1 s2.0 S0020722516304773 Main

International Journal of Engineering Science 109 (2016) 115164
Contents lists available at ScienceDirect
International Journal of Engineering Science

journal homepage: www.elsevier.com/locate/ijengsci
On the mechanics of laminated doubly-curved shells

subjected to point and line loads
Francesco Tornabene, Nicholas Fantuzzi, Michele Bacciocchi
DICAM - Department, School of Engineering and Architecture, University of Bologna, Italy
a r t i c l e i n f o a b s t r a c t
Article history: It is well-known that the implementation of concentrated forces, such as point and line
Received 14 July 2016 loads, represents a challenging task, especially from the computational point of view, since
Revised 10 August 2016
a strong discontinuity has to be inserted in the structural model. The present paper aims
Accepted 1 September 2016
to solve the static problem of laminated composite doubly-curved shell structures sub-
Available online 28 September 2016
jected to concentrated loads employing the Generalized Differential Quadrature (GDQ) as
Keywords: numerical tool, according to what has been shown by the authors in their previous works.
Laminated composite doubly-curved shells Its accuracy and reliability features are proven for several grid distributions when the con-
Point and line loads centrated loads are modeled through the Dirac-delta function. The theoretical framework
Generalized differential quadrature method on which this approach is based is a Unied Formulation, which allows to investigate sev-
Higher-order equivalent single layer eral Higher-order Shear Deformation Theories (HSDTs). The differential geometry is used
theories to describe accurately the reference surface of various doubly-curved shell structures. The
validity of the current approach is shown comparing the GDQ results with the exact and
semi-analytical ones available in the literature. A posteriori recovery procedure based on
the three-dimensional equilibrium equations for a shell structure is introduced to com-
pute the through-the-thickness variation of strain, stress and displacement components by
means of the GDQ method.
2016 Elsevier Ltd. All rights reserved.
1. Introduction
In the last decades, composite materials have considerably affected the way of manufacturing and analyzing several kinds
of engineering structures employed in the automotive, aerospace, biomedical, architectural, mechanical, and civil sectors.
Their increasing use is proven by the huge number of research papers published in the last years. All these aspects are
discussed in depth in the book by Reddy (Reddy, 2004).
One of the most exploited types of composite materials is represented by brous composites. For the sake of complete-
ness, it should be recalled that a ber-reinforced medium consists of several bers embedded in a matrix according to
prearranged paths. Due to this micromechanical conguration, high level performance can be reached if compared with
an isotropic material. It is clear that the mechanical properties of these composites depend on the orientation of the re-
inforcing bers and their volume fraction, as well as their strength and stiffness (Tornabene, Fantuzzi, Bacciocchi, & Viola,
2016c,d). The remarkable mechanical properties of brous composites can be further improved by combining together dif-
ferent ber-reinforced layers in order to obtain the so-called laminated composite. The layers can be arranged in several
manners, according to the lamination scheme, or stacking sequence. The way these composites are assembled, as well as the

Corresponding author.
E-mail address: francesco.tornabene@unibo.it (F. Tornabene).
http://dx.doi.org/10.1016/j.ijengsci.2016.09.001
0020-7225/ 2016 Elsevier Ltd. All rights reserved.
116 F. Tornabene et al. / International Journal of Engineering Science 109 (2016) 115164
orientation and the mechanical properties of each single ber-reinforced layer, must be considered to obtain a peculiar struc-
tural response (Brischetto, 2014; Brischetto, & Torre, 2014; Eftekhari, & Jafari, 2014; Groh, & Weaver, 2015a,b; Jin, Ye, Wang,
& Miao, 2016; Kulikov, Mamontov, Plotnikova, & Mamontov, 2016; Le, & Yi, 2016; Librescu, & Reddy, 1989; Mohammad-
Abadi, & Daneshmehr, 2015; Mohammad-Abadi, Daneshmehr, & Homayounfard, 2015; Mohazzab, & Dozio, 2015; Piskunov,
Verijenko, Adali, & Summers, 1993; Vescovini, & Dozio, 2016; Wu, & Hung, 1999; Yifeng, Wenzheng, Yu, Wenwen, & Lichao,
2015).
It is evident that laminated composites are heterogeneous mediums. As a consequence, the well-known phenomenon
of delamination can happen due to the high values of shear stresses that can be registered among the laminae. For this
purpose, the class of granular composites (also known as Functionally Graded Materials or FGMs), has been developed to
avoid this problem (Alibeigloo, 2016; Behravan Rad, & Shariyat, 2015; Fazzolari, 2014, 2015, 2016a,b; Guo, Chen, & Pan,
2016; Jooybar, Malekzadeh, Fiouz, & Vaghe, 2016; Kim, 2015; Liu, Cheng, & Liu, 2015; Mantari, 2015; Mercan, Demir, &
Civalek, 2016; Quan, Tran, Tuan, & Duc, 2015; Shaei, Kazemi, Sa, & Ghadiri, 2016; Soyev, 2015, 2016; Soyev, & Kuruoglu,
2015, 2016). Recently, due to the advancements in the manufacturing process, innovative materials such as piezoelectric
ones can be inserted in the lamination scheme as distributed sensors and actuators for an active control on the structural
conditions, as highlighted in the work by Ray and Reddy (Ray, & Reddy, 2005). This idea has contributed to the development
of the so-called smart structures (Brischetto, & Carrera, 2013; Cheng, Lim, & Kitipornchai, 1999; Dehghan, Nejad, & Moosaie,
2016; Hadjiloizi, Kalamkarov, Metti, & Georgiades, 2014b,a). Another way to enhance the structural response of composite
materials that is attracting recently the interest of many researchers consists in inserting carbon nanoparticles in a polymeric
matrix as reinforcing phase (Brischetto, Tornabene, Fantuzzi, & Bacciocchi, 2015; Kamarian, Salim, Dimitri, & Tornabene,
2016; Tornabene, Fantuzzi, Bacciocchi, & Viola, 2016). As a consequence, more and more complex constitutive laws must
be introduced to take into account these mechanical congurations. Simultaneously, even the fundamental assumptions and
the kinematic model must be adequate to capture the effective structural response. For these purposes, several Higher-
order Shear Deformation Theories (HSDTs) have been developed recently by many researchers. One of the most ecient
approaches to deal with these HSDTs can be found in the works by Carrera (Carrera, 2002, 2003, 2004), in which a Unied
Formulation was presented. Its ecacy is given by its generality. In other words, many different higher-order theories can
be developed and investigated by means of only one formulation. In addition, the order of kinematic expansion which
characterizes the enrichment of the displacement eld can be chosen arbitrarily.
In the current paper, the aforementioned theoretical formulation is employed to investigate the static behavior of lami-
nated structures subjected to concentrated forces, starting from the preliminary results and considerations presented by the
authors in their previous work (Tornabene, Fantuzzi, Bacciocchi, & Viola, 2015c). In general, concentrated loads represent
a strong discontinuity for those researchers who aim to nd a solution to these kinds of structural problems employing a
numerical approach. At this point, a brief and partially complete literature review concerning concentrated load problems
is presented. The classical laminated plate theory was employed by Becker in his work (Becker, 1995) to study the static
behavior of innitely extended unsymmetrical laminates subjected to concentrated forces and moments, developing a com-
plex potential approach. Kim and Swanson (Kim, & Swanson, 2001) presented some analyses to design sandwich structures
in order to bear concentrated loading, highlighting the typical failure modes that characterize these peculiar laminates. The
issue of composite laminates with elliptic elastic inclusions subjected to concentrated forces and moments was considered
in the paper by Hwu and Tan (Hwu, & Tan, 2007), basing on the Kirchhoffs assumptions for thin plate. The Boundary Ele-
ment Method was used by Tsamasphyros et al. (Tsamasphyros, Theotokoglou, & Filopoulos, 2013) to solve some problems of
structures loaded by concentrated forces and moments. Abali et al. (Abali et al., 2014) combined an analytical model with
computational means to study the three-dimensional elastic deformations of isotropic functionally graded plates subjected
to point loads, putting in evidence some limitations typical of nite element approaches. An experimental analysis was car-
ried out by Nikopour and Selvadurai (Nikopour & Selvadurai, 2014) to investigate the exural behavior of ber-reinforced
rectangular plates with different boundary conditions. The results related to the application of a central load were compared
to the numerical solutions obtained through a nite element model. The same kind of analysis was realized by Cernescu and
Romanoff in their paper (Cernescu, & Romanoff, 2015) to study the bending deections of sandwich beams in three-point
bending. Finally, Eftekhari introduced the Dirac-delta function in order to apply a concentrated load combining differential
quadrature and integral quadrature methods (Eftekhari, 2015). In particular, the accuracy and the reliability of the present
approach were both proven by the analysis of beams and plates under concentrated loads. It can be easily noticed that only
at structures have been analyzed in these papers. The work by Reddy (Reddy, 1984) represents one of the few examples
related to laminated shells subjected to central point loads. In his paper, several benchmark solutions were obtained using
a semi-analytical approach, starting from the Sanders theory for doubly-curved shallow shells. A typical problem of shell
structures subjected to concentrated forces is represented by the problem of the pinched cylinders (Beiro da Veiga, 2005;
Jones, 1998b,a; Morley, 1960; Ting, & Yuan, 1958; Yuan, 1946; Yuan, & Ting, 1957; Zhang, 1991). For the sake of complete-
ness, it should be recalled that the pinched problem consists in a structure subjected to concentrated, equal and opposite
loads, applied on the diameter. In the works by Yuan (Yuan, 1946) and Yuan and Ting (Yuan, & Ting, 1957), the radial de-
ections of innitely long and nite-length cylinders were computed, respectively. The loading function was modeled by a
Fourier integral in the longitudinal direction and by a Fourier series along the circumferential direction. The radial deforma-
tions of thin-walled pinched circular cylinder with simply supported and free edges were evaluated by Ting and Yuan (Ting,
& Yuan, 1958). Morley (Morley, 1960) obtained an approximate solution of the pinched cylinder problem solving a simplied
eighth order equation. A set of parametric studies concerning directional properties was developed by Zhang in his paper
F. Tornabene et al. / International Journal of Engineering Science 109 (2016) 115164 117
(Zhang, 1991) to investigate the mechanical behavior of pinched orthotropic cylindrical shells in orthogonal curvilinear co-
ordinates using a nite element code. In the works (Jones, 1998b,a), Jones provided a solution for orthotropic cylindrical
shells under pinching loads with nite length starting from the corresponding solution valid for innitely long cylinders. It
is important to notice that Jones highlighted the diculties related to the implementation of a concentrated load. A dis-
placement analysis was carried out numerically for a pinched cylindrical shell by Beiro da Veiga (Beiro da Veiga, 2005).
The concentrated force was modeled through a point wise Dirac-delta function acting orthogonally to the shell surface. On
the other hand, the works (Belytschko, Liu, Ong, & Lam, 1985; Bonisoli, Delprete, Sesana, Tamburro, & Tornincasa, 2015;
Hughes, & Liu, 1981; Liu & Liew, 1998; Liu & Liew, 1999; Rao, 1978; Vlasov, 1964; Yang, 1971) must be mentioned as exam-
ples of structural elements subjected to concentrated actions by means of nite element methods. It is worth noting that
some authors developed different numerical strategies to deal with concentrated forces, especially for solving moving load
problems. For this purpose, the papers by Eftekhari (Eftekhari, 2016b,a; 2015a) and the book by Wang (Wang, 2015) must
be cited.
The present approach has its roots in the works (Eftekhari, 2016b,a; 2015a; Tornabene et al., 2015; Wang, 2015). Its main
aim is to investigate the static behavior in terms of both displacements and stress and strain proles of laminated doubly-
curved shell structures subjected to point and line loads. The solutions are obtained numerically employing the Generalized
Differential Quadrature (GDQ) method, presented in the nineties by Shu and Richards (Shu, & Richards, 1992b,a), due to its
accuracy and stability features. For completeness purposes, the main characteristics of the numerical technique in hand can
be found in the review paper by Tornabene et al. (Tornabene, Fantuzzi, Ubertini, & Viola, 2015). The outstanding features
of the GDQ method for the static and dynamic analyses of laminated composite doubly-curved shell structures are proven
in the works (Ferreira et al., 2014; Tornabene, & Reddy, 2013; Tornabene et al., 2013; Tornabene et al., 2015, Tornabene,
Fantuzzi, Bacciocchi, & Dimitri, 2015, Tornabene et al., 2015a, Tornabene, Brischetto, Fantuzzi, & Viola, 2015, Tornabene et al.,
2015a, Tornabene, Fantuzzi, & Bacciocchi, 2016, Tornabene, 2016, Tornabene, Fantuzzi, & Bacciocchi, 2016, Tornabene et al.,
2016, Bacciocchi, Eisenberger, Fantuzzi, Tornabene, & Viola, 2016, Tornabene et al., 2016; Tornabene, Fantuzzi, & Bacciocchi,
2014; Tornabene, Fantuzzi, Viola, & Batra, 2015; Tornabene, Fantuzzi, Viola, & Carrera, 2014; Tornabene, Fantuzzi, Viola, &
Ferreira, 2013; Tornabene, Fantuzzi, Viola, & Reddy, 2014; Viola et al., 2013b; Viola, Tornabene, & Fantuzzi, 2013). In the same
manner, excellent numerical results via GDQ method can be obtained also for the structural response of structures with
one or more layers made of FGMs (Brischetto, Tornabene, Fantuzzi, & Viola, 2016, Tornabene, & Ceruti, 2013, Tornabene, &
Viola, 2009, Tornabene, & Viola, 2013, Tornabene et al., 2014, Viola et al., 2016, Viola, Rossetti, Fantuzzi, & Tornabene, 2014).
Finally, an advanced application of the GDQ technique is used in the works (Fantuzzi, 2014, Fantuzzi, & Tornabene, 2014,
Fantuzzi, & Tornabene, 2014, Fantuzzi, & Tornabene, 2016, Fantuzzi et al., 2014a,b Fantuzzi et al., 2014, Fantuzzi, Dimitri,
& Tornabene, 2016, Fantuzzi, Bacciocchi, Tornabene, Viola, & Ferreira, 2015, Fantuzzi, Tornabene, & Viola, 2016, Tornabene,
Fantuzzi, & Bacciocchi, 2014a, Tornabene, Dimitri, & Viola, 2016, Tornabene, Liverani, & Caligiana, 2012, Viola et al., 2013,
Viola, Miniaci, Fantuzzi, & Marzani, 2015, Viola, Tornabene, & Fantuzzi, 2013, Viola, Tornabene, Ferretti, & Fantuzzi, 2013,
Viola, Tornabene, Ferretti, & Fantuzzi, 2013) to deal with structural problems characterized by geometric and mechanical
discontinuities solving the strong form of the governing equations and subdividing the reference domain in nite elements.
Finally, the outline of the present paper is summarized at this point. First, the differential geometry principles are illus-
trated to dene the doubly-curved surfaces which represent the middle surfaces of the shell structures under consideration.
Second, the general framework of the Unied Formulation is explained. Third, the computational implementation of both
the GDQ method and the concentrated forces are presented. Then, a posteriori recovery procedure is illustrated to compute
the through-the-thickness variation of strain, stress and displacement components. Thereafter, many numerical investiga-
tions are shown. The validity of the present approach is proven by means of several comparisons with the results available
in the literature.
2. Shell geometry
The starting point for studying several kinds of structural problems related to doubly-curved shells is the denition of the
reference domain in which the fundamental system of governing equations is written. Assuming a two-dimensional model
as structural theory, the reference domain in hand is given by the shell middle surface. As a consequence, each geometric
and mechanical parameter must be referred to the reference surface. This assumption is valid even if the structure under
consideration is made of several layers, as the one depicted in Fig. 1. Thus, the structural response of a shell structure under
dynamic or static loads is represented by the mechanical behavior of its middle surface. As highlighted in many previous
works recently presented by the authors (Bacciocchi et al., 2016, Tornabene, 2016, Tornabene, & Ceruti, 2013, Tornabene, &
Viola, 2009, Tornabene, & Viola, 2013, Tornabene et al., 2014, Tornabene et al., 2015, Tornabene et al., 2015b, Tornabene et
al., 2015, Tornabene et al., 2015, Tornabene et al., 2015,Tornabene et al., 2016, Tornabene et al., 2016, Tornabene et al., 2016,
Tornabene et al., 2016, Tornabene et al., 2014, Tornabene, Fantuzzi, Bacciocchi, & Viola, 2015, Viola et al., 2014), the principles
of the differential geometry are essential to deal with doubly-curved surfaces that can be characterized by variable radii of
curvature. It must be recalled that a generic shell is a three-dimensional solid. Each point P within the structural domain is
identied through the position vector R( 1 , 2 , ) introduced below
h
R(1 , 2 , ) = r(1 , 2 ) + zn(1 , 2 ) (1)
2
Fig. 1. Representation of a generic doubly-curved laminated panel and its stacking sequence. The parameters , , , denote the orientation of the
various ber-reinforced layers.
in which z = 2 /h [1, 1] is a non-dimensional parameter that stands for the distance of the point P from its projection
P on the middle surface. For the sake of clarity, the reader can refer to the picture depicted in Fig. 1. On the other hand,
h represents the total thickness of the shell. If a laminated composite structure made of l layers (or plies) in which each
lamina has thickness hk is considered, the overall thickness can be dened as follows

