Composite Structures 325 (2023) 117585

Contents lists available at ScienceDirect

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

Efficient zigzag theory-based spectral element model for guided waves in

composite structures containing delaminations
Mayank Jain, Santosh Kapuria ∗
Department of Applied Mechanics, Indian Institute of Technology Delhi, New Delhi 110016, India

ARTICLE INFO ABSTRACT

Keywords: Lamb wave-based structural health monitoring offers excellent potential for real-time delamination detection
Efficient layerwise theory in laminated structures. It necessitates fast and accurate numerical simulation of guided wave interaction
Time-domain spectral element with delaminations. Towards this objective, we present an efficient layerwise zigzag theory (ZIGT) based
Delamination
time-domain spectral element for wave propagation analysis of composite beam- and panel-type structures
Lamb wave
(strips) containing delamination at arbitrary locations. The ZIGT delivers accuracy like layerwise theories
Composite
Structural health monitoring
by allowing slope discontinuities in the in-plane displacement field at layer interfaces while maintaining
computational efficiency like equivalent single-layer (ESL) theories, making it ideal for such analyses. The
high-order elemental nodes at Lobatto points have only four displacement variables, irrespective of the number
of layers in the laminate. Delamination is modelled using the region approach, adapting the recently proposed
hybrid point-least squares continuity method to satisfy the nonlinear displacement field continuity at the
delamination fronts. The model’s efficacy is examined vis-à-vis elasticity-based elements, ESL theory-based
elements, and the ZIGT-based standard finite element for free vibration and guided wave propagation responses
of composite strips featuring multiple delaminations. The present model shows superior overall efficiency,
accuracy, and convergence. An application of the model is also explored for identifying delamination locations
from Lamb wave velocity fields.

1. Introduction ability to travel long distances with little attenuation and detect both
surface and internal defects, being highly sensitive, and being easy to
Fibre-reinforced composite materials have several superior at- actuate and sense, they have been used for both off-line NDE [6–
tributes over metals, such as high specific strength, stiffness, fatigue, 12] and online SHM [13–15] for identifying the sub-surface defects in
and corrosion resistance [1], which make them suitable for various laminated structures by exploiting their interactions with the defects.
high-performance structures including biomedical devices, aircraft, Understanding and predicting features of guided waves interacting
spacecraft, automobiles, marine vehicles, turbine blades, and sports with delaminations require accurate simulation tools which must be
products. However, their highly anisotropic and inhomogeneous nature computationally efficient and fast for deployment for the design of
makes them prone to a rather serious mode of invisible sub-surface effective NDE/SHM systems. Damage detection is essentially an inverse
damage namely delamination at ply interfaces, which can arise during problem requiring many forward simulations when model-assisted or
manufacturing or service under thermomechanical loading condi- physics-informed data-driven strategies are employed. The computing
tions [2,3]. In addition to reducing stiffness and strength, delamination
efficiency becomes even more important in such cases apart from
may undergo an unstable growth under in-plane compressive loading,
contributing to the sustainable computational mechanics by energy
leading to catastrophic failure of the structure. Thus, early detection
saving [16]. Although there has been notable progress on modelling
of delaminations through periodic non-destructive evaluation (NDE) or
delamination in laminated structures [17–19], the development of fast
even their real-time monitoring using onboard transducers, known as
and accurate prediction tools for wave propagation in delaminated
structural health monitoring (SHM), becomes imperative for ensuring
structures still needs much attention. The present paper is aimed to
the integrity and reliability of composite structures.
Lamb waves are ultrasonic-guided elastic waves propagating be- cover this critical gap in the literature.
tween the two bounding surfaces of thin-walled structures whose thick- The elasticity theory-based three-dimensional (3D) solid [20] and
ness is much less than the wavelength of the wave modes [4,5]. Due 2D plane stress/strain [7] finite elements offer arguably the most
to their several favourable attributes such as high group velocity, the amenable choice for modelling delaminations in composite structures

∗ Corresponding author.
E-mail address: [email protected] (S. Kapuria).

https://doi.org/10.1016/j.compstruct.2023.117585
Received 17 May 2023; Received in revised form 26 August 2023; Accepted 28 September 2023
Available online 4 October 2023
0263-8223/© 2023 Elsevier Ltd. All rights reserved.
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

by way of introducing two sets of coincident nodes at the delaminated and shear traction-free conditions at the top and bottom surfaces. These
interfaces, which are either unconnected or connected by contact ele- theories are also called zigzag theories (ZIGTs). Comparisons with 3D
ments [21]. Thus, they have been widely used to study the effect of de- elasticity theory-based solutions have established that the ZIGTs have
lamination on the free vibration response [20,22,23] and guided wave indeed attained their objective of high accuracy at less computational
propagation [7,21,24–31] in laminated structures. Another favoured cost for the static, dynamic, and buckling response of composite and
approach for modelling delaminated structures is based on the layer- sandwich beams [56,57], plates [58–62], and shells [63]. There have
wise theories (LWTs) in which additional degrees of freedom (DOFs) in been early attempts to model delaminations using the ZIGTs, by incor-
terms of jump discontinuities are introduced in the displacement field porating additional displacement jumps variables in the displacement
for modelling delaminations [32–36]. Both the continuity-based and field approximation for static and dynamic analyses of delaminated
LWT-based models, though accurate, are inherently computationally composite plates [64–66]. However, this approach would satisfy nei-
inefficient with the number of DOFs increasing with the number of ther the displacement field continuity at the delamination fronts nor the
laminas, which makes them unsuitable, particularly for transient wave shear traction-free conditions at the delaminated interfaces. Further,
propagation problems in multi-layered beam, plate, and shell-type delaminations have been modelled using the ZIGTs by introducing
structures. weak interfaces [67] or additional weak layers at the delaminated
On the other end of the spectrum lie the equivalent single layer interfaces [68–70]. In these cases too, the continuity of the displace-
(ESL) laminate theory-based models which assume a global variation ment field would not be satisfied at the delamination fronts. Kapuria
of the displacement field across the laminate thickness and are compu- and Ahmed [71,72] have been able to satisfy both these conditions in
tationally efficient with DOFs independent of the lamina count [37]. FE models for free vibration analysis of composite plates and shells
Models based on the ESL theories usually adopt the so-called region containing multiple delaminations by employing the region approach
approach to incorporate delaminations [38,39]. In this approach, the in the context of ZIGTs for the first time. The HPLSC method [50]
delaminated region is divided into sub-laminates separated by the was extended for the ZIGT to ensure displacement field continuity at
delaminations, which are modelled independently using the ESL theory. delamination fronts.
The sub-laminates are connected to the integral regions by applying The conventional FEM employed in the aforementioned studies,
the appropriate continuity conditions at the junctions between the however, is known to suffer from several disadvantages such as nu-
delaminated and adjacent integral regions, called the delamination merical dispersion, dissipation errors, and slow convergence rate [73],
fronts. which make it computationally inefficient and even unreliable for high-
Several FE models based on the region approach have been devel-
frequency wave propagation problems. It has spurred efforts towards
oped for free vibration and dynamic stability analyses of delaminated
developing various non-conventional strategies to improve accuracy
composite beams and plates using the ESL theories such as the first-
and convergence. Details of such methods can be seen in Refs. . [74–
order shear deformation theory (FSDT [40–44] and the refined third-
76], although only a few of them have been implemented or are indeed
order shear deformable theory (TOT) [45–49]. In these studies, the
implementable for wave propagation analysis of laminated structures
continuity at the delamination front is enforced by equating the in-
featuring delaminations [77–79].
plane and rotational displacements of a sub-laminate to those of intact
One of the most effective ones among the non-conventional methods
laminate at the former’s reference plane (mid-plane). It is called here
which enjoys all the generalities of the conventional FEM in solving
the point-continuity (PC) model. For a linear variation of the in-plane
complex geometry problems and yet has faster convergence and higher
displacements, as in the FSDT, this condition automatically satisfies the
accuracy is the time-domain spectral element method (TDSEM). Mo-
continuity of displacements at all points on the delamination fronts
tivated by the exceptional accuracy of spectral methods [80], Patera
in the thickness direction. However, in the case of the higher order
developed the TDSEM in the context of fluid dynamics using higher-
theories such as the TOT, wherein the in-plane displacement variation
order orthogonal polynomials, namely, Chebyshev polynomials [81]
is nonlinear, the model would fail to maintain the displacement con-
and Legendre polynomials [82] as interpolation functions in the FE
tinuity through the sub-laminate thickness at the delamination front.
formulation. It yields exponential convergence (∝ ( 1𝑝 )𝑝 ) with the inter-
For such theories, a new continuity displacement model developed
by Kapuria and Ahmed [50] satisfies the displacement continuity at polation polynomial order 𝑝, which is known as spectral convergence.
the delamination front by equating the in-plane displacements at the Further, these functions do not suffer from undesirable oscillations near
sub-laminate’s mid-plane and minimizing the deviations between the the element ends, known as the Runge phenomenon, encountered by
in-plane displacement fields of the adjacent intact laminate and the the high-order Lagrange functions used in the standard p-FEM [83].
sub-laminate with respect to the rotation variables in the delaminated Two-dimensional [84,85] and 3D [86,87] elasticity-based TDSEs
region using the least squares method [51]. Since this method combines have been employed to study the interaction of guided waves with
the point-continuity and least squares methods, they termed it the delaminations in laminated structures. Rekatsinas et al. [88] have
hybrid point-least square continuity (HPLSC) model. The developed extended their third-order LWT-based spectral plate element [89] by
model is generic for an arbitrary number of delaminations occurring introducing displacement jump DOFs for modelling delamination in
at arbitrary interfaces of the laminate. As the ESL theories do not multi-layered plates. Among the ESL theories, only FSDT-based TDSEs
account for the slope discontinuities generally seen in the displacement have been developed in conjunction with the region approach for
field’s variation at the interfaces of laminas of different material prop- simulation of guided waves in composite structures in the presence of
erties [52,53], they are susceptible to yielding erroneous predictions, delaminations [77,90–92].
particularly for ultrasonic guided wave propagation problems featuring The authors [93,94] have recently developed TDSEs based on the
very short wavelengths (which act as the effective span of the structure TOT for wave propagation analysis of isotropic and laminated compos-
resulting in a rather low span-to-thickness ratio). ite strips, but no study has been reported so far on using the TOT-based
The efficient layerwise theories (ELTs) were pioneered by Ambart- SEs for wave propagation analysis of delaminated structures. So far as
sumyan [54] and Rath and Das [55] to combine the high accuracy of ELTs are concerned, a spectral strip element has been developed very
the LWTs and the high computational efficiency of the ELTs. Thus, the recently based on the third-order ZIGT [95], which has shown excel-
in-plane displacements are initially considered to follow a layerwise lent accuracy, convergence rate, and efficiency in predicting guided
variation, allowing for the slope discontinuities at layer interfaces like wave propagation response of single- and multi-material laminated
the LWTs, but the number of primary variables is made independent of panels [96]. Despite these overwhelming advantages, no TDSE model
the number of layers like the ESL theories, with the help of interfacial based on an ELT has been developed to date for wave propagation
continuity conditions of displacements and transverse shear stresses analysis of delaminated multi-layered structures to the best of the

