. Kulkarni, An improved discrete Kirchhoff quadrilateral element based on third-order zigzag theory for static analysis of
composite and sandwich plates, Internat. J. Numer. Methods Engrg. 69 (2007) 1948–1981.
[39] S. Kapuria, P. Dumir, N. Jain, Assessment of zigzag theory for static loading, buckling, free and forced response of composite and sandwich
beams, Compos. Struct. 64 (2004) 317–327.
[40] R. Schmidt, L. Librescu, Geometrically nonlinear theory of laminated anisotropic composite plates featuring interlayer slips, Nova J. Math
Game Theory Algebra 5 (1996) 131–147.
[41] U.I. M Di Sciuva, L. Librescu, Effects of interfacial damage on the global and local local static response of cross-ply laminates, Int. J. Fract.
96 (1999) 17–35.
[42] M.D.S. U Icardi, L. Librescu, Dynamic response of adaptive cross-ply cantilevers featuring interlaminar bonding imperfections, AIAA J. 38
(2000) 499–506.
[43] M. Cho, J.-S. Kim, Higher-order zig-zag theory for laminated composites with multiple delaminations, J. Appl. Mech. 68 (2001) 869–877.
[44] H.S. Kim, A. Chattopadhyay, A. Ghoshal, Characterization of delamination effect on composite laminates using a new generalized layerwise
approach, Comput. Struct. 81 (2003) 1555–1566.
[45] J. Oh, M. Cho, J.-S. Kim, Dynamic analysis of composite plate with multiple delaminations based on higher-order zigzag theory, Int. J. Solids
Struct. 42 (2005) 6122–6140.
[46] C. Jeyachandrabose, J. Kirkhope, L. Meekisho, An improved discrete Kirchhoff quadrilateral thin-plate bending element, Internat. J. Numer.
Methods Engrg. 24 (1987) 635–654.
[47] O.C. Zienkiewicz, R.L. Taylor, The Finite Element Method, Vol. 2: Solid Mechanics, Butterworth-Heinemann, Oxford, 2000.
[48] S. Kapuria, A coupled zig-zag third-order theory for hybrid cross-ply plates, ASME J. Appl. Mech. 71 (2004) 604–614.
[49] S. Srinivas, A. Rao, Bending, vibration and buckling of simply supported thick orthotropic rectangular plates and laminates, Int. J. Solids
Struct. 6 (11) (1970) 1463–1481.

View publication stats