l
h= hk (2)
k=1
The shell middle surface is described by the corresponding position vector r( 1 , 2 ), which can be specialized to dene
the proper geometry. At this point, it should be highlighted that the local reference system O 1 2 has been introduced
upon the reference surface (Fig. 1). By hypotheses, the curvilinear orthogonal coordinates 1 , 2 coincide with the lines of
principal curvature of the shell reference surface. In other words, 1 , 2 represent a couple of principal curvilinear coor-
dinates. The third coordinate , instead, represents the normal direction which is identied in each point of the reference
domain by the outward unit normal vector n( 1 , 2 ) dened as follows
r,1 r,2
n= (3)
|r,1 r,2 |
in which r,i = r/ i , for i = 1, 2. For clarity purposes, it should be specied that the notation stands for the vector
product. The curvilinear coordinates 1 , 2 have different meanings according to the structure that has to be studied. Due
to the generality of the present formulation, the mechanical behavior of doubly-curved and singly curved shells, as well as
annular and rectangular plates, can be investigated. For instance, the couple of principal coordinates can be assumed equal
to 1 = , 2 = for a doubly-curved revolution shell or taken as 1 = , 2 = y for a singly-curved translational cylinder.
If a at structure is considered, which can be classied as a degenerate shell, the curvilinear coordinates at issue are set
equal to 1 = x, 2 = and 1 = x, 2 = y for circular and rectangular plates, respectively. In any case, since a shell structure
has specic limits that identify the three-dimensional solid, the coordinates in hand are limited according to the following
ranges

1 10 , 11 , 2 20 , 21 , [h/2, h/2] (4)
The denition of the position vector r( 1 , 2 ) allows to compute also the well-known Lam parameters A 1 ( 1 , 2 ) and
A 2 ( 1 , 2 )

A1 = r,1 r,1 , A2 = r,2 r,2 (5)
where the scalar product is specied through the symbol . Finally, the last quantities that have to be introduced are
the principal radii of curvature of the shell reference surface R 1 ( 1 , 2 ) and R 2 ( 1 , 2 ). In the most general case, such
quantities are different point by point. The differential geometry allows to dene the geometric parameters in hand through
the following expressions
r,1 r,1 r,2 r,2
R1 = , R2 = (6)
r,11 n r,22 n
if the hypothesis of principal curvilinear coordinates is respected. Once the principal radii of curvature are dened, the
quantities H1 = 1 + /R1 and H2 = 1 + /R2 , which represent the parameters linked to the shell curvature, can be introduced
too.
3. Higher-order shear deformation theories
The well-known works by Carrera (Carrera, 2002, 2003, 2004) have provided a general framework for the development
of a unied formulation, known as Carrera Unied Formulation (CUF), able to study several Higher-order Shear Deformation
Theories (HSDTs). In general, such theories can be classied as Layer-Wise (LW) theories, which consider each layer of
a laminated structure separately from the others (Tornabene, 2016, Tornabene, Fantuzzi, Bacciocchi, & Viola, 2015), and
Equivalent Single Layer (ESL) models, in which all the mechanical and geometric parameters are evaluated on the shell
middle surface (Viola et al., 2013a,b). In the present paper, only the ESL approach is taken into account. According to this
theory, the 3D displacement eld for a generic laminated composite shell structure assumes the following aspect
U1 = F0 u1( ) + F1 u1( ) + F2 u1( ) + F3 u1( ) + ... + FN u1( ) + FN+1 u1( )

0 1 2 3 N N+1

0 1 2 3 N N+1

0 1 2 3 N N+1
(7)
in which U1 ( 1 , 2 , ), U2 ( 1 , 2 , ), U3 ( 1 , 2 , ) represent the three-dimensional displacement components. As it can be

easily noticed from the denition (7), the kinematic model can be arbitrarily expanded. Thus, the choice of the order of the
kinematic expansion is completely free and different HSDTs can be generated. It is evident that the degrees of freedom
of the problem in hand are given by the generalized displacements u1( ) (1 , 2 ), u2( ) (1 , 2 ), u3( ) (1 , 2 ), for = 0, 1, 2, ...,
N, N + 1, dened on the shell middle surface. The vector u( ) = u( ) ( 1 , 2 ) can be conveniently introduced to collect each
triad of displacements for every order of kinematic expansion. Each generalized displacement is multiplied by the so-called
thickness function (or shear function) F = F ( ). The selection of these functions is completely free and allows to establish a
huge variety of kinematic models. Several examples of thickness function are listed in the works (Viola et al., 2013a,b). For a
proper choice of both the order of expansion and the thickness functions, the classic Reissner-Mindlin theory or First-order
Shear Deformation Theory (FSDT) is obtained, too. In general, the development of peculiar composite materials or particular
laminated structures has caused the need of enriched kinematic models able to capture some characteristic effects. Among
them, the zig-zag effect is quite typical when a sandwich structure must be analyzed. Thus, the Murakamis function Z = Z( )
is introduced for these purposes. In general, it is related to the (N + 1)-th order of expansion and it is dened as follows for
a laminated composite structure

2 k+1 + k
Z = (1 )k (8)
k+1 k k+1 k
where k denotes the coordinate of the k-th ply along the direction (Fig. 1). A complete treatise concerning the use and
the properties of the Murakamis function can be found in the works by Carrera (Carrera, 2003, 2004). In the present paper,
the power function , for = 0, 1, 2, ..., N, is chosen as thickness function for each order of kinematic expansion. As a
consequence, a specic nomenclature can be introduced now to identify univocally any HSDT. In general, the maximum
order of kinematic expansion = N denes the model itself. Thus, the following structural theories can be obtained
N =1 ED1
N =2 ED2
(9)
N =3 ED3
N =4 ED4
where the letter E means that an ESL model is considered, whereas D signies that the unknown variables of the struc-
tural problem are the generalized displacements. If the Murakamis function is embedded in the model, the letter Z is
added to the previous nomenclature
N =1 EDZ1
N =2 EDZ2
(10)
N =3 EDZ3
N =4 EDZ4
Similarly, the Reissner-Mindlin theory can be supplemented with the Murakamis function and the notation FSDTZ is
dened for this aim.
When a higher-order ESL model is considered, the generalized strain components dened on the shell middle surface
must be written for each order of kinematic expansion. The corresponding algebraic vector ( ) = ( ) ( 1 , 2 ) is introduced
to collect these quantities as follows
T
( ) = 1( ) 2( ) 1( ) 2( ) 13( ) ( )
23 ( )
13 ( )
23 3( ) (11)
In compact notation, the th order generalized strains ( ) can be related to the generalized displacements u( ) , linked
to same order of kinematic expansion, according to the following expression
( ) = D u ( ) (12)
where the differential operator D is dened as

1 1 A2 1 A1 1 1
T
0 1 0 0
A 1 1 A 1 A 2 1 A 1 A2 2 A2 2 R1

1 A1 1 1 1 A2 1
D =
A 1 A 2 2 0 0 1 0
(13)
A2 2 A1 1 A 1 A2 1 R2

1 1 1 1
0 0 0 0 1
R1 R2 A1 1 A2 2
Similarly, the th order stress resultants collected in the vector S( ) =S( ) ( 1 , 2 ) are related to the generalized displace-
ments of the same kinematic expansion order as follows

N+1
S ( ) = A ( ) D u ( ) (14)
=0
Having in mind the duality between the strains and the stresses, the extended form of the algebraic vector S( ) is given
by
T
S( ) = N1( )
N2( )

( )
N12

( )
N21

T1( )

T2( )

P1( )

P2( )

S3( ) (15)
The matrix A( ) , for each order of kinematic expansion , = 0, 1, 2, ..., N, N + 1, represents the stiffness matrix. When
a laminated shell structure composed of l orthotropic plies made of linear elastic material is taken into account (Fig. 1), the
stiffness matrix assumes the aspect shown below
( )
( ) ( )
( )
( )

A11 (20 ) A12 (11 ) A16 (20 ) A16 (11 ) 0 0 0 0 A13 (10 )

A ( )
( ) ( )
( )
( )
12(11) A22 (02 ) A26 (11 ) A26 (02 ) 0 0 0 0 A23 (01 )

A ( )
( ) ( )
( )
( )
16(20) A26 (11 ) A66 (20 ) A66 (11 ) 0 0 0 0 A36 (10 )

A ( )
( ) ( )
( )
( )
16(11) A26 (02 ) A66 (11 ) A66 (02 ) 0 0 0 0 A36 (01 )

( ) ( ) ( ) ( ) ( )
A = 0 0 0 0 A44 (20 ) A45 (11 ) A44 (10 ) A45 (10 ) 0 (16)

( ) ( ) ( ) ( )
0 0 0 0 A45 A55 A45 A55 0
(11 ) (02 ) (01 ) (01 )
( ) ( ) ( ) ( )
0 0 0 0 A44 A45 A44 A45 0
(10 ) (01 ) (00 ) (00 )
( ) ( ) ( ) ( )
0 0 0 0 A45 A55 A45 A55 0
(10 ) (01 ) (00 ) (00 )
( )
( ) ( )
( )
( )
A13 (10 ) A23 (01 ) A36 (10 ) A36 (01 ) 0 0 0 0 A33 (00 )
Each elastic coecient included in the matrix A( ) can be dened according to the following expressions, for , = 0, 1,
2, ..., N, N + 1, n, m = 1, 2, 3, 4, 5, 6 and p, q = 0, 1, 2

l k+1

( )
(k ) H1 H2
Anm ( pq ) = B nm F F p q d
k=1
H H
1 2
k
k+1
(k ) F H1 H2
l
( )
Anm = B nm F d
( pq )
k=1
H1p H2q
k

l k+1
(k ) F
( )
H1 H2
Anm ( pq ) = B nm F p q d
k=1
H H
1 2
k

l k+1
(k ) F
( )
F H1 H2
Anm ( pq ) = B nm d (17)
k=1
H1p H2q
k
The indices , and , indicate the thickness functions F , F and the correspondent derivatives with respect to ,
(k )
respectively. On the other hand, the quantities B nm are dened as follows
(k ) (k )
B nm = Enm for n, m = 1, 2, 3, 6
(18)
(k )
Bnm = Enm (k ) for n, m = 4, 5
(k )
where Enm denotes the elastic constant of the material. For the sake of generality, this parameter can have two different
(k ) (k )
meanings. In particular, one gets Enm = Q nm when the reduced elastic coecients are needed as in those structural models
(k ) (k )
that neglect the stretching effect. On the other hand, the relation Enm = Cnm means that the non-reduced elastic coecients
are used. In this last circumstance, the strain along the normal direction is included in the structural theory. The subscript
RS is added to specify the use of the reduced elastic coecients. More details about these aspects can be found in the
works (Bacciocchi et al., 2016, Tornabene et al., 2015, Tornabene et al., 2016). With reference to expression (18), the symbol
= 1/ represents the well-known shear correction factor that must be included in those structural models that do not
take into account a parabolic through-the-thickness shear stress prole. On the contrary, its value will be set equal to the
unity. In general, different values of have been proposed in the literature when the shear correction factor is needed
(Bacciocchi et al., 2016, Tornabene & Reddy, 2013), especially for the FSDT. Commonly, a constant value of 5/6 (or in other
words = 1.2) is set. Nevertheless, a different choice can be performed. In fact, the contribution of shear stresses is neglected
if the shear stiffness tends to innity. Thus, for = 10 6 the Reissner-Mindlin theory tends to the Kirchhoff-Love theory,
even though the degrees of freedom are still ve. In any case, the adopted value of will be specied time by time if
required. Finally, it should be mentioned that a numerical technique must be used to compute the integrals in (17). For this
purpose, the Generalized Integral Quadrature (GIQ) method is employed by the authors in the present investigation, due to
its accuracy and reliability. The reader can nd a complete treatise about this technique in the recent paper by Tornabene
et al. (Tornabene et al., 2015). Obviously, several numerical approaches exist to deal with this problem.
At this point, the Hamiltons variational principle can be applied to obtain the governing equation system. The mechanical
behavior of a shell structure subjected to concentrated load is ruled by the following equation, for = 0, 1, 2, ..., N, N + 1
D S( ) + q( ) = 0 (19)
where D is the equilibrium operator dened as follows

T
1 1 A2 1 A1 1
+
A1
1 A1 A2 1 A1 A2 2 R1

1 A2 1 1 A1 1
+
A1 A2 1 A2 2 A1 A2 2 R2

1 A1 1 1 A2
+ 0
A1 A2 2 A1 1 A1 A2 1

1 1 A1 1 A2
+ 0
A2 2 A1 A2 2 A1 A2 1
D = (20)
1 1 1 A2
0
1
+
A1 A2 1
R1 A1
1 A1
1 1
+
0
2 A1 A2 2
R2 A2
1

0 0

1

0 0
0 0 1
On the other hand, q( ) represents the static load vector which collects three load components for each order of
kinematic expansion. It allows the following denition
T
q( ) = q1( )
q2( )

qn( ) (21)
Fig. 2. External loads applied on the top surface of a generic shell element: a) Point load applied in ( 1n , 2m ); b) Line load applied along 2 direction for
1n [10 , 11 ].
In general, the generalized external force components are given by the expressions written below for = 0, 1, 2, ..., N,
N+1

q1( ) = q1( ) F( ) H1( ) H2( ) + q1( ) F( ) H1( ) H2( )
+ + + +

q2( ) = q2( ) F( ) H1( ) H2( ) + q2( ) F( ) H1( ) H2( )
+ + + +

qn( ) = qn( ) F( ) H1( ) H2( ) + qn( ) F( ) H1( ) H2( )
+ + + +
(22)
(+ ) (+ ) (+ ) ( ) ( ) ( )
where q1 , q2 , qn are the load applied on the top surface of the shell. Conversely, q1 , q2 , qn are the external load
applied on the bottom surface. Due to the fact that the external forces are enforced on the external surfaces, the curvature
parameters H1( ) , H2( ) and the shear functions F( ) , for = 0, 1, 2, ..., N, N + 1, must be computed assuming respectively =
h/2. It should be noticed that denitions (22) are general and allow to deal with different kinds of actions. In the present
work, both point loads (Fig. 2a) and line loads (Fig. 2b) are taken into account. Let us consider the case of point forces
rst. In general, an external force can be oriented arbitrarily with respect to the principal direction. The angle parameters
1() , 2() , 3() noticeable in Fig. 2a are introduced for this purpose. In particular, i() represents the angle between the
principal direction i and the force itself. Thus, a generic point load can be inserted in relations (22) following the expression
written below
qi( ) = Qi( ) (1 , 2 )