2
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

authors’ knowledge. This gap applies to both displacement jump and The angle between the principal material direction 𝑥1 of the 𝑘th layer
region approaches of modelling delaminations. and the structural axis 𝑥 is denoted as 𝜃𝑘 .
In this article, we present for the first time an TDSE model based In line with the majority of 1D and 2D laminate theories, the
on an ELT for wave propagation analysis of laminated beam- and transverse normal stress 𝜎𝑧 is assumed to be negligible. The present
panel-type structures (strips) containing multiple delaminations. The study considers the strip either in the plane stress state (𝜎𝑦 = 𝜏𝑦𝑧 = 𝜏𝑥𝑦 =
formulation employs the C1 -continuous spectral element based on the 0) for a beam of small width or the plane strain state (𝜀𝑦 = 𝛾𝑦𝑧 = 𝛾𝑥𝑦 = 0)
third-order ZIGT recently developed by the authors [96] to model the for a panel of infinite width along 𝑦-direction. There is no restriction on
intact laminate and delaminated sub-laminates, which automatically the lamina fibre orientations for the beams of small width. However,
ensures the satisfaction of traction-free conditions at the delaminated for the panel of infinite width, the current work is restricted to cross-
interfaces. The delaminations can occur at arbitrary interfacial and ply laminates only, although the formulation can be easily extended to
longitudinal locations and are modelled using the region approach. angle-ply panels by considering the displacement 𝑢𝑦 (𝑥, 𝑧) along the 𝑦-
The high-order 𝑛-node spectral element has only four DOFs per node, direction. As a result of these assumptions, the 3D constitutive relations
namely, in-plane displacement 𝑢0 , transverse displacement 𝑤0 , slope of a lamina lead to two nonzero independent stress components, the
d𝑤0 normal axial stress 𝜎𝑥 and the transverse shear stress 𝜏𝑧𝑥 , being related
, and shear rotation 𝜓0 at the mid-surface of the laminate/sub-
d𝑥 to the normal axial strain 𝜀𝑥 and transverse shear strain 𝛾𝑧𝑥 as
laminate, irrespective of the number of layers in the respective lam-
inate. The deflection 𝑤0 is interpolated using C1 -continuous Lobatto 𝜎𝑥 = 𝑄̂ 11 𝜀𝑥 , 𝜏𝑧𝑥 = 𝑄̂ 55 𝛾𝑧𝑥 (1)
basis functions [97], while the C0 -continuous Lobatto basis functions
are used to interpolate 𝑢0 and 𝜓0 . The HPLSC method of Ref. [71] is through the reduced stiffness coefficients 𝑄̂ 11 and 𝑄̂ 55 in the structural
adapted for the spectral strip element to satisfy the continuity of the coordinate system (𝑥, 𝑦, 𝑧). The expressions of these coefficients for
nonlinear in-plane displacement field across the laminate thickness at plane stress and plane strain strips are available in [96] in terms of
the delamination fronts. These conditions are imposed using a trans- Young’s moduli 𝑌𝑖 , shear moduli 𝐺𝑖𝑗 , and Poisson’s ratio 𝜈𝑖𝑗 for the
formation matrix that relates the nodal DOFs of the sub-laminate at lamina.
the delamination front to those of the intact laminate. The solution The displacement fields in the intact laminate and each of the sub-
based on the related ESL theory, TOT, is obtained as a special case laminates are approximated using the efficient ZIGT [56], which will
of the present model. A comprehensive numerical study has been be briefly recalled below. It needs to ensure the continuity of the
displacement fields in the intact laminate and sub-laminates across the
performed to validate and assess the TDSE model for free vibration
delamination front, which is non-trivial for the nonlinear and layerwise
and guided wave propagation responses of composite strips featuring
distribution of the in-plane displacement and will be discussed later.
single and multiple delaminations. The Lamb wave propagation exam-
ple is selected so as to represent actual SHM applications. To critically
2.1. Zigzag laminate theory
examine the present results, we compare them with the previously
published results and the elasticity-based FE solution from ABAQUS.
Consider a sub-laminate (or an intact laminate) of 𝐿 number of
The performance of the conventional two-node FE solution based on the
layers of total thickness ℎ. The layers are perfectly bonded at the in-
same laminate theory is also compared with the proposed SE solution.
terfaces. For the purpose of describing the displacement field variations
Further, we delineate the role of the slope discontinuities in the in-
across the sub-laminate’s thickness, the mid-surface of the sub-laminate
plane displacement approximation by comparing it with the TOT-based
is considered as the reference (𝑧 = 0) plane. The zigzag theory assumes
SE solution. Finally, we use the present simulation to explore the the in-plane displacement 𝑢 to follow a global third-order variation in
identification of the longitudinal as well as interfacial locations of 𝑧 superimposed with a linear variation in each layer and the trans-
delaminations using Lamb wave velocity field measurements. verse displacement 𝑤 independent of 𝑧, i.e., uniform throughout the
laminate thickness. For both plane stress and plane strain conditions,
2. Efficient layerwise modelling of composite laminate with mul- the displacement field does not depend on 𝑦-direction. Thus, it can be
tiple delaminations expressed for any layer 𝑘 as a function of 𝑥, 𝑧, and 𝑡 as

Consider a laminated composite strip featuring multiple delamina- 𝑢(𝑥, 𝑧, 𝑡) = 𝑢𝑘 (𝑥, 𝑡) − 𝑧𝑤0,𝑥 + 𝑧𝜓𝑘 (𝑥, 𝑡) + 𝑧2 𝜉(𝑥, 𝑡) + 𝑧3 𝜂(𝑥, 𝑡),
tions at arbitrary interfacial and axial locations (Fig. 1(a)). The laminas 𝑤(𝑥, 𝑧, 𝑡) = 𝑤0 (𝑥, 𝑡) (2)
are made of unidirectional fibre-reinforced composite materials, which
exhibit orthotropic material symmetry. The delaminations are of known where 𝑤0 denotes the deflection at the mid-plane (𝑧 = 0) and □,𝑥
lengths which do not change during deformations. The planes normal represents 𝜕□𝜕𝑥
. In Eq. (2), the layerwise linear variation for layer 𝑘 is
to a delaminated interface at the start and the end of the delamination represented by the axial displacement 𝑢𝑘 and shear rotation 𝜓𝑘 , whereas
are called delamination fronts (Fig. 1(b)). The region approach [38, the global variation of third order in 𝑧 by 𝜉 and 𝜂, the second and
39] is popular for modelling delaminated structures using ESL theo- third-order coefficients of 𝑧, respectively.
ries [40,43,45,48] and has been recently developed in the context of For 𝐿 number of layers in the laminated strip, 𝑢𝑘 (𝑢1 , 𝑢2 , …, 𝑢𝐿 )
the ZIGTs in Refs. [71,72]. In this approach, the plate is divided into and 𝜓𝑘 (𝜓1 , 𝜓2 , …, 𝜓𝐿 ) yield 2𝐿 number of unknown displacement
regions separated by the delamination fronts and delaminated inter- variables. On adding the global field variables 𝜉, 𝜂, and 𝑤0 , the total
faces. Thus, the regions feature either the intact laminate (regions 𝐼𝑖 in number of unknown variables is 2𝐿 + 3, which increases with 𝐿 like in
Fig. 1(b)) or sub-laminates created by the delaminations (regions 𝐷𝑖 in the LWTs and would result in high computation efforts for many layers.
Fig. 1(b)). Each of the regions 𝐷𝑖 between two delamination fronts is Therefore, to make the number of primary field variables independent
of 𝐿 and thereby the model computationally efficient, the following
modelled independently using a separate set of finite/spectral elements
conditions are imposed: (i) shear traction-free condition (𝜏𝑧𝑥 = 0) at
(Fig. 1(c)) to allow both the in-plane and transverse displacements to
the bottom and top surfaces of the sub-laminate, yielding 2 equations
be discontinuous across the delaminated interfaces.
and (ii) continuity of 𝑢 and 𝜏𝑧𝑥 at each of the (𝐿 − 1) interfaces in the
The structure axes 𝑥 and 𝑧 are along the longitudinal and thickness
sub-laminate, yielding (2𝐿 − 2) conditions. These 2𝐿 conditions help us
directions, respectively. The numbering of layers starts from the bottom
to eliminate 2𝐿 unknowns and express 𝑢 in Eq. (2) in terms of only three
of a sub-laminate (or the intact laminate), as shown in Fig. 2. 𝑧0 denotes
primary displacement variables, namely, axial displacement 𝑢0 , slope
the 𝑧-location of the bottom surface of the first layer (𝑘 = 1) and 𝑧𝑘 the
𝑤0,𝑥 , and shear rotation 𝜓0 at the reference plane of the sub-laminate
top surface of the 𝑘th layer (𝑘 = 1, 2, … , 𝐿) of the sub-laminate. The
as
layer through which the mid-surface (𝑧 = 0) passes is denoted by 𝑘0 . If
the mid-surface lies on an interface, the layer above the interface is 𝑘0 . 𝑢(𝑥, 𝑧, 𝑡) = 𝑢0 (𝑥, 𝑡) − 𝑧𝑤0,𝑥 (𝑥, 𝑡) + 𝑅𝑘 (𝑧)𝜓0 (𝑥, 𝑡) (3)

3
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 1. (a) Configuration of a laminated strip containing multiple delaminations at arbitrary interfacial and longitudinal locations, showing structural axes 𝑥 and 𝑧. (b) Regions
containing intact laminates 𝐼𝑖 and sub-laminates 𝐷𝑖 separated by delaminated interfaces around a delamination zone. (c) Spectral element model of these regions having a separate
set of elements for each sub-laminate with common nodes at delamination fronts.

material in layer 𝑘 and 𝛿() denotes the first-order variational operator.

The symbol ⟨…⟩ denotes the integration over the sub-laminate thick-
∑𝐿 𝑧−
𝑘
ness consisting of 𝐿 layers, ⟨…⟩ = (…) 𝑑𝑧 and □̄ represents
∫𝑧+
𝑘=1 𝑘−1
specified values of □ at the transverse boundaries of the system.
We now substitute the 𝑧-variations of displacement and strains,
associate the functions of 𝑧 with the stress components 𝜎𝑥 and 𝜏𝑧𝑥
and material density 𝜌𝑘 , and apply the integration ⟨…⟩ over the sub-
laminate thickness. This process leads to expressing the elemental
contribution to the integral in Eq. (4) in terms of the generalized
Fig. 2. Geometry details and reference plane of a sub-laminate. displacement vector 𝒖̄ and strain vector 𝜺̄ as

𝑇𝑒 = [𝛿 𝐮̄ T 𝐈̃𝐮̄̈ + 𝛿 𝜺̄ T 𝐃̄
̄ 𝜺 − 𝛿 𝐮̄ T 𝐟𝑢 ]𝑏(𝑥) 𝑑𝑥 (5)
∫𝑥𝑒
where 𝑅𝑘 (𝑧)
is a cubic function of 𝑧 for layer 𝑘, which depends on
where
the laminate lay-up and material properties of all layers. Its explicit { }T { }T
expression can be seen in [96]. 𝐮̄ = 𝑢0 −𝑤0,𝑥 𝜓0 𝑤0 , 𝜺̄ = 𝑢0,𝑥 −𝑤0,𝑥𝑥 𝜓0,𝑥 𝜓0 ,
Note that the above conditions ensure that at the delaminated { }T
interfaces, the shear tractions vanish identically, which is not satisfied 𝐟𝑢 = 0 0 0 𝑞𝑧 ,
by the ELT models that employ Heavyside’s function to account for the (6)
displacement discontinuity at delaminated interfaces [64–66].
𝐈̃ and 𝐃
̄ correspond to the inertia and stiffness coefficients matrices,
2.2. Weak formulation via Hamilton’s variational principle whose expressions are available in [96].

The development of the spectral element would require a weak 3. High-order spectral strip element
formulation of the governing equations of motion of the structure based
on the above ZIGT approximations. Hamilton’s variational principle is The weak form of governing equations in Eq. (4) is solved by
employed to directly obtain the Ritz–Galerkin type weak formulation discretizing the structural domain using the spectral (finite) element
for the strip structure based on the ZIGT. A normal traction 𝑞𝑧 (𝑥) per method [81]. SEs commonly employ Lagrange interpolations defined
unit area is considered to act on the bottom or top surface of the sub- on non-uniformly spaced nodal grids [83]. These shape functions, how-
laminate/intact laminate. Since 𝑤 is uniform over the thickness and the ever, are only C0 -continuous, which would not suffice while dealing
strip is flat, the 𝑧-location of 𝑞𝑧 would not matter. The strip’s width is with many beam, plate, and shell elements such as the ones based
𝑏(𝑥) which can be a function of 𝑥 for the beam and equals to unity for on the classical lamination theory (CLT), TOT, and the present ZIGT.
infinite panels. Considering that the 𝑧-variations of the field variables The highest orders of derivatives of displacement variables appear-
are known as per the ZIGT and there is no 𝑦-dependence, the volume ing in Eq. (5) suggest that the present ZIGT requires C0 -continuous
interpolations N for 𝑢0 and 𝜓0 and C1 -continuous interpolations 𝐍 ̄
integral in Hamilton’s principle can be reduced to an integral in 𝑥 as
follows: for 𝑤0 and 𝜃 (= 𝑤0,𝑥 ). The authors [96] have developed a spectral
𝑛𝑒 element (Fig. 3) for laminated strips based on the ZIGT satisfying these
∑
[𝑏(𝑥)⟨𝜌𝑘 (𝑥)𝑢𝛿𝑢
̈ + 𝜌𝑘 (𝑥)𝑤𝛿𝑤
̈ + 𝜎𝑥 𝛿𝜀𝑥 + 𝜏𝑧𝑥 𝛿𝛾𝑧𝑥 ⟩ + 𝑏(𝑥)𝑞𝑧 (𝑥)𝛿𝑤] d𝑥 continuity requirements. The same interpolations are employed here in
∫𝑥𝑒 developing the spectral element for the sub-laminates.
𝑒=1
| Accordingly, the field variables 𝑢0 , 𝜓0 , and 𝑤0 are interpolated with
−𝑏(𝑥)⟨𝜎̄ 𝑥 𝛿𝑢 + 𝜏̄𝑧𝑥 𝛿𝑤⟩| = 0, ∀ 𝛿𝑢0 , 𝛿𝑤 (4)
|𝑥 element nodes located at the Gauss–Lobatto–Legendre (GLL) points as
In the above expression, we have considered the strip structure to be [ ]{ }T
an assembly of 𝑛𝑒 elements of domain 𝑥𝑒 . 𝜌𝑘 (𝑥) is the mass density of 𝑢0 = 𝑁 1 𝑁 2 … 𝑁 𝑛 𝑢01 𝑢02 … 𝑢0𝑛 = 𝐍𝒖𝒆𝟎 ,

4
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 3. Spectral strip element based on the ZIGT, showing 𝑛 nodes and four DOFs per node: Distribution of nodes is unequal with GLL points spacing.

