Acta Mech

DOI 10.1007/s00707-017-1825-4

O R I G I NA L PA P E R

Gennadi Mikhasev · Marina Botogova

Effect of edge shears and diaphragms on buckling

of thin laminated medium-length cylindrical shells with low
effective shear modulus under external pressure

Received: 4 October 2016 / Revised: 19 January 2017

© Springer-Verlag Wien 2017

Abstract Governing equations based on the generalized kinematic hypotheses of Timoshenko and including
the effect of transverse shears are used to predict the buckling of a medium-length thin laminated cylindrical
shell under action of the external normal pressure. It is assumed that some of layers are made of a “soft”
material so that the effective shear modulus turns out to be too less than the effective Young’s modulus for
the laminate. Of all possible variants of boundary conditions, the boundary conditions corresponding to the
simple support of edges with and without diaphragms in their planes are considered. For the case of the simply
supported edges with diaphragms, the critical buckling pressure as well as the modes of buckling are found
in an explicit form. If one of the edges is free from the diaphragm, the boundary-value problem is solved
by using the asymptotic approach, a solution being constructed in the form of the superposition of functions
describing the main stress state and the edge effect integrals. It is shown that the absence of the edge diaphragm
accounts for the appearance of the edge transverse shears (non-classical edge effect integrals) whose decay
rate is lower than that of the classical simple edge effect integrals. The effect of edge shears and diaphragms
as well on both the critical buckling pressure and eigenform is studied for a laminated cylindrical shell with
any number of layers regardless of materials used for laminae. As an example, the buckling of a cylindrical
sandwich assembled from the ABS-plastic and magnetorheological elastomer under different levels of an
applied magnetic field is examined.

1 Introduction

The use of new and non-traditional materials with wide range of physical properties permits to design durable
thin-walled multilayered elements (cylindrical shells, panels, plates) [1] satisfying up-to-date requirements
such as lightness, high-specific stiffness, good buckling resistance, safety, and noiselessness. For thin-walled
constructions (such as airborne/spaceborne vehicles, underwater objects) experiencing heavy external load,
the bearing capacity becomes a crucial characteristic. The critical buckling load, and particularly the critical
external pressure is significant because it provides a benchmark for assessing the bearing capacity and total
performance of thin laminated shell-like structures.
Often a multilayered shell is assembled from laminae whose elastic properties differ significantly. So, a
sandwich construction may consist of two thin high-elasticity faces and a thick soft core made of foam [2,3],
epoxy [4], or smart materials like an magnetorheological elastomer [5,6] or electrorheological composite [7].
In these structures, the soft core carries mainly shear stresses, while face sheets carry the normal stresses
[2]. Sometimes the reduced shear modulus for all packages becomes much smaller than the reduced elastic
modulus that promotes the shear effects. In turn, the significant transverse shears result in a reduction of the

G. Mikhasev (B) · M. Botogova

Belarusian State University, 4 Nezavisimosti Avenue, 220030 Minsk, Belarus
E-mail: Mikhasev@bsu.by
 G. Mikhasev, M. Botogova

critical buckling load. So, the need for the accurate prediction of critical loads for similar laminated or sandwich
structures with reduced shear resistance becomes a very important problem.
The mathematical formulation of problems on elastic nonlinear behavior and buckling of multilayered
shells is rather complicated, because in the accurate statement these problems are reduced to the analysis
of nonlinear behavior of each layer with required satisfaction of the boundary conditions on both edges and
interfaces. Currently, the finite-element simulation based on the 3D models of elasticity became a commonly
used mathematical tool for solving similar problems [3,8–10]. Although the FEM method is universal, it permits
to predict the buckling of only a specific laminated shell with fixed number of laminae and given distribution
of mechanical properties and does not allow to reveal qualitative results for a wide class of problems. New
advanced theories based on 3D stress analysis and rigid-body motions (e.g., see [11]) as well as available
high-accuracy layer-wise theories (see, among many others, [12–18,20–22]) are also rather sophisticated in
theoretical formulations and numerical computations, thus preventing their general use in the modeling of
buckling of laminated shells.
Among different theories developed for the buckling analysis of laminated shells, the equivalent single-
layer (ESL) models seem to be more simple. They may be subdivided into the classical lamination theory
[23,24], the first-order shear deformation theory [12,25–27], and higher-order shear deformation theories
[18,19,28–32] (see also the detailed classification in references [21,33]). The last ones assume quadratic,
cubic, or higher variations of the tangential displacements through the entire thickness of a laminated shell
and are more accurate. The accuracy of the ESL models depends on correlations of thicknesses and elastic
properties of all layers. Even though some layers are more soft than others, the ESL model gives an accurate
result in the estimation of the critical buckling pressure if the total thickness of a shell is sufficiently small and
the stiffness of all layers is approximately of the same order [3,29,34,35]. So, studying the buckling of the thin
sandwich cylinder with face sheets made of aluminum and an epoxy core, Mikhasev et al. [35] showed that
the divergence of values of the critical pressure obtained by using the ESL model of Grigoliuk and Kulikov
[29] and the FEM as well varied from 1 to 4% for very thin and moderately thin shells, respectively. And Han
et al. [3] analyzed the buckling of cylindrical sandwiches of different total thicknesses with alloy–foam core
and face sheets made of different materials (boron/epoxy, graphite/epoxy, and kevlar/epoxy) in three ways:
a) considering the sandwich as a three-dimensional (3D) elastic body, b) applying the ESL model accounting
for the transverse shear effects, and c) performing the finite-element simulation. The comparative analysis of
different approaches has revealed that an error of the ESL model versus the 3D model has varied (depending
on the material of the face layers) from 3.1 to 16.6% for moderately thin shells (R/ h = 30), and between 0.13
and 3.3% for thin and very thin shells (R/ h = 60 and R/ h = 120).
The ESL models turned out very fruitful and promoted a further development of higher-order shear defor-
mation theories as well as numerous studies on the pressure induced buckling of laminated and sandwich
cylindrical shells in different statements and under various complicating factors (see, among many others, Refs.
[36–40]). In particular, in paper [37], the governing equations based on the higher-order shear deformation the-
ory with von Kármán-Donnell-type of kinematic nonlinearity have been used to study nonlinear buckling and
postbuckling of a moderately thick anisotropic laminated cylindrical shell of finite length subjected to lateral
pressure, hydrostatic pressure, and external liquid pressure. In Ref. [38], an improved high-order theory has
been presented for a biaxial buckling analysis of sandwich plates with soft orthotropic core. And Grover et al.
[39] proposed a new inverse hyperbolic shear deformation theory satisfying traction-free boundary conditions
for the buckling response of laminated shells.
In the majority of papers on buckling of laminated shells accounting for the transverse shears, the authors
considered the traditional variants of boundary conditions corresponding to simply supported or clamped
edges, but did not distinguished between the cases when an edge has a diaphragm preventing transverse shears
in its plane or when it has not. As follows from the book of Grigoliuk and Kulikov [29], the presence of the
diaphragm on a clamped edge or its absence on a simply supported edge results in the edge effects which are
characterized by the transverse shears localized in the vicinity of the edge. It is obvious that for a very thin
laminated shell the effect of these edge shears on the buckling load is weak, but for a moderately thin shell
(with the dimensionless thickness satisfying the inequality 20 < R/ h < 60), this influence may be essential.
The general aim of this paper is to eliminate the above-mentioned gap and study the effect of the edge
transverse shears and diaphragms on the critical buckling hydrostatic pressure for a thin and moderately thin
laminated cylindrical shell of a medium length. It is assumed that the shell is assembled from different materials
resulting in the low effective shear modulus for the laminated package in whole. Of all variants of boundary
conditions, we consider only the two variants of conditions belonging to the simple support group but being
different for the transverse shears in the edge plane. To predict the buckling of a laminated shell, we apply
Buckling of thin cylindrical shells

to the ESL model the system of differential equations derived by Grigoliuk and Kulikov [29] and based on
the generalized kinematic hypothesis of Timoshenko being used as the governing one. The buckling problem
is reduced to the singularly perturbed boundary-value problem whose asymptotic solution is constructed in
the form of the superposition of functions describing the main stress–strain state and the special edge effect
integrals. It is shown that the special edge effect integrals taking into account the edge shears do not coincide
with the classical simple edge effect integrals [41] and have a small decay rate in comparison with the simple
edge effect integrals. Regardless of the number of layers and their mechanical properties, the effect of the
dimensionless reduced shear parameter on the critical buckling pressure is studied in detail. As an example,
the buckling of a cylindrical sandwich with a soft core made of a magnetorheological elastomer (MRE) and
face sheets fabricated from the ABS-plastic is analyzed for different levels of an applied magnetic field. The
buckling of multilayered thin cylinders assembled from different materials (silicon nitrate, alloy–foam, epoxy
and ABS-plastic) is also considered. In all examples, it is shown that the effect of the edge shears on the critical
pressure increases with the total thickness of the shell and should be taken into account when predicting the
buckling of laminated shells with low effective shear modulus.