(23)
( )
where Qi , for i = 1, 2, 3, represents the entity of the applied load along each principal direction. In extended notation, one
gets
Q1( ) = Q () cos 1( )

Q2( ) = Q () cos 2( )

Q3( ) = Q () cos 3( )

(24)
in which Q ( ) is the magnitude of the applied point load measured in unit of force, e.g. Newton [N]. The quantity ( 1 , 2 )
in (23) denotes a two-dimensional function that, generally speaking, can assume two different meanings: the normalized
Gaussian and the Dirac-delta functions (Tornabene et al., 2015). It should be recalled that the Dirac-delta functions is a
generalized function that is zero everywhere in the domain except in the point in which is applied. As a consequence, it
has been employed in different scientic elds, such as engineering, mathematics and physics. Since in their previous work
(Tornabene et al., 2015) the authors have proven that the Dirac-delta function is more reliable than the normalized Gaussian
function, in the present paper only the former is considered. Nevertheless, the reader can refer to the aforementioned paper
for further details about the application of the latter. Let now assume that the point load is applied in ( 1n , 2m ), with
1n [10 , 11 ] and 2m [20 , 21 ], the function ( 1 , 2 ) becomes
(1 , 2 ) = (1 1n ) (2 2m ) (25)
As denition, the relations written below must be followed
+ + 1 1
(1 , 2 )A1 A2 H1() H2() d1 d2 = (1 , 2 )A1 A2 H1() H2() d1 d2 = 1
1 2
(26)
10 20
+ + 1 1
Qi( ) (1 , 2 )A1 A2 H1( ) H2( ) d1 d2 = Qi( ) (1 , 2 )A1 A2 H1( ) H2( ) d1 d2 = Qi( )
1 2

(27)
10 20
The rst relation (26) indicates that the volume covered by ( 1 , 2 ) over the whole domain is equal to the unity. In
this way, it is assured that the load resultant dened by (23) coincides with the concentrated load of value Qi( ) , as specied
by relation (27). Let us consider a line load now. Expression (23) must be rewritten as follows

Pi( ) (1 )

( ) line load along 2
qi = (28)
( )
Pi (2 ) line load along 1
in which Pi( ) is the magnitude of the applied line load measured in force by unit length, e.g. Newton per unit of meters
[N/m]. The meaning of these expressions is obvious. In fact, only the coordinate 2 must be specied to indicate a line load
applied along the principal coordinate 1 and vice versa. Even in this circumstance, both ( 1 ) and ( 2 ) are dened by
means of the Dirac-delta function, which becomes a one-dimensional function
(1 ) = (1 1n )
(29)
(2 ) = (2 2m )
where 1n , 2m are the coordinates in which the line loads at issue are applied. Similarly to what has been shown before
for the point load case, the following properties must be satised if a line load applied along 2 is considered
+ 1
(1 )A2 H2() d2 = (1 )A2 H2() d2 = 1
2
(30)
20
+ 1
Pi( ) (1 )A2 H2( ) d2 = Pi( ) (1 )A2 H2( ) d2 = Pi( )
2

(31)
20
On the other hand, for a line load applied along 1 the relations above become
+ 1
(2 )A1 H1() d1 = (2 )A1 H1() d1 = 1
1
(32)
10
+ 1
Pi( ) (2 )A1 H1( ) d1 = Pi( ) (2 )A1 H1( ) d1 = Pi( )
1

(33)
10
At this point, the fundamental system of equations which governs the static problem in hand can be established. Sum-
marizing the kinematics, the constitutive model and the equilibrium equations in only one relation valid for each order
of kinematic expansion, one gets

N+1
L ( ) u ( ) + q ( ) = 0 (34)
=0
This equation written in matrix form represents the so-called fundamental nucleus of the CUF. It is evident that this
set of three equilibrium equations in terms of generalized displacements must be solved for each order of expansion. If
N represents the maximum order of kinematic expansion, 3 (N + 2) equilibrium equations must be written to solve the
problem at issue. The matrix L( ) = D A( ) D is the fundamental operator. In extended form, it assumes the following
aspect
( ) ( ) ( )
L11 L12 L13
L( ) = L21
( ) ( )
L22 ( )
L23 (35)
( ) ( ) ( )
L31 L32 L33
for , = 0, 1, 2, ..., N, N + 1. For conciseness purposes, the meaning of each differential operator is omitted in the present
paper. Nevertheless, the reader can refer to the work (Tornabene, Viola, & Fantuzzi, 2013) for a complete description of these
terms. To complete the treatise, the boundary conditions must be introduced. In general, an external edge can be completely
clamped (C), simply-supported (S) or free (F). With reference to Fig. 1, if the edge in which the boundary conditions must
be enforced is identied by 1 = 10 , or 1 = 11 , and 2 [20 , 21 ], the conditions in hand assume the following aspect

C u1( ) = u2( ) = u3( ) = 0

S N1( ) = 0, u2( ) = u3( ) = 0
( )
F N1( ) = N12 = T1( ) = 0 (36)
for each order of kinematic expansion. On the other hand, if the edge is dened by 2 = 20 , or 2 = 21 , and 1 [10 , 11 ],
the conditions in hand become

C u1( ) = u2( ) = u3( ) = 0

S u1( ) = 0, N2( ) = 0, u3( ) = 0
( )
F N21 = N2( ) = T2( ) = 0 (37)
for each order of kinematic expansion. In general, if the shell structure under consideration has four distinct edges, the
boundary conditions are specied along each side following the order WSEN, where
West edge(W ) 10 1 11 , 2 = 20
South edge(S ) 1 = 11 , 20 2 21
East edge(E ) 10 1 11 , 2 = 21
North edge(N ) 1 = 10 , 20 2 21 (38)
For instance, the notation CCCF represents a panel in which the West, the South and the East edges are clamped, whereas
only the North one is free.
4. Numerical implementation
In this section, the bases of the numerical techniques used to evaluate both the derivatives that appear in the governing
equations and the integrals needed to compute the elastic stiffness coecients are presented. In particular, the fundamental
aspects of the GDQ and the GIQ methods are shown. In addition, a numerical approach is presented to implement the
concentrated loads.
4.1. Discretization of the fundamental equations
The governing system of Eq. (34) is solved numerically through the GDQ method (Shu & Richards, 1992b,a, Tornabene
et al., 2015). Its application in many structural problems is becoming more and more common due to its accuracy, stability
and reliability features, as proven by the huge number of papers published in the recent years (Bacciocchi et al., 2016,
Brischetto et al., 2016, Fantuzzi, 2014, Fantuzzi, & Tornabene, 2014, Fantuzzi, & Tornabene, 2014, Fantuzzi, & Tornabene,
2016, Fantuzzi et al., 2014, Fantuzzi et al., 2014, Fantuzzi et al., 2015, Fantuzzi et al., 2016, Fantuzzi et al., 2016, Ferreira
et al., 2014, Tornabene, 2016, Tornabene, & Ceruti, 2013, Tornabene, & Reddy, 2013, Tornabene, & Viola, 2009, Tornabene,
& Viola, 2013, Tornabene et al., 2013, Tornabene et al., 2013, Tornabene et al., 2014, Tornabene et al., 2014, Tornabene et
al., 2014, Tornabene et al., 2015, Tornabene et al., 2015, Tornabene et al., 2015, Tornabene et al., 2015, Tornabene et al.,
2015, Tornabene et al., 2015, Tornabene et al., 2016, Tornabene et al., 2016, Tornabene et al., 2016, Tornabene et al., 2016,
Tornabene et al., 2014, Tornabene et al., 2014, Tornabene et al., 2012, Tornabene et al., 2016, Viola et al., 2013, Viola et al.,
2013, Viola et al., 2014, Viola et al., 2016, Viola et al., 2013, Viola et al., 2013, Viola et al., 2013, Viola et al., 2013, Viola et
al., 2015). For the sake of completeness, only its fundamental aspects are now addressed. For a complete treatise about this
technique, the reader can consult the review paper by Tornabene et al. (Tornabene et al., 2015). In general, the GDQ approach
allows to compute numerically the nth derivative at a generic point of a suciently smooth function, as a weighted linear
sum of the function values at some chosen grid points. If a two-dimensional domain is considered, three different kinds of
derivative of a smooth function f( 1 , 2 ) must be evaluated at a generic point ( 1i , 2j ): the nth derivative with respect to
1 coordinate

n f (1 , 2 )
IN

= (n )
1 (ik ) f 1 k , 2 j (39)
1n =
1 1i , 2 = 2 j k=1
the m-th derivative with respect to 2 coordinate

m f (1 , 2 )
IM

= (m()jl ) f (1 i , 2 l ) (40)
2m =
2
1 1i, 2= 2 j l=1
and the mixed partial derivative of order (n + m)

n+m f (1 , 2 )
IN
IM

= (n1 ()ik) (m2 ()jl ) f (1 k , 2 l ) (41)
1n 2m =
1 1i, 2= 2 j k=1 l=1
for i= 1, 2,...,IN and j= 1, 2,...,IM , where IN , IM represent the total number of discrete points along 1 and 2 directions,
respectively. On the other hand, (n )(ik) and (m()jl ) are the weighting coecients for the derivatives. For completeness pur-
1 2
poses, it should be emphasized the fact that the derivative with respect to 1 involves each point of the reference domain
along a line for 2 = 2j constant. Vice versa, each point along a line for 1 = 1i xed is taken into account when the
derivative with respect to 2 is computed. Finally, the mixed derivative considers each point within the reference domain,
since a double summation appears in its denition. It is clear that a discrete grid point distribution must be chosen along
each principal direction according to the following expressions
11 10
1i = (ri r1 ) + 10 (42)
rIN r1
Fig. 3. Convergence features of the central deection for a SSSS square plate as a function of the number of points IN = IM varying the grid distributions, for
two lamination schemes and two different kinds of load: a)-b) point load Q3(+) = 10 kN applied in (0.5 Lx , 0.5 Ly ); c)-d) line load P3(+) = 10 kN/m applied
along 2 direction in 1 = 0.5 Lx ; e)-f) cross load P3(+) = 10 kN/m centered in (0.5 Lx , 0.5 Ly ). The reference solutions are taken from the book by Reddy
Reddy, 2004).
21 20
2 j = r j r1 + 20 (43)
rIM r1
for i = 1, 2, ..., IN and j = 1, 2, ..., IM , respectively. Table 1 shows some grid distributions that can be used for this purpose. A
more complete list of discretizations can be found in the review paper by Tornabene et al. (Tornabene et al., 2015).
The numerical technique just illustrated in brief allows to obtain the solution of the static problem at issue. Once a
discrete grid distribution is applied along the two principal directions and the GDQ method is used to approximate each
derivative, the fundamental system (34) can be written in its discrete form
K = f (44)
where K is the stiffness matrix, is the displacement vector, and f is the vector of external loads. By applying the static
condensation, the size of the problem is reduced since the degrees of freedom of the internal points of the domain (d) are
separated from the ones related to the boundaries (b). As a consequence, Eq. (44) can be rewritten as follows
Kbb b + Kbd d = fb
Fig. 4. Comparison of radial deection for a pinched SS orthotropic cylinder for two lamination schemes and two different lengths: a)-b) Ly = 200 mm;
c)-d) Ly = 400 mm. The reference solutions are taken from the work by Jones (Jones, 1998).
Table 1
Several kinds of grid point distributions. The indices k and N have different
meanings according to the principal direction: if 1 is considered, they respec-
tively become i and IN ; along 2 they are assumed equal to j and IM ; if is
taken into account, these indices have to be set as m and IT .
Chebyshev-Gauss-Lobatto (Che-Gau-Lob)
rk = cos( N1
Nk
) for k = 1, 2, ..., N, r [1, 1]
Legendre-Gauss (Leg-Gau)
rk = roots of (1 r 2 )LN1 (r ) for k = 1, 2, ..., N, r [1, 1]
k1 k1
(1 ) d k1
with Lk (r ) = 2k1 (k1 )! dr k1
( (1 r 2 ) )
Chebyshev-Gauss (Che-Gau)
r 1 = 1, r N = 1, rk = cos( 2(2Nk )1
(N2 ) ) for k = 2, 3, ..., N 1, r [1, 1]
Quadratic (Quad)

rk = 2( N1 ) for k = 1, 2, ...,
k1 2 N+1
2
rk = 2( N1 ) + 4( N1
k1 2 k1
)1 for k = N+1
2
+ 1, ..., N 1, N
Fig. 5. Comparison of the through-the-thickness variation of stress components [Pa] for a simply supported laminated beam subjected to a point load
applied in the beam midpoint. The reference solutions are taken from the book by (Reddy, 2004).
Kdb b + Kdd d = fd (45)

Since no boundary loads are applied, the corresponding term fb is equal to zero. Thus, the rst equation in (45) leads to
the following one
b = K1 K
bb bd d
(46)
If this result is inserted in the second equation in (45), the unknown variable vector d can be obtained as follows
1
d = Kdd Kdb K1 K
bb bd
fd (47)
4.2. Evaluation of the stiffness coecients
As explained briey in the previous sections, a numerical integration must be performed to compute the value of the
( ) ( ) ( ) ( )
elastic constants Anm ( pq) , Anm ( pq ) , Anm ( pq) , Anm ( pq ) dened in (17). For this purpose, the GIQ method is employed. In this
paragraph, only the fundamental aspects are shown. For further details, the reader can refer to the review paper by Torn-
abene et al. (Tornabene et al., 2015). In brief, the GIQ technique allows to compute numerically the integral of a smooth
function f( ) over a limited domain [ l , e ] as a weighted linear sum of the function values in each point of the domain.
Mathematically speaking, one gets
e
IT

f ( )d = wk f (k ) (48)
l k=1

in which wk is the weighting coecient for the integral. They can be evaluated starting from the weighting coecients for
the rst order derivatives, taking into account the same considerations valid for the GDQ method as shown in the work
(Tornabene et al., 2015). In this case, IT is the total number of points of the one-dimensional domain. It is clear that in this
paper the domain in which the integrals must be computed coincides with the shell thickness. In general, the following
expression can be employed to apply an arbitrary grid distribution through the shell thickness
h h
m = ( rm r1 ) (49)
rIT r1 2
Fig. 6. Comparison of the through-the-thickness variation of membrane stress components [Pa] for a SSSS laminated square plate for two lamination
schemes and two different kinds of load: a)-c)-e) point load Q3(+) = 10 kN applied in (0.5 Lx , 0.5 Ly ); b)-d)-f) line load P3(+) = 10 kN/m applied along 2
direction in 1 = 0.5 Lx . The reference solutions are taken from the book by Reddy (Reddy, 2004). The Leg-Gau grid distribution is used to discretize the
reference domain with IN = 51, IM = 51 and the stresses are evaluated at the point P = (0.25691 Lx , 0.25691 Ly ).
assuming m = 1, 2, ..., IT . For this purpose, one of the discretizations listed in Table 1 can be chosen. In the present paper,
the Chebyshev-Gauss-Lobatto (Che-Gau-Lob) grid is used along this direction, assuming IT = 31 in any case.
4.3. Implementation of the concentrated loads
A peculiar procedure has to be developed to compute numerically both Eq. (23) for point forces and (28) for line loads.
From the computational point of view, an integral equation must be solved in the point (or along the line) in which the
forces under consideration are applied. Let us consider rst the case of a point load Qi( ) applied in ( 1n , 2m ). Starting
from the Eftekharis approach (Eftekhari, 2016b,a, 2015) and considering the more general case of a doubly-curved shell,
denition (23) has to be written in discrete form. Thus, relation (23) must be specied in each point of the domain, except
for those nodes that are located along the external edges. In other words, one gets

