Engineering Analysis with Boundary Elements 87 (2018) 90–110

Contents lists available at ScienceDirect

Engineering Analysis with Boundary Elements

journal homepage: www.elsevier.com/locate/enganabound

Nonlinear bending analysis of nonlocal nanoplates with general shapes and

boundary conditions by the boundary-only method
Monchai Panyatong a,∗, Boonme Chinnaboon b, Somchai Chucheepsakul b
a
Department of Civil and Environmental Engineering, Faculty of Engineering, Rajamangala University of Technology Lanna, Chiang Rai 57120, Thailand
b
Department of Civil Engineering, Faculty of Engineering, King Mongkut’s University of Technology Thonburi, Bangkok 10140, Thailand

a r t i c l e i n f o a b s t r a c t

Keywords: In this paper, the geometrically nonlinear bending analysis of nanoplates with general shapes and boundary con-
Nanoplates ditions is highlighted. The governing equations are derived based on the classical plate theory using nonlocal
Nonlinear bending differential constitutive relation of Eringen and von Kármán’s nonlinear strains. The boundary-only method is
Boundary element method
developed by using the principle of the analog equation (PAE). According to the PAE, the original governing
Analog equation method
differential equations are replaced by three uncoupled equations with fictitious sources under the same bound-
Meshless
Nonlocal elasticity ary conditions, namely two Poisson equations and one biharmonic equation. Subsequently, the fictitious sources
are established using a technique based on the boundary element method and approximated by using the radial
basis functions. The solution of the actual problem is attained from the known integral representations of the
potential and plate problems. Therefore, the kernels of the boundary integral equations are conveniently estab-
lished and readily calculated that the complex nanoplates can be easily analyzed. The accuracy of the proposed
methodology is evaluated by comparing the obtained results with available solutions. Moreover, the influences of
nonlocal parameter on the various characteristics of effective distributed loads are elucidated. Finally, the effects
of nonlocal parameter, von Kármán’s nonlinearity and aspect ratio on nonlinear bending responses are studied.
© 2017 Elsevier Ltd. All rights reserved.

1. Introduction neighbor points in the continuum body. The influence of size-effect is

In recent years, nanoplates had been widely applied in many ad-
vanced devices such as resonators, mass sensors, transistors, sensitive
gas detectors and nanoelectromechanical systems (NEMS) since their
excellent mechanical, chemical, thermal and electronic properties. Dur-
ing manufacturing, testing or operating, nanoplates have a chance of
excessive deformation, which can affect the performance of nanode-
vices and might lead to failure. Therefore, the precise evaluation and
proper understanding of the bending behavior are very important for
engineering design and manufacture. Although the experiments provide
valuable insights, conducting at nanoscale is so delicate and expensive.
Hence, the development of appropriate mathematic models at nanoscale
is an essential key issue for the evaluation of mechanical responses of
nanostructures. It has been well known that nanostructures appear to
be a significant “size-effect” in mechanical properties due to the influ-
ences of inter-atomic and inter-molecular cohesive forces and this con-
sequence cannot be neglected. One of the continuum mechanics prob-
lems including small scale effects and providing an acceptable accuracy
is the nonlocal elasticity of Eringen [1,2]. This theory is based on the
assumption that the stress at a point is a function of the strain field at all
directly considered in the constitutive equations as a material parameter
called “the nonlocal parameter”. Consequently, the nonlocal elasticity
was extensively applied to analyze various performances of nanobeams
[3–6] and nanoplates [7–23]. Recently, some researchers have focused
on investigating the reinforcement for nanocomposites plates [24,25] in
order to improve theirs physical, mechanical and electrical properties.
This is already achieved by adoption of reinforcing carbon nanotubes
(CNTs) because they have excellent mechanical, thermal and electronic
properties.
In literature, the nonlinear bending responses of nanoplates have
been given a very attention by the researchers. Reddy [26] formu-
lated the nonlocal nonlinear equilibrium equations of nanobeams and
nanoplates by using the nonlocal constitutive relation of Eringen in com-
bination with the von Kármán’s strain. However, numerical results were
not provided in his work. Shen et al. [27] studied the nonlinear bending
of simply supported rectangular Single Layer Graphene Sheets (SLGS)
subjected to transverse loads in thermal environment and a perturba-
tion technique was used to solve the problem. The nonlinear bending
of orthotropic rectangular nanoplates embedded in an elastic matrix
was investigated by Golmakani and Rezatalab [28] using the differential

∗
Corresponding author.
E-mail address: monchai_art@rmutl.ac.th (M. Panyatong).

https://doi.org/10.1016/j.enganabound.2017.12.003
Received 8 March 2017; Received in revised form 3 November 2017; Accepted 4 December 2017
0955-7997/© 2017 Elsevier Ltd. All rights reserved.
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

quadrature method (DQM). Recently, Dastjerdi et al. [29] studied non-

linear behavior of sector graphene sheets resting on a Winkler Pasternak
elastic foundation based on the first shear deformation theory and the
nonlocal elasticity. Also, Dastjerdi et al. [30] analyzed the orthotropic
annular/circular graphene sheets by considering the first shear deforma-
tion theory in conjunction with the von Kármán’s nonlinear strain. The
nonlinear thermo-elastic bending behavior of orthotropic SLGS embed-
ded in elastic medium was investigated by Golmakani and Far [31] who
solved the problem by using the differential quadrature method. More-
over, the nonlinear large deformation of SLGS was analyzed by using
the element-free kp-Ritz method presented by Zhang et al. [32].
Actually, the nanodevices encounter the complexities of both geome-
tries and boundary supports. Nevertheless, the analytical solutions of
continuum models can be only obtained from the simple cases. To date,
the nonlinear bending of nanoplates with general shapes and boundary
conditions has still not been systematically investigated. The effective Fig. 1. Nanoplate geometry and notation.
numerical methods are, therefore, required to be applicable to the afore-
mentioned issues. Although the finite element method (FEM) is already a
well-established numerical technique to deal with an enormous number by demonstrating the application of the proposed method to analyze the
of complex engineering problems. However, it is ineffective and trob- nonlinear bending of the nanoplate with a complex shape and boundary
lesome when the geometric bodies are not simple, e.g., there are holes condition as generally employed in nanodevices.
or corners. The fine meshing and high density elements are expected at
these critical regions. Additionally, it is also inefficient in finding their 2. Problem formulation
derivatives of the field functions (e.g., fluxes, strains, stresses, moments)
and thus less accuracy are obtained when the problems have a large gra- 2.1. Nonlocal elasticity
dient. At present, the boundary element method (BEM) is realized as a
powerful method for the analysis of engineering problems, and an al- According to the nonlocal elasticity theory of Eringen [1,2], the cru-
ternative numerical technique to the domain methods. Discretization is cial concept is that the stress at a reference point x in the body depends
only over the boundary of the body, making the numerical modeling not only on the strains at the point x but also on strains at all other
with the BEM convenient and, hence, reducing the number of the prob- points in the body. Based on aforementioned theory, the constitutive
lem dimension by one order. Consequently, a preparative data input for relations of homogeneous material can be represented by the following
a computer program is a simple procedure and also the BEM is more ef- differential equations:
fective in evaluation of the derivatives of the field functions. However, 𝑁𝐿 𝐿
⎧𝜎𝑥𝑥 ⎫ ⎧𝜎𝑥𝑥 ⎫ ⎡1 𝜐 0 ⎤⎧𝜀𝑥𝑥 ⎫
application of the BEM can be only applied for problems whose the fun- ⎪ 𝑁𝐿 ⎪ ⎪ 𝐿 ⎪ 𝐸 ⎢ ⎥⎪ ⎪
(1 − 𝜇∇ )⎨𝜎𝑦𝑦 ⎬ = ⎨ 𝜎𝑦𝑦 ⎬ =
2
𝜐 ⎥⎨ 𝜀𝑦𝑦 ⎬,
2 ⎢
damental solution of the differential equation can be determined. The 1 0 (1)
⎪ 𝑁𝐿 ⎪ ⎪ 𝐿 ⎪ (1 − 𝜐 ) ⎢ ⎥⎪ ⎪
method is obviously not applicable to nonlinear problems for which the 𝜎 𝜎
⎩ 𝑥𝑦 ⎭ ⎩ 𝑥𝑦 ⎭ ⎣ 0 0 (1 − 𝜐)⎦⎩𝜀𝑥𝑦 ⎭
fundamental solution cannot be established.
To alleviate the restriction as mentioned above, Katsikadelis where 𝜎𝑖𝑗𝑁𝐿 , 𝜎𝑖𝑗𝐿 and 𝜀ij , (i, j = x, y) are the nonlocal stresses, the local
[33] proposed the analog equation method (AEM) that the original prob- stresses (classical stresses) and the strains, respectively. Meanwhile, E
lem is converted to a substitute problem (so-called “the analog equa- is the modulus of elasticity and 𝜐 is Poisson’s ratio. Moreover, the scale
tion”) under the fictitious sources with the same boundary and initial factor 𝜇 = (e0 li )2 is the nonlocal parameter which expresses the effects
conditions. A linear equation with simple known fundamental solution of the inter-atomic forces and atomic length scales. This is in accordance
can be chosen as the analog equation. Then the standard BEM is readily with the atomic theory of lattice dynamics and experimental observa-
employed to solve the substitute problem and the solution of the orig- tions on phonon dispersion. li is an internal characteristic length (such
inal problem can be obtained from the integral representations of the as lattice spacing, granular distance, distance between C–C bonds) and
substitute problem. At present, the AEM has been already successfully e0 is a material constant which is determined to calibrate the nonlocal
employed to solve a variety of complex engineering problems [34–45]. model with experimental results or the results of molecular dynamics
The aim of this work is to develop the boundary-only method to (MD) simulations. Additionally, the nonlocal stresses can be expressed
solve the geometrically nonlinear bending responses of the nanoplates in terms of the local stress components as the following [17,20]:
with general shapes and boundary conditions. The principle of the ana- ∑
∞
log equation for the BEM is applied to solve this complex problem. 𝜎𝑖𝑗𝑁𝐿 = 𝜇 𝑚 ∇2𝑚 𝜎𝑖𝑗𝐿 . (2)
The methodology presented in this study retains all advantages of the 𝑚=0

BEM since the discretization and integration are operated only on the
2.2. Nonlocal nonlinear formulation of nanoplates
boundary. The analog equations of the original problem are comprised
of the Poisson and Bihamonic equations which are linear equations hav-
Consider a nanoplate with a uniform thickness h occupying the two-
ing simple known fundamental solutions. The ficititous sources are ap-
dimensional multiply connected domain Ω of the xy plane whose the
proximated by using Thin Plate Splines (TPSs) as the radial basis func-
boundary Γ = ∪𝐾 Γ may be piecewise, i.e., the boundary may have a
𝑖=0 𝑖
tions and established using a BEM-based procedure. Thus, the solution
finite number of corners. The nanoplate is subjected to a distributed
of the original problem is conveniently established and easily evalu-
transverse load f(x, y) (see Fig. 1). In this study, the governing equa-
ated. Comparisons of the obtained results with those of other available
tions of nanoplates are formulated based on the classical plate theory in
solutions are performed to verify the reliability and accuracy of the pro-
conjunction with the nonlinear strain-displacement relationships of von
posed method. Then, the influences of the nonlocal parameter on the
Kármán. According to above mentions, the governing equations can be
various characteristics of the effective distributed loads are explained.
established by application of the principle of virtual displacements [26],
Moreover, the effects of the nonlocal parameter and the von Kármán
as follows:
nonlinearity as well as the aspect ratio on nonlinear bending responses
𝑁𝐿 𝑁𝐿
are thoroughly studied. Finally, the present study fills a gap in literature 𝑁𝑥𝑥,𝑥 + 𝑁𝑥𝑦,𝑦 = 0, (3a)