2 Setting the problem

Consider a thin laminated medium-length cylindrical shell (see Fig. 1) consisting of N transversely isotropic
layers characterized by the following parameters: length L, thickness h k , density ρk , Young’s modulus E k ,
shear modulus G k , and Poisson’s ratio νk , where k = 1, 2, . . . , N is a number of layers. It is assumed that
each layer has a constant thickness. The middle surface of any fixed layer is taken as the reference surface.
A coordinate system α1 , α2 is shown in Fig. 1, where α1 , α2 are the axial and circumferential coordinates,
respectively. The radius of curvature of the reference surface is R.
At the edges α1 = 0, α1 = L, we consider only the group of the simple support which includes two
different types of boundary conditions. The first one is characterized by the presence of an edge diaphragm
preventing the transverse edge shears in the edge plane, and for the second one, this diaphragm is absent. An
accurate statement of these conditions will be given below.
Let the shell be under the action of the external constant hydrostatic pressure qn . The basic goal of our
paper is to estimate the critical buckling pressure qn∗ for the two variants of boundary conditions and to show
the effect of the edge shears or diaphragms on the critical pressure.

2.1 Principal hypotheses

Let δk be the distance between the reference surface of the cross section and the upper bound of the kth layer,
δ0 is the distance from the reference surface to the inner bound of the shell, u i and w are the tangential and
N (k)
normal displacements of the shell, respectively, h = k=1 h k is the total thickness of the laminate, u i are

hN
n
hk e1
hk-1
e2
h1

Fig. 1 Laminated cylindrical shell with a curvilinear coordinate system

 G. Mikhasev, M. Botogova

Fig. 2 Infitesimal element of the laminated shell, reference surface, and stresses

the tangential displacements of points of the k th layer, σi3 are the transverse shear stresses (see Fig. 2), and θi
are the angles of rotation of the normal n about the vectors ei (see Fig. 1). Here i = 1, 2; k = 1, 2, . . . , N .
We assume the following hypotheses of the laminated shell theory stated in [29]:
(i) The distribution law of the transverse tangent stresses across the thickness of the k th layer is assumed to
be of the form
(0) (k)
σi3 = f 0 (z)μi (α1 , α2 ) + f k (z)μi (α1 , α2 ) (1)

where f 0 (z), f k (z) are continuous functions introduced by

1
f 0 (z) = (z − δ0 )(δ N − z) z ∈ [δ0 , δ N ],
h2
1 (2)
f k (z) = 2 (z − δk−1 )(δk − z) ∈ [δk−1 , δk ],
hk
f k (z) = 0 z ∈/ [δk−1 , δk ].
(ii) Normal stresses acting on the area elements parallel to the original one are negligible with respect to the
other components of the stress tensor.
(iii) The normal deflection w(α1 , α2 , t) does not depend on the coordinate z.
(iv) The tangential displacements are distributed across the thickness of the layer package according to the
generalized kinematic Timoshenko hypothesis,
(k)
u i (α1 , α2 , z) = u i (α1 , α2 ) + zθi (α1 , α2 ) + g(z)ψi (α1 , α2 ), (3)
z
where g(z) = 0 f 0 (x)dx.
In Eq. (3), ψi are required parameters characterizing the transverse shears in the shell. Hypothesis (iii)
permits to describe the nonlinear dependence of the tangential displacements on z; at g ≡ 0, it turns into the
linear Timoshenko hypothesis coinciding with the classical Kirchhoff–Love hypothesis since θi (i = 1, 2) are
functions of the tangential displacements and derivatives of the normal deflection.

2.2 Nonlinear equilibrium equations

Based on the above listed hypotheses Grigoluk and Kulikov [29] have derived the system of five nonlinear
differential equations with respect to the displacements u i , w, ψi (because of inconvenience, they are not
written down here). In paper [5], these equations have been generalized for the more common case when
E k , G k are functions of curvilinear coordinates and time. If vibrations or buckling occur with the formation
of a large number of waves or dents only in one direction at the shell surface, then these equations may be
essentially simplified.
We introduce the index of variation ι of stress–strain state as
   
 ∂Z   
ι = max {ι1 , ι2 } ,   ∼ h −ι1 Z ,  ∂ Z  ∼ h −ι2 Z (4)
∂α1  ∗  ∂α2  ∗
Buckling of thin cylindrical shells

where h ∗ = h/R is the dimensional thickness which is assumed as a small parameter, and Z is any unknown
function which determines this state. Here and below, the symbol ∼ means that two quantities have the same
asymptotic orders at h ∗ → 0. Depending on the value of ι and orders of all required functions in the governing
equations, one can deduce simplified equations corresponding to different stress–strain states of a shell. The
classification of the characteristic stress states of a thin single-layer isotropic shell considered within the
scope of the Kirchhoff–Love theory has been proposed by Gol’denveizer [41] and Novozhilov [42]. The same
classification can be made for the full system of differential equations derived by Grigoluk and Kulikov [29]
for laminated shells.
As follows from paper [35], the buckling mode of a radially pressurized medium-length thin laminated
cylinder is characterized by the one semi-wave along the shell generatrix and a large number of dents in the
circumferential direction with the indexes ι1 = 0, ι2 = 1/4 introduced by (4). The corresponding stress state
is called the main state [41], it does not reflect a real pattern of the buckling mode near the shell edges. In what
follows, we will take into account the edge effects with the index [41] ι1 = 1/2. Thus, we assume that the
total index of variation of all unknown functions is equal to ι ≤ 1/2. Let also the tangential displacements be
negligible quantities in comparison with the normal displacements, u i  w. Then, when introducing functions
ψi appearing in (3) by

∂a ∂φ ∂a ∂φ
ψ1 = + , ψ2 = − (5)
∂α1 ∂α2 ∂α2 ∂α1

where a, φ are the shear functions defined from equations

η2 h 2
a=− Δχ (6)
η1 β

and

1 − ν h2
Δφ = φ, (7)
2 β

the following compact system of nonlinear differential equations with respect to the displacement and stress
functions, χ and F, may be derived [29]:
   
θ h2 ∂ 2 F ∂ 2w ∂2 F ∂ 2w ∂2 F 1 ∂ 2w
D 1− χ−
2
+2 + − = qn , (8)
β ∂α22 ∂α12 ∂α1 ∂α2 ∂α1 ∂α2 ∂α12 R ∂α22
 2 2

1 ∂ 2w ∂ w ∂ 2w ∂ 2w
F − Eh
2
+ − =0 (9)
R ∂α12 ∂α1 ∂α2 ∂α12 ∂α22

where

Eh 3
D= η3 (10)
12(1 − ν 2 )

is the effective (or reduced) bending stiffness of the laminated cylindrical shell.
The displacement function χ and the normal displacement w are coupled by the equation
 
h2
w = 1− Δ χ (11)
β

and the stress function F permits to find the specific membrane stress resultants

∂2 F
Ti j = δi j Δ F − (12)
∂αi ∂α j
 G. Mikhasev, M. Botogova

where Δ = ∂ 2 /∂α12 + ∂ 2 /∂α22 is the Laplace operator in the curvilinear coordinates α1 , α2 , and δi j is Kro-
necker’s symbol (δii = 1; δi j = 0, i
= j). The reduced membrane stress resultants for the sandwich are
expressed in the standard way as

N δk

Ti j = σi j dz. (13)
k=1δ
k−1

In Eqs. (6)–(11), E, ν are the reduced modulus of elasticity and Poisson’s ratio, respectively, and parameters
ηi , θ, β characterize the reduced shear stiffness of the multilayered shell, β being the principle shear parameter
taking into account the transverse shears integrally over the shell thickness (see [43]). These parameters are
calculated as follows:

 −1
N
E k h k νk N
Ek hk 1 − ν 2  Ek hk
N
ν= , E= ,
1 − νk2 1 − νk2 h 1 − νk2
k=1 k=1  2
k=1
N λ2ko
12(1 − ν 2 ) k=1 λk − λkk N
λ2k0
β= q44 , q44 =   + Gk ,
Ehη1 N λ2ko −1 λkk
k=1 λk − λkk G k
k=1

δk δk
λk = f 02 (z)dz, λkn = f k (z) f n (z)dz (n = 0, k), θ = 1 − η22 /(η1 η3 ),
δk−1 δk−1
N 
N
η1 = ξk−1 π1k γk − 3c12
2
, η2 = ξk−1 π2k γk − 3c12 c13 ,
k=1 k=1 (14)
N
η3 = 4 (ξk2 + 3ζk−1 ζk )γk − 3c13
2
, hξk = h k , hζn = δn (n = 0, k),
k=1
δk δk
1 2 1 2
h π1k = g (z) dz,
2
h π2k = z g(z) dz,
12 12
δk−1 δk−1
δk 
N 
N
1 2
h π3k = g(z) dz, c13 = (ζk−1 + ζk )γk , c12 = ξk−1 π3k γk ,
12
δk−1 k=1 k=1
 −1
Ek hk N
Ek hk
γk = .
1 − νk2 k=1
1 − νk2