Qi( )

if f = n, g = m
wf 1 wg 2 A1( f g) H1( ( f)g) A2( f g) H2( ( f)g)

qi(( f )g) =

(50)

0 otherwise
Fig. 7. Comparison of the through-the-thickness variation of membrane stress components [Pa] for a SSSS laminated square plate subjected to a cross
load P3(+) = 10 kN/m centered in (0.5 Lx , 0.5 Ly ) for two lamination schemes: a)-c)-e) (0/90/90/0); b)-d)-f) (0). The reference solutions are obtained by
superimposing the solutions taken from the book by Reddy (Reddy, 2004). The Leg-Gau grid distribution is used to discretize the reference domain with
IN = 51, IM = 51 and the stresses are evaluated at the point P = (0.34399 Lx , 0.34399 Ly ).

for f = 2, 3, ..., IN 1 and g = 2, 3, ..., IM 1. The parameters w f 1 , wg 2 represent the GIQ weighting coecients for the eval-
uation of the integrals along 1 , 2 , respectively. It should be noticed that both the Lam parameters A1(fg) , A2(fg) and the
curvature terms H1( )
( f g)
, H2( )
( f g)
dened in the previous sections must be evaluated in each point of the reference domain,
for = h/2. This approach is evidently valid only if the force application point ( 1n , 2m ) coincides with a node of the
chosen grid. Nevertheless, the work by Wang (Wang, 2015) can be taken into account to extend this procedure even to a
point load applied in a generic point within the reference domain. If ( 1 , 2 ) represent the geometric point in which the
force is applied, the following interpolation procedure can be used to generalize expression (50) as follows
Qi( ) l f 1 ( 1 )lg2 ( 2 )

qi(( f )g) =

(51)
wf 1 wg 2 A1( f g) H1( ( f)g) A2( f g) H2( ( f)g)

for f = 2, 3, ..., IN 1 and g = 2, 3, ..., IM 1. The following expressions can be used to dene the Lagrange interpolation

polynomials l f 1 ( 1 ), lg 2 ( 2 )

IN
1 1k
l f 1 ( 1 ) =
k=1, k = f
1 f 1k

IM
2 2l
lg2 ( 2 ) = (52)
l =1, l =g
2g 2l
Fig. 8. Laminated shell structures and discrete grid representation.
for f = 2, 3, ..., IN 1 and g = 2, 3, ..., IM 1. In other words, when the force is not applied in a specic grid point, its value is
spread in its neighborhood using the Lagrange distribution. It should be noticed that the current approach aims to illustrate
in a clearer way what has been already explained by the authors in their previous work (Tornabene et al., 2015). As far as
the line load implementation is concerned, a similar procedure can be developed. Let us assume now that a line load Pi( )
is applied along 2 in 1n . As shown before, the rst equation of (28) must be specied in each node within the domain,
except for those points that are placed on the external edges. Thus, one gets
Pi( )

if f = n
w f 1 A1( f g) H1( ( f)g)

qi(( f )g) =

(53)

0 otherwise
Fig. 9. Through-the-thickness variation of strain components for a CCCF laminated annular plate with (0/90/0/90/0/90) as lamination scheme, evaluated at
the point P = (0.23581 L, 0.25691 ( 1 0 )) for several higher-order ESL theories. The structure is subjected to a point load Q( + ) = 10 kN applied in (0.5 L,
0.5 ( 1 0 )) on the top surface, oriented with 1(+) = 45 , 2(+) = 90 , 3(+) = 45 . The Leg-Gau grid distribution is used to discretize the reference domain
with IN = 41, IM = 51.
for f = 2, 3, ..., IN 1 and g = 2, 3, ..., IM 1. Assuming that the line load is applied along 2 in a generic point 1 that does
not coincide with a node of the discretization, the interpolation procedure shown above can be adapted also to the present
case
Pi( ) lg2 ( 2 )

qi(( f )g) =

(54)
wf 1 A1( f g) H1( ( f)g)

for f = 2, 3, ..., IN 1 and g = 2, 3, ..., IM 1. The Lagrange interpolation polynomials l f 1 ( 1 ) is dened according to expression
(52). Analogously, it is possible to deal with a line load applied along 1 . In particular, relation (53) becomes
Pi( )

if g = m
wg 2 A2( f g) H2( ( f)g)

qi(( f )g) =

(55)

0 otherwise
for f = 2, 3, ..., IN 1 and g = 2, 3, ..., IM 1. On the other hand, expression (54) is changed as follows
Pi( ) l f 1 ( 1 )

qi(( f )g) =

(56)
wg 2 A2( f g) H2( ( f)g)

Fig. 10. Through-the-thickness variation of stress components [Pa] for a CCCF laminated annular plate with (0/90/0/90/0/90) as lamination scheme, evalu-
ated at the point P = (0.23581 L, 0.25691 ( 1 0 )) for several higher-order ESL theories. The structure is subjected to a point load Q( + ) = 10 kN applied in
(0.5 L, 0.5 ( 1 0 )) on the top surface, oriented with 1(+) = 45 , 2(+) = 90 , 3(+) = 45 . The Leg-Gau grid distribution is used to discretize the reference
domain with IN = 41, IM = 51.
for f = 2, 3, ..., IN 1 and g = 2, 3, ..., IM 1. A third load case can be obtained superimposing the effect related to the appli-
cation of two line loads applied orthogonally between them. Thus, the cross load case centered in the intersection of the
two contributions in hand can be achieved. It should be recalled that the superimposition at issue is valid since the static
analysis is linear, as well as the mechanical behavior of the materials of the structures in hand.
5. Strain and stress recovery procedure
In the previous section, the fact that a shell structure is a three-dimensional structure has been already emphasized. As
a consequence, its mechanical behavior can be described through a set of three 3D equilibrium equations written in prin-
cipal curvilinear coordinates. These equations represent the starting point of the a posteriori recovery procedure presented
below. The main aim of this approach is to evaluate the through-the-thickness variation of stress and strain components.
For completeness purposes, the 3D equilibrium equations for a generic shell structure can be written as follows
1 n 2 1
1 1 2 1 A2 1 12
+ 1 n + = +
R1 + R2 + A 1 (1 + /R 1 ) 1 A 1 A 2 (1 + /R 2 ) 1 A 2 (1 + /R 2 ) 2
212 A1
(57)
A 1 A 2 (1 + /R 1 ) 2
Fig. 11. Through-the-thickness variation of displacement components [m] for a CCCF laminated annular plate with (0/90/0/90/0/90) as lamination scheme,
evaluated at the point P = (0.23581 L, 0.25691 ( 1 0 )) for several higher-order ESL theories. The structure is subjected to a point load Q( + ) = 10 kN
applied in (0.5 L, 0.5 ( 1 0 )) on the top surface, oriented with 1(+) = 45 , 2(+) = 90 , 3(+) = 45 . The Leg-Gau grid distribution is used to discretize the
reference domain with IN = 41, IM = 51.
Fig. 12. Through-the-thickness variation of strain components for a FCFC laminated cylindrical surface with parabolic prole made of four plies arranged
as (0/30/60/90), evaluated at the point P = (0.25691 ( 1 0 ), 0.25691 L) for several higher-order ESL theories. The structure is subjected to a point load
Q( + ) = 10 kN applied in (0.5 ( 1 0 ), 0.5 L) on the top surface, oriented with 1(+) = 90 , 2(+) = 0 , 3(+) = 90 . The Leg-Gau grid distribution is used to
discretize the reference domain with IN = 51, IM = 51.
2 n 1 2
1 2 1 2 A1 1 12
+ 2 n + = +
R1 + R2 + A 2 (1 + /R 2 ) 2 A 1 A 2 (1 + /R 1 ) 2 A 1 (1 + /R 1 ) 1
212 A2
(58)
A 1 A 2 (1 + /R 2 ) 1
n 1 1
1 1 n 1 n A2 1 2 n
+ n + =
R1 + R2 + A 1 (1 + /R 1 ) 1 A 1 A 2 (1 + /R 2 ) 1 A 2 (1 + /R 2 ) 2
2 n A1 1 2
+ + (59)
A 1 A 2 (1 + /R 1 ) 2 R1 + R2 +
The GDQ method allows to transform these equations in their discrete form. Thus, the discrete system of these relations
can be written in each point ( 1i , 2j ) of the shell middle surface, for every discrete point along the shell thickness for
m = 1, 2, ..., IT

IT
(1 ) 2 1
mk 1n(i jk) + 1n(i jm) +
R 1(i j ) + m R 2(i j ) + m
k=1
Fig. 13. Through-the-thickness variation of stress components [Pa] for a FCFC laminated cylindrical surface with parabolic prole made of four plies ar-
ranged as (0/30/60/90), evaluated at the point P = (0.25691 ( 1 0 ), 0.25691 L) for several higher-order ESL theories. The structure is subjected to a point
load Q( + ) = 10 kN applied in (0.5 ( 1 0 ), 0.5 L) on the top surface, oriented with 1(+) = 90 , 2(+) = 0 , 3(+) = 90 . The Leg-Gau grid distribution is used
to discretize the reference domain with IN = 51, IM = 51.

1 1 2(i jm) 1(i jm) A 2
= +
A 1(i j ) 1 + m /R 1(i j ) 1 (i jm) A 1(i j ) A 2(i j ) 1 + m /R 2(i j ) 1 (i j )

1 12 212(i jm) A 1
(60)

IT
(1 ) 1 2
mk 2n(i jk) + 2n(i jm) +
R 1(i j ) + m R 2(i j ) + m
k=1

1 2 1(i jm) 2(i jm) A 1
= +

1 212(i jm) A 2
12 (61)
Fig. 14. Through-the-thickness variation of displacement components [m] for a FCFC laminated cylindrical surface with parabolic prole made of four plies
arranged as (0/30/60/90), evaluated at the point P = (0.25691 ( 1 0 ), 0.25691 L) for several higher-order ESL theories. The structure is subjected to a
point load Q( + ) = 10 kN applied in (0.5 ( 1 0 ), 0.5 L) on the top surface, oriented with 1(+) = 90 , 2(+) = 0 , 3(+) = 90 . The Leg-Gau grid distribution
is used to discretize the reference domain with IN = 51, IM = 51.
Fig. 15. Through-the-thickness variation of strain components for a CFCF laminated doubly-curved panel of translation with parabolic prole made of eight
plies arranged as (30/60/ 45/0/0/45/ 60/ 30), evaluated at the point P = (0.74309 (11 10 ), 0.74309 (21 20 )) for several higher-order ESL theories.
The structure is subjected to a point load Q( + ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the top surface, oriented with 1(+) = 60 , 2(+) = 60 ,
3(+) = 45 . The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.

IT
(1 ) 1 1
mk n(i jk) + n(i jm) +
R 1(i j ) + m R 2(i j ) + m
k=1

1(i jm) 2(i jm) 1 1n 1n(i jm) A 2
= +
R 1(i j ) + m R 2(i j ) + m A 1(i j ) 1 + m /R 1(i j ) 1 (i jm) A 1(i j ) A 2(i j ) 1 + m /R 2(i j ) 1 (i j )

1 2n(i jm) A 1
2n (62)
At this point, relations (60)-(61) can be written as a function of the shear stresses 1n and 2n . Their effective through-
the-thickness proles can be carried out enforcing the proper boundary conditions on the shell bottom surface
1n(i j1) = q1((i)j )

2n(i j1) = q2((i)j ) (63)
Fig. 16. Through-the-thickness variation of stress components [Pa] for a CFCF laminated doubly-curved panel of translation with parabolic prole made
of eight plies arranged as (30/60/ 45/0/0/45/ 60/ 30), evaluated at the point P = (0.74309 (11 10 ), 0.74309 (21 20 )) for several higher-order ESL
theories. The structure is subjected to a point load Q( + ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the top surface, oriented with 1(+) = 60 ,
2(+) = 60 , 3(+) = 45 . The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
and on the top one
1n(i jIT ) = q1(+(i)j )

2n(i jIT ) = q2(+(i)j ) (64)
As a result, one obtains the following relations valid in each grid point located along the shell thickness (for m = 2, 3, ...,
IT )

q1( (i)j ) 1n(i jIT )
+
h
1n(i jm) = 1n(i jm) + m + (65)
h 2

q2( (i)j ) 2n(i jIT )
+
h
2n(i jm) = 2n(i jm) + m + (66)
h 2
in which 1n and 2n are the shear stresses that must be modied imposing the boundary conditions on the top surface of
the shell. Eq. (62) can be now taken into account to compute the normal stress n , enforcing the corresponding boundary
Fig. 17. Through-the-thickness variation of displacement components [m] for a CFCF laminated doubly-curved panel of translation with parabolic prole
made of eight plies arranged as (30/60/ 45/0/0/45/ 60/ 30), evaluated at the point P = (0.74309 (11 10 ), 0.74309 (21 20 )) for several higher-order
ESL theories. The structure is subjected to a point load Q( + ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the top surface, oriented with 1(+) = 60 ,
Fig. 18. Through-the-thickness variation of strain components for a CCCC doubly-curved sandwich panel of translation dened by a parabolic prole with
lamination scheme (Aluminum/Foam/Aluminum), evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) for several higher-order ESL theories.
The structure is subjected to a point load Q( ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface, oriented with 1() = 90 , 2() =
90 , 3() = 0 . The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
conditions on the external surfaces
n(i j1) = qn((i)j )

n(i jIT ) = qn(+(i)j ) (67)
Consequently, the effective prole of the normal stress along the shell thickness is established

qn( (i)j ) n(i jIT )
+
h
n(i jm) = n(i jm) + m + (68)
h 2
for m = 2, 3, ..., IT , where n is the normal stress that must be adjusted enforcing the boundary condition on the top surface
of the structure. The three-dimensional constitutive relations must be now recalled to evaluate the shear strains 1n , 2n
Fig. 19. Through-the-thickness variation of stress components [Pa] for a CCCC doubly-curved sandwich panel of translation dened by a parabolic prole
with lamination scheme (Aluminum/Foam/Aluminum), evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) for several higher-order ESL the-
ories. The structure is subjected to a point load Q( ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface, oriented with 1() = 90 ,
2() = 90 , 3() = 0 . The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
and the normal strain n . If an orthotropic medium is considered, they assume the following aspect