91
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

𝑁𝐿 𝑁𝐿 where nx and ny are the components of the unit vector n which is normal
𝑁𝑥𝑦,𝑥 + 𝑁𝑦𝑦,𝑦 = 0, (3b)
to the boundary Γ (see Fig. 1). The coefficients 𝛾 1 , 𝛾 2 and 𝛿 1 , 𝛿 2 in Eqs.
𝑁𝐿 𝑁𝐿 𝑁𝐿
(7a) and (7b) take the values 0 or 1, depending on the type of boundary
𝑀𝑥𝑥,𝑥𝑥 − 2𝑀𝑥𝑦,𝑥𝑦 + 𝑀𝑦𝑦,𝑦𝑦 + 𝑁 𝑁𝐿 (𝑢, 𝑣, 𝑤) + 𝑓 = 0, (3c) conditions. Meanwhile, the coefficient 𝛾 3 is the prescribed displacement
where, in Eq. (3c), 𝑁 𝑁𝐿 (𝑢, 𝑣, 𝑤) = 𝑁𝑥𝑥 𝑁𝐿 𝑤 𝑁𝐿 𝑁𝐿 u∗ or nonlocal boundary traction 𝑡𝑁𝐿 ∗ , and the coefficient 𝛿 is the pre-
,𝑥𝑥 + 2𝑁𝑥𝑦 𝑤,𝑥𝑦 + 𝑁𝑦𝑦 𝑤,𝑦𝑦 𝑥 3
and the nonlocal stress resultants are described by scribed displacement v or nonlocal boundary traction 𝑡𝑁𝐿
∗
𝑦
∗ .

{ ( )} (b) for the transverse displacement

⎧ 𝑁𝑥𝑥
𝑁𝐿 ⎫ ⎧ 𝑁 𝐿 ⎫ ⎧𝐶 𝑢 + 1 𝑤 2 + 𝜐 𝑣 + 1 𝑤 2 ⎫
⎪ ⎪ ⎪ 𝑥𝑥 ⎪ ⎪ { ,𝑥 2 ,𝑥 (
,𝑦 2 ,𝑦 )}⎪ ∑
∞
( )
𝑁𝐿 𝐿
⎪ 𝑁𝑦𝑦 ⎪ ⎪ 𝑁𝑦𝑦 ⎪ ⎪𝐶 𝑣 + 1 𝑤2 + 𝜐 𝑢 + 1 𝑤2 ⎪ 𝛼1 𝑤 + 𝛼2 𝜇 𝑚 ∇2𝑚 𝑉 𝐿 (𝑤) + 𝑁𝑛𝐿 𝑤,𝑛 + 𝑁𝑡𝐿 𝑤,𝑡 = 𝛼3 on Γ, (9a)
⎪ 𝑁𝐿 ⎪ ⎪ 𝐿 ⎪ ⎪ ,𝑦 ,𝑥
⎪ 𝑁 ⎪ ⎪ 𝑁 ⎪ ⎪ 𝐶(1− 𝜐)
2 ,𝑦
( 2 ,𝑥
) ⎪⎪
𝑚=0
𝑥𝑦 𝑥𝑦
(1 − 𝜇∇2 )⎨ 𝑁𝐿 ⎬ = ⎨ 𝐿 ⎬ = ⎨ 𝑢,𝑦 + 𝑣,𝑥 + 𝑤,𝑥 𝑤,𝑦
2
( ) ⎬, (4)
𝑀 𝑀
⎪ 𝑥𝑥 ⎪ ⎪ 𝑥𝑥 ⎪ ⎪ ⎪ ∑
∞
−𝐷 𝑤,𝑥𝑥 + 𝜐𝑤,𝑦𝑦 𝛽1 𝑤,𝑛 + 𝛽2 𝜇 𝑚 ∇ 2 𝑚 𝑀 𝐿 ( 𝑤 ) = 𝛽3 on Γ, (9b)
⎪𝑀 𝑁𝐿 ⎪ ⎪ 𝑀 𝐿 ⎪ ⎪ ( ) ⎪
⎪ 𝑦𝑦 ⎪ ⎪ 𝑦𝑦 ⎪ ⎪ −𝐷 𝑤,𝑦𝑦 + 𝜐𝑤,𝑥𝑥 ⎪ 𝑚=0
⎪𝑀 𝑁𝐿 ⎪ ⎪𝑀 𝐿 ⎪ ⎪ ⎪
⎩ 𝑥𝑦 ⎭ ⎩ 𝑥𝑦 ⎭ ⎩ 𝐷(1 − 𝜐)𝑤,𝑥𝑦 ⎭ where the coefficients 𝛼 i , 𝛽 i (i = 1, 2, 3) are functions specified on the
where 𝐶 = 𝐸ℎ∕(1 − 𝜐2 )
is the membrane stiffness of the nanoplate and boundary Γ; ML (w) and VL (w) are the local normal bending moment
𝐷 = 𝐸 ℎ3 ∕12(1 − 𝜐2 ) is the bending or flexural stiffness of the nanoplate. and the local effective shear force on the boundary Γ, respectively; 𝑁𝑛𝐿
Moreover, by the relation of Eq. (2), the nonlocal stress resultants can and 𝑁𝑡𝐿 are the local membrane loads in the normal and tangential di-
be written in terms of the local stress resultants series as rections on the boundary Γ, respectively. Additionally, for the case of
free or transverse elastically restrained edges, the boundary conditions
⎧ 𝑁𝑥𝑥
𝑁𝐿 ⎫ ⎧ 𝑁𝑥𝑥
𝐿 ⎫
⎪ ⎪ ⎪ ⎪ (9a) and (9b) must be supplemented with the following corner condition
𝑁𝐿 𝐿
⎪ 𝑁𝑦𝑦 ⎪ ⎪ 𝑁𝑦𝑦 ⎪ [47]:
⎪ 𝑁𝐿 ⎪ ⎪ 𝐿
⎪
⎪ 𝑁𝑥𝑦 ⎪ ∑ 𝑚 2𝑚 ⎪ 𝑁𝑥𝑦 ⎪
∞
∑
∞
[[ ]]
⎨ 𝑁𝐿 ⎬ = 𝜇 ∇ ⎨ 𝐿 ⎬. (5) 𝑐1𝑘 𝑤(𝑘) + 𝑐 2𝑘 𝜇 𝑚 ∇2𝑚 𝑇 𝐿 (𝑤) 𝑘 = 𝑐 3𝑘 at corner 𝑘, (9c)
⎪𝑀𝑥𝑥 ⎪ 𝑚=0 ⎪𝑀𝑥𝑥 ⎪ 𝑚=0
⎪𝑀 𝑁𝐿 ⎪ ⎪𝑀 𝐿 ⎪
⎪ 𝑦𝑦 ⎪ ⎪ 𝑦𝑦 ⎪ where the coefficients cik , (i = 1, 2, 3) are the constants specified at the
⎪𝑀 𝑁𝐿 ⎪ ⎪𝑀 𝐿 ⎪
⎩ 𝑥𝑦 ⎭ ⎩ 𝑥𝑦 ⎭ corner point k and [[𝑇 𝐿 (𝑤)]]𝑘 is the jump of discontinuity of the local
twisting moment at the corner k. The operators ML , VL and TL appearing
Substitution of Eq. (4) into Eqs. (3a)–(3c) yields the governing equa-
in Eqs. (9a)–(9c) are defined with the boundary curvilinear coordinates
tions in terms of the displacement fields (u, v, w) as follows:
n and s (see Fig. 1) as follows [47]:
(a) for the membrane displacements
( ( 2 ))
( ) 𝜕 𝜕
(1 + 𝜐) ( ) 2 𝑀 𝐿 = −𝐷 ∇2 + (𝜐 − 1) +𝜅 , (10a)
∇2 𝑢 + 𝑢,𝑥𝑥 + 𝑣,𝑥𝑦 + 𝑤,𝑥𝑦 𝑤,𝑦 + 𝑤,𝑥𝑥 + 𝑤,𝑦𝑦 𝑤,𝑥 = 0 in Ω, 𝜕𝑠 2 𝜕𝑛
(1 − 𝜐) (1 − 𝜐)
(6a) ( ( ))
𝜕 2 𝜕 𝜕2 𝜕
𝑉 𝐿 = −𝐷 ∇ + (1 − 𝜐) −𝜅 , (10b)
𝜕𝑛 𝜕𝑠 𝜕 𝑛𝜕 𝑠 𝜕𝑠
( )
(1 + 𝜐) ( ) 2
∇2 𝑣 + 𝑣 + 𝑢,𝑥𝑦 + 𝑤,𝑥𝑦 𝑤,𝑥 + 𝑤,𝑦𝑦 + 𝑤,𝑥𝑥 𝑤,𝑦 = 0 in Ω, ( )
(1 − 𝜐) ,𝑦𝑦 (1 − 𝜐) 𝜕2 𝜕
𝑇 𝐿 = 𝐷(1 − 𝜐) −𝜅 , (10c)
(6b) 𝜕 𝑛𝜕 𝑠 𝜕𝑠

(b) for the transverse displacement where 𝜅 = 𝜅(s) is the curvature of the boundary at point s.
𝑁𝐿 Note that when the value of m is set to be zero, the boundary condi-
𝐷∇ 𝑤 − (1 − 𝜇∇ )𝑁
4 2
(𝑢, 𝑣, 𝑤) = (1 − 𝜇∇ )𝑓 2
in Ω . (6c)
tions (7a), (7b) and (9a)−(9c) are the most general boundary conditions
Furthermore, the displacement fields (u, v, w) and the nonlocal stress for the plates. Evidently, the conventional boundary conditions can be
resultants must satisfy the following boundary conditions on the bound- obtained by specifying the coefficients 𝛾 i , 𝛿 i , 𝛼 i , 𝛽 i and cik , appropriately.
ary Γ = ∪𝐾 Γ:
𝑖=0 𝑖
(a) for the membrane displacements
3. The boundary-only method for nonlinear bending of nonlocal
∑
∞
( ) nanoplates
𝑚 2𝑚
𝛾1 𝑢 + 𝛾2 𝜇 ∇ 𝑡𝐿𝑙
𝑥 + 𝑡𝐿𝑛𝑙
𝑥 = 𝛾3 on Γ, (7a)
𝑚=0