The accuracy of the ESL model represented by Eqs. (6)–(11) depends on many factors. In particular, the
stiffness characteristics of all layers composing a thin-walled multilayered structure have to be approximately
of the same order. One of the principle parameters affecting the error of the ESL model is the dimensionless
stiffness γk introduced in (14). To minimize the total error, the geometrical and physical parameters of layers
should be chosen in such a manner that parameters γk were values of the same order for all k = 1, 2, . . . , N ,
where N is a number of layers. This condition becomes essential for shells assembled from elastic and more
soft viscoelastic layers.
We shall introduce also a parameter G = q44 / h which will be called here the reduced or effective shear
modulus for the entire package of the laminated shell. This magnitude will play the determinant part in our
study. When G → ∞ or 1/β → 0, Eqs. (8), (9) degenerate into the well-known nonlinear equations based
on the classical Kirchhoff–Love hypothesis [44]. For 1/R = 0, Eq. (8) becomes the plate nonlinear bending
equation [45,46].
Buckling of thin cylindrical shells

2.3 Boundary conditions

Consider the group of boundary conditions corresponding to the simply supported edges α1 = 0, α1 = L.
These conditions are [29]
w = M̂11 = L̂ 11 = ψ2 = 0 (15)
for the edge with a diaphragm preventing transverse shears in the edge plane, and
w = M̂11 = L̂ 11 = L̂ 12 = 0 (16)
for the edge without a diaphragm. Here, M̂11 is the generalized stress couple, and L̂ i j are the high-order
generalized stress couples defined by
Eh 3
M̂11 = 12(1−ν 2 )
(η3 Ξ κ11 + η2 Ξ ψ11 )
Eh 3
  (17)
[3ex] L̂ i j = 12(1−ν 2 ) η2 Ξ κi j + η1 Ξ ψi j
where
 
1 ∂ψi ∂ψ j ∂ 2w
ψi j = + , κi j = − , (18)
2 ∂α j ∂αi ∂αi α j
and the operator Ξ is introduced as
Ξ xi j = (1 − ν)xi j + νδi j (x11 + x22 ). (19)

If the edge is clamped, then the second and third equations for M̂11 and L̂ 11 in Eqs. (15) and (16) should
∂w
be replaced by ∂α 1
= 0 and ψ1 = 0, respectively.
The above listed boundary conditions are directly following from the variational principle [29], and, hereby,
the natural boundary conditions. The analysis of available literature shows that the boundary conditions for
the simply supported edges without diaphragms have never been considered before. As a rule, assuming the
simple support or clamping of edges, authors constrained shears in the edge planes.
Each variant of boundary conditions, (15) or (16), is incomplete because they do not contain conditions for
the tangential displacements and/or membrane stress resultants. For instance, the conditions of free support,
T11 = ê22 = 0, where ê22 is the normal component of the strain tensor in the α2 -direction, results in the
additional conditions for the stress function [29]:
F = F = 0. (20)
If the edges are free in both the axial and circumferential directions, then conditions (20) are substituted for
the following equations:
∂2 F
F = 0, = 0. (21)
∂α1 α2
The boundary conditions (15) and (16) may be rewritten in terms of the displacement and shear functions,
χ and φ. For the simply supported edge with the diaphragm, they are the following [29]:
∂φ
χ = Δχ = Δ2 χ = = 0, (22)
∂α1
and for the edge without a diaphragm, they become more complicated:
   
h2 ∂2 h2
1 − Δ χ = 0, 1 − Δ χ = 0,
β ∂α12 β
 
∂2 ∂2 ∂ 2φ
+ ν 2 χ − (1 − ν) = 0, (23)
∂α12 ∂α2 ∂α1 α2
∂ 2χ ∂ 2φ ∂ 2φ
2 + 2 − 2 = 0.
∂α1 ∂α2 ∂α1 ∂α2
 G. Mikhasev, M. Botogova

As seen, the boundary conditions (22) and (23) for the required functions χ and φ differ essentially. In
what follows, conditions (22) and (23), with appropriate conditions for F, will be called as the SSD and SSF
boundary conditions, respectively.

2.4 Buckling equations

Let F ◦ , χ ◦ , φ ◦ be the functions describing the initial (pre-buckling) stress state of the shell. Following Euler,
we consider the adjacent stress state which is infinitesimally close to the pre-buckling one and characterized
by required functions

F ◦ + F, χ ◦ + χ , φ ◦ + φ. (24)

If the external pressure qn is uniform, then the initial displacement w ◦ , displacement function χ ◦ , and
the membrane stress resultants Ti◦j characterizing this state are weakly varying functions of the curvilinear
coordinates α1 , α2 . Furthermore, the pre-buckling normal deflection w ◦ is small so that the neutral surface
before and after deformation may be identified [47]. Let us substitute functions (24) into the nonlinear Eqs. (8),
(9) and perform linearization in the neighborhood of the stress state characterized by the functions F ◦ , χ ◦ , φ ◦ .
Then, omitting terms with the pre-buckling functions w ◦ , χ ◦ , one obtains the following homogeneous buckling
equations:
 
θ h2 1 ∂2 F
D 1− Δ Δ2 χ + − ΔT w = 0,
β R2 ∂α12
  (25)
Eh ∂ 2 w h2
Δ F=
2
, w = 1 − Δ χ,
R2 ∂α12 β
1 − ν h2
Δφ = φ (26)
2 β
where

◦ ∂ 2w ◦ ∂ w
2
◦ ∂ w
2
ΔT w = T11 + 2T12 + T22 , (27)
∂α12 ∂α1 ∂α2 ∂α22

and the membrane stress resultants Ti0j are found from equations of the membrane shell theory. In our case
[29],
◦ ◦ ◦
T22 = Rqn , T11 = T12 = 0. (28)

The problem is reduced to an eigenvalue problem for Eqs. (25), (26) with appropriate boundary conditions
◦ | for which this problem has a nontrivial solution.
listed above, which is to find the least |T22
−1
At β → 0, one can set φ = 0, and Eq. (25) degenerate into the well-known buckling equations of the
technical theory of thin isotropic single-layer shells which are based on the original Kirchhoff–Love hypotheses
and were widely utilized by many researchers for investigation of the enormous number of problems. It should
be noted that Eqs. (25), (26) give additional integrals of the shear edge effects which are absent in the classical,
Kirchhoff–Love hypotheses-based, theory.
The buckling equations (25), (26) were obtained after significant simplifications which introduced the error
∗ . As shown by Gol’denveizer [41] and Koiter [48], the index of variation ι of an expected solution
of order h 2ι
may give the conclusive contribution in the estimation of an error. If ι < 1, then the total error of Eqs. (25),
(26) will be estimated as
 
δe ∼ max h 2ι ∗ , h∗
2−2ι
. (29)

The peculiarity of Eqs. (25), (26) is that due to shears they have solutions with very high index of variation.
So, for an isotropic and homogeneous shell with Young’s and shear moduli E, G having the same asymptotic
order (E ∼ G), additional integrals accounting for shears have the index of variation ι = 1 [49]. Then δe ∼ 1,
and Eqs. (25), (26) become asymptotically incorrect.
Buckling of thin cylindrical shells

Remark Let
ζ
G ∼ h ∗G E (30)
where ζG > 0. We note that Eqs. (25), (26) were derived under the assumptions that ι ≤ 1/2, ζG ≥ 1. On the
other hand, the additional integrals accounting for shears have the index of variation [49] ι = 1 − ζG /2 < 1.
Then ζG < 2. Hence, one has the inequality 1 ≤ ζG < 2 restricting the order of the reduced shear modulus G.
In what follows, we will show that for a medium-length cylindrical shell with a low reduced shear modulus
(at ζG = 3/2) the pressure induced buckling occurs with the formation of a large number of dents in the
circumferential direction with the index of variation ι = 1/4. Then the error of the expected solution becomes
1/2
δe ∼ h ∗ .
Let us study buckling of the shell using Eqs. (25), (26) for the two variants of boundary conditions, the
SSD and SSF conditions.