1 C11 C12 C16 0 0 C13 1

C23 2
2 C12 C22 C26 0 0

12 C16 C26 C66 0 0 C36 12

= (69)
13 0 0 0 C44 C45 0 13

23 0 0 0 C45 C55 0 23
3 C13 C23 C36 0 0 C33 3
Since the actual value of both the shear stresses 1n , 2n and the normal stress n is computed, the strain components
at issue can be evaluated too. Therefore, the through-the-thickness variations of these quantities can be dened according
to the following relations
(m ) (m )
C55 1n(i jm) C45 2n(i jm)
1n (i jm) = 2 (70)
(m ) (m ) (m )
C55 C44 C45
Fig. 20. Through-the-thickness variation of displacement components [m] for a CCCC doubly-curved sandwich panel of translation dened by a parabolic
prole with lamination scheme (Aluminum/Foam/Aluminum), evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) for several higher-order
ESL theories. The structure is subjected to a point load Q( ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface, oriented with 1() =
90 , 2() = 90 , 3() = 0 . The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
Fig. 21. Through-the-thickness variation of strain components for a CFCC Funnel surface made of six layers oriented as (0/90/0/0/90/0), evaluated at the
point P = (0.25691 (11 10 ), 0.25691 (21 20 )), considering only the ED3 theory. The structure is subjected to a point load Q( + ) = 10 kN applied in
(0.5 (11 10 ), 0.5 (21 20 )) on the top surface, varying the angle 1(+) , for 2(+) = 90 and 3(+) = 90 1(+) . The Leg-Gau grid distribution is used to
(m ) (m )
C44 2n(i jm) C45 1n(i jm)
2n(i jm) = 2 (71)
(m ) (m ) (m )
C55 C44 C45
(m ) (m ) (m )
n(i jm) C13 1(i jm) C23 2(i jm) C36 12(i jm)
n(i jm) = (m )
(72)
C33
Finally, the through-the-thickness proles of membrane stresses 1(ijm) , 2(ijm) , 12(ijm) can be corrected employing the
actual value of the normal strain n just evaluated, by means of the constitutive law (69).
6. Results and discussion
The static response of several structures subjected to point and line loads is investigated in the present section. The
solutions are achieved numerically by means of the GDQ method just introduced and the recovery procedure is applied to
obtain the through-the-thickness proles of stress and strain components. The validity and the accuracy of this approach,
Fig. 22. Through-the-thickness variation of stress components [Pa] for a CFCC Funnel surface made of six layers oriented as (0/90/0/0/90/0), evaluated at
the point P = (0.25691 (11 10 ), 0.25691 (21 20 )), considering only the ED3 theory. The structure is subjected to a point load Q( + ) = 10 kN applied
in (0.5 (11 10 ), 0.5 (21 20 )) on the top surface, varying the angle 1(+) , for 2(+) = 90 and 3(+) = 90 1(+) . The Leg-Gau grid distribution is used to
Table 2
Mechanical properties of the materials.
Aluminum
E = 70 GPa, = 0.3, = 2707 kg/m3
Ceramic Foam
E = 0.232 GPa, = 0.2, = 320 kg/m3
Glass-Epoxy
E1 = 53.78 GPa, E2 = E3 = 17.93 GPa, G12 = G13 = 8.96 GPa, G23 = 3.45 GPa, 12 = 13 = 0.25, 23 = 0.34, = 1900 kg/m3
Graphite-Epoxy
E1 = 137.9 GPa, E2 = E3 = 8.96 GPa, G12 = G13 = 7.1 GPa, G23 = 6.21 GPa, 12 = 13 = 0.3, 23 = 0.49, = 1450 kg/m3
which has been implemented in a MATLAB code (Viola, Tornabene, Fantuzzi, & Bacciocchi, 2016), are proven by the compar-
ison with the results available in the literature. Thus, the rst part of this section is devoted to the validation of both the
structural model and the numerical technique, whereas new results are presented subsequently for several kinds of shell
structures loaded by concentrated forces. For conciseness purposes, all the mechanical properties of the materials employed
in the following applications are listed in Table 2.
Fig. 23. Through-the-thickness variation of displacement components [m] for a CFCC Funnel surface made of six layers oriented as (0/90/0/0/90/0), eval-
uated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )), considering only the ED3 theory. The structure is subjected to a point load Q( + ) = 10 kN
applied in (0.5 (11 10 ), 0.5 (21 20 )) on the top surface, varying the angle 1(+) , for 2(+) = 90 and 3(+) = 90 1(+) . The Leg-Gau grid distribution is
used to discretize the reference domain with IN = 51, IM = 51.
Fig. 24. Through-the-thickness variation of strain components for a CCCC helicoidal surface made of three layers oriented as (45/ (2) / 45), evaluated at
the point P = (0.25691 (11 10 ), 0.25691 (21 20 )), considering only the ED3 and EDZ3 theories and varying the orientation of the 2nd ply (2) . The
structure is subjected to a point load Q( ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface, oriented with 1() = 90 , 2() = 90 ,
3() = 0 . The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
6.1. Comparison with the literature
As a preliminary test, a convergence analysis is carried out to establish the most accurate grid point distribution among
the ones listed in Table 1. For this purpose, a SSSS square plate (Lx = Ly = 1 m) is taken into account and its central deection
is taken as control parameter. Since the reference solutions are valid for the Kirchhoff-Love theory, the present solution
is obtained by using the corresponding structural model according to what has been explained in the previous sections.
=106
In other words, the FSDTRS theory is used. For each load case, two different lamination schemes are contemplated.
In particular, a single layer made of Graphite-Epoxy oriented of 0 and four plies made all of the same material arranged
as (0/90/90/0) are considered. In each circumstance, the total thickness of the structure is h = 0.1 m. The number of grid
point along each principal direction for every grid is increased from IN = IM = 11 to IN = IM = 51, as it can be noticed in Table
3 for a point load Q3(+ ) = 10 kN applied in (0.5 Lx , 0.5 Ly ), in Table 4 for a line load P3(+ ) = 10 kN/m applied along 2
direction in 1 = 0.5 Lx , and in Table 5 for a cross load P3(+ ) = 10 kN/m centered in (0.5 Lx , 0.5 Ly ). It should be pointed out
that the GDQ solutions are in excellent agreement with the semi-analytical ones that can be found in the book by Reddy
(Reddy, 2004). As far as the third case is concerned, the reference solution is obtained by superimposing the semi-analytical
results obtained considering two line loads applied along 1 and 2 , respectively. In any case, the numerical values tend
to the reference displacements increasing the number of grid points. For the sake of completeness, the results of these
convergence tests are represented also in their graphic form in Fig. 3. It is evident that the Leg-Gau discretization is stable
Fig. 25. Through-the-thickness variation of stress components [Pa] for a CCCC helicoidal surface made of three layers oriented as (45/ (2) / 45), evaluated
at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )), considering only the ED3 and EDZ3 theories and varying the orientation of the 2nd ply (2) . The
structure is subjected to a point load Q( ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface, oriented with 1(+) = 90 , 2(+) = 90 ,
3(+) = 0 . The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
and accurate when employed to model these concentrated loads, if compared to the other grid distributions. In particular, it
can be observed that the Che-Gau-Lob and the Che-Gau grids show an oscillating behavior increasing the number of points.
For these reasons, the Leg-Gau distribution is used in all the following applications, even though the Che-Gau grid could
reach a greater accuracy, but characterized by oscillations. Finally, it should be noticed that the Quad grid is stable, but the
level of accuracy that can be reached is lower.
The next analysis in terms of displacements involves the static response of an orthotropic pinched circular cylinder with
simply-supported ends, which is known to be a challenging test of shell behavior. This example is taken from the work
by Jones (Jones, 1998), in which two geometric congurations and two material orientations are considered. In particular,
the length of the cylinder is taken equal to Ly = 200 mm rst, and then equal to Ly = 400 mm. It should be specied that
the length is set as twice the diameter and four times the diameter, respectively. A single orthotropic layer of thickness
h = 1 mm is taken into account in both cases (E1 = 100 GPa, E2 = E3 = 10 GPa, G12 = G13 = G23 = 15 GPa). If it is oriented of 90,
the Poissons ratios are assumed to be 12 = 13 = 23 = 0.3, whereas if the orientation is 0, the values of these Poissons
ratios must be divided by ten, according to what has been illustrated in the reference paper (Jones, 1998). It is important to
underline the fact that in the present approach the rst principal coordinate denes the circumferential direction, whereas
the second one is set along the shell meridian. These considerations are useful to describe univocally the material orien-
tation. Due to the symmetry features of the cylinder, only half structure is modeled. The pinching load (Q3(+ ) = 200 N) is
applied in the median circumference. For the shorter one, the Leg-Gau distribution is employed with IN = IM = 51, whereas
Fig. 26. Through-the-thickness variation of displacement components [m] for a CCCC helicoidal surface made of three layers oriented as (45/ (2) / 45),
evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )), considering only the ED3 and EDZ3 theories and varying the orientation of the 2nd ply
(2) . The structure is subjected to a point load Q( ) = 10 kN applied in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface, oriented with 1(+) = 90 ,
Fig. 27. Through-the-thickness variation of strain components for a FCFC laminated cylindrical surface with parabolic prole made of four plies arranged
as (0/30/60/90), evaluated at the point P = (0.25691 ( 1 0 ), 0.25691 L) for several higher-order ESL theories. The structure is subjected to a line load
P3(+) = 10 kN/m applied along 2 y direction in = 0.5 ( 1 0 ) on the top surface. The Leg-Gau grid distribution is used to discretize the reference
domain with IN = 51, IM = 51.
for the longer one the same grid is applied with IN = 41 and IM = 61. The well-known Reissner-Mindlin theory (FSDT) is con-
sidered in the analyses, with = 5/6 as shear correction factor. The deections along the cylinder are depicted in Fig. 4. In
each graph, these displacements are measured directly under the load and perpendicularly to the load itself. It should be
noticed that the present approach is perfectly in agreement in any case with the exact solution. Thus, the accuracy of the
present approach in terms of displacements is guaranteed.
At this point, the comparison is performed in terms of through-the-thickness stress proles, in order to prove the validity
also of the stress recovery procedure when point and line loads are applied. First, a laminated beam subjected to a concen-
trated force (Q3(+ ) = 1 N) with simply supported ends is considered. The following mechanical properties are employed
E1
= 25, G12 = G13 = 0.5E2 , G23 = 0.2E2 , 12 = 13 = 23 = 0.25 (73)
E2
with E2 = 1 Pa. The beam has length a = 1 m and thickness h = 0.1 m, whereas the cross section width is b = 0.2 m. Two differ-
ent stacking sequences are considered: (0/45/ 45/90)s and (90/45/ 45/0)s , in which the subscript s stands for symmetric
laminate. Thus, the beam is made of eight layers of equal thickness. In addition, the case of single orthotropic ply oriented
of 0 is also considered to emphasize the effect of the lamination scheme on the stress proles. For the sake of complete-
ness, it should be specied that this example is taken from the book by Reddy (Reddy, 2004). Fig. 5 shows the in-plane and
transverse shear stress proles, denoted respectively as 1 and 1n . The former is evaluated in the midpoint of the struc-
ture, whereas the latter is computed at the beam end. Good agreement can be noticed in any circumstance. It is important
Fig. 28. Through-the-thickness variation of stress components [Pa] for a FCFC laminated cylindrical surface with parabolic prole made of four plies
line load P3(+) = 10 kN/m applied along 2 y direction in = 0.5 ( 1 0 ) on the top surface. The Leg-Gau grid distribution is used to discretize the
to specify that the beam is modeled as a plate to obtain the GDQ solution. Thus, a two-dimensional grid of discrete points
is applied. In particular, the Leg-Gau distribution is applied along the two principal directions with IN = 31 and IM = 51.
The through-the-thickness variation of membrane stress components are now computed for a SSSS square plate
(Lx = Ly = 1 m) of thickness h = 0.1 m. The laminated composite plate is subjected to three different load cases: a point load
Q3(+ ) = 10 kN applied in (0.5 Lx , 0.5 Ly ), a line load P3(+ ) = 10 kN/m applied along 2 direction in 1 = 0.5 Lx , and a cross
load P3(+ ) = 10 kN/m centered in (0.5 Lx , 0.5 Ly ). The reference solution is taken once again from the book by Reddy (Reddy,
2004), in which the Navier solution is employed to solve the static problems under consideration. As stated before, the ref-
erence solution for the third case is achieved through the superimposition of the two corresponding line load cases. The
GDQ solution, instead, is obtained considering the Leg-Gau discretization with IN = IM = 51. Even in this case, the stress pro-
les obtained for the symmetric lamination scheme (0/90/90/0) is made parallel with the ones obtained for a single layer
oriented of 0. Each ply is made of Graphite-Epoxy. As explained before, the semi-analytical solution is achieved for the
Kirchhoff-Love theory. Thus, the shear correction factor is set equal to = 106 . In other words, one gets = 1/ = 106
and the FSDT, consequently, tends to the Kirchhoff-Love theory since the contributions of the transverse shear stresses are
neglected. It is evident form Fig. 6 that a perfect overlap exists between the two approaches, the Reddys semi-analytical
one and the numerical method herein presented. The through-the-thickness variation of membrane stresses are computed
at P = (0.25691 Lx , 0.25691 Ly ). At this point, the accuracy in terms of stress proles is proven too and new results concern-
ing several shell structures can be introduced. The through-the-thickness proles of membrane stresses for the cross load
case are shown in Fig. 7 and are evaluated at the point P = (0.34399 Lx , 0.34399 Ly ). A perfect matching between the two
solutions can be noticed for both the considered lamination schemes.
Fig. 29. Through-the-thickness variation of displacement components [m] for a FCFC laminated cylindrical surface with parabolic prole made of four plies
line load P3(+) = 10 kN/m applied along 2 y direction in = 0.5 ( 1 0 ) on the top surface. The Leg-Gau grid distribution is used to discretize the
Table 3
Central deection [m] of a SSSS square plate subjected to a point load Q3(+) = 10 kN applied in (0.5 Lx , 0.5 Ly ) for
several grid point distributions, varying the number of discrete points IN = IM . Two different lamination schemes
are taken into account.
Lamination scheme: (0)
IN = 11 IN = 21 IN = 31 IN = 41 IN = 51 Reference (Reddy, 2004)
Che-Gau-Lob
-3.26598e-05 -3.40204e-05 -3.39808e-05 -3.41070e-05 -3.40788e-05
Leg-Gau
-3.29947e-05 -3.38894e-05 -3.40296e-05 -3.40760e-05 -3.40969e-05 -3.41200e-05
Che-Gau
-3.28714e-05 -3.38689e-05 -3.40275e-05 -3.40738e-05 -3.40965e-05
Quad
-3.19758e-05 -3.36337e-05 -3.39189e-05 -3.40182e-05 -3.40619e-05
Lamination scheme: (0/90/90/0)
Che-Gau-Lob
-3.0 0 042e-05 -3.10756e-05 -3.09676e-05 -3.10791e-05 -3.10393e-05
Leg-Gau
-3.04382e-05 -3.09425e-05 -3.10211e-05 -3.10471e-05 -3.10588e-05 -3.10717e-05
Che-Gau
-3.03956e-05 -3.09278e-05 -3.10208e-05 -3.10455e-05 -3.10588e-05
Quad
-2.97296e-05 -3.07923e-05 -3.09557e-05 -3.10135e-05 -3.10385e-05
Table 4
Central deection [m] of a SSSS square plate subjected to a line load P3(+) = 10 kN/m applied along 2 direction
in 1 = 0.5 Lx for several grid point distributions, varying the number of discrete points IN = IM . Two different
lamination schemes are taken into account.
Che-Gau-Lob
-1.65428e-05 -1.67697e-05 -1.67117e-05 -1.67401e-05 -1.67237e-05
Leg-Gau
-1.67097e-05 -1.67284e-05 -1.67297e-05 -1.67300e-05 -1.67301e-05 -1.67302e-05
Che-Gau
-1.67235e-05 -1.67260e-05 -1.67303e-05 -1.67298e-05 -1.67303e-05
Quad
-1.66006e-05 -1.67215e-05 -1.67269e-05 -1.67290e-05 -1.67296e-05
Che-Gau-Lob
-1.69940e-05 -1.72263e-05 -1.71672e-05 -1.71963e-05 -1.71795e-05
Leg-Gau
-1.71642e-05 -1.71841e-05 -1.71856e-05 -1.71859e-05 -1.71861e-05 -1.71861e-05
Che-Gau
-1.71781e-05 -1.71817e-05 -1.71862e-05 -1.71857e-05 -1.71862e-05
Quad
-1.70498e-05 -1.71764e-05 -1.71825e-05 -1.71848e-05 -1.71855e-05
6.2. Stress and strain recovery procedure for shell structures
In this section, the stress and strain recovery procedure illustrated above is employed to characterize the static behavior
of several laminated shell structures subjected to point and line loads. The shells under consideration are shown in Fig.
8, together with the global and local reference systems, the discrete grid representation and the edge identication. On
the other hand, the position vectors that describe the reference surfaces of the structures in hand, as well as their main
geometric parameters, are listed in Table 6. The principles of differential geometry presented in the previous sections can
be used to evaluate all the geometric quantities needed to write the fundamental governing equations. The main aim of
this section is to compare the strain, stress and displacement component variations along the shell thickness varying the
structural theories. In particular, several higher-order ESL models are taken into account, by varying the order of kinematic
expansions. In addition, the Murakamis function is also included to show the inuence of the zig-zag effect on the static
response.
Table 5
Central deection [m] of a SSSS square plate subjected to a cross load P3(+) = 10 kN/m centered in (0.5 Lx , 0.5 Ly )
for several grid point distributions, varying the number of discrete points IN = IM . Two different lamination schemes
are taken into account.
Che-Gau-Lob
-3.66434e-05 -3.72139e-05 -3.71358e-05 -3.71873e-05 -3.71612e-05
Leg-Gau
-3.68598e-05 -3.71437e-05 -3.71655e-05 -3.71703e-05 -3.71720e-05 -3.71733e-05
Che-Gau
-3.68142e-05 -3.71375e-05 -3.71660e-05 -3.71698e-05 -3.71721e-05
Quad
-3.63446e-05 -3.70621e-05 -3.71389e-05 -3.71595e-05 -3.71666e-05
Che-Gau-Lob
-3.53856e-05 -3.58763e-05 -3.57720e-05 -3.58275e-05 -3.57966e-05
Leg-Gau
-3.56806e-05 -3.57974e-05 -3.58062e-05 -3.58082e-05 -3.58088e-05 -3.58093e-05
Che-Gau
-3.56829e-05 -3.57922e-05 -3.58072e-05 -3.58076e-05 -3.58091e-05
Quad
-3.53415e-05 -3.57593e-05 -3.57931e-05 -3.58031e-05 -3.58063e-05
Table 6
Position vectors r( 1 , 2 ) and geometric characterization of the shell structures depicted in Fig. 8.
Annular plate
r(x, ) = (R + x ) cos e1 (R + x) sin e2
x [0, L] = [0, 5m], [0 , 1 ] = [ /4, /4], R = 2m, h = 0.12m
Cylindrical surface with parabolic prole (singly-curved panel of translation)