( ) In this section, the governing equations of nonlinear nonlocal

∑
∞
𝛿1 𝑣 + 𝛿2 𝜇 𝑚 ∇2𝑚 𝑡𝐿𝑙 𝐿𝑛𝑙
𝑦 + 𝑡𝑦 = 𝛿3 on Γ, (7b) nanoplates as described by Eqs. (6a)–(6c) with the associated bound-
𝑚=0 ary conditions, Eqs. (7a) and (7b) and Eqs. (9a)–(9c), will be solved
by using the boundary-only method. However, in order to develop the
where 𝑡𝐿𝑙 𝐿𝑙
𝑥 and 𝑡𝑦 are linear terms of the local boundary tractions; 𝑡𝐿𝑛𝑙
𝑥
𝐿𝑛𝑙
boundary-only method to solve the problem expediently, the value of
and 𝑡𝑦 are nonlinear terms of the local boundary tractions. The local
m in Eqs. (7a) and (7b) and Eqs. (9a)–(9c) is set to be zero. The proce-
boundary tractions are provided by following equations (see more de-
dure of boundary-only method will be applied to the problem at hand
tails in [46–48]):
( ) as following.
𝐶 1−𝜐 1−𝜐
𝑡𝐿𝑙
𝑥 = ℎ 𝑢,𝑥 𝑛𝑥 + 𝑢,𝑦 𝑛𝑦 + 𝑣,𝑥 𝑛𝑦 + 𝜐 𝑣,𝑦 𝑛𝑥 , (8a)
2 2
3.1. The membrane displacements
( )
𝐶 1−𝜐 1−𝜐
𝑡𝐿𝑙
𝑦 = ℎ 𝜐 𝑢,𝑥 𝑛𝑦 + 𝑢 𝑛 + 𝑣 𝑛 + 𝑣,𝑦 𝑛𝑦 , (8b)
2 ,𝑦 𝑥 2 ,𝑥 𝑥 Let u and v be the sought solutions of Eqs. (6a)–(6c). These functions
(( ) ) are two times continuously differentiable with respect to the spatial co-
𝐶 1 2 1 1−𝜐
𝑡𝐿𝑛𝑙
𝑥 = 𝑤,𝑥 + 𝜐𝑤2,𝑦 𝑛𝑥 + 𝑤,𝑥 𝑤,𝑦 𝑛𝑦 , (8c) ordinate x, y in Ω and one time on its boundary Γ. According to the
ℎ 2 2 2
principle of the analog equation (PAE) [33], the differential operator
( ( ) )
𝐶 1−𝜐 1 2 1 that produces the analog equation should be of the second order. The
𝑡𝐿𝑛𝑙 = 𝑤,𝑥 𝑤,𝑦 𝑛𝑥 + 𝑤 + 𝜐𝑤2 𝑛 , (8d)
𝑦 ℎ 2 2 ,𝑦 2 ,𝑥 𝑦 simplest operator of this order with known fundamental solution is the

92
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

field point q with r = |q − p| (see Fig. 1), 𝜀(p) is a free coefficient which
depends on the position of point p and it is defined as
⎧ 1 for 𝑝 = 𝑃 inside Ω,
⎪
𝜀(𝑝) = ⎨𝛼∕2𝜋 for 𝑝 on the boundary Γ, (13)
⎪
⎩ 0 for 𝑝 outside Ω.
Note that Eqs. (12a) and (12b) are domain-boundary integral equa-
tions, since they include the unknown fictitious sources in the domain in-
tegrals. They could be solved by using domain discretization for domain
integrals. However, in order to the method maintains its pure bound-
ary character, the aforementioned domain integrals can be converted to
boundary integrals by approximating the fictitious sources by using the
radial basis functions. Thus, they can be expressed as
𝑀
∑
𝑏(1) = 𝑎(1)
𝑗 𝑓𝑗 , (14a)
𝑗=1

𝑀
∑
𝑏(2) = 𝑎(2)
𝑗 𝑓𝑗 , (14b)
Fig. 2. Discretization of the nanoplate.
𝑗=1

Laplace operator. Therefore, the analog equations of Eqs. (6a)–(6c) can where fj = fj (rij ) are Mradial-basis approximation functions, 𝑎(𝑗𝑖) , (𝑖 = 1, 2)
be written as: are M (in total 2M) coefficients to be determined and rij = |pj − pi | is the
distance between the collocation point pj (xj , yj ) and any point pi (xi , yi )
∇2 𝑢 = 𝑏(1) (𝑥, 𝑦), 𝑖𝑛 Ω, (11a) ∈ Ω∪Γ (see Fig. 2). Then, using Green’s second identity, the domain
integrals in Eqs. (12a) and (12b) become
∇2 𝑣 = 𝑏(2) (𝑥, 𝑦), 𝑖𝑛 Ω, (11b) ⎧ ⎫
𝑀
∑ ⎪ ( ) ⎪
where b(1) (x,
y) and b(2) (x,
y) are the two fictitious sources, unknown in 𝑣 𝑏 𝑑Ω =
∗ (1)
𝑎(1)
𝑗 ⎨𝜀(𝑝)𝑢̂ 𝑗 + 𝑣 𝑢̂ 𝑗,𝑛 −
∗
𝑢̂ 𝑗 𝑣∗,𝑛 𝑑𝑠⎬, (15a)
∫ ⎪ ∫ ⎪
𝑗=1
the first instance. They can be solved by using the BEM. Therefore, the Ω ⎩ Γ ⎭
integral representations of the solution of Eqs. (11a) and (11b) are
( ) ⎧ ⎫
𝑀
∑ ⎪ ( ) ⎪
𝜀(𝑝)𝑢 (𝑝) = − 𝑣∗ 𝑢,𝑛 − 𝑢𝑣∗,𝑛 𝑑𝑠 + 𝑣∗ 𝑏(1) 𝑑Ω, 𝑝 ∈ Ω ∪ Γ, (12a) 𝑣∗ 𝑏(2) 𝑑Ω = 𝑎(2)
𝑗 ⎨ 𝜀 ( 𝑝 ) 𝑢
̂ + 𝑣 ∗
𝑢
̂ − 𝑢
̂ 𝑣 ∗
𝑑𝑠⎬, (15b)
∫ ∫ ∫ 𝑗 ∫ 𝑗,𝑛 𝑗 ,𝑛
𝑗=1 ⎪ ⎪
Γ Ω Ω ⎩ Γ ⎭
( )
𝜀(𝑝)𝑣 (𝑝) = − 𝑣 𝑣,𝑛 −
∗
𝑣𝑣∗,𝑛 𝑑𝑠 + 𝑣 𝑏 𝑑Ω,
∗ (2)
𝑝 ∈ Ω ∪ Γ, (12b) where 𝑢̂ 𝑗 = 𝑢̂ 𝑗 (𝑟𝑖𝑗 ) is the particular solution of
∫ ∫
Γ Ω ∇2 𝑢̂ 𝑗 = 𝑓𝑗 , 𝑗 = 1, 2, … , 𝑀. (16)
where = (ln r)/2𝜋 is the fundamental solution of the Laplace equa-
v∗ The solution 𝑢̂ 𝑗 = 𝑢̂ 𝑗 (𝑟𝑖𝑗 ) of Eq. (16) can always be established, if fj is
tion and 𝑣∗,𝑛 = 𝑟,𝑛 ∕2𝜋𝑟 is its derivative normal to the boundary at the specified. By substituting Eqs. (15a) and (15b) into Eqs. (12a) and (12b),

Fig. 3. The non-dimensional deflection at the center of plates with various numbers of boundary elements N.

93
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 4. The non-dimensional deflection at the center of plates with various numbers of internal collocation points M.

Fig. 5. Variation of the non-dimensional deflections at the center of the clamped circular plate versus the uniform loads.

94
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

the solutions at a point p ∈ Γ for smooth boundary can be expressed

as
⎧ ⎫
( ) ∑𝑀
⎪1 ( ) ⎪
1
𝑢(𝑝) = − 𝑣∗ 𝑢,𝑛 − 𝑢𝑣∗,𝑛 𝑑𝑠 + 𝑎(1)
𝑗 ⎨2 𝑗 𝑢
̂ + 𝑣 ∗
𝑢
̂ 𝑗,𝑛 − 𝑢
̂ 𝑣 ∗
𝑗 ,𝑛 𝑑𝑠⎬,
2 ∫ ⎪ ∫ ⎪
𝑗=1
Γ ⎩ Γ ⎭
(17a)

⎧ ⎫
( ) ∑𝑀
⎪1 ( ) ⎪
1
𝑣 (𝑝) = − 𝑣∗ 𝑣,𝑛 − 𝑣𝑣∗,𝑛 𝑑𝑠 + 𝑎(2)
𝑗 ⎨2 𝑗 𝑢
̂ + 𝑣 ∗
𝑢
̂ 𝑗,𝑛 − 𝑢
̂ 𝑣 ∗
𝑗 ,𝑛 𝑑𝑠 ⎬.
2 ∫ ⎪ ∫ ⎪
𝑗=1
Γ ⎩ Γ ⎭
(17b)
Additionally, the derivatives of the displacements u and v inside the
domain Ω(𝜀(p) = 1) can be obtained by direct differentiation of Eqs.
(12a) and (12b), respectively. Thus, for the sake of conciseness, we can
write the integral representations of the displacements u and v and their
derivatives up to the second order as
( ) Fig. 6. Plate with complex geometry and boundary conditions [47] (C = Clamped sup-
𝑢,𝑎𝑏 (𝑃 ) = − 𝑣∗,𝑎𝑏 𝑢,𝑛 − 𝑢𝑣∗,𝑛𝑎𝑏 𝑑𝑠 port, F = Free support).
∫
Γ

⎧ ⎫
𝑀
∑ ⎪ ( ) ⎪ coordinate x, y in Ω and three times on its boundary Γ. By using the
+ 𝑎(1)
𝑗 ⎨𝑢̂ 𝑗,𝑎𝑏 + 𝑣 ∗
,𝑎𝑏 𝑢
̂ 𝑗,𝑛 − 𝑢
̂ 𝑗 𝑣 ∗
,𝑛𝑎𝑏 𝑑𝑠⎬, (18a) PAE, the biharmonic operator is applied. Thus, the analog equation can
⎪ ∫ ⎪
𝑗=1 be written as
⎩ Γ ⎭
( ) ∇4 𝑤 = 𝑏(3) (𝑥, 𝑦), 𝑖𝑛 Ω, (19)
𝑣,𝑎𝑏 (𝑃 ) = − 𝑣,𝑎𝑏 𝑣,𝑛 − 𝑣𝑣,𝑛𝑎𝑏 𝑑𝑠
∗ ∗
∫
Γ where b(3) (x,
y) is the fictitious source. By using the direct BEM for the
⎧ ⎫ biharmonic equation or plate equation with D = 1, the integral represen-
∑𝑀
⎪ ( ) ⎪ tation of the solution [47] of Eq. (19) is
+ 𝑎(2) 𝑢
̂
𝑗 ⎨ 𝑗,𝑎𝑏 + 𝑣 ∗
𝑢
̂
,𝑎𝑏 𝑗,𝑛 − 𝑢
̂ 𝑣 ∗
𝑗 ,𝑛𝑎𝑏 𝑑𝑠⎬, (18b)
⎪ ∫ ⎪
𝑗=1 𝜀 (𝑝)𝑤(𝑝)
⎩ Γ ⎭
( )
where a, b = 0, x, y and point P ∈ Ω . = 𝑤∗ 𝑉 𝐿 (𝑤) − 𝑤𝑉 𝐿 (𝑤∗ ) − 𝑤∗,𝑛 𝑀 𝐿 (𝑤) + 𝑤,𝑛 𝑀 𝐿 (𝑤∗ ) 𝑑𝑠
∫
Γ
3.2. The transverse displacement ∑( [[ ]] [[ ]] )
− 𝑤∗ 𝑇 𝐿 (𝑤) 𝑘 − 𝑤 𝑇 𝐿 (𝑤∗ ) 𝑘 + 𝑤∗ 𝑏(3) 𝑑Ω, 𝑝 ∈ Ω ∪ Γ,
∫
𝑘
Let w be the sought solution of Eqs. (6a)–(6c). Then, the function Ω
w is four times continuously differentiable with respect to the spatial (20)

Fig. 7. Non-dimensional deflections of the complex plate at point A.