[ ]{ }T
̄ 𝑒, 𝐊
̄ 𝑒 , and 𝐏̄ 𝑒 are
𝜓0 = 𝑁1 𝑁2 … 𝑁𝑛 𝜓01 𝜓02 … 𝜓0𝑛 = 𝐍𝝍 𝒆𝟎 (7) The integrals in Eq. (13) for element matrices 𝐌
[ ]{ }T computed using the Lobatto quadrature described in [96]. For exact
𝑤0 = 𝑁̄ 1 𝑁̄ 2 … 𝑁̄ 2𝑛 𝑤01 𝜃1 𝑤02 𝜃2 … 𝑤0𝑛 𝜃𝑛 integration of these integrals, we consider (2𝑛 − 1), (2𝑛 + 1), and (𝑛 + 1)
̄ 𝒆
= 𝐍𝒘 integration points for 𝐊 ̄ 𝑒, 𝐌
̄ 𝑒 , and 𝐏̄ 𝑒 , respectively. In this case, of
𝟎
̄ 𝐈,
course, 𝐃, ̃ 𝑏, and 𝑞𝑧 are assumed to be uniform over the element.
Both C0 -continuous and C1 -continuous spectral interpolations have
These numbers will change if the above parameters vary along the
been shown to eliminate Runge oscillations near the element ends and
element.
are guaranteed to have a maximum value of unity [97].
The element displacement vector of element 𝑒 is introduced as
4. Satisfying continuity at delamination fronts
⎧ ⎫
⎪ 𝒖𝒆𝟎 ⎪
⎪ ⎪
𝐔∗ 𝑒 = ⎨𝒘𝒆 ⎬ , (8) Various models employed in the literature for modelling the inter-
⎪ 𝟎⎪ action of delaminated interfaces can be classified into (i) free-mode
⎪𝝍 𝒆 ⎪ model [98], which considers no interaction between the adjacent in-
⎩ 𝟎⎭
terfaces during the deformation allowing both opening and interpene-
and 𝐮̄ and 𝜺̄ in Eq. (6) are expressed in terms of 𝐔∗ 𝑒 using the
tration, (ii) constrained mode model [39] which considers the trans-
interpolations defined in Eq. (7) as
verse displacements of the interfaces as equal, thereby allowing no
⎧ ⎫ ⎡𝐍 𝟎 𝟎⎤ ⎧ 𝒆 ⎫ opening, and (iii) the contact element model [21], which disallows
⎪ 𝑢0 ⎪ ⎢ ⎪ 𝒖𝟎 ⎪
− 𝑙2𝑒 d𝐍 𝟎 ⎥⎥ ⎪ ⎪
̄ interpenetration. In the current study, the free-mode model is adopted.
⎪−𝑤0,𝑥 ⎪ ⎢ 𝟎 ̂ ∗ 𝐔∗ 𝑒 ,
𝒖̄ = ⎨ ⎬= d𝜉 ⎨𝒘𝒆 ⎬ ⇒ 𝒖̄ = 𝐍 Although, this model allows physically inadmissible interpenetration,
⎪ 𝜓0 ⎪ ⎢⎢ 𝟎 𝟎 𝐍⎥ ⎪ 𝟎 ⎪
⎥ it has shown excellent accuracy in predicting natural frequencies of
⎪ 𝑤0 ⎪ ⎣ 𝟎 𝐍̄ 𝟎⎦ ⎪ 𝒆⎪
⎩ ⎭ ⎩𝝍 𝟎 ⎭ both opening and non-opening modes of composite beams compared
⎫ ⎡ 𝑙𝑒 𝟎 ⎤⎧ ⎫
2 d𝐍 to the experimental results [51]. Further, this assumption will not
⎧ 𝟎
⎪ 𝑢0,𝑥 ⎪ ⎢⎢ d𝜉 2
⎥ ⎪ 𝒖𝒆 ⎪
⎥ 𝟎
alter the main conclusions drawn from the ensuing numerical study.
⎪−𝑤 ⎪ 𝟎 − 4 d 𝐍̄ 𝟎 ⎥ ⎪ 𝒆⎪
𝜺̄ = ⎨ 0,𝑥𝑥 ⎬ = ⎢ 𝑙𝑒 2 d𝜉 2 ̂ ∗ 𝐔∗ 𝑒 , (9) The intact regions (denoted by 𝐼) of the laminate are modelled along
⎢ ⎨𝒘 ⎬ ⇒ 𝜺̄ = 𝐁
⎪ 𝜓0,𝑥 ⎪ 2 d𝐍 ⎥ ⎪ 𝟎 ⎪ its mid-surface with strip elements having the entire laminate as the
𝟎 𝟎
⎪ 𝜓0 ⎪ ⎢⎢ 𝑙𝑒 d𝜉 ⎥ ⎪ 𝒆 ⎪ cross-section. Thus, the nodal DOFs of these elements correspond to
𝐍 ⎥⎦ ⎩ 𝟎 ⎭
⎩ ⎭ ⎣ 𝟎 𝝍
𝟎
the mid-surface of the intact laminate. On the other hand, in the
2𝑥 − (𝑥𝑒1 + 𝑥𝑒𝑛 ) delaminated regions (denoted by 𝐷), the sub-laminates generated by
where 𝜉 is the local natural coordinate given by 𝜉 = . the delaminated interfaces are modelled with independent sets of strip
𝑙
Having obtained 𝐍 ̂ ∗ matrices, the DOFs in 𝐔∗ 𝑒 is reordered to
̂ ∗ and 𝐁 elements having the respective sub-laminates as the cross sections (see
obtain the redefined element displacement vector 𝐔𝑒 as Figs. 1(b) and (c)). The nodal DOFs of these elements correspond to the
[ ]T mid-surface of the respective sub-laminate. Since the 𝑧-coordinates with
𝐔𝑒 = 𝑢01 𝑤01 𝑤0,𝑥1 𝜓 01 𝑢 02 𝑤02 𝑤0,𝑥2 𝜓 02 … 𝑢0𝑛 𝑤0𝑛 𝑤0,𝑥𝑛 𝜓 0𝑛 respect to the reference surface of the intact laminate are not modelled
(10) explicitly, all the meshes for the sub-laminates in a delaminated region
physically lie on the same plane (although they are shown separately in
Accordingly 𝐍̂ ∗ and 𝐁
̂ ∗ matrices are rearranged to obtain 𝐍
̂ and 𝐁
̂
Fig. 1(c) for illustration purpose). At the delamination entry and exit
matrices such that fronts, the stress resultants in the intact laminate and the connected
̂ 𝑒,
𝒖̄ = 𝐍𝐔 ̂ 𝑒,
𝜺̄ = 𝐁𝐔 (11) sub-laminates must satisfy the equilibrium conditions. This condition
is satisfied by making the adjoining elements in the intact and delami-
Substituting 𝐮̄ and 𝜺̄ from Eq. (11) into Eq. (5), the contribution 𝑇 𝑒 of nated zones share a common global node number at the delamination
element 𝑒 to the variational integral is obtained as front (e.g. 𝑛2 and 𝑛3 in Fig. 1(c)).
1
2 ̂ T 𝐈̃𝐍 ̂ T𝐃
̂ 𝐔̈ 𝑒 + 𝐁 ̄ 𝐁𝐔 ̂ T 𝒇 𝒖 ] 𝑏(𝑥(𝜉)) d𝜉
̂ 𝑒−𝐍 Further, the displacement field along each sub-laminate thickness
𝑇𝑒 = 𝛿𝐔𝑒T [𝐍
𝑙𝑒 ∫−1 at the delamination front must ideally match the field in the intact
̄ 𝑒 𝐔̈ 𝑒 + 𝐊̄ 𝑒 𝐔𝑒 − 𝐏̄ 𝑒 ] laminate point-wise. This can be challenging for higher-order theories
= 𝛿𝐔𝑒T [𝐌 (12)
wherein the variation of the displacement field is non-linear as in the
with present case. In the following, we discuss two models for this purpose:
𝑒 1 (1) the point continuity (PC) model, which has been used in most
̄𝑒 = 𝑙
𝐌 ̂ T (𝜉)𝐈(𝜉)
𝐍 ̃ 𝐍(𝜉)
̂ 𝑏(𝑥(𝜉)) d𝜉,
2 ∫−1 studies on delaminated beams and plates [40–43,45–48], and (2) the
𝑒 1 hybrid point-least squares continuity (HPLSC) model, a new approach
̄𝑒 = 𝑙
𝐊 ̂ T (𝜉)𝐃(𝜉)
𝐁 ̄ 𝐁(𝜉)
̂ 𝑏(𝑥(𝜉)) d𝜉, (13) developed for laminated shells recently by Kapuria and Ahmed [50,71]
2 ∫−1
1 using four-node facet shell elements based on higher-order theories.
𝑙𝑒 ̂ T (𝜉)𝒇 𝒖 𝑏(𝑥(𝜉)) d𝜉.
𝐏̄ 𝑒 = 𝐍 We present these two models for the current spectral strip element
2 ∫−1 based on the ZIGT. To develop these models, we consider two adjacent
𝐌̄ 𝑒 and 𝐊
̄ 𝑒 define the element mass and stiffness matrices, respectively, spectral strip elements at a delamination front, one in the intact region
of the strip element and 𝐏̄ 𝑒 represents the element force vector. (𝐼) and the other in the delaminated region (𝐷) for a sub-laminate.

5
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 4. (a) A delamination front connecting an intact laminate (𝐼) and a delaminated sub-laminate (𝐷) having 𝑝 number of layers starting from 𝑏th interface of the intact laminate,
and (b) Variations of 𝑢 in laminates 𝐼 and 𝐷 at the delamination front. They should be ideally continuous at all points over sub-laminate thickness ℎ𝑑 .

The sub-laminate is made of 𝑝 layers and its thickness spans from where
the 𝑏th interface to the (𝑏 + 𝑝)th interface of the intact laminate as
⎧ 𝑢𝐷 ⎫ ⎧ 𝑢𝐷 ⎫ ⎡1
shown in Fig. 4. All variables are super- or subscripted with 𝐼 and 𝐷 to ⎪ 0𝑖 ⎪ ⎪ 0𝑖 ⎪ 0 𝜆̄ 𝜆̃⎤
⎪ 𝐷 ⎪ ⎪ 𝐷 ⎪ ⎢ ⎥
indicate their associations with the intact and delaminated segments, ⎢ ⎥
⎪𝑤 ⎪ ⎪𝑤 ⎪ 0 1 0 0⎥
respectively. 𝐔𝐷 = ⎨ 0𝑖 ⎬ , 𝐔𝐼𝑖 = ⎨ 0𝑖 ⎬ , 𝐓𝑖 = ⎢ (18)
𝑖
⎪𝑤𝐷 ⎪ ⎪𝑤𝐷 ⎪ ⎢ ⎥
⎪ 0,𝑥𝑖 ⎪ ⎪ 0,𝑥𝑖 ⎪ ⎢0 0 1 0⎥
⎪ 𝐷⎪ ⎪ 𝐷⎪ ⎢ ⎥
4.1. Point continuity method
⎩ 𝜓0 ⎭ ⎩ 𝜓0 ⎭ ⎣0 0 0 1⎦
𝑖 𝑖