3 Buckling of the shell with SSD boundary conditions

For a start, we consider the shell with simply supported edges having the diaphragms. The corresponding
SSD boundary conditions at α1 = 0, L are specified by Eqs. (20), (22). This is the unique case when the
boundary-value problem (20), (22), (25), (26) permits a solution in explicit form. Indeed, Eq. (25) for χ and
F are not coupled with Eq. (26) for φ, and the boundary condition (20) for the function φ is independent of
the residual conditions. It is obvious that the unique solution of Eq. (26) satisfying (22) is the trivial solution
φ = 0. That means that the diaphragms at both simply supported edges prevent appearing of the edge shears.
The residual required functions χ , F satisfying the SSD boundary conditions (20), (22) are readily found
as
πnα1 mα2 πnα1 mα2
χ = χ0 sin sin , F = F0 sin sin (31)
L R L R
where n, m are positive integers. The substitution of Eqs. (31) into Eqs. (25) yields the following equation for
the hoop stress resultant:

◦ ε8 π 4 h EΔnm h 2 η3
T22 =− , ε8 = ,
m 2 12(1 − ν 2 )R 2
 
1 + θ K δnm n4 π 2h2
Δnm = δnm
2
+ 4 4 8 2 , K = , (32)
1 + K δnm l π ε δnm β R2
 2 
n m2 L
δnm = + , l=
l2 π2 R
where K is the dimensionless shear parameter, and ε is a small parameter characterizing the shell thickness.
◦ over integers n and m results in the critical value for pressure
The minimization of T22
qn∗ = T2∗ /R, T2∗ = min |T22
◦ ◦
(n, m)| = |T22 (n ∗ , m ∗ )|. (33)
n,m

For a single-layer thin isotropic cylinder, one has the following relations and estimates [29,47]:

η3 = 1, θ = 1/85, n ∗ = 1, m ∗ ∼ (R/ h)1/4 , n ∗ /l  m ∗ /π. (34)

When omitting the transverse shears (at K = 0) and assuming that π 2 /l 2 and 1 can be neglected compared to
(m ∗ )2 , then Eq. (33) degenerates into the well-known Southwell–Papkovich formula [50]
 1/2
0.856E h 5
qn∗ = . (35)
(1 − ν 2 )3/4 L 2 R 3

In Eq. (33), the principal mechanical characteristics influencing the buckling pressure are the reduced modulus
of elasticity E and the reduced shear parameter K . However, the effect of these parameters on the critical
pressure is also different and depends strongly upon the correlation of the geometrical and physical parameters
 G. Mikhasev, M. Botogova

Fig. 3 Load parameter P ∗ × 108 versus shear parameter K at fixed θ = 0.05, l = 2 and different values of parameter ε:
1 − ε = 0.1; 2 − ε = 0.13; 3 − ε = 0.15

Fig. 4 Load parameter P ∗ × 108 versus shear parameter K at fixed θ = 0.05, ε = 0.1 and different values of dimensionless
length l: 1 − l = 4; 2 − l = 2; 3 − l = 1

of layers composing the shell as well as on a number of waves. For instance, if n, m ∼ 1, and also R/L ∼ 1,
then the influence of K is negligibly small, but on the other hand the reduced modulus E will be the main
−1/4
parameter. But if we study buckling of a very thin medium-length cylinder (when n ∼ 1, m ∼ h ∗ [47]),
then the parameter K becomes main, and the influence of the reduced parameter E decreases.
Specifying neither a number of layers nor materials, we will calculate the dimensionless critical load
parameter

T2∗
P∗ = (36)
π 4 Eh

for different values of the shear parameter K . Figure 3 shows the load parameter P ∗ versus K at the fixed
θ = 0.05, l = 2, and different values of ε. It may be seen that taking into account shears results in decreasing
the critical buckling pressure. At that, the drop in the critical buckling pressure turns out to be more noticeable
for very thin shells. And Fig. 4 demonstrates the effect of K on the buckling parameter P ∗ at the fixed θ = 0.05,
ε = 0.1 and different values of the dimensionless length l = L/R. As expected, the increase in the shell length
reduces the effect of the shear parameter on the buckling pressure. Indeed, a lengthy cylindrical shell under
lateral pressure buckles with formation of one semi-wave in the axial direction and a small number of waves
in the circumferential direction. But as follows from Eq. (32), the influence of K on the buckling pressure
becomes negligibly small at n ∗ , m ∗ ∼ 1.
We note that the reducing effect of shears illustrated in Figs. 3, 4 is not associated with the boundary
conditions, it reflects introducing shears (additional degrees of freedom) into the shell model.
Buckling of thin cylindrical shells

4 Buckling of the shell with SSF boundary conditions

Let simply supported edges be free of diaphragms. In this case the boundary-value problem (21), (23), (25),
(26) does not admit the explicit form of a solution.
To estimate the effect of the SSF boundary conditions (21), (23) on the critical buckling pressure we will
apply the asymptotic approach. This effect depends on the correlation between the reduced Young’s and shear
moduli. As follows from the remark, Eqs. (25), (26) are asymptotically correct if G  E and 1 ≤ ζG < 2. The
case ζG = 1 corresponding to a sandwich with a moderately “soft” core has been considered in paper [51]. In
this case the effect of the SSF boundary conditions is weak. In particular, the account of the edge shears gives
a correction to the classical critical pressure (35), which does not exceed a value of order O(ε2 ).
Let us consider the case when some of the layers are made of a sufficiently "soft" material (ζG > 1). If
ζG > 3/2, then the error of an expected solution becomes very large, δe ∼ h 2ι ∗ , where ι < 1/4. And for ζG
approaching 2, we obtain δe ∼ 1. Thus, the variant when 3/2 < ζG ≤ 2 is excluded from our consideration.
1/2
For the sake of definiteness, we assume that ζG = 3/2. Here, the error δe ∼ h ∗ . For instance, estimation (30)
at ζG = 3/2 holds when a sandwich shell is composed of the ABS-plastic and magnetorheological core
(treated as elastic materials) whose properties are affected by a magnetic field [43]. Estimation (30) implies
the following asymptotic correlation for the shear dimensionless parameter K :
K
= ε2 κ, κ ∼ 1 (37)
π2
where ε is a small parameter determined above.
Let us introduce dimensionless coordinates x, ϕ and a load parameter Λ,
◦
α1 = Rx, α2 = Rϕ, T22 = −ε6 EhΛ, (38)
where 0 ≤ x ≤ l = L/R.
It is known [47] that the buckling of a medium-length thin cylindrical shell under external pressure occurs
−1/4
with formation of a large number m of dents/bulges in the circumferential direction so that m ∼ h ∗ ∼ ε−1 .
Then the required functions χ , F, φ may be sought in the form
   
χ = R X (x) sin ε−1 pϕ , F = ε4 Eh R 2 Φ(x) sin ε−1 pϕ ,
  (39)
φ = R S(x) cos ε−1 pϕ
where p ∼ 1.
The substitution of Eqs. (37)–(39) into Eqs. (25), (26) results in the differential equations written in the
dimensionless form
d2 Φ
ε4 (1 − ε2 κθ Δε )Δ2ε X + − Λp 2 (1 − ε2 κΔε )X = 0,
dx 2
(40)
d2
ε 4
Δ2ε Φ − 2 (1 − ε2 κΔε )X = 0,
dx
1−ν
κ1 ε2 Δε S = S (41)
2
where
d2
Δε = − ε−2 p 2 (42)
dx 2
is the differential operator, and κ1 ≡ κ is introduced to analyze the effect of shears in the neighborhood of the
edges.
The SSF boundary conditions (21), (23) are rewritten as
d2
(1 − ε2 κ1 Δε )X = 0, 2
(1 − ε2 κ1 Δε )X = 0, (43.1,2)
  dx
d2 dS
ε2 2 − νp 2 X + ε(1 − ν) p = 0, (44)
dx dx
 G. Mikhasev, M. Botogova

dX d2 S
2εp + ε2 2 + p 2 S = 0, (45)
dx dx

dΦ
Φ = 0, = 0. (46.1,2)
dx

The boundary-value problem (40)–(46.1,2) is a singularly perturbed one. Its solution may be presented in
the form of the superposition of the main stress–strain state and the integrals of the edge effects [41]

X = X (m) + X (e) , Φ = Φ (m) + Φ (e) (47)

where the superscripts (m) and (e) denote functions corresponding to the main stress state and the edge effect
solutions, respectively. Here, the main state is the semi-momentless one having the indexes of variation ι1 = 0
and ι2 = 1/4 in the axial and circumferential direction, respectively [41], and the edge effect solutions have
a large index of variation in the axial direction. Contrary to the classical Kirchhoff–Love theory, our problem
stated in terms of the displacement and shear functions, X (e) and S, has six edge integrals for X (e) and two
edge integrals for the shear function S.

4.1 Edge effect integrals

At first, we consider Eq. (41). It has the following general solution:

   
ϑs x ϑs (l − x)
S = εγ0 a1 exp − + a2 exp − (48)
ε ε
where a1 , a2 are unknown constants, γ0 is the index of intensity of the shear function, and

2
ϑs = + p2 . (49)
(1 − ν)κ1

The edge effect integrals X (e) may be found from the edge effect equations which are derived from the full
system of differential equations in terms of u i , ψi , and w. Another way is to obtain their asymptotic estimations
directly from Eqs. (40). Let

X (e) (x) = X̂ (e) eλx , Φ (e) (x) = Φ̂ (e) eλx . (50)

The substitution of Eqs. (50) into Eqs. (40) results in the characteristic equation

[1 − κ1 θ (ε2 λ2 − p 2 )](ε2 λ2 − p 2 )4 + λ4 [1 − κ1 (ε2 λ2 − p 2 )]

(51)
−Λp 2 (ε2 λ2 − p 2 )2 [1 − κ1 (ε2 λ2 − p 2 )] = 0
which has only the six roots