95
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 8. Non-dimensional deflections of the complex plate along thex-axis.

Fig. 9. Variation of the central deflections of SLGS with various boundary supports versus uniformly transverse loads.

where w∗ = (r2 ln r)/8𝜋 is the fundamental solution of the biharmonic 1

𝑇 𝐿 (𝑤∗ ) = (1 − 𝜐)𝑟,𝑛 𝑟,𝑡 . (21c)
equation, 𝑤∗,𝑛 = 𝑟𝑟,𝑛 (1 + 2 ln 𝑟)∕8𝜋 is its derivative normal to the bound- 4𝜋
ary at the field point q with r = |q − p|. Moreover, the ML (w∗ ), VL (w∗ )
and TL (w∗ ) are provided as [47]: The domain integral in Eq. (20) results from the fictitious source
( ) b(3) (x, y) which can be estimated by a radial basis function series, and
1
𝑀 𝐿 (𝑤∗ ) = − 2(1 + 𝜐) ln 𝑟 + (3 + 𝜐)𝑟2,𝑛 + (1 + 3𝜐)𝑟2,𝑡 , (21a) then it can be converted to a boundary integral on Γ. Thus, we have
8𝜋
( ) 𝑀
∑
1
𝑉 𝐿 (𝑤∗ ) = − 2𝑟,𝑛 + (1 − 𝜐)(𝑟,𝑛 − 𝜅𝑟)(𝑟2,𝑛 − 𝑟2,𝑡 ) , (21b) 𝑏(3) = 𝑎(3)
𝑗 𝑓𝑗 . (22)
4𝜋𝑟 𝑗=1

96
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 12. The characteristic distribution of the effective load type (2), f2 < f1 .
Fig. 10. Single layered graphene sheets with various boundary conditions
(S = Simple support, C = Clamped support, F = Free support).

Fig. 13. The characteristic distribution of the effective load type (3), f2 = f1 .

𝑀
⎧ ( ( ) ( ) ) ⎫
∑ ⎪1 𝑤∗ ∇2 𝑤̂ 𝑗 ,𝑛 − 𝑤̂ 𝑗 ∇2 𝑤∗ ,𝑛 ⎪
Fig. 11. The characteristic distribution of the effective load type (1), f2 > f1 . + 𝑎(3)
𝑗 ⎨2 𝑗
̂
𝑤 + 𝑑𝑠⎬
⎪ ∫ −𝑤∗,𝑛 ∇2 𝑤̂ 𝑗 + 𝑤̂ 𝑗,𝑛 ∇2 𝑤∗ ⎪
𝑗=1
⎩ Γ ⎭
∑ ( [[ ]] [[ ]] )
Using the Rayleigh-Green identity, the domain integral in Eq. − 𝑤∗ 𝑇 𝐿 (𝑤) 𝑘 − 𝑤 𝑇 𝐿 (𝑤∗ ) 𝑘 . (25)
𝑘
(20) can be written as
Additionally, at the corner point, the boundary integral solution can
𝑀
⎧ ( ( ) ( ) ) ⎫ be expressed as
∑ ⎪ 𝑤∗ ∇2 𝑤̂ 𝑗 ,𝑛 − 𝑤̂ 𝑗 ∇2 𝑤∗ ,𝑛 ⎪
𝑤∗ 𝑏(3) 𝑑Ω = 𝑎(3)
𝑗 ⎨ 𝜀 ( 𝑝 ) ̂
𝑤 𝑗 + 𝑑𝑠⎬,
∫ ⎪ ∫ − 𝑤 ̂
∗ ∇2 𝑤
𝑗 + ̂
𝑤 𝑗,𝑛 ∇ 2 𝑤∗
⎪ ( )
𝑗=1 ,𝑛 𝛼
Ω ⎩ Γ ⎭ 𝑤 (𝑝) = 𝑤∗ 𝑉 𝐿 (𝑤) − 𝑤𝑉 𝐿 (𝑤∗ ) − 𝑤∗,𝑛 𝑀 𝐿 (𝑤) + 𝑤,𝑛 𝑀 𝐿 (𝑤∗ ) 𝑑𝑠
2𝜋 𝑐 ∫
(23) Γ

𝑀
⎧ ( ( ) ( ) ) ⎫
where the function 𝑤̂ 𝑗 = 𝑤̂ 𝑗 (𝑟𝑖𝑗 ) is a particular solution of ∑ ⎪ 𝛼 𝑤∗ ∇2 𝑤̂ 𝑗 ,𝑛 − 𝑤̂ 𝑗 ∇2 𝑤∗ ,𝑛 ⎪
+ 𝑎(3)
𝑗 ⎨ 2𝜋 𝑗
̂
𝑤 + 𝑑𝑠⎬
⎪ ∫ − 𝑤 ,𝑛 ̂
∗ ∇2 𝑤
𝑗 + ̂
𝑤 𝑗,𝑛 ∇ 2 𝑤∗
⎪
𝑗=1
∇4 𝑤̂ 𝑗 = 𝑓𝑗 , 𝑗 = 1, 2, … , 𝑀. (24) ⎩ Γ ⎭
∑ ( [[ ]] [[ ]] )
− 𝑤∗ 𝑇 𝐿 (𝑤) 𝑘 − 𝑤 𝑇 𝐿 (𝑤∗ ) 𝑘 . (26)
Substituting Eq. (23) into Eq. (20) yields the boundary integral equa-
𝑘
tion for a point p ∈ Γ on the smooth boundary as
Moreover, the second boundary integral equation can be obtained
( )
1
𝑤(𝑝) = 𝑤∗ 𝑉 𝐿 (𝑤) − 𝑤𝑉 𝐿 (𝑤∗ ) − 𝑤∗,𝑛 𝑀 𝐿 (𝑤) + 𝑤,𝑛 𝑀 𝐿 (𝑤∗ ) 𝑑𝑠 by a direct differentiation of Eq. (25) with respect to the normal unit
2 ∫ vector v at a point p ∈ Γ (see Fig. 1). Thus, for the smooth boundary, we
Γ

97
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 14. The distribution of the effective load type (1) along the x-axis with different nonlocal parameters.

Fig. 15. The distribution of the effective load type (2) along the x-axis with different nonlocal parameters.

98
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 16. The distribution of the effective load type (3) along the x-axis with different
nonlocal parameters.

obtain
( )
1
𝑤 (𝑝) = 𝑤∗,𝑣 𝑉 𝐿 (𝑤) − 𝑤𝑉 𝐿 (𝑤∗,𝑣 ) − 𝑤∗,𝑛𝑣 𝑀 𝐿 (𝑤) + 𝑤,𝑛 𝑀 𝐿 (𝑤∗,𝑣 ) 𝑑𝑠
2 ,𝑣 ∫
Γ

𝑀
⎧ ( ( ) ( ) ) ⎫
∑ ⎪1 𝑤∗,𝑣 ∇2 𝑤̂ 𝑗 ,𝑛 − 𝑤̂ 𝑗 ∇2 𝑤∗ ,𝑛𝑣 ⎪
+ 𝑎(3)
𝑗 ⎨ 𝑤̂ 𝑗,𝑣 + 𝑑𝑠⎬
⎪2 ∫ −𝑤∗,𝑛𝑣 ∇2 𝑤̂ 𝑗 + 𝑤̂ 𝑗,𝑛 ∇2 𝑤∗,𝑣 ⎪
𝑗=1
⎩ Γ ⎭
∑ ( [[ 𝐿 ]] [[ ]] )
𝐿
− 𝑤,𝑣 𝑇 (𝑤) 𝑘 − 𝑤 𝑇 (𝑤,𝑣 )
∗ ∗
. (27)
𝑘
𝑘

Also, the integral representations of the derivatives of the displace-

ment w at a point P ∈ Ω(𝜀(p) = 1) can be written up to the fourth order
as:

𝑤,𝑎𝑏𝑐𝑑 (𝑃 )
( )
= 𝑤∗,𝑎𝑏𝑐𝑑 𝑉 𝐿 (𝑤) − 𝑤𝑉 𝐿 (𝑤∗,𝑎𝑏𝑐𝑑 ) − 𝑤∗,𝑛𝑎𝑏𝑐𝑑 𝑀 𝐿 (𝑤) + 𝑤,𝑛 𝑀 𝐿 (𝑤∗,𝑎𝑏𝑐𝑑 ) 𝑑𝑠
∫
Γ

𝑀
⎧ ( ∗ ( 2 ) ( ) ) ⎫
∑ ⎪ 𝑤,𝑎𝑏𝑐𝑑 ∇ 𝑤̂ 𝑗 ,𝑛 − 𝑤̂ 𝑗 ∇2 𝑤∗ ,𝑛𝑎𝑏𝑐𝑑 ⎪
+ 𝑎(3) 𝑤̂
𝑗 ⎨ 𝑗,𝑎𝑏𝑐𝑑 + 𝑑𝑠⎬
⎪ ∫ − 𝑤 ∗ ∇ 2𝑤̂ 𝑗 + 𝑤̂ 𝑗,𝑛 ∇ 2 𝑤∗
⎪
𝑗=1 ,𝑛𝑎𝑏𝑐𝑑 ,𝑎𝑏𝑐𝑑
⎩ Γ ⎭
∑( [[ ]] [[ ]] )
𝐿 𝐿
− 𝑤,𝑎𝑏𝑐𝑑 𝑇 (𝑤) 𝑘 − 𝑤 𝑇 (𝑤,𝑎𝑏𝑐𝑑 )
∗ ∗
, (28)
𝑘
𝑘

where a, b, c, d = 0, x, y and point P ∈ Ω .

Consider the boundary value problem described by Eqs. (6a)–(6c).
There are total ten variables: eight variables on the boundary that are u,
u, n , v, v, n , w, w, n , VL (w), ML (w) and two variables at the corner point
wc , TL (w). We have ten available equations comprised of: five boundary
integral equations, i.e., (17a), (17b), (25), (27) at a point on the bound-
ary and Eq. (26) at the corner point; four equations from the prescribed
boundary conditions (7a), (7b) and (9a), (9b); and one equation from the
corner condition (9c). Consequently, this problem can be solved to ob- Fig. 17. The non-dimensional central deflections of SLGS versus the distributed load for
tain all the boundary and corner quantities in terms of the 3M unknown the boundary condition case (a).

coefficients 𝑎(1) (2) (3)

𝑗 , 𝑎𝑗 and 𝑎𝑗 and the 2N unknown nonlinear terms of
local boundary tractions 𝑡𝐿𝑛𝑙 𝐿𝑛𝑙
𝑥 and 𝑡𝑦 . 𝑡𝐿𝑛𝑙
𝑥 and 𝑡𝐿𝑛𝑙
𝑦 . In the next section, the numerical implementation of the
The final step of the proposed method is to apply Eqs. (6a)–(6c) to the
presented method will be demonstrated in detail.
M domain collocation points, and Eqs. (8c) and (8d) to the N boundary
nodal points with the replacement of the involved values of the deflec-
tions u, v, w and their derivatives using Eqs. (18a), (18b) and (28) for 4. Numerical implementation
the M domain collocation points, and the first derivative of Eq. (25) for
the N boundary nodal points which can be solved to yield the coeffi- In this work, the BEM with constant element is employed to approx-
cients 𝑎(𝑗𝑙) (l = 1, 2, 3) and the nonlinear terms of local boundary tractions imate the boundary integrals in Eqs. (17a), (17b), (25), (26) and (27).