It is the most widely used method to model delaminations in lam- with 𝜆̄ = −𝑧𝑐 and 𝜆̃ = 𝑅𝑘𝐼 (𝑧𝑐 ). Note that for an 𝑛-node spectral element,
inated structures. As the name suggests, it satisfies the continuity of 𝑖 will take the value of 1 or 𝑛 in general.
the in-plane displacement 𝑢𝑥 at only one point across the sub-laminate The above conditions will ensure exact point-wise satisfaction of
thickness, which is its mid-plane. At a point 𝑃 (see Fig. 4(b)) on continuity of the displacement field in the intact and delaminated
the common interface between the intact laminate and delaminated segments at the common interface for laminate theories like the CLT
sub-laminate, the in-plane displacements of the two laminates can be and the FSDT, which assume a linear variation of 𝑢 across the laminate
expressed as thickness. However, the same will not happen when through-thickness
variation of 𝑢 is non-linear as in the present theory. It may lead to large
𝑢𝐼 (𝑥, 𝑧, 𝑡) = 𝑢𝐼0 (𝑥, 𝑡) − 𝑧𝑤𝐼0,𝑥 (𝑥, 𝑡) + 𝑅𝑘𝐼 (𝑧)𝜓0𝐼 (𝑥, 𝑡) discontinuity in 𝑢 at points other than the mid-plane of the sub-laminate
𝑢𝐷 (𝑥, 𝑧′ , 𝑡) = 𝑢𝐷 (𝑥, 𝑡) − 𝑧′ 𝑤𝐷 (𝑥, 𝑡) + 𝑅𝑘𝐷 (𝑧′ )𝜓0𝐷 (𝑥, 𝑡) (14) causing significant errors. The issue is elegantly addressed in the HPLSC
0 0,𝑥
model discussed in the following section.
In the above expressions, 𝑧 and 𝑧′
denote, respectively, the local
thickness coordinates of the 𝐼 and 𝐷 laminates with respect to their
mid-planes. The two coordinates are, of course, related by 𝑧′ = 𝑧 − 𝑧𝑐 , 4.2. Hybrid point-least squares method
with 𝑧𝑐 = (𝑧𝑏 + 𝑧𝑏+𝑝 )∕2. At the mid-plane of the sub-laminate, 𝑧′ = 0
and 𝑧 = 𝑧𝑐 . Equating 𝑢𝐼 and 𝑢𝐷 at this point would mean 𝑢𝐼 (𝑥, 𝑧𝑐 , 𝑡) = An alternative approach to satisfying the displacement continuity
𝑢𝐷 (𝑥, 0, 𝑡), which yields using Eq. (14) at the delamination front would be by minimizing the norm of the
𝑢𝐷 = 𝑢𝐼0 − 𝑧𝑐 𝑤𝐼0,𝑥 + 𝑅𝑘𝐼 (𝑧𝑐 )𝜓0𝐼 (15) deviation (𝑢𝐼 − 𝑢𝐷 ) over the sub-laminate thickness, known as the least
0
squares method. However, it would mean that the continuity of 𝑢 is
The rotation variables 𝑤0,𝑥 and 𝜓0 of both elements at the common not satisfied exactly at any predefined point within the sub-laminate. It
node are made equal: may result in unpredictable accuracy. The HPLSC method proposed by
Kapuria and Ahmed [71] attempts to combine the advantages of the PC
𝑤𝐷
0,𝑥
= 𝑤𝐼0,𝑥 , 𝜓0𝐷 = 𝜓0𝐼 . (16) and least squares methods. It satisfies the continuity of 𝑢 exactly at the
𝑥 𝑥

Relations (15) and (16) can be expressed in terms of a transformation midplane of the delaminated sub-laminate like the PC method. But, the
matrix 𝑇𝑖 relating the nodal DOF vector 𝐔𝐷 rotation variables 𝑤𝐷 and 𝜓0𝐷 are not equated with 𝑤𝐼0 and 𝜓0𝐼 unlike
𝑖 of the 𝑖th node of the 0,𝑥 ,𝑥
element in the delaminated segment to that of the adjacent element the PC method. Instead, they are determined by minimizing the norm of
in the intact segment (𝐔𝐼𝑖 ) as the deviation between 𝑢𝐼 and 𝑢𝐷 over the common interface. The norm
𝐹 of the deviation (𝑢𝐼 − 𝑢𝐷 ) over the sub-laminate thickness spanning
𝐔𝐷 𝐼
𝑖 = 𝐓𝑖 𝐔𝑖 (17) from the 𝑏th interface to the (𝑏 + 𝑝)th interface of the intact laminate is

6
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

given by Table 1
Essential and natural boundary conditions.
∑
𝑏+𝑝 𝑧𝑘
Essential 𝑢 = 𝑢̄ 0 𝑤0 = 𝑤̄ 0 𝑤0,𝑥 = 𝑤̄ 0,𝑥 𝜓0 = 𝜓̄ 0
𝐹 = (𝑢𝐼 − 𝑢𝐷 )2 𝑑𝑧 (19)
∫𝑧𝑘−1 Natural 𝑁𝑥 = 𝑁̄ 𝑥 𝑉𝑥 = 𝑉̄𝑥 𝑀𝑥 = 𝑀 ̄𝑥 𝑃𝑥 = 𝑃̄𝑥
𝑘=𝑏+1

To minimize 𝐹 with respect to 𝑤𝐼0 and 𝜓0𝐼 , we have

,𝑥

𝜕𝐹 𝜕𝐹 spectral element 𝑒 in the delaminated region, the first or 𝑛th or both

= 0, =0 (20)
𝜕𝑤𝐷
0,𝑥
𝜕𝜓0𝐷 nodes may lie on a delamination front. For such elements, we then
introduce an element displacement vector 𝐔𝑒𝐼 as given below
Eqs. (15) and (20) constitute the continuity conditions for the HPLS
method, which yields, upon using Eq. (19), the following relations: [ ]T
𝐔𝑒𝐼 = 𝑢𝐼0 𝑤𝐼0 𝑤𝐼0,𝑥 𝜓0𝐼 𝑢𝐷
0
𝑤𝐷
0
… 𝑤𝐷
0,𝑥
𝜓0𝐷 𝑢𝐼0 𝑤𝐼0 𝑤𝐼0,𝑥 𝜓0𝐼
1 1 1 1 2 2 𝑛−1 𝑛−1 𝑛 𝑛 𝑛 𝑛
𝑢𝐷
0
− 𝑢𝐼0 − 𝜆𝑤
̄ 𝐼 − 𝜆𝜓
0,𝑥
̃ 𝐼 =0
0
(27)
𝑎1 𝑤 𝐷
0,𝑥
− 𝑎1 𝑤𝐼0,𝑥 + 𝑎2 𝜓0𝐼 − 𝑎3 𝜓0𝐷 = 0 (21)
The original element displacement vector 𝐔𝑒𝐷 of the delaminated seg-
𝑏1 𝑢𝐼0 − 𝑏1 𝑢𝐷
0
− 𝑏2 𝑤𝐼0,𝑥 + 𝑏3 𝑤𝐷
0,𝑥
+ 𝑏4 𝜓0𝐼 − 𝑏5 𝜓0𝐷 = 0 ment element can be related to 𝐔𝑒𝐼 by using Eq. (17) as
with
𝐔𝑒𝐷 = 𝐓𝑒 𝐔𝑒𝐼 (28)
ℎ3𝑑 ∑𝑏+𝑝 𝑧
𝑎1 = 12
, 𝑎2 = ∫ 𝑘
𝑘=𝑏+1 𝑧𝑘−1
𝑧′ 𝑅𝑘𝐼 (𝑧) 𝑑𝑧, where
∑𝑏+𝑝 𝑧 ∑𝑏+𝑝 𝑧
𝑎3 = ∫ 𝑘 𝑧′ 𝑅𝑘𝐷 (𝑧′ ) 𝑑𝑧, 𝑏1 = ∫ 𝑘 𝑅𝑘𝐷 (𝑧′ ) 𝑑𝑧, ⎡𝐓1 ⎤
𝑘=𝑏+1 𝑧 𝑘=𝑏+1 𝑧
𝑘−1 𝑘−1 (22) ⎢ ⎥
∑𝑏+𝑝 𝑧 ∑𝑏+𝑝 𝑧 ⎢
𝐈4
⎥
𝑏2 = ∫ 𝑘 𝑧𝑅𝑘𝐷 (𝑧′ ) 𝑑𝑧, 𝑏3 = ∫ 𝑘 𝑧′ 𝑅𝑘𝐷 (𝑧′ ) 𝑑𝑧,
𝑘=𝑏+1 𝑧𝑘−1 𝑘=𝑏+1 𝑧𝑘−1 ⎢ 𝐈4 ⎥
∑𝑏+𝑝 𝑧 ∑𝑏+𝑝 𝑧 𝐓𝑒 = ⎢ . ⎥ (29)
𝑏4 = ∫ 𝑘
𝑘=𝑏+1 𝑧𝑘−1
𝑅𝑘𝐷 (𝑧′ )𝑅𝑘𝐼 (𝑧) 𝑑𝑧, 𝑏5 = ∫ 𝑘 {𝑅𝑘𝐷 (𝑧′ )}2 𝑑𝑧,
𝑘=𝑏+1 𝑧𝑘−1 ⎢ ⎥
⎢ . ⎥
where ℎ𝑑 = 𝑧𝑏+𝑝 − 𝑧𝑏 denotes the thickness of the delaminated sub- ⎢ 𝐈4 ⎥
laminate. Eq. (21) is rewritten in the matrix form and solved to express ⎢ ⎥
⎣ 𝐓𝑛 ⎦
the displacement variables of the delaminated segment in terms of
𝐈𝑛 denotes identity matrix of size 𝑛 × 𝑛. 𝐓1 and 𝐓𝑛 are obtained from
those of the intact segment as
Eqs. (18) or (26) or taken as equal to 𝐈4 depending on whether or not
⎡ 𝐷⎤ ⎡ 𝐼 ⎤ node 1 and 𝑛 lie on the delamination front. The HPLSC method will be
⎢ 𝑢0 ⎥ ⎢ 𝑢0 ⎥
⎢ ⎥ ⎢ ⎥ used in the present study unless specified otherwise.
̄ ̃
𝐂 ⎢𝑤 ⎥ = 𝐂 ⎢𝑤𝐼 ⎥
𝐷 (23)
0,𝑥 0,𝑥 For these elements, 𝑇 𝑒 given by Eq. (12) can be expressed upon
⎢ ⎥ ⎢ ⎥ applying Eq. (28) as
⎢ 𝜓𝐷 ⎥ ⎢ 𝜓𝐼 ⎥
⎣ 0 ⎦ ⎣ 0 ⎦
𝑇 𝑒 = 𝛿𝐔𝑒𝐷 T [𝐌
̄ 𝑒 𝐔̈ 𝑒 + 𝐊
̄ 𝑒 𝐔𝑒 − 𝐏̄ 𝑒 ]
⎡ 𝐷⎤ ⎡ 𝐼 ⎤ 𝐷 𝐷
⎢ 𝑢0 ⎥ ⎢ 𝑢0 ⎥ 𝑒T T ̄ 𝑒
= 𝛿𝐔𝐼 𝐓𝑒 [𝐌 𝐓𝑒 𝐔̈ 𝑒𝐼 + 𝐊 ̄ 𝑒 𝐓𝑒 𝐔𝑒 − 𝐏̄ 𝑒 ]
⎢ ⎥ ⎢ ⎥ 𝐼
⇒ ⎢𝑤𝐷 ⎥ = 𝝀 ⎢𝑤𝐼 ⎥ (24) ⇒ 𝑇 𝑒 = 𝛿𝐔𝑒𝐼 T [𝐌𝑒 𝐔̈ 𝑒𝐼 + 𝐊𝑒 𝐔𝑒𝐼 − 𝐏𝑒 ] (30)
0 0
⎢ ,𝑥 ⎥ ⎢ ,𝑥 ⎥
⎢ 𝜓𝐷 ⎥ ⎢ 𝜓𝐼 ⎥ where
⎣ 0 ⎦ ⎣ 0 ⎦
where 𝐌𝑒 = 𝐓T𝑒 𝐌
̄ 𝑒 𝐓𝑒 , 𝐊𝑒 = 𝐓T𝑒 𝐊
̄ 𝑒 𝐓𝑒 , 𝐏𝑒 = 𝐓T𝑒 𝐏̄ 𝑒 (31)
⎡1 0 0 ⎤ ⎡1 𝜆̄ 𝜆̃ ⎤
̄ = ⎢0 By summing up the contributions 𝑇 𝑒 given by Eq. (30) from the total
𝐂 𝑎1 −𝑎3 ⎥ , ̃ = ⎢0
𝐂 𝑎1 −𝑎2 ⎥ , ̄ −1 𝐂.
𝝀=𝐂 ̃ (25)
⎢ ⎥ ⎢ ⎥ 𝑛𝑒 number of spectral elements 𝑇 in Eq. (5) is obtained as
⎣ 𝑏1 −𝑏3 𝑏5 ⎦ ⎣𝑏1 −𝑏2 𝑏4 ⎦
𝑛𝑒 𝑛𝑒
∑ ∑
With the help of Eq. (24), the nodal displacement vector 𝐔𝐷
𝑖 of 𝑖th node 𝑇 = 𝑇𝑒 = 𝛿𝐔𝑒T [𝐌𝑒 𝐔̈ 𝑒 + 𝐊𝑒 𝐔𝑒 − 𝐏𝑒 ] (32)
of the sub-laminate element can be transformed into 𝐔𝐼𝑖 of the intact 𝑒=1 𝑒=1
laminate using the same relation as Eq. (17) with the transformation The summation in Eq. (32) is achieved by gathering the elements of
matrix 𝐓𝑖 given by matrices 𝐌𝑒 , 𝐊𝑒 , 𝐏𝑒 , and 𝐔𝑒 in their global counterparts 𝐌, 𝐊, 𝐏, and 𝐔.
⎡𝜆11 Since virtual displacements 𝛿𝐔 are arbitrary, for the variational integral
0 𝜆12 𝜆13 ⎤
⎢ ⎥ in Eq. (4) to vanish, the coefficient of 𝛿𝐔 must be identically zero. The
0 1 0 0 ⎥
𝐓𝑖 = ⎢ , 𝜆𝑖𝑗 ∈ 𝝀. (26) semi-discrete equation of motion of the spatially discretized system is
⎢𝜆21 0 𝜆22 𝜆23 ⎥
⎢𝜆 0 𝜆32 ⎥
𝜆33 ⎦
thus yielded as
⎣ 31
The continuity models discussed above are very generic and can 𝐌𝐔̈ + 𝐊𝐔 = 𝐏. (33)
be applied to multiple delaminations occurring at arbitrary axial and The material damping of the system is included by using the most
interfacial locations in the laminated strip-structure. widely used Rayleigh’s hypothesis, which estimates the proportional
damping matrix as 𝐂 = 𝛼1 𝐌 + 𝛼2 𝐊. 𝛼1 and 𝛼2 represent, respectively,
5. System equation of motion the mass and stiffness coefficients for damping. The global equation of
motion of the damped system is thus obtained as
For those spectral elements in the delaminated regions, which share
a common node with elements in the intact region, the assembly of the 𝐌𝐔̈ + 𝐂𝐔̇ + 𝐊𝐔 = 𝐏. (34)
element stiffness, mass, and force matrices would require that the DOFs
at the common node correspond to one reference system. The DOFs 𝐔𝐷 The variationally consistent essential and natural boundary condi-
𝑖
of such nodes 𝑖 are transformed to 𝐔𝐼𝑖 in the intact laminate reference tions can be derived based on the efficient layerwise ZIGT from the
system by using Eqs. (17) and (18) or (26) as per the chosen continuity boundary terms in Eq. (4), which are shown in Table 1.
model. The other nodes of the delaminated segment elements can retain Thus, for instance, we can express the boundary conditions for some
the DOFs in the sub-laminate reference system itself. For an 𝑛-node common supports as given in Table 2.