1 1
λ1,2 =± + p 2 + O(ε3 ), (52)
ε κ1

1 4 1 √
λ3,4,5,6 =± 2 (1 ± i) + O(1), i = −1 (53)
ε 4θ
with nonzero real parts, and the remaining four roots with zero real parts are not written down here.
The corresponding partial solutions of Eqs. (40) form two groups of functions:
r1 r1
(e) (e)
X 1 (x; ε) = e− ε x
[1 + O(ε)], X 2 (x; ε) = e− ε (l−x)
[1 + O(ε)],
1−θ (54)
(e) (e)
Φ1, 2 = −ε2 X ,
κ1 (1 + κ1 p 2 ) 1,2
Buckling of thin cylindrical shells

and
(e) −
r2  
cos ε−2 r2 x [1 + O(1)],
x
X 3 (x; ε) = e ε2

(e) −
r2  
sin ε−2 r2 x [1 + O(1)],
x
X 4 (x; ε) = e ε2

(e) −
r2
(l−x)  
X 5 (x; ε) = e ε2 cos ε−2 r2 (l − x) [1 + O(1)], (55)

(e) −
r2
(l−x)  
X 6 (x; ε) = e ε2 sin ε−2 r2 (l − x) [1 + O(1)],
(e) κ1 (e)
Φj = X , j = 3, 4, 5, 6
ε2 j
with the properties of the edge effect integrals where
 
1 4 1
r1 = + p , r2 =
2 . (56)
κ1 θ

It may be seen that functions (55) have the index of variation ι1 = 1/2, that is the same as in the classical
simple edge effect integrals [41], whereas functions (56) have the index ι1 = 1/4.
We compose the following superposition of the found integrals:


2
(e)

6
(e)
(e) γ1 γ2
X =ε bi X i +ε cj X j (57)
i=1 j=3

where bi , c j are constants, and γ1 , γ2 are the indexes of intensity of the edge effect integrals which remain
unknown at this step.

4.2 The main stress–strain state

The unknown functions X (m) , Φ (m) corresponding to the main stress–strain state and the eigenvalue Λ are
sought in the form of formal asymptotic series

X (m) = X 0 + ε X 1 + · · · , Φ (m) = Φ0 + εΦ1 + · · · , (58)

Λ = Λ0 + εΛ1 + · · · . (59)
We substitute Eqs. (58), (59) into Eqs. (40) and consider the first two approximations.
In the zero-order approximation, one has the homogeneous differential equation

d4 X 0 p 6 [ p 2 + θ κ p 4 − Λ0 (1 + κ p 2 )]
LX 0 ≡ + X0 = 0 (60)
dx 4 1 + κ p2
with respect to X 0 . And the next approximation produces the non-homogeneous equation

LX 1 = Λ1 p 6 X 0 . (61)
The stress and displacement functions are coupled by the formula

1 + κ p 2 d2 X j
Φj = , j = 1, 2. (62)
p4 dx 2
Equations (60), (61) have the fourth order. So, one needs to split the boundary conditions (43.1,2–46.1,2)
and assign the main two conditions for X j at each edge and the additional ones which will serve to determine
constants ai , bi , c j , and parameters γ0 , γ1 , γ2 as well. To this purpose, we substitute Eqs. (47), (57), (58) into
the boundary conditions (43.1,2–46.1,2), and, taking into account the indexes of variation of all functions (we
remind that dX 0 /dx ∼ X, dΦ0 /dx ∼ Φ0 ), demand fulfillment of the following conditions:
 G. Mikhasev, M. Botogova

− the boundary conditions for X 0 , Φ0 should be homogeneous;

− at each edge, there is an inhomogeneous condition coupling ai and X 0 or its derivatives;
− at each edge, there is an inhomogeneous equation for bi ;
− it is desirable to get even if one inhomogeneous equation for constants c j ;
− the boundary conditions for X 1 , Φ1 should be inhomogeneous and expressed in terms of ai , bi , c j .
When taking into account the above conditions, Eqs. (43.1) and (46.1) in the zero-order approximation
result in the main boundary conditions
X 0 = Φ0 = 0 at x = 0, l. (63)
The next approximation allows to determine parameters γ0 = γ1 = 1, γ2 = 3 and generates the main
conditions
(1 + κ1 p 2 )X 1 (0) − κ1r12 b1 = 0, (1 + κ1 p 2 )X 1 (l) − κ1r12 b2 = 0,
(64)
Φ1 (0) = 0, Φ1 (l) = 0
for X 1 , Φ1 , and the following additional equations for ai , bi , c j :

c j = 0, for j = 3, 4, 5, 6,

−νp 2 X 1 (0) + r12 b1 − (1 − ν) pϑs a1 = 0,

(65)
−νp 2 X 1 (l) + r12 b2 − (1 − ν) pϑs a2 = 0,
dX 0 (0) dX 0 (l)
2p + (ϑs2 + p 2 )a1 = 0, 2 p + (ϑs2 + p 2 )a2 = 0.
dx dx

The zero-order approximation Consider the boundary-value problem (60), (63) arising in the zero-order
approximation. It should be noted that it is the same within the group of the boundary conditions for simply
supported edges and does not depend on whether an edge has a diaphragm (SSD conditions) or not (SSF
conditions). This problem has the solution
X 0 = A sin(πnx/l) (66)
if
π 4n4 p 2 (1 + θ κ p 2 )
Λ0 ( p; n) = + (67)
l 4 p6 1 + κ p2
where n is a number of semi-waves in the axial direction of the shell. Minimizing the function Λ0 ( p) over p
and n, one obtains the zero-order approximation of the critical buckling load parameter,
Λ◦0 = min Λ0 ( p, n) = min Λ0 ( p, 1) = Λ0 ( p ◦ , 1), (68)
p,n p

and the corresponding eigenfunction

X 0 = A sin(π x/l). (69)
For κ = 0, one obtains

3π 4 4π
p◦ = , Λ◦0 = 3/4 ,
8
4
(70)
l 3 l
and the remaining equation in (38) results again in the known Southwell–Papkovich formula for the critical
value of the hoop stress resultants.

The first-order approximation Consider the non-homogeneous boundary-value problem arising in the first-
order approximation. From Eqs. (65),(69), (70), one obtains
Buckling of thin cylindrical shells

2π p ◦ A
a1 = −  ◦ 2  , a2 = −a1 ,
l ( p ) + (ϑs◦ )2
(71)
2π(1 − ν)ϑs◦ ( p ◦ )2 κ1 A
b1 = b2 = −   ,
l 1 + (1 − ν)( p ◦ )2 κ1 ( p ◦ )2 + (ϑs◦ )2
and the boundary conditions for Eq. (61) read

2π(1 − ν)κ1 ϑs◦ ( p ◦ )2 A

X 1 (0) = X 1 (l) = −   ,
l 1 + (1 − ν)( p ◦ )2 κ1 ( p ◦ )2 + (ϑs◦ )2 (72)
Φ1 (0) = 0, Φ1 (l) = 0

where ϑs◦ = ϑs ( p ◦ ). We have the non-homogeneous boundary-value problem (61), (72) on the “spectrum.”
The existence condition of a solution of this problem is written as

l
◦ 6
Λ1 ( p ) X 02 dx = X 0 (0)X 1 (0) − X 0 (l)X 1 (l). (73)
0

Hence, one obtains the formula for correction of the critical buckling parameter,

16π 4 (1 − ν)κ1 ϑs◦

Λ◦1 =   . (74)
l 5 ( p ◦ )4 1 + (1 − ν)( p ◦ )2 κ1 ( p ◦ )2 + (ϑs◦ )2
Then,
 
2π(1 − ν)κ1 ϑs◦ ( p ◦ )2 A 2x πx
X1 = −    1− cos . (75)
l 1 + (1 − ν)( p ◦ )2 κ1 ( p ◦ )2 + (ϑs◦ )2 l l

If the edge x = 0 has the diaphragm and the edge x = l has not, then c j = 0 for j = 3, 4, 5, 6, and
parameters a1 , b1 and X 1 (0) are defined by Eqs. (71), (72), but a2 = b2 = X 1 (l) = 0. Then the correction of
the critical buckling parameter becomes half the value determined by (74),

8π 4 (1 − ν)κ1 ϑs◦
Λ◦1 =   . (76)
l 5 ( p ◦ )4 1 + (1 − ν)( p ◦ )2 κ1 ( p ◦ )2 + (ϑs◦ )2

4.3 The critical buckling pressure and eigenform

Finally, we obtain the following equations for the critical buckling pressure for the simply supported shells
without diaphragms on the edges:

ε6 Eh ◦   Λ◦
qn◦ = − Λ , Λ◦ = Λ◦0 1 + εks + O(ε2 ) , ks = ◦1 (77)
R Λ0
where ks is the normalized correction depending on the shear parameter κ1 ≡ κ and taking into account shears
in the vicinity of the shell edges. These edge shears appear as a result of the absence of the edge diaphragms.
Indeed, assuming κ1 = 0, we ignore the edge effect equation (26). Then ks = 0, and Eqs. (77) give an
approximate value of the critical buckling pressure for the simply supported shells with diaphragms on all
edges. Note, that parameters p ◦ , Λ◦0 are also influenced by the shear parameter κ, but this effect is generated
by shears in the shell but not by boundary conditions.
The approximate formula for the buckling mode will be as follows:
  
  πx 2x πx
χ ≈ R sin ε−1 p ◦ ϕ sin − ε a0 1 − cos
l l l
 r◦x r ◦ (l−x) 
+ b1 e− ε − e− ε
1 1
(78)
 G. Mikhasev, M. Botogova

Fig. 5 Wave parameter p ◦ versus shear parameter κ at l = 2.5 and different θ: 1 − θ = 0.005; 2 − θ = 0.025; 3 − θ = 0.05

where


2π(1 − ν)κ1 ϑs◦ ( p ◦ )2 A 1
a0 =    , r1◦ = + ( p ◦ )2 , (79)
l 1 + (1 − ν)( p ◦ )2 κ1 ( p ◦ )2 + (ϑs◦ )2 κ1

and b1 is calculated by (71).