99
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Let N, M and Nc be the numbers of the boundary nodal points, domain

collocation points and corner points, respectively (see Fig. 2). The nu-
merical implementation is conducted as follows.
Firstly, for the membrane displacements u and v, collocating at the
N boundary nodes by using the boundary integral Eqs. (17a) and (17b)
as well as the boundary conditions (7a) and (7b), they can be expressed
in the matrix form as
𝐀 𝟏 𝐱 𝟏 =𝐛 𝟏 +𝐂 𝟏 𝐝 𝟏 , (29)
]T
where x1 = [u, u, n , v, v, n is the 4N × 1 vector of the boundary quan-
tities; d1 is the (2M + 2N) × 1 vector of the unknown variables which
comprises the coefficients 𝑎(1) (2)
𝑗 , 𝑎𝑗 and the nonlinear terms of local
𝐿𝑛𝑙 𝐿𝑛𝑙
boundary tractions 𝑡𝑥 , 𝑡𝑦 ; A1 is the 4N × 4N known matrix originating
from the boundary conditions (7a) and (7b) and the integration of the
kernels on the boundary elements; b1 is the 4N × 1 vector of 𝛾 3 and 𝛿 3
as defined in Eqs. (7a) and (7b); C1 is the 4N × (2M + 2N) known coef-
ficient matrix. Note that the local boundary tractions 𝑡𝐿𝑙 𝐿𝑙
𝑥 and 𝑡𝑦 in the
boundary conditions (7a) and (7b) can be written in terms of u, u, n , v
and v, n by using the transformation matrix due to the rotation of the
axes and the finite difference procedure.
Similarly, for the case of the transverse displacement w, applying the
boundary integral Eqs. (25) and (27) and the boundary conditions (9a)
and (9b) at the N boundary nodes as well as Eq. (26) and the corner
condition (9c) at the Nc corner points, one obtains the matrix equation
as
𝐀 2 𝐱 2 =𝐛 2 +𝐂 2 𝐝 2 , (30)
where x2 = [w, w, n , VL (w),ML (w),w
c ,TL (w)]T
is the (4N + 2Nc ) × 1 vec-
tor of the boundary and corner quantities; d2 is the M × 1 vector of the
unknown coefficients for 𝑎(3) 𝑗 ; A2 is the (4N + 2Nc ) × (4N + 2Nc ) known
matrix originating from the boundary and corner conditions (9a)–(9c)
and the integration of the kernels on the boundary elements; b2 is the
(4N + 2Nc ) × 1 vector of 𝛼 3 , 𝛽 3 and c3k as defined in Eqs. (9a)–(9c); C2
is the (4N + 2Nc ) × M known coefficient matrix. Note that the tangential
derivative w, t in Eq. (9a) can be expressed in terms of w by employing
the transformation matrix due to the rotation of axes and the finite dif-
ference scheme. Then, Eqs. (29) and (30) are exploited to express the
vectors x1 and x2 in terms of the unknown variable d1 and d2 , respec-
tively, as the following:
𝟏
( )
𝐱𝟏 = 𝐀−𝟏 𝐛𝟏 +𝐂𝟏 𝐝𝟏 , (31)

( )
𝐱2 = 𝐀−1
2 𝐛2 +𝐂2 𝐝2 . (32)

Subsequently, collocating at the M interior nodal points by using Eqs.

(18a) and (18b) for the membrane displacements u, v and Eq. (28) for
the transverse displacement w. Hence, we obtain the displacement fields
inside the domain Ω, respectively, as follows:
𝐮̄ ,𝑎𝑏 = 𝐃(1) 𝐱 +𝐂(1)
,𝑎𝑏 𝟏𝐮
𝐝 ,
,𝑎𝑏 𝟏
(33)

𝐯̄ ,𝑎𝑏 = 𝐃(2) 𝐱 + 𝐂(2)

,𝑎𝑏 𝟏𝐯
𝐝 ,
,𝑎𝑏 𝟏
(34)

𝐰̄ ,𝑎𝑏𝑐𝑑 = 𝐃(3) 𝐱 + 𝐂(3)

,𝑎𝑏𝑐𝑑 𝟐
𝐝 .
,𝑎𝑏𝑐𝑑 𝟐
(35)
Fig. 18. The non-dimensional central deflections of SLGS versus the distributed load for
And then applying the first derivative of Eq. (25) at the N boundary the boundary condition case (b).
nodes, one gets
1
𝐰 = 𝐃(4) ,𝑎 𝐱𝟐 + 𝐂,𝑎 𝐝𝟐 ,
(4)
(36)
2 ,𝑎 first derivative of the integration of the kernels on the boundary ele-
where 𝐮̄ ,𝑎𝑏 , 𝐯̄ ,𝑎𝑏 , 𝐰̄ ,𝑎𝑏𝑐𝑑 are the M × 1 vectors of the involved derivatives of ments in Eq. (25); x1u = [u, u, n ]T and x1v = [v, v, n ]T are the 2N × 1
the membrane displacement u, v and the deflection w inside the domain vectors of the boundary quantities which are the complement of the
Ω, respectively; w, a is the N × 1 vector of the involved first derivative of vector x1 ; 𝐂(1)
,𝑎𝑏
, 𝐂(2)
,𝑎𝑏
are the M × (2M + 2N) known coefficient matrices
of 𝑎(1) (2) 𝐿𝑛𝑙 𝐿𝑛𝑙 (3) (4)
the deflection w on the boundary Γ; 𝐃(1) , 𝐃(2) and 𝐃(3) are the M × 2N 𝑗 , 𝑎𝑗 and 𝑡𝑥 , 𝑡𝑦 ; 𝐂,𝑎𝑏𝑐𝑑 and 𝐂,𝑎 are the M × M and N × M known
,𝑎𝑏 ,𝑎𝑏 ,𝑎𝑏𝑐𝑑
and M × (4N + 2Nc ) known matrices origination from the integration of coefficient matrices of 𝑎(3)
𝑗 , respectively.
the kernels on the boundary elements in Eqs. (18a), (18b) and (28), re- Finally, by substituting Eqs. (33)–(35) into Eqs. (6a)–(6c) and Eq.
spectively; 𝐃(4)
,𝑎 is N × (4N + 2Nc ) known matrices origination from the (36) into Eqs. (8c) and (8d), we obtain the following system of the

100
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 19. The displacement ratio R at the center of SLGS with various nonlocal parameters for the boundary condition case (a).

Fig. 20. The displacement ratio R at the center of SLGS with various nonlocal parameters for the boundary condition case (b).

3M + 2N nonlinear algebraic equations for 𝐚(1) (2) (3) 𝐿𝑛𝑙 𝐿𝑛𝑙

𝑗 , 𝐚𝑗 , 𝐚𝑗 and 𝐭𝑥 , 𝐭 𝑦 , 𝐭𝑦𝐿𝑛𝑙 = 𝐋2 (𝐚(3)
𝑗 ), (37e)
namely,
𝐁(1) 𝐚(1) + 𝐁(1) 𝐚(2) + 𝐅(1) 𝐭 𝐿𝑛𝑙 + 𝐅(1) 𝐭 𝐿𝑛𝑙 = 𝐉𝟏 (𝐚(3) where 𝐁(1) (2) (1) (2)
𝑖 , 𝐁𝑖 and 𝐅𝑖 , 𝐅𝑖 , (𝑖 = 1, 2) are the M × M and M × N known
𝟏 𝑗 𝟐 𝑗 𝟏 𝑥 𝟐 𝑦 𝑗 ), (37a)
matrices, respectively; J1 , J2 and L1 , L2 are the M × 1 and N × 1 vector
depending on the unknown vectors 𝐚(3) 𝑗 ; K is the M × 1 vector nonlin-
𝐁(2)
𝟏 𝑗
𝐚(1) + 𝐁(2)
𝟐 𝑗
𝐚(2) + 𝐅(2)
𝟏 𝑥
𝐭 𝐿𝑛𝑙 + 𝐅(2)
𝟐 𝑦
𝐭 𝐿𝑛𝑙 = 𝐉𝟐 (𝐚(3)
𝑗 ), (37b) (2) (3) 𝐿𝑛𝑙 𝐿𝑛𝑙
early depending on the components of 𝐚(1)
𝑗 , 𝐚𝑗 , 𝐚𝑗 , 𝐭𝑥 and 𝐭𝑦 . The sys-
(2) (3) 𝐿𝑛𝑙 𝐿𝑛𝑙 tem of the nonlinear algebraic Eqs. (37a)–(37e) can be solved to obtain
𝐊(𝐚(1)
𝑗 , 𝐚𝑗 , 𝐚𝑗 , 𝐭𝑥 , 𝐭𝑦 ) = 𝟎, (37c)
(2) (3) 𝐿𝑛𝑙
3M + 2N variables (𝐚(1)
𝑗 , 𝐚𝑗 , 𝐚𝑗 , 𝐭𝑥 and 𝐭𝑦𝐿𝑛𝑙 ) by using the numerical
𝐭𝑥𝐿𝑛𝑙 = 𝐋𝟏 (𝐚(3)
𝑗 ), (37d) method.

101
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 21. The displacement ratio R at the center of SLGS with various nonlocal parameters for the boundary condition case (c).

Fig. 22. The displacement ratio R at the center of SLGS with various nonlocal parameters for the boundary condition case (d).

However, Eqs. (37a) and (37b) can be solved for 𝐚(1) (2)
𝑗 and 𝐚𝑗 and
from Eqs. (33)–(35). For the point P not coinciding with the collocation
domain points, the respective quantities can be established from the
expressed in terms of the variable 𝐚(3)
𝑗 . Consequently, in order to solve
discretized counterparts of Eqs. (18a), (18b) and (28).
the nonlinear system expediently, Eqs. (37a)–(37e) can be combined to
the system of Mnonlinear equations as
5. Numerical results and discussion
𝐒(𝐚(3)
𝑗 ) = 𝟎, (38)
5.1. Convergence and validation
Eq. (38) can be solved to evaluate the M unknown variables of 𝐚(3)
𝑗 .
Then, the values of the membrane displacements u, v and the transverse In order to demonstrate the efficiency, accuracy and reliability of the
deflection w and their derivatives inside the domain Ω are determined presented method, a computer program has been developed and used to

102
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 23. The displacement ratio R at the center of SLGS with various boundary conditions for 𝜇 = 2 nm2 .