7
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Table 2
Boundary conditions for some commonly used supports.
Support type Simply-supported Pinned Clamped Free
𝑁𝑥 = 0, 𝑢0 = 0, 𝑢0 = 0, 𝑁𝑥 = 0,
Boundary 𝑤0 = 0, 𝑤0 = 0, 𝑤0 = 0, 𝑉𝑥 = 0,
conditions 𝑀𝑥 = 0, 𝑀𝑥 = 0 𝑤0,𝑥 = 0, 𝑀𝑥 = 0,
𝑃𝑥 = 0. 𝑃𝑥 = 0. 𝜓0 = 0. 𝑃𝑥 = 0.

For time-history response, Eq. (34) is solved in time domain using the present results for all ten flexural modes are in excellent agreement
Newmark-𝛽 implicit time integration scheme with 𝛽 = 0.25 and 𝛾 = 0.5, with the benchmark continuum-based FE solution with a maximum
which provides unconditional stability. For free vibration response, we error of 0.9%. It establishes the accuracy of the ZIGT SE-based model
have used the subspace iteration technique [99] to solve the resulting with the hybrid continuity method for delamination fronts.
generalized eigenvalue problem for obtaining the natural frequencies Next, the proposed SE’s accuracy is evaluated for free vibration
and mode shapes of the structure. response of thick cantilever and simply-supported composite beams
with a span-to-thickness ratio of 𝑆 (= 𝑎∕ℎ) = 5. The width-to-span (𝑏∕𝑎)
6. Results and discussion of the beam is 0.1. It is made of a four-layer graphite/epoxy (Gr/Ep-
2, properties shown in Table 3) laminate [0◦ /90◦ ]𝑆 . All laminas are
In this section, we first validate the proposed formulation and considered to be of equal thickness. The beam contains a delamination
the developed computer program for free vibration and Lamb wave of the length of 10%, 25%, and 50% of its span, located at the mid-
propagation analyses of multi-layered beams and panels containing span and the middle interface. Table 6 presents a convergence study
single and multiple delaminations. These problems are chosen from on the mesh size of the present SE model for the natural frequencies of
the published literature in order to compare with them for validation. the first two flexural modes of cantilever and simply-supported beams
The elasticity-based FE solutions are employed to assess the accuracy with a delamination size of 50% of the beam span. In the table, symbols
of the SE-based delamination modelling, since such solutions do not 𝑛𝐿 , 𝑛𝐷 , and 𝑛𝑅 denote the number of eight-node SEs used to model
make any prior assumptions on the variations of the displacement field parts AB, each sub-laminate in BC, and CD, respectively. The results
across the laminate thickness. We also illustrate, in comparison with show that a mesh of 16 eight-node SEs with 4 elements in each part
continuum-based FE solutions, the relative performances of the PC and yields converged solutions. For comparison, the 3D elasticity-based FE
HPLSC methods in modelling the delamination front continuity. The solution is considered, which was earlier obtained in Ref. [50] using
next example evaluates the ZIGT-based SE’s accuracy for Lamb wave a twenty-node quadratic solid element (C3D20R) in ABAQUS. In this
propagation response in delaminated composite strips caused by the case, a mesh of 100 × 10 × 40 elements was used. We utilize this
actuation of a piezoelectric transducer mounted at the laminate’s top example also to assess the relative accuracies of the continuity models
surface. This problem represents what is encountered in real-world PC and HPLSC. Table 7 compares the present SE solutions for the two
SHM applications. Since the refined TOT has the same number of DOFs continuity models with the 3D FE solution. The comparison reveals
as the ZIGT, we also assess their relative accuracies with respect to that the HPLSC model yields a much superior prediction of the natural
the elasticity-based solutions. Next, we study the feasibility of local- frequencies than the PC model for both cantilever and simply-supported
izing delamination damage in composite strips from the guided wave beams featuring delaminations. The error in the PC becomes very
propagation response. large even for the fundamental frequency as the delamination length
increases, making it an undesirable choice for satisfying continuity. As
6.1. Free vibration of composite beams with multiple delaminations expected, the natural frequencies of the beam decrease with an increase
in delamination length.
In this example, we consider a cantilever beam of size 127 mm ×
12.7 mm × 1.02 mm (length × width × thickness), made of eight-layer
6.2. Lamb wave propagation problem of Ramadas et al. [24]
graphite/epoxy (Gr/Ep-1) composite having a lay-up of [0◦ /90◦ ]2𝑆 .
The beam has three cases of single to triple delaminations located at
the mid-span as shown in Fig. 5. Table 3 lists its material properties. Now, we consider the problem of guided wave propagation in a
This problem was analysed earlier using the FSDT-based analytical delaminated composite strip for further validation of the present SE.
solution [98] and the ZIGT-based FE (plate element) solution [71]. The This example was analysed earlier by Ramadas et al. [24] utilizing AN-
elasticity-based 2D FE solution is also obtained in ABAQUS utilizing SYS software employing an eight-node quadratic plane strain element
an eight-node quadrilateral plane stress element (CPS8R) to assess the (PLANE183). The laminate is made of three-layer glass epoxy (Gl/Ep,
accuracy of the present solution. The continuum-based FE solution is see Table 3 for its material properties) with a lay-up of [0◦ /90◦ /0◦ ]. Its
chosen as a benchmark for the assessment as it does not have any dimensions are 400 mm × 0.99 mm (length × total thickness), and the
pre-defined approximations on the through-thickness distributions of position of the delamination is as shown in Fig. 7. The guided wave is
displacements. excited by two equal transverse forces (𝐹 (𝑡)∕2) acting at the left end
We determine the natural frequencies (𝑓𝑛 = 𝜔𝑛 ∕2𝜋) for the first ten of the panel on its top and bottom surfaces, as shown in Fig. 7(a).
flexural modes, for which the present SE with the HPLSC model yielded The excitation is a modulated tone burst pulse expressed as 𝐹 (𝑡) =
𝐹0 2𝜋(𝑡−𝑡 )
converged results with 3, 4, and 3 eight-node elements for regions AB, 𝐻 (𝑡) sin{2𝜋𝑓 (𝑡 − 𝑡0 )}, where 𝐻𝑤 (𝑡) = 1 − cos 𝑇 0 is the Hann
2 𝑤 𝐻
each sub-laminate of BC, and CD for the case of 50.8 mm long delami- window used for modulation. 𝑇𝐻 denotes the width of the window,
nations. The corresponding numbers of elements for the three parts for which depends on the number of cycles 𝑁𝐵 and frequency 𝑓 of the
the case of 101.6 mm long delaminations are 1, 8, and 1, respectively. pure tone burst as 𝑇𝐻 = 𝑁𝐵 ∕𝑓 . For the current example, the excitation
The 2D FE model in ABAQUS is made of a very fine mesh of 2000 × 8 signal parameters are 𝐹0 = 42 kN, 𝑓 = 200 kHz, 𝑁𝐵 = 7, and 𝑡0 = 0 μs.
(length × thickness) elements to obtain the benchmark solution. Table 4 The signal and its transform are shown in Fig. 6, which indicates the
and 5 present these results for delaminations of 50.8 mm and 101.6 mm narrowband nature of the signal. Ramadas et al. [24] employed a mesh
in length, respectively. The present solution matches both the reference of 1600 × 6 (length × thickness) PLANE183 elements in the FE model
solutions [71,98] available for the first mode very well, which validates and used Newmark-𝛽 method for time-marching with a uniform time
the proposed modelling for single and multiple delaminations. Further, step 𝛥T of 50 ns in ANSYS.

8
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 5. Configuration of eight-layer composite beam showing three cases of delaminations. The response is computed for two boundary conditions — cantilever and simply-supported.

Table 3
Material properties.
Material 𝑌1 𝑌2 𝑌3 𝐺23 𝐺13 𝐺12 𝜈12 𝜈13 𝜈23 𝜌
(GPa) (kg/m3 )
Gr/Ep-1 [71] 134.0 10.3 10.3 3.28 5.0 5.0 0.33 0.33 0.33 1480
Gr/Ep-2 [50] 181.0 10.3 10.3 2.87 7.17 7.17 0.28 0.28 0.33 1578
Gr/Ep-3 [100] 132.5 10.8 10.8 3.40 5.7 5.7 0.24 0.24 0.588 1600
Gl/Ep [24] 44.68 6.9 6.9 2.546 2.54 2.54 0.28 0.28 0.355 1990

Table 4
Natural frequencies (Hz) of first ten flexural modes of a cantilever composite beam ([0◦ /90◦ ]2𝑆 ) having multiple delaminations of 50.8 mm
length.
No. of del. Flexural mode Present 2D FE % error FSDT [98] Ref. [71]
1 1 80.689 80.485 0.25 80.383 80.945
2 443.53 441.11 0.55 – –
3 570.41 567.37 0.54 – –
4 1135.0 1127.0 0.71 – –
5 1408.2 1397.4 0.77 – –
6 2495.7 2473.1 0.91 – –
7 2654.3 2635.1 0.73 – –
8 4018.4 4006.7 0.29 – –
9 4320.2 4294.0 0.61 – –
10 5815.9 5781.8 0.59 – –
2 1 80.154 79.921 0.29 – 80.394
2 135.38 135.49 −0.09 – –
3 368.82 369.02 −0.05 – –
4 447.56 444.06 0.79 – –
5 585.48 580.92 0.78 – –
6 733.14 733.50 −0.05 – –
7 1109.5 1101.8 0.71 – –
8 1222.1 1217.8 0.36 – –
9 1416.3 1409.5 0.48 – –
10 1789.5 1790.7 −0.06 – –
3 1 73.498 73.090 0.56 – 70.904
2 135.29 135.40 −0.08 – –
3 367.13 367.29 −0.04 – –
4 429.10 425.86 0.76 – –
5 479.58 475.53 0.85 – –
6 598.27 594.06 0.71 – –
7 727.83 727.06 0.10 – –
8 761.78 755.44 0.84 – –
9 1203.4 1203.2 0.01 – –
10 1318.3 1310.4 0.60 – –