As can be seen, the edge integrals (55) with the index of variation ι1 = 1/2 do not make a contribution
in the first-order approximation. Their effect may be estimated by considering the higher approximations.
However, the accuracy of Eqs. (25) is not sufficient to determine a correction ε2 Λ2 . For this purpose, the full
system of nonlinear differential equations written in terms of the displacements u i , w, ψi should be used. It
is interesting to note that the construction of the second-order approximation for a thin single-layer simply
supported cylindrical shell being considered within the scope of the Kirchhoff–Love theory also results in the
zeroth coefficients c j in Eq. (57). Whereas for other variants of boundary conditions (particularly, for the case
of clamped edges), the edge effect integrals like (55) give a non-zeroth correction ε2 Λ2 [47].

4.4 Effect of edge shears

Equation (77) allows to study the influence of the transverse shears on the critical buckling pressure without
specification of a number of layers and their mechanical properties.
Figures 5, 6, and 7 show the effect of the shear parameter κ on the parameters p ◦ , Λ◦0 , and ks at l = 2.5
and different values of a parameter θ . It is seen that at small values of κ (less than 0.25) a parameter θ does
not affect the required dimensionless magnitudes p ◦ , Λ◦0 , and ks , and when increasing the shear parameter κ
this influence becomes considerable. The increase of κ results in the increase of the wave parameter p ◦ and
the decrease in the zero approximation Λ◦0 of the critical buckling parameter Λ∗ . The effect of parameters κ
and θ on the normalized correction ks turns out to be more complicated: for small θ (here θ = 0.005) the
correction ks growths together with κ, and, approaching a maximum value at κ ≈ 0.52, begins to fall, but
at θ > 0.025 it is a monotonically increasing function of κ. Clearly, this effect depends on the correlation
between the geometrical and mechanical characteristics of laminas.
Figure 8 displays the normalized correction ks versus the shear parameter κ for different values of the
dimensionless length l. As expected, the shorter the cylinder is, the larger the effect of the edge shears on the
critical buckling pressure becomes. But the total impact of the edge shears on the critical pressure is not high.
So, calculations performed at ε = 0.1, κ = 2, θ = 0.05 show that the edge effect integrals generated by the
edge shears give the positive buckling pressure increments of about 2.4 and 3.4 % for the lengths l = 1.5 and
l = 1, respectively.
Buckling of thin cylindrical shells

Fig. 6 Critical buckling parameter Λ◦0 versus shear parameter κ at l = 2.5 and different θ: 1 − θ = 0.005; 2 − θ = 0.025;
3 − θ = 0.05

Fig. 7 Normalized correction ks versus shear parameter κ at l = 2.5 and different θ: 1 − θ = 0.005; 2 − θ = 0.025; 3 − θ = 0.05

Fig. 8 Normalized correction ks versus shear parameter κ at θ = 0.05 and different values of dimensionless length l: 1 − l = 2.5;
2 − l = 1.5; 3 − l = 1

5 Examples

We shall consider two examples illustrating the effect of the edge shears on the buckling pressure for sandwich
and multilayered cylindrical shells having a soft core or layers with a low shear modulus.
Example 1 In the first example, we shall study the buckling of a sandwich thin cylinder of the radius R = 0.5 m
and length L = 0.5 m with a core made of a magnetorheological elastomer (MRE) and skins made of the
ABS-plastic SD-0170. Both materials are assumed here as elastic and isotropic with properties:

E 1 = E 3 = 1.5 · 109 Pa, ν1 = ν3 = 0.4 for the ABS-plastic,

(80)
E 2 = (13, 230 + 45, 040B) kPa, ν2 = 0.4 for the MRE [52]
 G. Mikhasev, M. Botogova

Table 1 Wave numbers m ∗ , m ◦ , wave parameter p ◦ , critical buckling pressures qn∗ , q0◦ , qn◦ (Pa) for the sandwich with the MRE
core of the thickness h 2 = 11 mm for two variants of boundary conditions (SSD, SSF) versus magnetic induction B (mT)

B m∗ qn∗ p◦ m◦ q0◦ qn◦ δ (%) δ  (%)

0 9 11,714 2.78 8 10,246 10,721 +4.64 −8.48
20 7 16,883 2.45 7 13,986 14,652 +4.76 −13.21
40 7 19,905 2.32 7 16,789 17,531 +4.40 −14.00
60 7 22,102 2.25 7 17,556 18,121 +3.22 −18.01
80 6 23,681 2.21 7 18,551 19,042 +2.65 −19.59
100 6 24,705 2.18 7 19,229 19,727 +2.60 −20.00
The edge shears induced corrections δ, δ  (%) for the critical buckling pressure versus magnetic induction B

Table 2 Wave numbers m ∗ , m ◦ , wave parameter p ◦ , critical buckling pressures qn∗ , q0◦ , qn◦ (Pa) for the sandwich with the MRE
core of the thickness h 2 = 12 mm for two variants of boundary conditions (SSD, SSF) versus magnetic induction B (mT)

B m∗ qn∗ p◦ m◦ q0◦ qn◦ δ (%) δ  (%)

0 10 10,872 3.00 9 9365 9549 +1.96 −12.17
20 8 17,545 2.53 7 14,561 14,929 +2.53 −14.91
40 7 21,593 2.37 7 16,985 17,360 +2.21 −19.60
60 7 24,420 2.28 7 19,333 19,629 +1.53 −19.62
80 6 26,355 2.23 7 20,320 20,620 +1.48 −21.76
100 6 27,684 2.20 6 21,344 21,607 +1.23 −21.95
The edge shears induced corrections δ, δ  (%) for the critical buckling pressure versus magnetic induction B

Table 3 Wave numbers m ∗ , m ◦ , wave parameter p ◦ , critical buckling pressures qn∗ , q0◦ , qn◦ (Pa) for the sandwich with the MRE
core of the thickness h 2 = 13 mm for two variants of boundary conditions (SSD, SSF) versus magnetic induction B (mT)

B m∗ qn∗ p◦ m◦ q0◦ qn◦ δ (%) δ  (%)

0 12 9970 3.25 9 9070 9354 +3.13 −6.18
20 8 18,196 2.60 7 15,352 16,145 +5.17 −11.27
40 7 23,282 2.40 7 18,905 19,792 +4.69 −14.99
60 7 26,782 2.30 7 21,144 21,970 +3.91 −17.97
80 6 29,124 2.25 6 22,691 23,426 +3.24 −19.56
100 6 30,790 2.21 6 23,836 24,483 +2.71 −20.48
The edge shears induced corrections δ, δ  (%) for the critical buckling pressure versus magnetic induction B

where B is the induction of a magnetic field. The skins have the same fixed thickness h 1 = h 3 = 0.5 mm,
and a thickness h 2 of the soft MRE core will be varied. The application of an external magnetic field leads to
changing the mechanical properties of the core and the whole sandwich as well. It is evident that the viscous
and rheological properties of the MRE are not taken here into consideration. Tables 1, 2, and 3 demonstrate
the dependence of the wave numbers m ∗ , m ◦ , parameter p ◦ , and the critical buckling pressures qn∗ , q0◦ , qn◦
on the magnetic induction B for the sandwiches with two variants of the boundary conditions (SSD and
SSF conditions) and the MRE core thicknesses h 2 = 11, 12, 13 mm, respectively. Here, parameters with the
superscripts ∗ and ◦ correspond to the sandwiches with and without the edge diaphragms, respectively; the
wave number m ◦ is defined as the integer part of ε−1 p ◦ , and q0◦ is the zeroth approximation of the critical
buckling pressure for the SSF sandwich determined by Eq. (77) at ks = 0. Tables 1, 3 also show the correction
q ◦ −q ◦
δ = nq ◦ 0 100% induced by the edge shears with respect to the zeroth approximation of the critical buckling
0
q ∗ −q ◦
pressure q0◦ for the shell with the SSF conditions, and δ  = nq ∗ n 100% is the divergence between the critical
n
buckling pressures qn∗ and qn◦ for the shells with the SSD and SSF conditions, respectively.
It may be seen that for any fixed values of the geometrical parameters, increasing the magnetic field
induction B results in decreasing the wave numbers m ∗ , m ◦ and the wave parameter p ◦ as well, increasing
the total stiffness and, as result, the buckling pressures qn∗ , qn◦ for the simply supported sandwiches with and
without diaphragms. The dependence of the critical buckling pressure on the thickness h 2 of the soft MRE core
is more complicated: at a low level of the applied magnetic field, or without it, the increase of B leads to the
drop of the sandwich stiffness and the critical buckling pressure, but at B ≥ 20 mT, the critical pressures qn∗ , qn◦
grow together with h 2 . It may be also concluded that the edge shears in the simply supported sandwich shells
without diaphragms have a weak supporting effect, the correction δ being maximum at about B = 20 mT.
When comparing the critical values of pressure for the sandwich shells with the SSD and SSF boundary
Buckling of thin cylindrical shells