Fig. 24. The displacement ratio R at the center of SLGS with various boundary conditions for 𝜇 = 4 nm2 .

analyze conventional plates and nanoplates with various geometries and
boundary conditions. The accuracy of the proposed technique is evalu-
ated by comparing the obtained results with those available from other
works. In this study, the Thin Plate Splines (TPSs) fj is employed as the
approximation functions, which is previously applied for analyzing vari-
ous plate problems. With TPSs, no shape parameter is required to obtain
the solution while the accurate numerical results have been successfully
achieved [34,37]. It is defined as
𝑓𝑗 (𝑟) = 𝑟2 ln 𝑟.
The thin plate spline fj and the derivatives of 𝑢̂ 𝑗 and 𝑤̂ 𝑗 involving
in Eqs. (17a), (17b), (18a), (18b) and (25)–(28) are presented in the
Appendix.
5.1.1. A simply supported square plate
Firstly, the effects of different numbers of boundary elements N
and internal collocation points M on nonlinear deflection responses
are investigated to study the convergence of the solution. Nonlinear
bending of a simply supported square plate subjected to a uniform
(39) load (𝑓̄ = 𝑓 𝑎4 ∕𝐸 ℎ4 = 50) has been analyzed. The properties of plate are

103
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 25. The effect of aspect ratio (a/b) on the non-dimensional central deflection of SLGS for the boundary condition case (a).

Fig. 26. The effect of aspect ratio (a/b) on the non-dimensional central deflection of SLGS for the boundary condition case (b).

Table 1 value when the numbers of boundary elements N and internal collo-
Properties of a square plate [50].
cation points M have more than 320 and 64, respectively. Therefore,
Properties Values these numbers of N and M are the optimum values for analyzing this
Modulus of Elasticity, E 7.8 × 106 psi
problem.
Poisson’s ratio, 𝜐 0.3 Subsequently, the numerical results have been evaluated by using
Thickness, h 1 in N = 320 constant boundary elements and M = 64 internal nodes and
compared with the results from the analytical solution [49] and the
contained in Table 1. The non-dimensional deflection (𝑤̄ 𝑐 = 𝑤𝑐 ∕ℎ) at the finite element method (FEM) [50]. In Table 2, it is apparent that the
center of plate are evaluated and plotted in Figs. 3 and 4 for variation present results are in good accordance with the available solutions and,
of the values N and M, respectively. As can be seen in Figs. 3 and 4, especially, they closely agree with the analytical solution that demon-
the computed results have small discrepancies and converge to a stable strates very good accuracy of the proposed methodology.

104
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Table 2
Non-dimensional deflection at the center of a simply supported square plate.

𝑓̄ = 𝑓 𝑎4
𝐸 ℎ4
𝑤̄ 𝑐 = 𝑤𝑐 ∕ℎ Difference, %

(1) Analytical solution [49] (2) FEM [50] (3) Present Between (2) and (1) Between (3) and (1)

6.25 0.2602 0.2790 0.2534 7.23 2.61

12.5 0.4452 0.4630 0.4443 4.00 0.20
25 0.6918 0.6911 0.6978 0.10 0.87
50 0.9794 0.9575 1.0008 2.24 2.19
75 1.1726 1.1333 1.2010 3.35 2.42
100 1.3228 1.2688 1.3546 4.08 2.40
125 1.4477 1.3809 1.4813 4.61 2.32
150 1.5558 1.4774 1.5900 5.04 2.20
175 1.6516 1.5628 1.6860 5.38 2.08
200 1.7382 1.6398 1.7724 5.66 1.97

Table 3 5.1.4. A simply supported single layered graphene sheet

Properties of a clamped circular plate.
Properties Values
Modulus of Elasticity, E 7 × 107
Poisson’s ratio, 𝜐 0.3
Thickness, h 1 first-order shear deformation plate theory (FSDT) and nonlocal elas-
Radius of plate, R 10
5.1.2. A clamped circular plate
In this example, a clamped circular plate subjected to uniform load
has been analyzed. The properties of a clamped circular plate are pro-
vided in Table 3. The results have been obtained with N = 300 constant
boundary elements and M = 60 internal nodes. The non-dimensional de-
flection (𝑤̄ 𝑐 = 𝑤𝑐 ∕ℎ) at the center of a clamped circular plate are listed
in Table 4 and also plotted in Fig. 5 to compare with those obtained from
work of Zhang and Kim [51] which used a new refined non-conforming
triangular plate/flat shell element for solving the nonlinear bending
problem. As can be seen in Table 4 and Fig. 5, all of the present results
are in good accordance with the finite element method and very close
agreement with the analytical solution, thus confirming the reliability
and applicability of the proposed technique.
5.1.3. Plate with complex geometry and boundary conditions
To illustrate the effectiveness of the presented methodology, the non-
linear bending response of plate with complex geometry and boundary
conditions as shown in Fig. 6 has been analyzed. The material prop-
erties are E = 3 × 107 kN/m2 , 𝜐 = 0.3 and h = 0.05 m. Fig. 7 shows the
non-dimensional deflection at point A versus the uniform load f. Fur-
thermore, Fig. 8 displays the non-dimensional deflection along the x-
axis when f = 10 kN/m2 . All the numerical results have been obtained
by using N = 400 constant boundary elements and M = 67 internal nodes
and also compared with those from the analog equation method (AEM)
of Katsikadelis [47]. As can be seen in Figs. 7 and 8, the present results
are almost coincident with the AEM that indicate the efficiency of the
present method.
To evaluate the correctness of the presented formulation and tech-
nique for nonlinear bending analysis of nanoplates, the simply supported
SLGS have been analyzed and the numerical results are compared with
those obtained from the differential quadrature method based on the
ticity of Eringen. Table 5 displays the properties of orthotropic SLGS
employed in work of Golmakani and Rezatalab [28]. Meanwhile, the
present analysis
√ is performed with approximation of the modulus of elas-
ticity 𝐸 = 𝐸11 𝐸22 for isotropic materials as conforming to the assump-
tion in this study. Table 6 shows the central deflections of nanoplates
which are calculated by using N = 320 constant boundary elements and
M = 64 internal nodes. It can be observed from Table 6 that the obtained
results are in a close agreement with their solutions. Thus, the validation
of nonlinear bending analysis of nanoplates in this study is confirmed.
5.1.5. The single layered graphene sheet with various boundary supports
To verify the reliability of the nonlocal nonlinear formulations at
present study and application of proposed methodology for analyzing
nanoplates with various boundary supports, the SLGS subjected to a uni-
form load is analyzed. The obtained results have been compared with
those from the atomistic–continuum coupled model which was proposed
by Singh and Patel [52]. The numerical results are obtained by using
N = 320 constant boundary elements and M = 64 internal nodes and the
material properties of Zigzag sheet IV (see Table 5) are employed for this
analysis. Fig. 9 shows the relationships between the central deflections
and the uniform loads for nanoplates with various boundary supports.
What is striking in Fig. 9 is that the nonlocal model in the present study
agrees with the results from the atomistic–continuum coupled model by
setting the proper nonlocal parameter. Thus, the nonlocal formulation
in this study has accuracy and reliability for investigating the nonlinear
bending response of nanoplates.
5.2. Parametric studies
This section aims to investigate the influences of the pertaining pa-
rameters on nonlinear bending responses of nanoplates. The material
properties of the graphene sheet in Table 7 are employed in this study

Table 4
Non-dimensional deflection at the center of a clamped circular plate.

𝑓̄ = 𝑓 𝑅4
𝐸 ℎ4
𝑤̄ 𝑐 = 𝑤𝑐 ∕ℎ Difference (%)

(1) Analytical solution [51] (2) FEM [51] (3) Present Between (2) and (1) Between (3) and (1)

1 0.169 0.171 0.167 1.18 1.07

2 0.323 0.333 0.322 3.10 0.19
3 0.457 0.475 0.460 3.94 0.61
6 0.761 0.796 0.777 4.60 2.13
10 1.035 1.084 1.063 4.73 2.74
15 1.279 1.339 1.313 4.69 2.68

105
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Table 5
Material properties of the simply supported SLGS [27,28].

Material Properties E11 (GPa) E22 (GPa) 𝜐12 lx (nm) ly (nm) h (nm)

Armchair sheet I 2434 2473 0.197 9.519 4.844 0.129

√
Approximate Armchair I 𝐸 = 𝐸11 𝐸22 = 2453
Zigzag sheet IV 2145 2097 0.223 9.496 4.877 0.145
√
Approximate Zigzag sheet IV 𝐸 = 𝐸11 𝐸22 = 2121

Table 6
The central deflection of the simply supported SLGS with 𝜇 = 0.64 nm2 .

f (MPa) Golmakani and Rezatalab [28] Present Difference, %

Armchair sheet I 20 0.0953 0.0939 1.51

40 0.1269 0.1254 1.15
60 0.1483 0.1468 0.99
70 0.1571 0.1556 0.94
Zigzag sheet IV 20 0.0934 0.0912 2.38
40 0.1267 0.1239 2.20
60 0.1491 0.1459 2.14
80 0.1664 0.1631 1.99