In the current 1D model, the total resultant transverse force of 𝐹 (𝑡) As the panel has a symmetric lay-up [0◦ /90◦ /0◦ ], the given loading
is applied at the node at 𝑥 = 0 (Fig. 7(b)). We employ the eight-node excites only the antisymmetric Lamb wave mode A0 , which transmits
SEs and the HPLSC model for satisfying delamination front continuity through the sub-laminates and is seen at the sensor location S as the
in the present and all subsequent examples. The converged solution for first prominent waveform in Fig. 7 (c). Let us denote this waveform as
this case is achieved by taking 25 elements in segment AB, ten elements the A0 A0 mode. At the same time, the interaction at the delamination
in each sub-laminate in the delaminated region BC, and 45 elements in entry point B results in a symmetric mode S0 due to mode conversion.
CD and a time step of 100 ns. The deflection 𝑤 is recorded at the sensor However, being symmetric, this mode, called A0 S0 mode, is not visible
location S at the top surface of the laminate at 15 mm from the entry in the 𝑤0 response at the sensor location (S). But, when it reaches
point of delamination (Fig. 7). the exit point C of the delamination, it transforms into an A0 mode,

9
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Table 5
Natural frequencies (Hz) of first ten flexural modes of a cantilever composite beam ([0◦ /90◦ ]2𝑆 ) having multiple delaminations of 101.6 mm
length.
No. of del. Flexural mode Present 2D FE % error FSDT [98] Ref. [71]
1 1 73.092 72.756 0.46 73.147 73.303
2 123.93 122.97 0.78 – –
3 333.93 331.29 0.80 – –
4 460.87 457.52 0.73 – –
5 653.83 648.63 0.80 – –
6 1066.4 1059.6 0.64 – –
7 1142.1 1137.1 0.45 – –
8 1605.9 1593.4 0.78 – –
9 2226.6 2219.5 0.32 – –
10 2245.4 2227.8 0.79 – –
2 1 33.332 33.345 −0.04 – 33.361
2 72.980 72.574 0.56 – –
3 93.752 93.796 −0.05 – –
4 124.24 123.24 0.81 – –
5 182.01 182.09 −0.04 – –
6 299.79 299.91 −0.04 – –
7 334.26 331.51 0.83 – –
8 447.45 447.45 0.00 – –
9 478.78 474.82 0.83 – –
10 625.19 625.38 −0.03 – –
3 1 32.974 32.975 0.00 – 33.022
2 50.841 50.490 0.70 – –
3 93.144 93.180 −0.04 – –
4 120.78 119.84 0.78 – –
5 126.46 125.59 0.69 – –
6 181.91 181.99 −0.04 – –
7 298.99 298.96 0.01 – –
8 318.55 317.32 0.39 – –
9 332.89 330.33 0.78 – –
10 341.90 339.36 0.75 – –

Table 6
Convergence study on mesh size of the SE model for natural frequencies of thick (𝑆=5) composite beams ([0◦ /90◦ ]𝑆 ) with
delamination size of 50% of the span.
Mesh Mode Natural frequency (Hz)
(𝑛𝐿 , 𝑛𝐷 , 𝑛𝑅 ) number Cantilever Simply supported
PC HPLS PC HPLS
4, 2, 4 1 212.00 191.13 510.76 526.64
2 827.71 817.35 940.51 910.89

4, 4, 4 1 212.00 191.29 510.77 526.64

2 827.11 817.34 940.51 910.87

8, 8, 8 1 212.00 191.29 510.77 526.64

2 827.11 817.34 940.51 910.87

Table 7
Natural frequencies (Hz) of thick (𝑆=5) cantilever and simply-supported delaminated composite beams ([0◦ /90◦ ]𝑆 ): Comparison of the two
continuity models with 3D FE solution.
Delamination Mode Cantilever Simply supported
size (%) number PC Hybrid 3D FE [50] PC Hybrid 3D FE [50]
10 1 248.05 242.12 239.36 526.63 552.39 553.49
2 845.32 858.23 854.82 1443.56 1325.47 1305.30

25 1 247.11 231.77 224.69 515.36 549.04 551.91

2 829.01 857.85 846.90 1352.11 1189.14 1125.40

50 1 212.00 191.29 179.55 510.77 526.64 526.84

2 827.11 817.34 785.19 940.51 910.87 836.87

now to be called A0 S0 A0 mode, due to mode conversion and travels 6.3. Surface-bonded transducer induced lamb wave propagation for SHM
back to the sensor location S. It is seen as the second wave packet in
the response shown in Fig. 7 (c). The present ZIGT-based SE model is In Lamb wave-based SHM techniques, the guided waves are com-
able to predict both the direct A0 A0 and reflected A0 S0 A0 waveforms in monly generated using thin piezoelectric wafer active sensors (PWAS)
close agreement with the continuum-based reference solution [24]. It surface-bonded to the host structure. In the current example, we take
validates the current SE model in conjunction with the HPLSC method such a problem to assess the present SE’s accuracy in predicting Lamb
as well as shows its accuracy in simulating Lamb wave propagation and wave propagation and its interaction with delamination in composite
its interaction with delamination damage in composite panels. panels. The composite panel (state of plane strain along 𝑦-direction)

10
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 6. Hann window modulated tone burst signal having seven cycles with 𝑓 = 200 kHz showing its (a) time-history and (b) frequency spectrum (Fourier transform).

Fig. 7. Schematic configuration of the composite panel [0◦ /90◦ /0◦ ] containing a delamination (BC) of length 50 mm at the first interface from bottom for the problem of Ref. [24],
showing the transverse force application in (a) the 2D FE model, (b) the present 1D SE model, and (c) deflection response at the sensor location S due to excitation of seven-cycle
Hann window modulated signal of central frequency 𝑓 = 200 kHz at A.

considered in the study is 1.8 m in length and is made of a six-layer discretize the length for these two excitation cases. We try to keep the
graphite/epoxy (Gr/Ep 3, see Table 3) laminate [0◦ /90◦ /0◦ ]𝑆 with element length almost equal in each section. However, the transducer
each layer of 0.5 mm thickness. The PWAS transducer is not modelled section (BC) and delamination section (E1 E2 ) have 1 and 2 elements
physically, but the stress transfer from the piezoelectric patch of length for both excitation frequencies. For clarity, the sections AB, CE1 , E2 S,
𝑎 is represented by an in-plane force pair ∓𝐹 (𝑡) applied at the two and SD are discretized respectively with 28, 5, 5, and 29 eight-node
transducer edges B′ and C′ , separated by distance 𝑎, on the plate surface elements for the 70 kHz excitation and 37, 6, 6, and 38 elements for
(see Fig. 8(a)). Such excitations generate both symmetric (S0 ) and the 100 kHz excitation. The time step is taken as 0.4 μs and 0.2 μs for
antisymmetric (A0 ) Lamb wave modes in the un-delaminated strip. 70 and 100 kHz frequencies, respectively.
We consider the PWAS of size 𝑎 = 7 mm and record the output at The smeared refined TOT [93] assumes the same third-order varia-
the sensor location S at 300 mm from the centre of the PWAS patch. tion as the ZIGT for the in-plane displacement 𝑢 globally in the thick-
The delamination is of length 𝑙𝑑𝑒𝑙 = 50 mm, located at the first interface ness direction but does not have the layerwise terms. Both theories,
from the bottom (Fig. 8(c)). The centre-to-centre distance between the however, have the same number of primary displacement variables,
delamination and actuator (𝑥𝑑𝑒𝑙
𝑐 ) is 150 mm. The force 𝐹 (𝑡) is the Hann i.e., DOFs per node in the elements. So, a comparison with the TOT-
window modulated tone burst signal defined in the previous section based SE would allow us to quantify the contribution of these layerwise
with 𝐹0 = 1 N, 𝑁𝐵 = 5, 𝑡0 = 0, and 𝑓 = 70 and 100 kHz. terms to the solution accuracy. We keep the same mesh in the TOT
The benchmark solution for the problem is developed using a model as in the ZIGT for comparison, leading to the same number of
continuum-based 2D FE model employing the plane-strain eight-node total DOFs in both models.
quadratic element with reduced integration (CPE8R) in the commercial In Fig. 10, the in-plane displacement 𝑢(𝑡) is plotted at the sensor
software ABAQUS. After checking for convergence, the panel length location S obtained from the present ZIGT, TOT, and continuum-based
and total thickness are discretized with 2000 and 12 elements, respec- 2D FE models. The surface shear-type actuation induces both in-plane
tively, to obtain the reference solution. Fig. 9 shows the FE mesh of force and moments in the plate, exciting the fundamental symmetric
a portion of the panel showing the intact segment and the entry of mode (S0 ) and antisymmetric mode (A0 ) of Lamb wave, respectively, at
delamination in undeformed and deformed configurations. the given frequencies. These wave modes get transmitted through the
For the 1D element model in which only the length direction is sub-laminates and are seen in the response at S. In addition to these
discretized, the forces ∓𝐹 (𝑡) at B′ and C′ are transferred to the mid- two waveforms, an extra mode also appears between them. This extra
plane of the laminate with additional bending moments ±𝑀(𝑡) = ℎ∕2 × mode results from the interaction of the S0 and A0 modes with the entry
𝐹 (𝑡), and higher order moments ∓𝑃 = 𝑅𝐿 (ℎ∕2) × 𝐹 (𝑡) at locations B and exit ends of delamination, causing mode conversion. It is a distinct
and C (shown in Fig. 8(b)). The response of the delaminated panel is damage feature that can be used to identify delamination damage in
obtained for two excitation frequencies, 70 and 100 kHz. For converged composite structures. The results show that the present ZIGT-based
solutions, two different mesh sizes of 72 and 92 SEs are employed to SE yields an accurate solution for both the fundamental modes and

11
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 8. Schematic configuration of the six-layer composite panel containing delamination for studying its effect on Lamb wave propagation induced by a PWAS patch pasted at
B′ C′ . (a) Actuator force application in the 2D FE model according to the pin-force method, (b) application of equivalent force system in the 1D model, and (c) laminate lay-up
showing the location of delamination interface. The response is recorded at S.

Fig. 9. Portion of 2D FE model in ABAQUS for the delaminated six-layer composite panel showing the mesh: (a) undeformed and (b) deformed configurations.

the extra mode in comparison with the continuum-based benchmark obtained for the displacement time-history 𝑢(𝑡) at S by using a mesh of
solution. However, the TOT-based solution yields a good match only a total of 155 SEs with sections AB, BC, CE1 , E1 E2 , E2 S, and SD having
for the S0 mode and exhibits significant deviations in the time of arrival 62, 1, 10, 5, 10, and 62 elements. To assess its relative performance, we
(ToA) and peak amplitude (PA) for both the extra mode and the A0 develop standard FE models of the equivalent number of DOFs using
mode. It shows that layerwise description of 𝑢 incorporated in the ZIGT 645 and 1075 two-node elements. These models employ 261, 7, 42,
without any additional cost over in the TOT is essential for accurate 15, 44, and 261 elements and 435, 7, 71, 25, 73, and 439 elements in
solution of the A0 mode and the extra mode caused by delamination the aforementioned segments for the 100 kHz and 200 kHz excitations,
damage in the laminated composite panels. respectively.
Fig. 11 compares the present SE and the two-node FE solutions
6.4. Spectral element versus standard FE with the reference solution. It is seen that while the current solutions
practically coincide with the reference solutions for all modes in the
In this section, the above-mentioned wave propagation problem is response, the standard FE solution with an equivalent number of DOFs
utilized to verify the superiority or otherwise of the present SE with the has large errors in the ToA and PA for the A0 and extra modes. For
standard FE based on the ZIGT [101,102]. By standard FE, we mean the higher frequency of 200 kHz, significant numerical dispersion and
the two-node FE employing Hermite shape functions for 𝑤0 and linear spurious oscillations can be seen in both A0 and extra modes, a well-
Lagrange shape functions for 𝑢0 and 𝜓0 . This analysis is performed for known feature of the standard FEM. This example establishes that the
the five-cycle tone burst pulse excitations with 100 and 200 kHz central present SE is far more computationally efficient and accurate than the
frequencies. The time steps are taken as 0.2 μs and 0.1 μs for the two standard FE for simulating wave propagation and its interaction with
frequencies, respectively. delamination damage in composite structures.
Since analytical solutions of the ZIGT are not available for this
problem, an accurate numerical solution based on the ZIGT is needed 6.5. Localization of delamination damage using lamb wave response
as the reference solution for this assessment. Accordingly, the standard
FE obtained employing a very fine mesh of 6000 two-node elements is In this section, we use the present SE solution to study the local-
taken as the reference. The time steps for the time marching are the ization of delamination defects using guided wave response. For this
same as specified above. investigation, the same composite panel as given in Section 6.3 and
For the 100 kHz pulse input, the mesh in the SE model is the same shown in Fig. 8 is considered unless otherwise stated below. The central
as described in the preceding section, with a total of 92 eight-node frequency of the pulse excitation is taken as 100 kHz. The idea here
elements. For the 200 kHz pulse excitation, the converged solution is is whether the location of delamination can be ascertained from the