e1 y e2 + ktan4 e3 , with k = a d
2
r( , y ) = k tan
2 2
2 b
[0 , 1 ] = [1.10715, 1.10715], y [0, L] = [0, 2m], a = 2 m, b = 2 m, c = 2 m, d = 0 m, h = 0.1m
Doubly-curved panel of translation (a parabola slides over a parabola)
r(1 , 2 ) = (R01 (1 ) x3 2 (2 ) sin 1 ) e1 + R02 (2 ) e2 + (x3 1 (1 ) + x3 2 (2 ) cos 1 ) e3
2
ki tan i (R 0i (i ))
R0i (i ) = 2
, x3 i (i ) = k i
, k i = a2 d 2
b
, a = 3 m, b = 1 m, c = 3 m, d = 1 m, for i = 1, 2
1 [ = [0.64350, 0.64350], 2 [ 21 ] = [0.64350, 0.64350], h = 0.2 m
1,
0 1
1] 2,
0
Funnel surface (doubly-curved panel of revolution)

r(1 , 2 ) = 1 cos 2 e1 1 sin 2 e2 + log 1 e3
1 [10 , 11 ] = [0.3, 1], 2 [20 , 21 ] = [ /3, /3], h = 0.06 m
Helicoid (doubly-curved panel)
r(1 , 2 ) = cos(1 + 2 ) sinh(1 2 )e1 sin(1 + 2 ) sinh(1 2 )e2 + (1 2 )e3
1 [10 , 11 ] = [0, /3], 2 [20 , 21 ] = [0, /3], h = 0.07 m
The rst example is related to a CCCF laminated annular plate made of six laminae with (0/90/0/90/0/90) as stacking
sequence depicted in Fig. 8a. Each layer is made of Graphite-Epoxy and has the same thickness (hk = 0.02 m). The structure
is loaded on the top surface by a point force Q( + ) = 10 kN applied in (0.5 L, 0.5 ( 1 0 )), oriented with 1(+ ) = 45 , 2(+ ) =
90 and 3(+ ) = 45 . According to the graphical representation of Fig. 2, it is clear that the force in hand does not give any
contribution along 2 . The reference domain is discretize by means of the Leg-Gau distribution, with IN = 41 and IM = 51.
The strain, stress and displacement components are depicted respectively in Figs. 911. Except for the second order theories,
no signicant differences can be noticed among the various theories. In addition, the Murakamis function does not implicate
any benet as expected, since no soft-core effect is present. On the other hand, relevant discrepancies can be observed in the
normal stress n prole. Nevertheless, it must be pointed out that its value is many orders of magnitude lower compared
to the other components. Moreover, since the proles in hand are evaluated away from the application point of the force,
the transverse shear stresses and the normal stress must be equal to zero due to the boundary conditions. This aspect is
certainly evident in Fig. 10. The same consideration is valid also in each following application related to other structures
subjected to different load cases.
A FCFC laminated cylindrical surface with parabolic prole made of four plies of Glass-Epoxy oriented according to the
following scheme (0/30/60/90) is investigated in the current analysis (Fig. 8b). The structure is subjected to a point load
Q( + ) = 10 kN applied on the top surface in (0.5 ( 1 0 ), 0.5 L), oriented with 1(+ ) = 90 , 2(+ ) = 0 and 3(+ ) = 90 . Thus,
this concentrated force gives only a tangential contribution along 2 y, whereas the normal component is null. The Leg-
Fig. 30. Through-the-thickness variation of strain components for a CCCC laminated doubly-curved panel of translation with parabolic prole made of eight
plies with (30/60/ 45/0/0/45/ 60/ 30) as lamination scheme, evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) for several higher-order
ESL theories. The structure is subjected to a line load P3(+) = 10 kN/m applied along 1 direction in 2 = 0.5 (21 20 ) on the top surface. The Leg-Gau
grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
Gau distribution is employed to discretize the shell middle surface with IN = IM = 51. The through-the-thickness variations of
strain, stress and displacement components computed at the point P = (0.25691 ( 1 0 ), 0.25691 L) are depicted respec-
tively in Figs. 1214. In this circumstance, the various curves related to different ESL theories detach from themselves in
some quantities, especially as far as the stress components are concerned. All the previous considerations are still valid also
in this example. It is clear that the stress discontinuities are caused by the different orientations of the orthotropic layers.
A CFCF doubly-curved laminated panel of translation is the shell structure considered in the next numerical application
(Fig. 8c). Its geometry is obtained by sliding a parabola over another identical parabola, whose geometric features are listed
in Table 6. The structure consists in eight plies of equal thickness (hk = 0.025 m) made of Graphite-Epoxy and the lamina-
tion scheme is given by (30/60/ 45/0/0/45/ 60/ 30). The doubly-curved shell, discretized through the Leg-Gau grid with
IN = IM = 51, is subjected to a point load Q( + ) = 10 kN applied on the top surface in (0.5 (11 10 ), 0.5 (21 20 )), oriented
with 1(+ ) = 60 , 2(+ ) = 60 , 3(+ ) = 45 . The through-the-thickness variation of strain, stress and displacement components
are evaluated at the point P = (0.74309 (11 10 ), 0.74309 (21 20 )), taking into account several higher-order ESL theo-
ries, and are depicted respectively in Figs. 1517. In general, no signicant differences can be observed among the various
structural models. Even in this case, the Murakamis function is not required. In addition, the choice of the shear correction
factor does not affect the solution when the FSDTZ and the EDZ2 are considered.
The same position vector and geometric features can be used to dene also the next structure. With reference to Fig.
8d, it is easy to notice that the doubly-curved panel of translation under consideration can be obtained from the pre-
vious one just inverting the concavity of the two parabolas. The structure is made of three isotropic layers stacked as
Fig. 31. Through-the-thickness variation of stress components [Pa] for a CCCC laminated doubly-curved panel of translation with parabolic prole made
of eight plies with (30/60/ 45/0/0/45/ 60/ 30) as lamination scheme, evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) for several
higher-order ESL theories. The structure is subjected to a line load P3(+) = 10 kN/m applied along 1 direction in 2 = 0.5 (21 20 ) on the top surface.
The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
(Aluminum/Foam/Aluminum). Since the mechanical properties of the Foam are considerably different from the ones of
the Aluminum (Table 2), a soft-core effect is expected. The external sheets have thickness h1 = h3 = 0.03 m, whereas the
thickness of the core is h2 = 0.14 m. The CCCC sandwich structure is subjected to a point load Q( ) = 10 kN applied in
(0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface, oriented with 1() = 90 , 2() = 90 , 3() = 0 . The Leg-Gau grid
distribution is used to discretize the reference domain with IN = 51, IM = 51. In this circumstance, every ESL theory related
to each order of kinematic expansion is taken with and without the Murakamis function. As it can be easily observed
from Figs. 1820, in which the through-the-thickness variation of strain, stress and displacement components evaluated at
the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) are depicted, the structural models that embed the Murakamis func-
tion give similar proles, whereas the others are quite different among them. With reference to the previous works by the
authors (Bacciocchi et al., 2016, Tornabene et al., 2016), it should be noticed that the Murakamis function must be intro-
duced in the kinematic model to capture the real mechanical behavior of sandwich structures, in which a soft-core effect is
noticeable.
Some parametric investigations are presented in the two following applications. The rst one aims to compute
the through-the-thickness variation of strain, stress and displacement components of a CFCC Funnel surface (Fig. 8e)
made of six layers oriented according to the stacking sequence (0/90/0/0/90/0). Each ply is made of Glass-Epoxy and
is dened by a constant thickness of hk = 0.01 m. The structure is subjected to a point load Q( + ) = 10 kN applied in
(0.5 (11 10 ), 0.5 (21 20 )) on the top surface, varying the angle 1(+) that denes the angle of incidence of the ap-
plied force. The other two angles are given respectively by 2(+ ) = 90 and 3(+ ) = 90 1(+ ) . As in the previous cases, the
Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51. The strain, stress and displacement
Fig. 32. Through-the-thickness variation of displacement components [m] for a CCCC laminated doubly-curved panel of translation with parabolic pro-
le made of eight plies with (30/60/ 45/0/0/45/ 60/ 30) as lamination scheme, evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) for
several higher-order ESL theories. The structure is subjected to a line load P3(+) = 10 kN/m applied along 1 direction in 2 = 0.5 (21 20 ) on the top
surface. The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
Fig. 33. Through-the-thickness variation of strain components for a CCCC laminated doubly-curved panel of translation with parabolic prole made of eight
plies with (30/60/ 45/0/0/45/ 60/ 30) as lamination scheme, evaluated at the point P = (0.34399 (11 10 ), 0.34399 (21 20 )) for several higher-order
ESL theories. The structure is subjected to a cross load P3() = 10 kN/m centered in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface. The Leg-Gau grid
distribution is used to discretize the reference domain with IN = 51, IM = 51.
proles are computed at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )) and are shown respectively in Figs. 2123.
Observing these graphs, it is easy to notice that the through-the-thickness variations of the quantities in hand are consider-
ably affected by the orientation of the applied load. The ED3 is considered as structural model.
The main purpose of the next test is to show the effect of the material orientation on the strain, stress and displace-
ment proles. The structure under consideration is a CCCC helicoidal surface (Fig. 8f) made of three laminae. The stack-
ing sequence is given by (45/ (2) / 45), in which the orientation of the central layer is variable. The laminated structure
is subjected to a point load Q( ) = 10 kN applied on the bottom surface in (0.5 (11 10 ), 0.5 (21 20 )), oriented with
1() = 90 , 2() = 90 , 3() = 0 . The results are evaluated at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )), tak-
ing into account only the ED3 and EDZ3 theories. As it can be observed in Figs. 2426, the Murakamis function does not
provide signicant improvements.
The FCFC laminated cylindrical surface with parabolic prole of Fig. 8b is employed once again in the following example.
The mechanical and geometric characteristics are unchanged. On the contrary, the structure is now subjected to a line load
P3(+ ) = 10 kN/m applied along 2 y direction in = 0.5 ( 1 0 ) on the top surface. The through-the-thickness variation
of strain, stress and displacement components, evaluated at P = (0.25691 ( 1 0 ), 0.25691 L), are depicted in Figs. 2729 for
several ESL structural theories, with and without the Murakamis function. In general, no signicant differences are observed
among the various curves. In some occasions, the FSDT curves detach from the others.
The next application is related to a CCCC laminated doubly-curved panel of translation with parabolic prole made of
eight layers with (30/60/ 45/0/0/45/ 60/ 30) as stacking sequence (Fig. 8c). The geometric features and the mechanical
properties are identical to the previous corresponding example. The structure is subjected to a line load P3(+ ) = 10 kN/m
Fig. 34. Through-the-thickness variation of stress [Pa] components for a CCCC laminated doubly-curved panel of translation with parabolic prole made
of eight plies with (30/60/ 45/0/0/45/ 60/ 30) as lamination scheme, evaluated at the point P = (0.34399 (11 10 ), 0.34399 (21 20 )) for several
higher-order ESL theories. The structure is subjected to a cross load P3() = 10 kN/m centered in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface. The
Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
applied along 2 direction in 2 = 0.5 (21 20 ) on the top surface. The through-the-thickness variation of strains, stresses
and displacements, computed at the point P = (0.25691 (11 10 ), 0.25691 (21 20 )), are depicted in Figs. 3032 vary-
ing the ESL structural model. The various theories give comparable proles. Some discrepancy can be observed in those
quantities that have lower order of magnitude if compared to the others (for instance, the strain 12 ). Even in this case,
the Murakamis function does not introduce any difference in the through-the-thickness plots. With reference to Fig. 32, it
should be highlighted the fact that higher-order theories show non-linear transverse displacements, differently from rst-
order (linear) models.
In the last application, the through-the-thickness variation of strain, stress and displacement components is evalu-
ated for the doubly-curved panel of translation depicted in Fig. 8d, but considering the previous lamination scheme
(30/60/ 45/0/0/45/ 60/ 30). Each ply has the same thickness and is made of Graphite-Epoxy. The structure is subjected
to a cross load P3( ) = 10 kN/m centered in (0.5 (11 10 ), 0.5 (21 20 )) applied on the bottom surface. The through-the-
thickness proles of strains, stresses and displacements, computed at the point P = (0.34399 (11 10 ), 0.34399 (21 20 )),
are shown in Figs. 3335 for various ESL structural models, with and without the Murakamis functions. The same consid-
erations mentioned in the previous cases are still valid in this circumstance.
Finally, the deformed shapes of all the shell structures analyzed in the present section are depicted in Fig. 36. The dif-
ference between a point load, line load and cross load is extremely evident, even though they have all a local effect on the
deformed shapes. This aspect justies the use of many discrete points to capture the correct mechanical behavior. In other
words, the grid points must be placed closer to represent the effective structural response.
Fig. 35. Through-the-thickness variation of displacement components [m] for a CCCC laminated doubly-curved panel of translation with parabolic prole
made of eight plies with (30/60/ 45/0/0/45/ 60/ 30) as lamination scheme, evaluated at the point P = (0.34399 (11 10 ), 0.34399 (21 20 )) for sev-
eral higher-order ESL theories. The structure is subjected to a cross load P3() = 10 kN/m centered in (0.5 (11 10 ), 0.5 (21 20 )) on the bottom surface.
The Leg-Gau grid distribution is used to discretize the reference domain with IN = 51, IM = 51.
Fig. 36. Deformed shapes of shell structures subjected to different types of actions: a)-f) point load; g)-h) line load; i) cross load.
7. Conclusion
The Generalized Differential Quadrature (GDQ) method has been employed to solve numerically the static problem of
shell structures subjected to concentrated forces. In particular, point and line loads have been taken into account and mod-
eled by means of the application of the Dirac-delta function. If the Legendre-Gauss grid distribution is used, the numerical
technique has proven to be an accurate, stable and reliable tool to deal with these kinds of structural problems. Several
comparisons in terms of both deections and through-the-thickness stress proles have been performed. Indeed, the present
approach has given results that are in excellent agreement with the semi-analytical ones that can be found in the literature.
For the rst time, a general kinematic model has been employed to compute the static response of laminated composites
plates and shells through different higher-order structural theories in a unied manner. A posteriori stress and strain re-
covery procedure based on the three-dimensional equilibrium equations has allowed to evaluate the through-the-thickness
variation of strain, stress and displacement components. Load discontinuities such as point and line loads lead to stress
patterns with strong discontinuities which are not easy to approximate with classic numerical technique especially near the
applied load area. Such concentrations have to be studied accurately in particular when composite structures are investi-
gated, because high inter-laminar stresses can occur. High stress concentrations among the plies in the stacking sequence
can result in delamination and premature structural failure. The present work demonstrated that such discontinuities can
be easily investigated by strong form numerical methods without the need of decomposing the physical domain into nite
elements.
Acknowledgments
The research topic is one of the subjects of the Centre of Study and Research for the Identication of Materials and
Structures (CIMEST) - M. Capurso of the University of Bologna (Italy).
References
Abali, B. E., Vllmecke, C., Woodward, B., Kashtalyan, M., Guz, I., & Mller, W. H. (2014). Three-dimensional elastic deformation of functionally graded
isotropic plates under point loading, 118, 367376.
Alibeigloo, A. (2016). Thermo elasticity solution of sandwich circular plate with functionally graded core using generalized differential quadrature method.
Composite Structures, 136, 229240.
Bacciocchi, M., Eisenberger, M., Fantuzzi, N., Tornabene, F., & Viola, E. (2016). Vibration analysis of variable thickness plates and shells by the generalized
differential quadrature method. Composite Structures. In Press. doi:10.1016/j.compstruct.2015.12.004.
Becker, W. (1995). Concentrated forces and moments on laminates with bending extension coupling. Composite Structures., 30, 111.
Behravan Rad, A., & Shariyat, M. (2015). Three-dimensional magneto-elastic analysis of asymmetric variable thickness porous FGM circular plates with
non-uniform tractions and Kerr elastic foundations. Composite Structures, 125, 558574.
Beiro da Veiga, L. (2005). Asymptotic study of the solution for pinched cylindrical shells. Computer Methods in Applied Mechanics and Engineering., 194,
11131139.
Belytschko, T., Liu, W.-K., Ong, J. S.-J., & Lam, D. (1985). Implementation and application of a 9-node Lagrange shell element with spurious mode control.
Bonisoli, E., Delprete, C., Sesana, R., Tamburro, A., & Tornincasa, S. (2015). Testing and simulation of the three point bending anisotropic behaviour of
hazelnut shells. Biosystems Engineering., 129, 134141.
Brischetto, S. (2014). An exact 3d solution for free vibrations of multilayered cross-ply composite and sandwich plates and shells. International Journal of
Applied Mechanics, 6, 1450076.
Brischetto, S., & Carrera, E. (2013). Static analysis of multilayered smart shells subjected to mechanical, thermal and electrical loads. Meccanica, 48,
12631287.
Brischetto, S., Tornabene, F., Fantuzzi, N., & Bacciocchi, M. (2015). Rened 2D and exact 3D shell models for the free vibration analysis of single- and
double-walled carbon nanotubes. Technologies, 3, 259284.
Brischetto, S., Tornabene, F., Fantuzzi, N., & Viola, E. (2016). 3D Exact and 2D generalized differential quadrature models for free vibration analysis of
functionally graded plates and cylinders. Meccanica, 51, 20592098.
Brischetto, S., & Torre, R. (2014). Exact 3D solutions and nite element 2D models for free vibration analysis of plates and cylinders. Curved and Layered
Structures, 1, 5992.
Carrera, E. (2002). Theories and Finite Elements for Multilayered, Anisotropic, Composite Plates and Shells. Archives of Computational Methods in Engineer-
ing, 9, 87140.
Carrera, E. (2003). Historical review of zig-zag theories for multilayered plates and shells. Applied Mechanics Reviews, 56, 287308.
Carrera, E. (2004). On the use of the Murakamis zig-zag function in the modeling of layered plates and shells. Composite Structures ., 82, 541554.
Cernescu, A., & Romanoff, J. (2015). Bending deection of sandwich beams considering local effect of concentrated force. Composite Structures , 134,
169175.
Cheng, Z.-Q., Lim, C. W., & Kitipornchai, S. (1999). Three-dimensional exact solution for inhomogeneous and laminated piezoelectric plates. International
Journal of Engineering Science., 37, 14251439.
Dehghan, M., Nejad, M. Z., & Moosaie, A. (2016). Thermo-electro-elastic analysis of functionally graded piezoelectric shells of revolution: Governing equa-
tions and solutions for some simple cases. International Journal of Engineering Science., 104, 3461.
Eftekhari, S. A. (2015a). A Differential Quadrature Procedure with Regularization of the Dirac-delta Function for Numerical Solution of Moving Load Problem.
Latin American Journal of Solids and Structures, 12, 12411265.
Eftekhari, S. A. (2015b). A note on mathematical treatment of the Dirac-delta function in the differential quadrature bending and forced vibration analysis
of beams and rectangular plates subjected to concentrated loads. Appl. Math. Model., 39, 62236242.
Eftekhari, S. A. (2016a). A differential quadrature procedure with direct projection of the heaviside function for numerical solution of moving load problem.
Latin American Journal of Solids and Structures, 13, 17631781.
Eftekhari, S. A. (2016b). A modied differential quadrature procedure for numerical solution of moving load problem. In Proceedings of the institution of