Table 7 Moreover, in order to assess the relationships between the non-

Material properties of a graphene sheet [19].
dimensional defection at the center of SLGS and each type of distributed
Properties Values loads, Figs. 17 and 18 are provided. The square nanoplates (a × a) with
Modulus of Elasticity, E 1.06 TPa boundary conditions cases (a) and (b) as shown in Fig. 10 are investi-
Poisson’s ratio, Ʋ 0.25 gated by performing both linear and nonlinear analyses. As can be seen
Density, 𝜌 2250 kg/m3 in Figs. 17 and 18, the linear analysis indicates that an increasing of
Thickness, h 0.34 nm the nonlocal parameter results in a decreasing of the slope of the load-
defection in load type (1). In contrast, the opposite results are obtained
for load type (2). And there is no effect of the nonlocal parameter on the
and it is considered to be an isotropic material [19]. Moreover, in or- linear bending response for load type (3), that is a uniform load. Further-
der to study effects of various boundary conditions, the four cases of more, the nonlinear analysis gives similar consequences with the linear
boundary conditions as shown in Fig. 10 are analyzed. analysis when the small defection problem is considered. Interestingly,
for the large deflection (e.g. 𝑤̄ 𝑐 > 0.2), the increasing of the nonlocal pa-
5.2.1. The effect of nonlocal parameters on the effective distributed load
rameter causes a significant decrease in the non-dimensional deflection
feff = (1 − 𝜇∇2 )f
for all type of loads. What stands out in this finding is the dominance of
Consider feff = (1 − 𝜇∇2 )f in the right hand side of Eq. (6c). Variations
the membrane force for the large deflection in nonlinear bending anal-
of nonlocal parameters 𝜇 can lead to interesting effects on an intensity of
ysis.
the effective distributed load, depending on the characteristic function f.
To clarify this point, assume that a nanoplate is subjected to a transverse
5.2.2. The effect of the nonlocal parameter and the von Kármán
load f which is defined by a cubic polynomial function as
( ) nonlinearity on bending responses
𝑓2 − 𝑓1 In order to investigate the influence of nonlocal parameters and the
𝑓 = 𝑥3 + 𝑓1 , (40)
𝑎3 von Kármán nonlinearity on bending responses, a square SLGS (a × a)
when f1 and f2 are the intensities of the distributed load at x = 0 and subjected to a uniform load is analyzed. All obtained results are com-
x = a, respectively. Therefore, the effective distributed load can be writ- puted by using N = 320 constant boundary elements and M = 64 internal
ten as collocation points, and they are presented in the form of the displace-
( ) ment ratio R at the central SLGS, which is defined as
𝑓2 − 𝑓1 ( 3 )
𝑓𝑒𝑓 𝑓 = (1 − 𝜇∇2 )𝑓 = 𝑥 − 6𝜇𝑥 + 𝑓1 . (41) 𝑤
𝑎 3
𝑅 = nonlinear , (42)
𝑤linear
Eq. (41) shows that the characteristic distributions of the effective
load depend on the values of f1 and f2 and they can be divided into where wnonlinear and wlinear are the deflections computed by nonlinear
three types, namely; type (1) for f2 > f1 , type (2) for f2 < f1 and type and linear analyses, respectively.
(3) for f2 = f1 . The load distributions of each type are illustrated in Figs. Figs. 19–22 show the displacement ratio R at the center of a square
11–13, respectively. Then, Figs. 14–16 are established in order to eval- SLGS corresponding to each case of boundary conditions illustrated in
uate the influence of nonlocal parameters on the variant distributions Fig. 10 with different values of nonlocal parameters. It is obvious that
of three effective loads along the x-axis, respectively. As a result of load all figures show a similar trend, i.e., the increasing value of nonlo-
type (1) in Fig. 14, it is indicated that the increasing value of the non- cal parameters causes a significant decrease in the displacement ratio.
local parameter produces a decreasing of the effective load intensity. This implies that an increasing nonlocal parameter escalates the mem-
Meanwhile, as shown in Fig. 15, the opposite result is given for load brane force or the effect of the von Kármán nonlinearity. Additionally,
type (2), i.e., a rise in nonlocal parameter causes also an increase in the it is worth to notice that, for the small displacement (e.g., wlinear /h <
effective load intensity. Finally, for load type (3) in Fig. 16 when f2 = f1 , 0.1), the displacement ratios R are almost equal to unity. Therefore,
Eq. (41) becomes a constant function (feff = f1 = f2 ) that the nonlocal pa- nonlinear bending analysis of nanoplates may not be required for the
rameter has no effects on an intensity of the effectively distributed load. small deflection. Moreover, Figs. 23 and 24 display the central displace-
Therefore, these findings point out the effects of nonlocal parameters on ment ratio R of a square SLGS with various boundary conditions for
the effective loads which directly depend on the characteristic function 𝜇 = 2 nm2 and 𝜇 = 4 nm2 , respectively. As a result, the von Kármán ge-
of a distributed load. ometric nonlinearity has more effect on the nanoplates with flexible

106
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

Fig. 27. The effect of aspect ratio (a/b) on the non-dimensional central deflection of SLGS for the boundary condition case (c).

Fig. 28. The effect of aspect ratio (a/b) on the non-dimensional central deflection of SLGS for boundary condition in case (d).

supports (e.g., the FSFS nanoplate) than those with rigid supports (e.g.,
the CCCC nanoplate). Therefore, the nanoplates with flexible supports
require nonlinear bending analysis at lower wlinear /h than those with
rigid supports.
5.2.3. The effect of the aspect ratio on nonlinear bending responses
The influences of the aspect ratio (a/b) on the non-dimensional de-
flection (𝑤̄ 𝑐 = 𝑤𝑐 ∕ℎ) at the center of SLGS are illustrated in Figs. 25–
28. The SLGS with different boundary conditions (see Fig. 10) under a
uniform load (𝑓̄ = 𝑓 𝑎4 ∕𝐸 ℎ4 = 100) is analyzed. All results indicate that
the increasing of the aspect ratios leads to the reducing of the non-
dimensional deflections at the center of the nanoplates. According to
Section 5.2.2, it can be noticed that the central displacements also de-
crease when the nonlocal parameters increase. However, the effect of
nonlocal parameter has very small domination when the aspect ratios
a/b are more than 3, Therefore, the classical continuum model may be
used for nonlinear bending analysis of nonlocal nanoplates when the
aspect ratios a/b > 3.

107
M. Panyatong et al.
Fig. 29. The T-shaped nanoplate with 𝜇 = 1 nm2 and variation of the non-dimensional deflections along the x-axis with different uniform loads.
5.3. Application of the boundary-only method to a nanoplate in
nanodevices
This example demonstrates an application of the proposed method
for analyzing the nonlinear bending of the T-shaped nanoplate, as shown
in Fig. 29, which is often found in nanodevices such as resonators and
mass sensors. The edge AB is clamped. Meanwhile, the other edges are
free. The numerical results have been obtained using N = 220 constant
boundary elements and M = 58 internal nodes. The material properties
in Table 7 are employed for this analysis. The variations of the com-
puted deflections along the x-axis at y = 0 with the different uniform
loads are presented in Fig. 29. Hence, the powerful methodology pre-
sented in this article provides the first numerical investigation into the
nonlinear bending of the nanoplates with complex shapes and boundary
conditions.
6. Conclusions
In this paper, the geometrically nonlinear bending of the nonlo-
cal nanoplate has been studied by the boundary-only method. The
108
Engineering Analysis with Boundary Elements 87 (2018) 90–110
presented method is based on the principle of the analog equation,
which the original differential equations are replaced by three uncou-
pled equations with fictitious sources under the same boundary condi-
tions. The fictitious sources are established by using a technique based
on the BEM and approximated by using Thin Plate Splines (TPSs) as
the radial basis functions. The solution of the actual problem is ob-
tained from the known integral representations of the potential and
plate problems which are derived by using the fundamental solutions
of the Laplace and biharmonic operators. The following conclusion can
be drawn from this study:
1. The presented method remains all advantages of the BEM, i.e., the
discretizations and integrations are performed only on the boundary.
The nanoplates with general shapes and boundary conditions can be
analyzed effectively.
2. The evaluated results from the nonlocal formulation agree with
those from the atomistic–continuum coupled model by specifying
the proper nonlocal parameter.
M. Panyatong et al. Engineering Analysis with Boundary Elements 87 (2018) 90–110

3. The influence of the nonlocal parameter on an intensity of the effec-

1 3
tive distributed load depends on the characteristic of the transverse (∇2 𝑤̂ 𝑗 ),𝑛 = 𝑢̂ 𝑗,𝑛 = 𝑟 𝑟 (4 ln 𝑟 − 1), (A12)
16 ,𝑛
load function.
4. Within the small displacement range, the effect of the nonlocal pa- 1 5
𝑤̂ 𝑗,𝑥 = 𝑟 𝑟 (3 ln 𝑟 − 2), (A13)
rameter reveals the similar behavior in the bending responses for 288 ,𝑥
both linear and nonlinear analyses.
1 5
5. Within the large displacement range, an increasing of the nonlocal 𝑤̂ 𝑗,𝑦 = 𝑟 𝑟 (3 ln 𝑟 − 2), (A14)
parameters always results in a decrease in the non-dimensional de- 288 ,𝑦
flection for the geometrically nonlinear analysis due to the domi- ( )
1 4
nance of the membrane force. 𝑤̂ 𝑗,𝑥𝑥 = 𝑟 15𝑟2,𝑥 ln 𝑟 + 3𝑟2,𝑦 ln 𝑟 − 7𝑟2,𝑥 − 2𝑟2,𝑦 , (A15)
288
6. The nonlocal parameter has more influence on nanoplates with flex- ( )
ible supports than those with rigid supports. 1 4
𝑤̂ 𝑗,𝑦𝑦 = 𝑟 15𝑟2,𝑦 ln 𝑟 + 3𝑟2,𝑥 ln 𝑟 − 7𝑟2,𝑦 − 2𝑟2,𝑥 , (A16)
7. The proposed methodology can be regarded as an efficient numeri- 288
cal tool for analyzing the nonlinear bending behavior of nanoplates 1 4
𝑤̂ 𝑗,𝑥𝑦 = 𝑟 𝑟 𝑟 (12 ln 𝑟 − 5), (A17)
in nanodevices. Furthermore, this methodology can be applied to 288 ,𝑥 ,𝑦
various problems in nanomechanics with complicated mathematical
formulations yielding the reliable numerical solutions. 1 3
(∇2 𝑤̂ 𝑗 ),𝑥 = 𝑢̂ 𝑗,𝑥 = 𝑟 𝑟 (4 ln 𝑟 − 1), (A18)
16 ,𝑥
Appendix 1 3
(∇2 𝑤̂ 𝑗 ),𝑦 = 𝑢̂ 𝑗,𝑦 = 𝑟 𝑟 (4 ln 𝑟 − 1), (A19)
16 ,𝑦
The Thin Plate Splines (TPSs) is defined by
1 2 2 1 2 2
𝑓𝑗 (𝑟) = 𝑟2 ln 𝑟, (A1) (∇2 𝑤̂ 𝑗 ),𝑥𝑥 = 𝑢̂ 𝑗,𝑥𝑥 = 𝑟 𝑟 (4 ln 𝑟 − 1) + 𝑟 𝑟 (12 ln 𝑟 + 1), (A20)
16 ,𝑦 16 ,𝑥
√
where 𝑟 = (𝑥 − 𝑥𝑗 )2 + (𝑦 − 𝑦𝑗 )2 𝑗 = 1, 2, … , 𝑀, 1 2 2 1 2 2
(∇2 𝑤̂ 𝑗 ),𝑦𝑦 = 𝑢̂ 𝑗,𝑦𝑦 = 𝑟 𝑟 (4 ln 𝑟 − 1) + 𝑟 𝑟 (12 ln 𝑟 + 1), (A21)
and (xj ,yj ) are the collocation points inside Ω. 16 ,𝑥 16 ,𝑦

1 2
A.1. The particular solution of ∇2 𝑢̂ 𝑗 = 𝑓𝑗 and the derivatives of 𝑢̂ 𝑗 (∇2 𝑤̂ 𝑗 ),𝑥𝑦 = 𝑢̂ 𝑗,𝑥𝑦 = 𝑟 𝑟 𝑟 (8 ln 𝑟 + 2). (A22)
16 ,𝑥 ,𝑦
Moreover, it can be shown that
The particular solution of ∇2 𝑢̂ 𝑗 = 𝑓𝑗 is
lim 𝑤̂ 𝑗 = 0, lim 𝑤̂ 𝑗,𝑛 = 0, lim ∇2 𝑤̂ 𝑗 = 0, lim (∇2 𝑤̂ 𝑗 ),𝑛 = 0,
1 4 𝑟→0 𝑟→0 𝑟→0 𝑟→0
𝑢̂ 𝑗 = 𝑟 (2 ln 𝑟 − 1), (A2)
32 lim 𝑤̂ 𝑗,𝑥 = 0, lim 𝑤̂ 𝑗,𝑦 = 0,
𝑟→0 𝑟→0
and by differentiation, it results in
lim 𝑤̂ 𝑗,𝑥𝑥 = 0, lim 𝑤̂ 𝑗,𝑦𝑦 = 0, lim 𝑤̂ 𝑗,𝑥𝑦 = 0,
1 3 𝑟→0 𝑟→0 𝑟→0
𝑢̂ 𝑗,𝑛 = 𝑟 𝑟 (4 ln 𝑟 − 1), (A3)
16 ,𝑛 lim (∇2 𝑤̂ 𝑗 ),𝑥 = 0, lim (∇2 𝑤̂ 𝑗 ),𝑦 = 0,
𝑟→0 𝑟→0

1 3 lim (∇2 𝑤̂ 𝑗 ),𝑥𝑥 = 0, lim (∇2 𝑤̂ 𝑗 ),𝑦𝑦 = 0, lim (∇2 𝑤̂ 𝑗 ),𝑥𝑦 = 0.