12
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 10. Comparison of ZIGT- and TOT-SE solutions with continuum-based 2D FE Fig. 11. Comparison of solution accuracy of present SE and standard FE models of
simulation in ABAQUS for 𝑢(𝑡) due to surface-bonded transducer actuation at B′ C′ with the same size for the A0 mode and the delamination induced extra mode of the
a modulated five-cycle tone burst signal. The output is recorded at sensor location S axial displacement 𝑢(𝑡) recorded at S on panel’s top surface (Fig. 8) due to modulated
on the top surface of the delaminated panel. five-cycle tone burst excitation at B′ C′ .

responses measured at a surface of the panel. Note that the actual three delamination locations 𝑥𝑑𝑒𝑙
𝑐 = 0.0860 m, 0.1500 m, and 0.2175 m
damage localization can be achieved only by physical measurements (i.e., 𝑥 = 0.8360 m, 0.9 m, and 0.9675 m). To ascertain the sensitivity
of response. Nevertheless, the simulation can help interpretation of the of this variation to the delaminated interface location, it is plotted
measurement data as well as in training a data-driven machine learning for three interfacial locations (interface 1, 3, and 5 from the bottom).
model. For all interfacial locations, we see a distinct spike in the energy in
With this objective, the time-history response at an individual sensor the delamination region. Notably, its peak occurs at the delamination
position on the top surface of panel is studied first to identify possible centre, giving a clear feature for accurately localizing the damage. The
damage signature. We consider a delamination of length 𝑙𝑑𝑒𝑙 = 25 mm peak decreases as the distance of the delamination from the actuator
at a centre-to-centre distance of 𝑥𝑑𝑒𝑙 𝑐 = 150 mm from the actuator. increases for the delaminated interfaces not lying on the mid-plane.
The interfacial location of delamination is varied as the first, second, For the delamination at the middle interface, we do not see such a
fourth, and fifth interface from the bottom (see Fig. 8). Fig. 12 depicts decreasing trend of the peak energy with the delamination’s distance
the in-plane displacement response 𝑢(𝑡) at the sensor location S for the from the actuator, which is due to the absence of the extra mode in
four cases. The S0 and A0 mode waveforms do not seem to be affected the response signal. For delamination interfaces away from the sensor
by the interfacial location of the delamination defect. However, the surface, other smaller spikes are also observed behind the main spike in
extra mode bears distinct signatures of the interfacial location. Noting the delaminated region, but they almost vanish for interface-5, which
that the laminate is symmetric, for the delaminated interface pairs is closest to the sensor-side surface.
that are equidistant from the mid-plane, the extra mode has the same
amplitude but opposite phases which can be used to differentiate on 7. Conclusions
which side of the mid-plane the interface is located. Further, the extra
mode amplitude and dispersion (waveform length change) are higher A time-domain high-order spectral strip element has been developed
for delaminations nearer to the laminate’s mid-plane. Thus, the extra based on the efficient layerwise zigzag theory for free vibration and
mode’s amplitude, dispersion, and phase can be used as viable features wave propagation analyses of laminated beams and panels (strips)
to identify the interfacial location of delamination. containing multiple delaminations at arbitrary axial and interfacial lo-
Next, the delamination location is varied along the length direction. cations. It represents the first spectral element based on an ELT, which
To identify the spatial location, we compute the square sum of the combines high accuracy with the computational efficiency required for
transverse velocity 𝑤̇ 0 (𝑡) as observed at the panel’s top surface, denoted fast simulation of high-frequency guided wave propagation problems.
∑𝑇𝑠 2
as 𝐸𝑤 = 𝑡=0 𝑤̇ 0 (𝑡), for the time history up to the A0 mode, i.e., 𝑇𝑠 = Using the region approach of modelling delaminations, the formulation
230 μs. The transverse velocity profile along the panel’s surface can be incorporates the conventional PC model and the recently developed
measured in practice by non-contact laser vibrometers [103]. 𝐸𝑤 gives advanced HPLSC model in the context of the high-order SE for satis-
the measure of the transverse kinetic energy of the panel. In Fig. 13, 𝐸𝑤 fying the continuity of the displacement field at delamination fronts.
is plotted against the 𝑥-coordinate from the actuator to the sensor for It does so using the exact transformation matrix approach instead of

13
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 12. Effect of interfacial location of delamination on the in-plane displacement 𝑢(𝑡) measured at sensor location S on the top surface of the panel due to Hann windowed
five-count tone burst excitation with 𝑓 = 100 kHz.

the widely used approximate penalty approach. A detailed numerical A0 mode and the mode-converted extra mode, displaying its inferior
study has been performed to evaluate the performance of the proposed performance over the latter in terms of computational efficiency and
element for analysis of delaminated strips, from which the following accuracy.
inferences can be drawn. The amplitude, dispersion, and phase of the extra mode generated
The proposed ZIGT SE in conjunction with the HPLSC model for by Lamb wave’s interaction with delamination can be used as viable
continuity gives very accurate results for natural frequencies of thin and features to identify the interfacial location of delaminations. On the
thick composite beams containing multiple delaminations with refer- other hand, its longitudinal location can be identified from the plot of
ence to continuum-based FE solutions. The continuity model employed the square sum of transverse velocity, which shows a distinct spike in
at the delamination fronts has a large influence on the accuracy for the delamination region.
thick laminated structures. The standard PC model may yield large In summary, the proposed ZIGT SE-based model with the HPLSC
errors even for the fundamental frequency for such structures, particu- method for ensuring continuity at the delamination front fills up an
larly when the delaminations are large and hence is not recommended important void in the development of an accurate and computation-
for use. ally efficient technique for simulating guided wave propagation in
The proposed SE exhibits excellent accuracy in simulating Lamb delaminated composite structures. This development will benefit the
wave propagation and its interactions with delaminations, which are model-based SHM and physics-assisted data-driven SHM of laminated
used as damage features for SHM applications. Its predictions match composite structures featuring delamination damage. The current for-
the continuum-based FE solutions excellently well, not only for the mulation will be extended for the delaminated composite strips with
direct S0 and A0 modes but also for the extra mode generated due to physically modelled surface-bonded piezoelectric sensors and actuators
mode conversion at the entry and exit fronts of delaminations. This with or without debonding.
extra mode can be used as a distinguishing feature to identify the
presence of damage in the structure. However, the TOT, which has the
same number of DOFs as the ZIGT, produces significant errors in the CRediT authorship contribution statement
prediction of the time of arrival and the peak amplitude for A0 and
extra modes. It shows the high importance of the layerwise description
incorporated in the in-plane displacement field 𝑢 in the ZIGT without Mayank Jain: Methodology, Software, Investigation, Visualization,
additional computational cost over the TOT. The standard FE model Validation, Writing – original draft. Santosh Kapuria: Conceptualiza-
with a similar number of DOFs as the present SE model produces tion, Methodology, Software, Supervision, Writing – review & editing,
significant numerical dispersion and spurious oscillations in the direct Resources, Funding acquisition.

14
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

Fig. 13. Variation of transverse kinetic energy parameter 𝐸𝑤 along the top surface of the panel between the actuator and sensor for different longitudinal locations of delamination
introduced at (a) bottommost, (b) middle, and (c) topmost interfaces. The up-arrows on the 𝑥-axis indicate the positions of the peaks of the 𝐸𝑤 profiles, which are seen to almost
coincide with the centres of delaminations, clearly identifying their presence.

Declaration of competing interest Acknowledgements

The Science & Engineering Research Board, Government of India,

The authors declare that they have no known competing finan- has provided financial support for this work to S. Kapuria through J.
cial interests or personal relationships that could have appeared to C. Bose National Fellowship (Grant No. JCB/2018/000025) and Core
influence the work reported in this paper. Research Grant (CRG-2021-001237).

References
Data availability
[1] Gibson RF. Principles of composite material mechanics. Boca Raton: CRC Press;
2016.
[2] Garg AC. Delamination – A damage mode in composite structures. Eng Fract
Data will be made available on request. Mech 1988;29(5):557–84.