mechanical engineers, Part c: journal of mechanical engineering science: 230 (pp. 715731).
Eftekhari, S. A., & Jafari, A. A. (2014). Accurate variational approach for free vibration of simply supported anisotropic rectangular plates. Archive of Applied
Mechanics, 84, 607614.
Fantuzzi, N. (2014). New insights into the strong formulation nite element method for solving elastostatic and elastodynamic problems. Curved and
Layered Structures, 1, 93126.
Fantuzzi, N., Bacciocchi, M., Tornabene, F., Viola, E., & Ferreira, A. J. M. (2015). Radial Basis Functions Based on Differential Quadrature Method for the Free
Vibration of Laminated Composite Arbitrary Shaped Plates. Composites Part B Engineering, 78, 6578.
Fantuzzi, N., Dimitri, R., & Tornabene, F. (2016). A SFEM-Based Evaluation of Mode-I Stress Intensity Factor in Composite Structures. Composite Structures,
145, 162185.
Fantuzzi, N., & Tornabene, F. (2014). Strong formulation nite element method for arbitrarily shaped laminated plates - I. Theoretical analysis. Advances in
Aircraft and Spacecraft Science , 1, 124142.
Fantuzzi, N., & Tornabene, F. (2014). Strong formulation nite element method for arbitrarily shaped laminated plates - II. Numerical analysis. Advances in
Aircraft and Spacecraft Science , 1, 143173.
Fantuzzi, N., & Tornabene, F. (2016). Strong Formulation Isogeometric Analysis (SFIGA) for Laminated Composite Arbitrarily Shaped Plates. Composites Part
B Engineering, 96, 173203.
Fantuzzi, N., Tornabene, F., & Viola, E. (2014). Generalized Differential Quadrature Finite Element Method for Vibration Analysis of Arbitrarily Shaped Mem-
branes. International Journal of Mechanical Sciences , 79, 216251.
Fantuzzi, N., Tornabene, F., & Viola, E. (2016). Four-parameter functionally graded cracked plates of arbitrary shape: A GDQFEM solution for free vibrations.
Mechanics of Advanced Materials and Structures , 23, 89107.
Fantuzzi, N., Tornabene, F., Viola, E., & Ferreira, A. J. M. (2014). A strong formulation nite element method (SFEM) based on RBF and GDQ techniques for
the static and dynamic analyses of laminated plates of arbitrary shape. Meccanica, 49, 25032542.
Fazzolari, F. A. (2014). A rened dynamic stiffness element for free vibration analysis of cross-ply laminated composite cylindrical and spherical shallow
shells. Composites Part B Engineering, 62, 143158.
Fazzolari, F. A. (2015). Natural frequencies and critical temperatures of functionally graded sandwich plates subjected to uniform and non-uniform temper-
ature distributions. Composite Structures, 121, 197210.
Fazzolari, F. A. (2016a). Reissners mixed variational theorem and variable kinematics in the modelling of laminated composite and fgm doubly-curved
shells. Composites Part B Engineering, 89, 408423.
Fazzolari, F. A. (2016b). Stability analysis of FGM sandwich plates by using variable-kinematics ritz models. Mechanics of Advanced Materials and Structures,
23, 11041113.
Ferreira, A. J. M., Carrera, E., Cinefra, M., Viola, E., Tornabene, F., & Fantuzzi, N. (2014). Analysis of thick isotropic and cross-ply laminated plates by gener-
alized differential quadrature method and a unied formulation. Composites Part B Engineering, 58, 544552.
Groh, R. M. J., & Weaver, P. M. (2015a). On displacement-based and mixed-variational equivalent single layer theories for modelling highly heterogeneous
laminated beams. International Journal of Engineering Science , 59, 147170.
Groh, R. M. J., & Weaver, P. M. (2015b). Static inconsistencies in certain axiomatic higher-order shear deformation theories for beams, plates and shells.
Guo, J., Chen, J., & Pan, E. (2016). Size-dependent behavior of functionally graded anisotropic composite plates. International Journal of Engineering Science,
106, 110124.
Hadjiloizi, D. A., Kalamkarov, A. L., Metti, Ch., & Georgiades, A. V. (2014a). Analysis of smart piezo-magneto-thermo-elastic composite and reinforced plates:
Part II Applications. Curved and Layered Structures , 1, 3258.
Hadjiloizi, D. A., Kalamkarov, A. L., Metti, Ch., & Georgiades, A. V. (2014b). Analysis of smart piezo-magneto-thermo-elastic composite and reinforced plates:
Part I Model Development. Curved and Layered Structures , 1, 1131.
Hughes, T. J. R., & Liu, W. K. (1981). Nonlinear nite element analysis of shells: Part I. Three dimensional shells. Computer Methods in Applied Mechanics
and Engineering, 26, 331362.
Hwu, C., & Tan, C. J. (2007). In-plane/out-of-plane concentrated forces and moments on composite laminates with elliptical elastic inclusions. International
Journal of Solids and Structures , 44, 65846606.
Jin, G., Ye, T., Wang, X., & Miao, X. (2016). A unied solution for the vibration analysis of FGM doubly-curved shells of revolution with arbitrary boundary
conditions. Composites Part B Engineering, 89, 230252.
Jones, I. A. (1998a). Approximate solution to the orthotropic pinched cylinder problem. Composite Structures , 42, 7391.
Jones, I. A. (1998b). Flgge shell theory and solution for orthotropic cylindrical shells under pinching loads. Composite Structures , 42, 5372.
Jooybar, N., Malekzadeh, P., Fiouz, A., & Vaghe, M. (2016). Thermal effect on free vibration of functionally graded truncated conical shell panels.
Thin-Walled Structures , 103, 4561.
Kamarian, S., Salim, M., Dimitri, R., & Tornabene, F. (2016). Free Vibration Analysis of Conical Shells Reinforced with Agglomerated Carbon Nanotubes.
International Journal of Mechanical Sciences, 108-109, 157165.
Kim, J., & Swanson, S. R. (2001). Design of sandwich structures for concentrated loading. Composite Structures , 52, 365373.
Kim, Y.-W. (2015). Free vibration analysis of FGM cylindrical shell partially resting on Pasternak elastic foundation with an oblique edge. Composites Part B
Engineering, 70, 263276.
Kulikov, G. M., Mamontov, A. A., Plotnikova, S. V., & Mamontov, S. A. (2016). Exact geometry solid-shell element based on a sampling surfaces technique for
3D stress analysis of doubly-curved composite shells. Curved and Layered Structures , 3, 116.
Le, K. C., & Yi, J.-H. (2016). An asymptotically exact theory of smart sandwich shells. International Journal of Engineering Science , 106, 179198.
Librescu, L., & Reddy, J. N. (1989). A few remarks concerning several rened theories of anisotropic composite laminated plates. International Journal of
Engineering Science , 27, 515527.
Liu, F.-L., & Liew, K. M. (1998). Static analysis of Reissner-Mindlin plates by differential quadrature element method. Journal of Applied Mechanics , 65,
705710.
Liu, F.-L., & Liew, K. M. (1999). Differential quadrature element method for static analysis of Reissner-Mindlin polar plates. International Journal of Solids
and Structures , 36, 51015123.
Liu, M., Cheng, Y., & Liu, J. (2015). High-order free vibration analysis of sandwich plates with both functionally graded face sheets and functionally graded
exible core. Composites Part B Engineering , 72, 97107.
Mantari, J. L. (2015). Rened and generalized hybrid type quasi-3D shear deformation theory for the bending analysis of functionally graded shells. Com-
posites Part B Engineering, 83, 142152.
Mercan, K., Demir, ., & Civalek, . (2016). Vibration analysis of FG cylindrical shells with power-law index using discrete singular convolution technique.
Curved and Layered Structures , 3, 8290.
Mohammad-Abadi, M., & Daneshmehr, A. R. (2015). Modied couple stress theory applied to dynamic analysis of composite laminated beams by considering
different beam theories. International Journal of Engineering Science , 87, 83102.
Mohammad-Abadi, M., Daneshmehr, A. R., & Homayounfard, M. (2015). Size-dependent thermal buckling analysis of micro composite laminated beams
using modied couple stress theory. International Journal of Engineering Science , 92, 4762.
Mohazzab, A. H., & Dozio, L. (2015). Prediction of natural frequencies of laminated curved panels using rened 2-D theories in the spectral collocation
method. Curved and Layered Structures, 2, 114.
Morley, L. S. D. (1960). The thin-walled circular cylinder subjected to concentrated radial loads. Quaterly Journal of Mechanics and Applied Math, 13, 2337.
Nikopour, H., & Selvadurai, A. P. S. (2014). Concentrated loading of a bre-reinforced composite plate: Experimental and computational modeling of bound-
ary xity. Composites Part B Engineering, 60, 297305.
Piskunov, V. G., Verijenko, V. E., Adali, S., & Summers, E. B. (1993). A Higher-order Theory for the Analysis of Laminated Plates and Shells with Shear and
Normal Deformation. International Journal of Engineering Science , 31, 967988.
Quan, T. Q., Tran, P., Tuan, N. D., & Duc, N. D. (2015). Nonlinear dynamic analysis and vibration of shear deformable eccentrically stiffened S-FGM cylindrical
panels with metal-ceramic-metal layers resting on elastic foundations. Composite Structures, 126, 1633.
Rao, K. P. (1978). A rectangular laminated anisotropic shallow thin shell nite element. Computer Methods in Applied Mechanics and Engineering , 15,
1333.
Ray, M. C., & Reddy, J. N. (2005). Active control of laminated cylindrical shells using piezoelectric ber reinforced composites. Composites Science and
Technology , 65, 12261236.
Reddy, J. N. (1984). Exact solutions of moderately thick laminated shells. Journal of Engineering Mechanics ., 110, 794809.
Reddy, J. N. (2004). Mechanics of laminated composite plates and shells. CRC Press.
Shaei, N., Kazemi, M., Sa, M., & Ghadiri, M. (2016). Nonlinear vibration of axially functionally graded non-uniform nanobeams. International Journal of
Engineering Science, 106, 7794.
Shu, C., & Richards, B. E. (1992a). Application of generalized differential quadrature to solve two-dimensional incompressible Navier-Stokes equations. In-
ternational Journal for Numerical Methods in Fluids , 15, 791798.
Shu, C., & Richards, B. E. (1992b). Parallel simulation of incompressible viscous ows by generalized differential quadrature. Computing Systems in Engi-
neering , 3, 271281.
Soyev, A. H. (2015). Inuences of shear stresses on the dynamic instability of exponentially graded sandwich cylindrical shells. Composites Part B Engi-
neering, 77, 349362.
Soyev, A. H. (2016). Large amplitude vibration of FGM orthotropic cylindrical shells interacting with the nonlinear Winkler elastic foundation. Composites
Part B Engineering , 98, 141150.
Soyev, A. H., & Kuruoglu, N. (2015). On a problem of the vibration of functionally graded conical shells with mixed boundary conditions. Composites Part
B Engineering ., 70, 122130.
Soyev, A. H., & Kuruoglu, N. (2016). Domains of dynamic instability of FGM conical shells under time dependent periodic loads. Composite Structures ,
136, 139148.
Ting, L., & Yuan, S. W. (1958). On radial deection of a cylinder of nite length with various end conditions. Journal of the Aerospace Sciences , 25, 230234.
Tornabene, F., Fantuzzi, N., Viola, E., & Reddy, J. N. (2014). Winkler-Pasternak Foundation Effect on the Static and Dynamic Analyses of Laminated Dou-
bly-Curved and Degenerate Shells and Panels. Composites Part B Engineering , 57, 269296.
Tornabene, F. (2016). General Higher Order Layer-Wise Theory for Free Vibrations of Doubly-Curved Laminated Composite Shells and Panels. Mechanics of
Advanced Materials and Structures , 23, 10461067.
Tornabene, F., Brischetto, S., Fantuzzi, N., & Viola, E. (2015a). Numerical and Exact Models for Free Vibration Analysis of Cylindrical and Spherical Shell
Panels. Composites Part B Engineering , 81, 231250.
Tornabene, F., & Ceruti, A. (2013). Free-form Laminated Doubly-Curved Shells and Panels of Revolution on Winkler-Pasternak Elastic Foundations: A 2D GDQ
Solution for Static and Free Vibration Analysis. World Journal of Mechanics ., 3, 125.
Tornabene, F., Dimitri, R., & Viola, E. (2016). Transient Dinamic Responce of Genarally-Shaped Arches Based on a GDQ-Time-Stepping Method. International
Journal of Mechanical Sciences In Press. doi:10.1016/j.ijmecsci.2016.05.005.
Tornabene, F., Fantuzzi, N., & Bacciocchi, M. (2014a). The strong formulation nite element method: stability and accuracy. Fracture and Structural Integrity,
29, 251265.
Tornabene, F., Fantuzzi, N., & Bacciocchi, M. (2014b). Free vibrations of free-form doubly-curved shells made of functionally graded materials using high-
er-order equivalent single layer theories. Composites Part B Engineering , 67, 490509.
Tornabene, F., Fantuzzi, N., & Bacciocchi, M. (2014c). The Local GDQ method applied to general higher-order theories of doubly-curved laminated composite
shells and panels: The free vibration analysis. Composite Structures , 116, 637660.
Tornabene, F., Fantuzzi, N., & Bacciocchi, M. (2016a). Higher-order structural theories for the static analysis of doubly-curved laminated composite panels
reinforced by curvilinear bers. Thin-Walled Structures , 102, 222245.
Tornabene, F., Fantuzzi, N., & Bacciocchi, M. (2016b). The Local GDQ method for the natural frequencies of doubly-curved shells with variable thickness: A
general formulation. Composites Part B Engineering , 92, 265289.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Viola, E. (2016c). Laminated composite doubly-curved shell structures. Differential and integral quadrature. Strong
formulation nite element method. Esculapio.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Viola, E. (2016d). Laminated composite doubly-curved shell structures. Differential Geometry. Higher-order struc-
tural theories. Esculapio.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Dimitri, R. (2015a). Free vibrations of composite oval and elliptic cylinders by the generalized differential
quadrature method. Thin-Walled Structures , 97, 114129.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Dimitri, R. (2015b). Dynamic analysis of thick and thin elliptic shell structures made of laminated composite
materials. Composite Structures , 133, 278299.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., Neves, A. M. A., & Ferreira, A. J. M. (2016). MLSDQ Based on RBFs for the free vibrations of laminated composite
doubly-curved shells. Composites Part B Engineering , 99, 3047.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Viola, E. (2015a). Higher-Order Theories for the Free Vibration of Doubly-Curved Laminated Panels with Curvi-
linear Reinforcing Fibers by Means of a Local Version of the GDQ Method. Composites Part B Engineering , 81, 196230.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Viola, E. (2015b). Accurate Inter-Laminar Recovery for Plates and Doubly-Curved Shells with Variable Radii of
Curvature Using Layer-Wise Theories. Composite Structures , 124, 368393.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Viola, E. (2015c). A new approach for treating concentrated loads in doubly-curved composite deep shells with
variable radii of curvature. Composite Structures , 131, 433452.
Tornabene, F., Fantuzzi, N., Bacciocchi, M., & Viola, E. (2016). Effect of agglomeration on the natural frequencies of functionally graded carbon nanotube-re-
inforced laminated composite doubly-curved shells. Composites Part B Engineering , 89, 187218.
Tornabene, F., Fantuzzi, N., Ubertini, F., & Viola, E. (2015). Strong formulation nite element method based on differential quadrature: A survey. Applied
Mechanics Reviews , 67(020801), 55.
Tornabene, F., Fantuzzi, N., & Viola, E. (2016). Inter-laminar stress recovery procedure for doubly-curved, singly-curved, revolution shells with variable radii
of curvature and plates using generalized higher-order theories and the local GDQ method. Mechanics of Advanced Materials and Structures , 23,
10191045.
Tornabene, F., Fantuzzi, N., Viola, E., & Batra, R. C. (2015). Stress and strain recovery for functionally graded free-form and doubly-curved sandwich shells
using higher-order equivalent single layer theory. Composite Structures , 119, 6789.
Tornabene, F., Fantuzzi, N., Viola, E., & Carrera, E. (2014). Static analysis of doubly-curved anisotropic shells and panels using CUF approach, differential
geometry and differential quadrature method. Composite Structures , 107, 675697.
Tornabene, F., Fantuzzi, N., Viola, E., & Ferreira, A. J. M. (2013). Radial Basis Function method applied to doubly-curved laminated composite shells and
panels with a general higher-order equivalent single layer theory. Composites Part B Engineering , 55, 642659.
Tornabene, F., Liverani, A., & Caligiana, G. (2012). Static Analysis of Laminated Composite Curved Shells and Panels of Revolution with a Posteriori Shear
and Normal Stress Recovery Using Generalized Differential Quadrature Method. International Journal of Mechanical Sciences , 61, 7187.
Tornabene, F., & Reddy, J. N. (2013). FGM and laminated doubly-curved and degenerate shells resting on nonlinear elastic foundations: A GDQ solution for
static analysis with a posteriori stress and strain recovery. Journal of the Indian Institute of Science, 93, 635688.
Tornabene, F., & Viola, E. (2009). Free Vibration analysis of functionally graded panels and shells of revolution. Meccanica, 44, 255281.
Tornabene, F., & Viola, E. (2013). Static Analysis of functionally graded doubly-curved shells and panels of revolution. Meccanica, 48, 901930.
Tornabene, F., Viola, E., & Fantuzzi, N. (2013). General higher-order equivalent single layer theory for free vibrations of doubly-curved laminated composite
shells and panels. Composite Structures , 104, 94117.
Tsamasphyros, G. J., Theotokoglou, E. E., & Filopoulos, S. P. (2013). Study and solution of BEM-singular integral equation method in the case of concentrated
loads. Int. J. International Journal of Solids and Structures, 50, 16341645.
Vescovini, R., & Dozio, L. (2016). A variable-kinematic model for variable stiffness plates: Vibration and buckling analysis. Composite Structures , 142, 1526.
Viola, E., Miniaci, M., Fantuzzi, N., & Marzani, A. (2015). Vibration analysis of multi-stepped and multi-damaged parabolic arches using GDQ. Curved and
Layered Structures, 2, 2849.
Viola, E., Rossetti, L., Fantuzzi, N., & Tornabene, F. (2014). Static analysis of functionally graded conical shells and panels using the generalized unconstrained
third order theory coupled with the stress recovery. Composite Structures, 112, 4465.
Viola, E., Rossetti, L., Fantuzzi, N., & Tornabene, F. (2016). Generalized stress-strain recovery formulation applied to functionally graded spherical shells and
panels under static loading. Composite Structures In Press. doi:10.1016/j.compstruct.2015.12.060.
Viola, E., Tornabene, F., & Fantuzzi, N. (2013a). Generalized differential quadrature nite element method for cracked composite structures of arbitrary
shape. Composite Structures, 106, 815834.
Viola, E., Tornabene, F., & Fantuzzi, N. (2013b). Static analysis of completely doubly-curved laminated shells and panels using general higher-order shear
deformation theories. Composite Structures, 101, 5993.
Viola, E., Tornabene, F., & Fantuzzi, N. (2013c). General higher-order shear deformation theories for the free vibration analysis of completely doubly-curved
laminated shells and panels. Composite Structures, 95, 639666.
Viola, E., Tornabene, F., Fantuzzi, N., & Bacciocchi, M. (2016). DiQuMASPAB software. Alma Mater Studiorum - University of Bologna DICAM Department
http:// software.dicam.unibo.it/ diqumaspab-project .
Viola, E., Tornabene, F., Ferretti, E., & Fantuzzi, N. (2013a). Soft core plane state structures under static loads using GDQFEM and Cell method. CMES, 94,
301329.
Viola, E., Tornabene, F., Ferretti, E., & Fantuzzi, N. (2013b). GDQFEM numerical simulations of continuous media with cracks and discontinuities. CMES, 94,
331368.
Viola, E., Tornabene, F., Ferretti, E., & Fantuzzi, N. (2013c). On static analysis of plane state structures via GDQFEM and Cell method. CMES, 94, 421458.
Vlasov, V. Z. (1964).General theory of shells and its applications in engineering, NASA TT F-99,
Wang, X. (2015). Differential quadrature and differential quadrature based element Methods: theory and applications. Butterworth-Heinemann.
Wu, C.-P., & Hung, Y.-C. (1999). Asymptotic theory of laminated circular conical shells. International Journal of Engineering Science, 37, 9771005.
Yang, T. Y. (1971). High order rectangular shallow shell nite element. Journal of Engineering Mechanics, 38, 400407.
Yifeng, Z., Wenzheng, Q., Yu, W., Wenwen, Y., & Lichao, J. (2015). A variational asymptotic theory of composite laminated plates: Hybrid transformation to
Reissner-Mindlin model. International Journal of Engineering Science, 97, 113125.
Yuan, S. W. (1946). Thin cylindrical shells subjected to concentrated loads. Quarterly Journal of Mechanics and Applied Mathematics, 4, 1326.
Yuan, S. W., & Ting, I. (1957). On radial deections of a cylinder subjected to equal and opposite concentrated radial loads innitely long cylinder and
nite-length cylinder with simply supported ends. Journal of Engineering Mechanics, 24, 278282.
Zhang, J.-W. (1991). A parametric study of pinched orthotropic cylindrical shells. Journal of Computational and Applied Mathematics, 35, 319330.