Table 4 Wave numbers m ∗ , m ◦ , wave parameter p ◦ , critical buckling pressures qn∗ , q0◦ , qn◦ (kPa) and the edge shears induced
corrections δ, δ  for the 5-layered cylindrical shell for two variants of boundary conditions (SSD, SSF) versus thickness h 3 (mm)
of the alloy–foam core

h3 m∗ qn∗ p◦ m◦ q0◦ qn◦ δ (%) δ  (%)

20 7 659.94 2.17 8 551,394 565,134 +2.49 −14.37
25 7 793.93 2.25 7 663,738 689,078 +3.82 −13.21
30 7 913.67 2.32 7 765,320 802,965 +4.92 −12.12
35 8 1010.00 2.39 7 861,574 910,589 +5.69 −9.84
38 8 1070.00 2.44 7 916,356 971,556 +6.02 −9.02

conditions, the critical buckling pressure qn∗ for the shell with the diaphragms is always more than the critical
pressure qn◦ for the same shell but without diaphragms. It is also seen that the correction δ  grows together with
the induction B but demonstrates the nonlinear behavior as a function of the core thickness h 2 .
Example 2 As the second example, we consider the five-layered cylindrical shell of the radius R = 0.9 m and
the length L = 1.0 m assembled from different laminas which are assumed to be isotropic:
− the first (innermost) layer of the thickness h 1 = 0.5 mm is the ABS-plastic SD-0170 with the elastic
properties specified above;
− the fifth (outermost) layer of the thickness h 5 = 0.5 mm is made of silicon nitrate (ceramic, Si3 N4 ) with
the elastic moduli [53] E 5 = 3.484 · 1011 Pa, ν5 = 0.24;
− the second and fourth layers with the same thicknesses h 2 = h 4 = 3.0 mm are made of epoxy for which
E 2 = E 4 = 3450 Pa, ν2 = ν4 = 0.3;
− the third soft layer of the thickness h 3 is alloy–foam for which [3] E 3 = 4.59 · 107 Pa, ν3 = 0.33.
Table 4 shows the effect of different thicknesses of the soft alloy–foam core on the parameters m ∗ , m ◦ , p ◦ , and
the critical buckling pressures qn∗ , q0◦ , qn◦ for the SSD and SSF boundary conditions. As expected, increasing the
thickness h 3 of the alloy form core at fixed thicknesses of other layers increases the critical buckling pressures
qn∗ and qn◦ for both variants of the boundary conditions. This effect is explained by increasing the reduced
bending stiffness of the laminated shells. Clearly, these trends may be easily changed if one or more material
or geometrical parameters are changed. For example, increasing the volume fraction of the alloy–foam core
will have another effect on the effective bending stiffness and buckling pressure. However, the basic results
of this example concerns the influence of the soft core thickness on the edge shears induced correction. The
increase in the thickness h 3 leads to the reduction of the effective shear modulus G, and this results in growing
the transverse shears near the simply supported edge without a diaphragm; in turn, rising edge shears with
minor “supporting effect” give the growing positive correction δ for the zeroth approximation of the critical
buckling pressure. As in the first example, the correction δ  is always negative, that is the edge diaphragm
reinforces the laminated structure. However, in this example, the value of the correction δ  decreases with the
increase in the soft core thickness.

6 Conclusions

The equivalent single-layer model was used to predict buckling of a laminated medium-length thin cylindrical
shell under external normal pressure. The buckling equations were derived from the governing nonlinear
equations based on the generalized kinematic hypotheses of Timoshenko and accounting for the transverse
shears. It is assumed that one of laminas or several ones are made of a "soft" material with a low shear
modulus so that the reduced (effective) shear modulus for the entire package is much less than the reduced
Young’s modulus. One unique feature of this paper is that, unlike previous available studies, it are considered
two different variants of boundary conditions corresponding to the simply supported edges but differing in
conditions for the transverse shears in the edge plane: In the first variant of the boundary conditions (SSD
conditions), there is a diaphragm hindering the occurrence of shears in the neighborhood of the edge, and
in the second variant (SSF conditions), a diaphragm is absent. In the case of the SSD boundary conditions,
equations for the critical buckling pressure and buckling mode as well are found in the explicit form. When the
simply supported edges are free of diaphragms, solutions of the boundary-value problem are constructed by
using the asymptotic method; all required functions are sought in the form of the superpositions of functions
pertaining to the main stress–strain state and the edge effect integrals. It is shown that if the reduced shear
 G. Mikhasev, M. Botogova

modulus is noticeably smaller than the reduced Young’s modulus, one arises the transverse shears localized in
the vicinity of the simply supported edge free of a diaphragm. Due to localization, the distribution of shears
has the properties of the edge effect integrals with the decay rate smaller than that in the classical simple edge
effect integrals.
Regardless of a number of layers and their mechanical properties, the effect of shears on the critical
buckling pressure was studied. Increasing transverse shears results in increasing the number of waves (in the
circumferential direction) in the buckling mode and decreasing the zeroth approximation (not taking the edge
shears into account) of the critical buckling pressure corresponding to the main stress state. However, account
of the edge shears in the neighborhood of the simply supported edge without a diaphragm gives a positive
normalized correction for the buckling parameter and so increases slightly the critical buckling pressure. The
magnitude of this correction depends on the geometrical parameters of a shell: In particular, it expectedly
increases when the shell length decreases and/or a thickness (as seen from the examples) decreases.
As an illustration, the buckling of a sandwich cylindrical shell with a core made of MRE under different
levels of an applied magnetic field was considered. The example has shown that the application of MRE as of
a soft core allows to regulate the shear compliance by changing the magnetic induction. The application of a
magnetic field leads to the reduction of shears, including the edge ones, increases the shell stiffness and so the
critical buckling pressure. In the second example, the prediction of the buckling of the 5-layered cylindrical
shell assembled from different materials (ABS-plastic, ceramic, epoxy) and having the soft alloy–foam core
was made. In spite of the fact that the innermost and outermost layers were made of very stiff materials, the
insertion of a very soft core experiencing large shears resulted in the reduction of the effective shear modulus
for the laminated structure. However, increasing the soft core thickness allows to rise the bearing capacity
of laminated shells. The analysis of numerical calculations performed in both examples has shown that the
account of the edge diaphragm is obligatory in the buckling analysis of simply supported laminated shells with
low effective shear modulus. The use of stiff edge diaphragms preventing the edge shears permits to increase
the critical buckling pressure.
Finally, it should be noted that a similar formal buckling analysis for a thin laminated cylinder with clamped
edges revealed that the divergence between buckling pressures for shells with and without diaphragms was
very small. An appropriate asymptotic estimation for the buckling pressure, that would take into account
diaphragms at the clamped edge, cannot be made on the base of the governing equations (25), (26) whose
accuracy turns out to be insufficient for this.

Acknowledgements The research leading to these results has received support from the People Programme (Marie Curie
Actions) of the European Union’s Seventh Framework Programme FP7/2007-2013/under REA grant agreement PIRSES-GA-
2013-610547-TAMER. The authors also acknowledge the support from the State Program of Scientific Researches in Belarus
“Physical materials science, new materials and technologies” (Assignment N 3.4.01).

References

1. Librescu, L., Hause, T.: Recent developments in the modeling and behavior of advanced sandwich constructions: a survey.
Compos. Struct. 48, 1–17 (2000)
2. Kardomateas, G.A.: Elasticity solutions for a sandwich orthotropic cylindrical shell under external pressure, internal pressure
and axial force. AIAA J. 39, 713–7199 (2001)
3. Han, J.-H., Kardomateas, G.A., Simitses, G.J.: Elasticity, shell theory and finite element results for the buckling of long
sandwich cylindrical shells under external pressure. Compos. Part B 35, 591–598 (2004)
4. Xie, Y.J., Yan, H.G., Liu, Z.M.: Buckling optimization of hybrid-fiber multilayer-sandwich cylindrical shells under external
lateral pressure. Compos. Sci. Technol. 56, 1349–1353 (1996)
5. Mikhasev, G.I., Botogova, M.G., Korobko, E.V.: Theory of thin adaptive laminated shells based on magnetorheological
materials and its application in problems on vibration suppression. In: Altenbach, H., Eremeyev, V.A. (eds.) Shell-like
Structures. Advanced Structured Materials, vol. 15, pp. 727–750. Springer, Berlin (2011)
6. Kozlowska, J., Boczkowska, A., Czulak, A., Przybyszewski, B., Holeczek, K., Stanik, R., Gude, M.: Novel MRE/CFRP
sandwich structures for adaptive vibration control. Smart Mater. Struct. 25, 035025 (2016)
7. Yeh, J.-Y.: Vibration and damping analysis of orthotropic cylindrical shells with electrorheological core layer. Aerosp. Sci.
Technol. 15, 293–303 (2011)
8. Tanguy, M., Pyrzb, M., Ginestec, B., Chauchotd, P.: Optimal laminations of thin underwater composite cylindrical vessels.
Compos. Struct. 58, 529–537 (2002)
9. Ohga, M., Wijenayaka, A.S., Croll, J.G.A.: Reduced stiffness buckling of sandwich cylindrical shells under uniform external
pressure. Thin. Wall. Struct. 43, 1188–1201 (2005)
10. Malinowski, M., Belica, T., Magnucki, K.: Buckling and post-buckling behaviour of elastic seven-layered cylindrical shells—
FEM study. Thin. Wall. Struct. 94, 478–484 (2015)
Buckling of thin cylindrical shells