15
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

[3] Wisnom M. The role of delamination in failure of fibre-reinforced composites. [35] Li D, Liu Y, Zhang X. An extended layerwise method for composite laminated
Phil Trans R Soc A 2012;370(1965):1850–70. beams with multiple delaminations and matrix cracks. Internat J Numer
[4] Lamb H. On waves in an elastic plate. Proc R Soc Lond Ser A Math Phys Eng Methods Engrg 2015;101(6):407–34.
Sci 1917;93(648):114–28. [36] Kharghani N, Soares CG. Behavior of composite laminates with embedded
[5] Rose JL. Ultrasonic waves in solid media. New York: Cambridge University delaminations. Compos Struct 2016;150:226–39.
Press; 1999. [37] Kapuria S, Kumari P, Nath J. Efficient modeling of smart piezoelectric composite
[6] Kinra VK, Dayal V. Ultrasonic nondestructive evaluation of fibre-reinforced laminates: A review. Acta Mech 2010;214(1–2):31–48.
composite materials–A review. Sadhana 1987;11(3–4):419–32. [38] Ramkumar R, Kulkarni S, Pipes R. Free vibration frequencies of a delaminated
[7] Guo N, Cawley P. Lamb wave propagation in composite laminates and its beam. Reinforcing Future 1979;(22E):1–5.
relationship with acousto-ultrasonics. NDT E Int 1993;26(2):75–84. [39] Mujumdar PM, Suryanarayan S. Flexural vibrations of beams with delamina-
[8] Seale MD, Smith BT, Prosser W. Lamb wave assessment of fatigue and thermal tions. J Sound Vib 1988;125(3):441–61.
damage in composites. J Acoust Soc Am 1998;103(5):2416–24. [40] Campanelli RW, Engblom JJ. The effect of aminations in graphite/PEEK com-
[9] Pierce SG, Culshaw B, Manson G, Worden K, Staszewski WJ. Application of posite plates on modal dynamic characteristics. Compos Struct 1995;31(3):195–
ultrasonic Lamb wave techniques to the evaluation of advanced composite 202.
structures. In: Smart structures and materials 2000: sensory phenomena and [41] Ju F, Lee H, Lee K. Finite element analysis of free vibration of delaminated
measurement instrumentation for smart structures and materials, vol. 3986. composite plates. Compos Eng 1995;5(2):195–209.
International Society for Optics and Photonics; 2000, p. 93–103. [42] Krawczuk M, Ostachowicz W, Zak A. Dynamics of cracked composite material
[10] Worden K, Pierce SG, Manson G, Philp W, Staszewski WJ, Culshaw B. Detection structures. Comput Mech 1997;20(1):79–83.
of defects in composite plates using Lamb waves and novelty detection. Internat [43] Żak A, Krawczuk M, Ostachowicz W. Numerical and experimental investigation
J Systems Sci 2000;31(11):1397–409. of free vibration of multilayer delaminated composite beams and plates. Comput
[11] Kessler SS, Spearing SM, Soutis C. Damage detection in composite materials Mech 2000;26(3):309–15.
using Lamb wave methods. Smart Mater Struct 2002;11(2):269. [44] Szekrényes A, Máté P, Hauck B. On the dynamic stability of delaminated
[12] Su Z, Ye L, Lu Y. Guided Lamb waves for identification of damage in composite composite beams under free vibration. Acta Mech 2022;233(4):1485–512.
structures: A review. J Sound Vib 2006;295(3–5):753–80. [45] Radu AG, Chattopadhyay A. Dynamic stability analysis of composite plates
[13] Ricci F, Monaco E, Boffa N, Maio L, Memmolo V. Guided waves for structural including delaminations using a higher order theory and transformation matrix
health monitoring in composites: A review and implementation strategies. Prog approach. Int J Solids Struct 2002;39(7):1949–65.
Aerosp Sci 2022;129:100790. [46] Kumar A, Shrivastava R. Free vibration of square laminates with delamination
[14] Ramalho GM, Lopes AM, da Silva LF. Structural health monitoring of ad- around a central cutout using HSDT. Compos Struct 2005;70(3):317–33.
hesive joints using Lamb waves: A review. Struct Control Health Monit [47] Park T, Lee S-Y, Voyiadjis G. Finite element vibration analysis of compos-
2022;29(1):e2849. ite skew laminates containing delaminations around quadrilateral cutouts.
[15] Gao D, Wu Z, Yang L, Zheng Y. Integrated impedance and Lamb wave–based
Composites B 2009;40(3):225–36.
structural health monitoring strategy for long-term cycle-loaded composite
[48] Noh M-H, Lee S-Y. Dynamic instability of delaminated composite skew plates
structure. Struct Health Monit 2018;17(4):763–76.
subjected to combined static and dynamic loads based on HSDT. Composites B
[16] Oishi A, Yagawa G. Sustainable computational mechanics assisted by deep
2014;58:113–21.
learning. Comput Methods Appl Mech Engrg 2022;402:115025.
[49] Pölöskei T, Szekrenyes A. Dynamic stability analysis of delaminated composite
[17] Tay T. Characterization and analysis of delamination fracture in composites: An
beams in frequency domain using a unified beam theory with higher order
overview of developments from 1990 to 2001. Appl Mech Rev 2003;56(1):1–32.
displacement continuity. Compos Struct 2021;272:114173.
[18] Della CN, Shu D. Vibration of delaminated composite laminates: A review. Appl
[50] Ahmed A, Kapuria S. Third order theory based quadrilateral element for
Mech Rev 2007;60(1):1–20.
delaminated composite plates with a hybrid method for satisfying continuity
[19] Nallim LG, Oller S, Oñate E, Flores FG. A hierarchical finite element for
at delamination fronts. Compos Struct 2017;181:84–95.
composite laminated beams using a refined zigzag theory. Compos Struct
[51] Hu N, Fukunaga H, Kameyama M, Aramaki Y, Chang F. Vibration analysis of
2017;163:168–84.
delaminated composite beams and plates using a higher-order finite element.
[20] Tenek LH, Henneke II EG, Gunzburger MD. Vibration of delaminated com-
Int J Mech Sci 2002;44(7):1479–503.
posite plates and some applications to non-destructive testing. Compos Struct
[52] Kapuria S, Dube G, Dumir P. Exact piezothermoelastic solution for simply
1993;23(3):253–62.
supported laminated flat panel in cylindrical bending. ZAMM-J Appl Math Mech
[21] Su Z, Ye L. Lamb wave-based quantitative identification of delamination in
1997;77(4):281–93.
CF/EP composite structures using artificial neural algorithm. Compos Struct
[53] Kapuria S, Kumari P. Multiterm extended Kantorovich method for three-
2004;66(1–4):627–37.
dimensional elasticity solution of laminated plates. J Appl Mech 2012;79(6).
[22] Yam L, Wei Z, Cheng L, Wong WO. Numerical analysis of multi-layer composite
[54] Ambartsumyan S. A new refined theory of anisotropic shells. Polym Mech
plates with internal delamination. Comput Struct 2004;82(7–8):627–37.
[23] Alnefaie K. Finite element modeling of composite plates with internal 1970;6(5):766–76.
delamination. Compos Struct 2009;90(1):21–7. [55] Rath B, Das Y. Vibration of layered shells. J Sound Vib 1973;28(4):737–57.
[24] Ramadas C, Balasubramaniam K, Joshi M, Krishnamurthy C. Interaction of [56] Kapuria S. An efficient coupled theory for multilayered beams with
guided Lamb waves with an asymmetrically located delamination in a laminated embedded piezoelectric sensory and active layers. Int J Solids Struct
composite plate. Smart Mater Struct 2010;19(6):065009. 2001;38(50–51):9179–99.
[25] Ramadas C, Hood A, Padiyar J, Balasubramaniam K, Joshi M. Sizing of [57] Kapuria S, Alam N. Zigzag theory for buckling of hybrid piezoelectric beams
delamination using time-of-flight of the fundamental symmetric Lamb modes. under electromechanical loads. Int J Mech Sci 2004;46(1):1–25.
J Reinf Plast Compos 2011;30(10):856–63. [58] Cho M, Parmerter RR. Efficient higher order composite plate theory for general
[26] Su Z, Yang C, Pan N, Ye L, Zhou L-M. Assessment of delamination in lamination configurations. AIAA J 1993;31(7):1299–306.
composite beams using shear horizontal (SH) wave mode. Compos Sci Technol [59] Cho M, Parmerter R. Finite element for composite plate bending based on
2007;67(2):244–51. efficient higher order theory. AIAA J 1994;32(11):2241–8.
[27] Delrue S, Van Den Abeele K. Three-dimensional finite element simulation of [60] Kapuria S, Nath J. On the accuracy of recent global–local theories for bending
closed delaminations in composite materials. Ultrasonics 2012;52(2):315–24. and vibration of laminated plates. Compos Struct 2013;95:163–72.
[28] Liu Z, Yu H, He C, Wu B. Delamination detection in composite beams using [61] Kapuria S, Achary G. An efficient higher order zigzag theory for laminated
pure Lamb mode generated by air-coupled ultrasonic transducer. J Intell Mater plates subjected to thermal loading. Int J Solids Struct 2004;41(16–17):4661–
Syst Struct 2014;25(5):541–50. 84.
[29] Murat BI, Khalili P, Fromme P. Scattering of guided waves at delaminations in [62] Kapuria S, Kulkarni S. An improved discrete Kirchhoff quadrilateral element
composite plates. J Acoust Soc Am 2016;139(6):3044–52. based on third-order zigzag theory for static analysis of composite and sandwich
[30] Hervin F, Maio L, Fromme P. Guided wave scattering at a delamination in a plates. Internat J Numer Methods Engrg 2007;69(9):1948–81.
quasi-isotropic composite laminate: Experiment and simulation. Compos Struct [63] Ahmed A, Kapuria S. A four-node facet shell element for laminated shells based
2021;275:114406. on the third order zigzag theory. Compos Struct 2016;158:112–27.
[31] Soleimanpour R, Ng C-T. Scattering analysis of nonlinear Lamb waves at [64] Cho M, Kim J-S. Higher-order zig-zag theory for laminated composites with
delaminations in composite laminates. J Vib Control 2022;28(11–12):1311–23. multiple delaminations. J Appl Mech 2001;68(6):869–77.
[32] Barbero E, Reddy J. Modeling of delamination in composite laminates using a [65] Kim HS, Chattopadhyay A, Ghoshal A. Characterization of delamination effect
layer-wise plate theory. Int J Solids Struct 1991;28(3):373–88. on composite laminates using a new generalized layerwise approach. Comput
[33] Marjanović M, Vuksanović D. Layerwise solution of free vibrations and buckling Struct 2003;81(15):1555–66.
of laminated composite and sandwich plates with embedded delaminations. [66] Oh J, Cho M, Kim J-S. Dynamic analysis of composite plate with multi-
Compos Struct 2014;108:9–20. ple delaminations based on higher-order zigzag theory. Int J Solids Struct
[34] Kumar SK, Cinefra M, Carrera E, Ganguli R, Harursampath D. Finite element 2005;42(23):6122–40.
analysis of free vibration of the delaminated composite plate with variable [67] Massabo R, Campi F. Assessment and correction of theories for multilayered
kinematic multilayered plate elements. Composites B 2014;66:453–65. plates with imperfect interfaces. Meccanica 2015;50:1045–71.

16
M. Jain and S. Kapuria Composite Structures 325 (2023) 117585

[68] Eijo A, Oñate E, Oller S. A numerical model of delamination in composite [86] Li F, Peng H, Sun X, Wang J, Meng G. Wave propagation analysis in composite
laminated beams using the LRZ beam element based on the refined zigzag laminates containing a delamination using a three-dimensional spectral element
theory. Compos Struct 2013;104:270–80. method. Math Probl Eng 2012;2012:1–19.
[69] Groh R, Tessler A. Computationally efficient beam elements for accurate stresses [87] Wandowski T, Kudela P, Ostachowicz W. Numerical analysis of elastic wave
in sandwich laminates and laminated composites with delaminations. Comput mode conversion on discontinuities. Compos Struct 2019;215:317–30.
Methods Appl Mech Engrg 2017;320:369–95. [88] Rekatsinas C, Chrysochoidis N, Saravanos D. Investigation of critical de-
[70] Flores FG, Oller S, Nallim LG. On the analysis of non-homogeneous laminates lamination characteristics in composite plates combining cubic spline piezo-
using the refined zigzag theory. Compos Struct 2018;204:791–802. layerwise mechanics and time domain spectral finite elements. Wave Motion
[71] Kapuria S, Ahmed A. An efficient zigzag theory based finite element modeling 2021;106:102752.
of composite and sandwich plates with multiple delaminations using a hybrid [89] Rekatsinas C, Saravanos D. A Hermite spline layerwise time domain spectral
continuity method. Comput Methods Appl Mech Engrg 2019;345:212–32. finite element for guided wave prediction in laminated composite and sandwich
[72] Kapuria S, Ahmed A. Delamination modeling in doubly curved laminated plates. J Vib Acoust 2017;139(3):1–14.
shells for free vibration analysis using zigzag theory-based facet shell el- [90] Kudela P, Ostachowicz W. A multilayer delaminated composite beam and plate
ement and hybrid continuity method. Internat J Numer Methods Engrg elements: Reflections of Lamb waves at delamination. Mech Adv Mater Struct
2019;120(9):1126–47. 2009;16(3):174–87.
[73] Lee R, Cangellaris AC. A study of discretization error in the finite [91] Schulte RT, Fritzen C-P. Simulation of wave propagation in damped
element approximation of wave solutions. IEEE Trans Antenna Propag Composite Structures with piezoelectric coupling. J Theoret Appl Mech
1992;40(5):542–9. 2011;49(3):879–903.
[74] Bond L. Numerical techniques and their use to study wave propagation and [92] Kudela P, Moll J, Fiborek P. Parallel spectral element method for guided wave
scattering- a review. In: Elastic waves and ultrasonic nondestructive evaluation. based structural health monitoring. Smart Mater Struct 2020;29(9):095010.
1990, p. 17–27. [93] Jain M, Kapuria S. Time-domain spectral finite element based on third-order
[75] Palacz M. Spectral methods for modelling of wave propagation in structures in theory for efficient modelling of guided wave propagation in beams and panels.
terms of damage detection–A review. Appl Sci 2018;8(7):1124. Acta Mech 2022;233(3):1187–212.
[76] Maio L, Fromme P. On ultrasound propagation in composite laminates: advances [94] Jain M, Kapuria S. C1 -continuous time-domain spectral finite element for
in numerical simulation. Prog Aerosp Sci 2022;129:100791. modeling guided wave propagation in laminated composite strips based on
[77] Ostachowicz W, Krawczuk M, Cartmell M, Gilchrist M. Wave propagation in third-order theory. Compos Struct 2022;289:115442.
delaminated beam. Comput Struct 2004;82(6):475–83. [95] Kapuria S, Dumir P, Jain N. Assessment of zigzag theory for static loading,
[78] Palacz M, Krawczuk M, Ostachowicz W. The spectral finite element model buckling, free and forced response of composite and sandwich beams. Compos
for analysis of flexural-shear coupled wave propagation. Part 2: Delaminated Struct 2004;64(3–4):317–27.
multilayer composite beam. Compos Struct 2005;68(1):45–51. [96] Jain M, Kapuria S, Pradyumna S. Efficient time-domain spectral element with
[79] Nanda N. Spectral finite element method for wave propagation analysis in zigzag kinematics for multilayered strips. Int J Mech Sci 2022;232:107603.
smart composite beams containing delamination. Aircr Eng Aerosp Technol [97] Kapuria S, Jain M. A C1 -continuous time domain spectral finite element for
2020;92(3):440–51. wave propagation analysis of Euler–Bernoulli beams. Internat J Numer Methods
[80] Gottlieb D, Orszag SA. Numerical analysis of spectral methods: theory and Engrg 2021;122(11):2631–52.
applications. SIAM; 1977. [98] Shen M-H, Grady J. Free vibrations of delaminated beams. AIAA J
[81] Patera AT. A spectral element method for fluid dynamics: Laminar flow in a 1992;30(5):1361–70.
channel expansion. J Comput Phys 1984;54:468–88. [99] Petyt M. Introduction to finite element vibration analysis. New York: Cambridge
[82] Lee-Wing H, Patera AT. A Legendre spectral element method for simulation University Press; 2010.
of unsteady incompressible viscous free-surface flows. Comput Methods Appl [100] Nam H, Hwang W, Han K. Stacking sequence design of fiber-metal laminate
Mech Engrg 1990;80(1–3):355–66. for maximum strength. J Compos Mater 2001;35(18):1654–83.
[83] Pozrikidis C. Introduction to finite and spectral element methods using [101] Kapuria S, Dumir P, Ahmed A, Alam N. Finite element model of efficient
MATLAB. Boca Raton: CRC Press; 2005. zig-zag theory for static analysis of hybrid piezoelectric beams. Comput Mech
[84] Peng H, Ye L, Meng G, Mustapha S, Li F. Concise analysis of wave propagation 2004;34:475–83.
using the spectral element method and identification of delamination in CF/EP [102] Kapuria S, Alam N. Efficient layerwise finite element model for dynamic
composite beams. Smart Mater Struct 2010;19(8):085018. analysis of laminated piezoelectric beams. Comput Methods Appl Mech Engrg
[85] Munian RK, Mahapatra DR, Gopalakrishnan S. Lamb wave interaction with 2006;195(19–22):2742–60.
composite delamination. Compos Struct 2018;206:484–98. [103] Kudela P, Wandowski T, Malinowski P, Ostachowicz W. Application of scanning
laser Doppler vibrometry for delamination detection in composite structures.
Opt Lasers Eng 2017;99:46–57.

17