1 s2.0 S0020722516304773 Main

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

1 s2.0 S0020722516304773 Main

Caricato da

Copyright:

Formati disponibili

International Journal of Engineering Science 109 (2016) 115164

Contents lists available at ScienceDirect

International Journal of Engineering Science

On the mechanics of laminated doubly-curved shells

3. Higher-order shear deformation theories

U1 = F0 u1( ) + F1 u1( ) + F2 u1( ) + F3 u1( ) + ... + FN u1( ) + FN+1 u1( )

U2 = F0 u2( ) + F1 u2( ) + F2 u2( ) + F3 u2( ) + ... + FN u2( ) + FN+1 u2( )

U3 = F0 u3( ) + F1 u3( ) + F2 u3( ) + F3 u3( ) + ... + FN u3( ) + FN+1 u3( )

in which U1 ( 1 , 2 , ), U2 ( 1 , 2 , ), U3 ( 1 , 2 , ) represent the three-dimensional displacement components. As it can be

where the differential operator D is dened as

where D is the equilibrium operator dened as follows

4.1. Discretization of the fundamental equations

the m-th derivative with respect to 2 coordinate

and the mixed partial derivative of order (n + m)

Kdb b + Kdd d = fd (45)

4.2. Evaluation of the stiffness coecients

4.3. Implementation of the concentrated loads

Fig. 8. Laminated shell structures and discrete grid representation.

5. Strain and stress recovery procedure

1n(i j1) = q1((i)j )

and on the top one

1n(i jIT ) = q1(+(i)j )

conditions on the external surfaces

n(i j1) = qn((i)j )

6. Results and discussion

6.1. Comparison with the literature

Lamination scheme: (0)

IN = 11 IN = 21 IN = 31 IN = 41 IN = 51 Reference (Reddy, 2004)

IN = 11 IN = 21 IN = 31 IN = 41 IN = 51 Reference (Reddy, 2004)

Lamination scheme: (0)

IN = 11 IN = 21 IN = 31 IN = 41 IN = 51 Reference (Reddy, 2004)

IN = 11 IN = 21 IN = 31 IN = 41 IN = 51 Reference (Reddy, 2004)

6.2. Stress and strain recovery procedure for shell structures

Lamination scheme: (0)

IN = 11 IN = 21 IN = 31 IN = 41 IN = 51 Reference (Reddy, 2004)

IN = 11 IN = 21 IN = 31 IN = 41 IN = 51 Reference (Reddy, 2004)

Funnel surface (doubly-curved panel of revolution)

Potrebbero piacerti anche