11. Kulikov, G.M., Plotnikova, S.V.: Advanced formulation for laminated composite shells: 3D stress analysis and rigid-body
motions. Compos. Struct. 95, 236–246 (2013)
12. Bolotin, V.V., Novichkov, Yu. N: Mechanics of Multilayer Structures. Mashinostroenie, Moscow (1980). [in Russian]
13. Reddy, J.N.: An evaluation of equivalent-single-layer and layerwise theories of composite laminates. Compos. Struct. 25,
21–35 (1993)
14. Reddy, J.N., Robbins, D.H.: Theories and computational models for composite laminates. Appl. Mech. Rev. 47, 147–165
(1994)
15. Altenbach, H.: Theories for laminated and sandwich plates. A review. Mech. Compos. Mater. 34, 243–252 (1998)
16. Carrera, E.: Multilayered shell theories accounting for layerwise mixed description. Part 1: governing equations. AIAA J.
37, 1107–1116 (1999)
17. Carrera, E.: Developments, ideas, and evaluations based upon Reissners mixed variational theorem in the modeling of
multilayered plates and shells. Appl. Mech. Rev. 54, 301–329 (2001)
18. Carrera, E.: Theories and finite elements for multilayered, anisotropic, composite plates and shells. Arch. Comput. Methods.
Eng. 9, 87–140 (2002)
19. Fiedler, L., Lacarbonara, W., Vestroni, F.: A generalized higher-order theory for multi-layered, shear-deformable composite
plates. Acta Mech. 209, 85–98 (2010)
20. Reddy, J.N., Arciniega, R.A.: Shear deformation plate and shell theories: from Stavsky to present. Mech. Adv. Mater. Struct.
11, 535–582 (2004)
21. Mantari, J.L., Oktem, A.S., Soares, C.G.: A new trigonometric shear deformation theory for isotropic, laminated composite
and sandwich plates. Int. J. Solids Struct. 49, 43–53 (2012)
22. Thai, C.N., Ferreira, A.J.M., Wahab, M.A., Nguyen-Xuan, H.: A generalized layerwise higher-order shear deformation theory
for laminated composite and sandwich plates based on isogeometric analysis. Acta Mech. 227, 1225–1250 (2016)
23. Reissner, E., Wan, F.Y.M.: On the equations of linear shallow shell theory. Stud. Appl. Math. 48, 133–145 (1969)
24. Whitney, J.M.: The effect of boundary conditions on the response of laminated composites. J. Compos. Mater. 4, 192–203
(1970)
25. Dong, S.B., Tso, F.K.W.: On a laminated orthotropic shell theory including transverse shear deformation. J. Appl. Mech. 39,
1091–1097 (1972)
26. Chou, P.C., Carleone, J.: Transverse shear in laminated plate theories. AIAA J. 11, 1333–1336 (1973)
27. Reissner, E.: A consistent treatment of transverse shear deformation in laminated anisotropic plates. AIAA J. 10, 716–718
(1972)
28. Reddy, J.N., Liu, C.F.: A higher-order shear deformation theory of laminated elastic shells. Int. J. Eng. Sci. 23, 319–330
(1985)
29. Grigolyuk, E.I., Kulikov, G.M.: Multilayer Reinforced Shells: Calculation of Pneumatic Tires. Mashinostroenie, Moscow
(1988). [in Russian]
30. Grigolyuk, E.I., Kulikov, G.M.: General direction of development of the theory of multilayered shells. Mech. Compos. Mater.
24, 231–241 (1988)
31. Kant, T., Swaminathan, K.: Analytical solutions for the static analysis of laminated composite and sandwich plates based on
a higher order refined theory. Compos. Struct. 56, 329–344 (2002)
32. Thakur, S.N., Ray, C., Chakraborty, S.: A new efficient higher-order shear deformation theory for a doubly curved laminated
composite shell. Acta Mech. (2016). doi:10.1007/s00707-016-1693-3
33. Toorani, M.H., Lakis, A.A.: General equations of anisotropic plates and shells including transverse shear deformations,
rotary inertia and initial curvature effects. J. Sound Vib. 237, 561–615 (2000)
34. Anastasiadis, J.S., Simitses, G.J.: Buckling of pressure-loaded, long, shear deformable, cylindrical laminated shells. Compos.
Struct. 23, 221–231 (1993)
35. Mikhasev, G.I., Seeger, F., Gabbert, U.: Comparison of analytical and numerical methods for the analysis of buckling and
vibrations of composite shell structures. In: (ed.) Proceedings of 5th Magdeburg Days of Mechanical Engineering. Otto-von-
Guericke-University Magdeburg, pp. 175–183. Logos, Berlin (2001)
36. Wu, Z., Cheung, Y.K., Lo, S.H., Chen, W.: Effects of higher-order globallocal shear deformations on bending, vibration and
buckling of multilayered plates. Compos. Struct. 82, 277–289 (2008)
37. Li, Z.-M., Lin, Z.-Q.: Non-linear buckling and postbuckling of shear deformable anisotropic laminated cylindrical shell
subjected to varying external pressure loads. Compos. Struct. 92, 553–567 (2010)
38. Kheirikhah, M.M., Khalili, S.M.R., Fard, K.M.: Biaxial buckling analysis of soft-core composite sandwich plates using
improved high-order theory. Eur. J. Mech. A Solids 31, 54–66 (2012)
39. Grover, N., Maiti, D.K., Singh, B.N.: A new inverse hyperbolic shear deformation theory for static and buckling analysis of
laminated composite and sandwich plates. Compos. Struct. 95, 667–675 (2013)
40. Nguyen, T.N., Thai, C.H., Nguyen-Xuan, H.: On the general framework of high order shear deformation theories for laminated
composite plate structures: a novel unified approach. Int. J. Mech. Sci. 110, 242–255 (2016)
41. Gol’denveizer, A.L.: Theory of Thin Elastic Shells. International Series of Monograph in Aeronautics and Astronautics.
Pergamon Press, New York (1961)
42. Novozhilov, V.: Theory of Thin Shells. Wolters-Noordhoff, Groningen (1970)
43. Mikhasev, G.I., Altenbach, H., Korchevskaya, E.A.: On the influence of the magnetic field on the eigenmodes of thin
laminated cylindrical shells containing magnetorheological elastomer. Compos. Struct. 113, 186–196 (2014)
44. Marguerre, K.: Zur Theorie der gekrümmten Platte grosser Formänderung. Jahrbuch 1939 der deutschen Akademie der
Luftfahrtforschung, Bewrlin, Adlershof Bücherei. 1, 413–426 (1939)
45. Föppl, A.: Vorlesungen über technische Mechanik. 5. Oldenburg Verlag, München, 132–144 (1907)
46. von Kàrmán, T.: Festigkeitsprobleme im Maschinenbau. In: Encyk. d. Math. Wiss., vol. 4, pp. 311–385, Teubner, Leipzig
(1910)
47. Tovstik, P.E., Smirnov, A.L.: Asymptotic Metods in the Buckling Theory of Elastic Shells. World Scientific, Singapore
(2001)
 G. Mikhasev, M. Botogova

48. Koiter, W.T.: On the nonlinear theory of thin elastic shells. Proc. Koninkl. Ned. Acad. Wetensch. 69, 1–54 (1966)
49. Mikhasev, G.I., Tovstik, P.E.: Localized Vibrations and Waves in Thin Shells. Asymptotic Methods. FIZMATLIT, Moscow
(2009) (in Russian)
50. Southwell, R.: On the collapse of tubes by external pressure. Parts 1, 2, 3. Philos. Mag. Ser. 6 25, 687–697 (1913)
51. Mikhasev, G.I., Mlechka, I.R.: On the influence of boundary conditions and transverse shear on buckling of thin laminated
cylindrical shells under external pressure. Facta Univ. Ser. Mech. Eng. 12, 95–106 (2014)
52. Korobko, E.V., Mikhasev, G.I., Novikova, Z.A., Zurauski, M.A.: On damping vibrations of three-layered beam containing
magnetorheological elastomer. J. Intell. Mater. Syst. Struct. 23, 1019–1023 (2012)
53. Reddy, J.N.: Mechanics of Laminated Composite Plates and Shells: Theory and Analysis. CRC Press, New York (2004)