𝑢̂ 𝑗,𝑥 = 𝑟 𝑟 (4 ln 𝑟 − 1), (A4) 𝑟→0 𝑟→0 𝑟→0
16 ,𝑥
References
1 3
𝑢̂ 𝑗,𝑦 = 𝑟 𝑟 (4 ln 𝑟 − 1), (A5)
16 ,𝑦 [1] Eringen AC. On differential equations of nonlocal elasticity and solutions of screw
dislocation and surface waves. J Appl Phys 1983;54:4703–10.
1 2 2 1 2 2 [2] Eringen AC. Nonlocal continuum field theories. New York (NY): Springer; 2002.
𝑢̂ 𝑗,𝑥𝑥 = 𝑟 𝑟 (4 ln 𝑟 − 1) + 𝑟 𝑟 (12 ln 𝑟 + 1), (A6)
16 ,𝑦 16 ,𝑥 [3] Juntarasaid C, Pulngern T, Chucheepsakul S. Bending and buckling of nanowires
including the effects of surface stress and nonlocal elasticity. Phys E 2012;46:68–76.
[4] Malekzadeh P, Shojaee M. Surface and nonlocal effects on the nonlinear free vibra-
1 2 2 1 2 2
𝑢̂ 𝑗,𝑦𝑦 = 𝑟 𝑟 (4 ln 𝑟 − 1) + 𝑟 𝑟 (12 ln 𝑟 + 1), (A7) tion of non-uniform nanobeams. Compos B Eng 2013;52:84–92.
16 ,𝑥 16 ,𝑦 [5] Thongyothee C, Chucheepsakul S. Postbuckling behaviors of nanorods includ-
ing the effects of nonlocal elasticity theory and surface stress. J Appl Phys
1 2 2013;114(24):243507.
𝑢̂ 𝑗,𝑥𝑦 = 𝑟 𝑟 𝑟 (8 ln 𝑟 + 2). (A8)
16 ,𝑥 ,𝑦 [6] Thongyothee C, Chucheepsakul S. Postbuckling of unknown-length nanobeam con-
sidering the effects of nonlocal elasticity and surface stress. Int J Appl Mech
Furthermore, it can be readily proved that 2015;7(3):1550042.
[7] Ahababaei R, Reddy JN. Nonlocal third-order shear deformation plate theory with
lim 𝑢̂ 𝑗 = 0, lim 𝑢̂ 𝑗,𝑛 = 0, lim 𝑢̂ 𝑗,𝑥 = 0, lim 𝑢̂ 𝑗,𝑦 = 0, application to bending and vibration of plate. J Sound Vib 2009;326:277–89.
𝑟→0 𝑟→0 𝑟→0 𝑟→0
[8] Pradhan SC, Murmu T. Small scale effect on the buckling of single-layered graphene
lim 𝑢̂ 𝑗,𝑥𝑥 = 0, lim 𝑢̂ 𝑗,𝑦𝑦 = 0, lim 𝑢̂ 𝑗,𝑥𝑦 = 0. sheets under biaxial compression via nonlocal continuum mechanics. Comput Mater
𝑟→0 𝑟→0 𝑟→0 Sci 2009;47:268–74.
[9] Ansari R, Sahmani S, Arash B. Nonlocal plate model for free vibrations of single-lay-
A.2. The particular solution of ∇4 𝑤̂ 𝑗 = 𝑓𝑗 and the derivatives of 𝑤̂ 𝑗 ered graphene sheets. Phys Lett A 2010;375:53–62.
[10] Murmu T, Adhikari S. Nonlocal vibration of bonded double-nanoplate-systems. Com-
pos B Eng 2011;42:1901–11.
The particular solution of ∇4 𝑤̂ 𝑗 = 𝑓𝑗 is [11] Prandhan SC, Phadikar JK. Nonlocal theory for buckling of nanoplates. Int J Struct
Stab Dyn 2011;11(3):411–29.
1 6 5 6
𝑤̂ 𝑗 = 𝑟 ln 𝑟 − 𝑟 , (A9) [12] Satish N, Narendar S, Gopalakrishnan S. Thermal vibration analysis of orthotropic
576 3456 nanoplates based on nonlocal continuum mechanics. Phys E 2012;44:1950–62.
[13] Hashemi SH, Zare M, Nazemnezhad R. An exact analytical approach for free vi-
and by differentiation, it results in bration of mindlin rectangular nano-plate via nonlocal elasticity. Compos Struct
1 5 2013;100:290–9.
𝑤̂ 𝑗,𝑛 = 𝑟 𝑟 (3 ln 𝑟 − 2), (A10) [14] Pouresmaeeli S, Ghavanloo E, Fazelzadeh SA. Vibration analysis of viscoelas-
288 ,𝑛 tic orthotropic nanoplates resting on viscoelastic medium. Compos Struct
2013;96:405–10.
1 4
∇2 𝑤̂ 𝑗 = 𝑢̂ 𝑗 = 𝑟 (2 ln 𝑟 − 1), (A11) [15] Zenkour AM, Sobhy M. Nonlocal elasticity theory for thermal buckling of nanoplates
32 lying on Winkler–Pasternak elastic substrate medium. Phys E 2013;53:251–9.

109
M. Panyatong et al.
[16] Asemi SR, Farajpour A. Decoupling the nonlocal elasticity equations for thermo-me-
chanical vibration of circular graphene sheets including surface effects. Phys E
2014;60:80–90.
[17] Naderi A, Saidi AR. Nonlocal postbuckling analysis of graphene sheets in a nonlinear
polymer medium. Int J Eng Sci 2014;81:49–65.
[18] Malekzadeh P, Golbahar Haghighi MR, Shojaee M. Nonlinear free vibration of skew
nanoplates with surface and small scale effects. Thin Walled Struct 2014;78:48–56.
[19] Zhang Y, Lei ZX, Zhang LW, Liew KM, Yu JL. Nonlocal continuum model for vibration
of single-layered graphene sheets based on the element-free kp-Ritz method. Eng
Anal Bound Elem 2015;56:90–7.
[20] Yan JW, Tong LH, Li C, Zhu Y, Wang ZW. Exact solutions of bending deflections
for nano-beams and nano-plates based on nonlocal elasticity theory. Compos Struct
2015;125:304–13.
[21] Panyatong M, Chinnaboon B, Chucheepsakul S. Incorporated effects of surface stress
and nonlocal elasticity on bending analysis of nanoplates embedded in an elastic
medium. Suranaree J Sci Technol 2015;22(1):21–33.
[22] Panyatong M, Chinnaboon B, Chucheepsakul S. Nonlocal second-order shear de-
formation plate theory for free vibration of nanoplates. Suranaree J Sci Technol
2015;22(4):339–48.
[23] Panyatong M, Chinnaboon B, Chucheepsakul S. Free vibration analysis of FG
nanoplates embedded in elastic medium based on second-order shear deformation
plate theory and nonlocal elasticity. Compos Struct 2016;153:428–41.
[24] Zhang LW, Liew KM, Reddy JN. Postbuckling analysis of bi-axially compressed lam-
inated nanocomposite plates using the first-order shear deformation theory. Compos
Struct 2016;152:418–31.
[25] Zhang LW. An element-free based IMLS-Ritz method for buckling analysis of
nanocomposite plates of polygonal planform. Eng Anal Bound Elem 2017;77:10–25.
[26] Reddy JN. Nonlocal nonlinear formulations for bending of classical and shear defor-
mation theories of beams and plates. Int J Eng Sci 2010;48:1507–18.
[27] Shen HS, Shen L, Zhang CL. Nonlocal plate model for nonlinear bending of sin-
gle-layer graphene sheets subjected to transverse loads in thermal environments.
Appl Phy A 2011;103:103–12.
[28] Golmakani ME, Rezatalab J. Nonlinear bending analysis of orthotropic nanoscale
plates in an elastic matrix based on nonlocal continuum mechanics. Compos Struct
2014;111:85–97.
[29] Dastjerdi S, Jabbarzadeh M, Tahani M. Nonlinear bending analysis of sector
graphene sheet embedded in elastic matrix based on nonlocal continuum mechanics.
IJE Trans B Appl 2015;28(5):802–11.
[30] Dastjerdi S, Jabbarzadeh M, Aliabadi S. Nonlinear static analysis of single layer an-
nular/circular graphene sheets embedded in Winkler–Pasternak elastic matrix based
on non-local theory of Eringen. Ain Shams Eng J 2016;7:873–84.
[31] Golmakani ME, Far MNS. Nonlinear thermo-elastic bending behavior of graphene
sheets embedded in an elastic medium based on nonlocal elasticity theory. Comput
Math Appl 2016;72:785–805.
[32] Zhang Y, Zhang LW, Liew KM, Yu JL. Nonlocal continuum model for large deforma-
tion analysis of SLGSs using the kp-Ritz element-free method. Int J Non Linear Mech
2016;79:1–9.
110
Engineering Analysis with Boundary Elements 87 (2018) 90–110
[33] Katsikadelis JT. The analog boundary integral equation method for nonlinear static
and dynamic problem in continuum mechanics. J Theor Appl Mech 2002;40:961–84.
[34] Chinnaboon B, Chucheepsakul S, Katsikadelis JT. A BEM-based meshless method for
elastic buckling analysis of plates. Int J Struct Stab Dyn 2007;7:81–99.
[35] Chinnaboon B, Katsikadelis JT, Chucheepsakul S. A BEM-based meshless method for
plates on biparametric elastic foundation with internal supports. Comput Meth Appl
Mech Eng 2007;197:3165–77.
[36] Katsikadelis JT, Babouskos NG. Post-buckling analysis of viscoelastic plates with
fractional derivative models. Eng Anal Bound Elem 2010;34:1038–48.
[37] Chinnaboon B, Chucheepsakul S, Katsikadelis JT. A BEM-based domain meshless
method for the analysis of Mindlin plates with general boundary conditions. Comput
Meth Appl Mech Eng 2011;200:1379–88.
[38] Katsikadelis JT. The BEM for numerical solution of partial fractional differential
equations. Comput Math Appl 2011;62:891–901.
[39] Tsiatas GC, Katsikadelis JT. Nonlinear analysis of elastic space cable-supported mem-
branes. Eng Anal Bound Elem 2011;35:1149–58.
[40] Nerantzaki MS, Babouskos NG. Analysis of inhomogeneous anisotropic viscoelas-
tic bodies described by multi-parameter fractional differential constitutive models.
Comput Math Appl 2011;62:945–60.
[41] Katsikadelis JT, Babouskos NG. Stiffness and buckling optimization of thin plates
with BEM. Arch Appl Mech 2012;82:1403–22.
[42] Nerantzaki MS, Babouskos NG. Vibrations of inhomogeneous anisotropic viscoelas-
tic bodies described with fractional derivative models. Eng Anal Bound Elem
2012;36:1894–907.
[43] Tsiatas GC, Yiotis AJ. A BEM-based meshless solution to buckling and vibration
problems of orthotropic plates. Eng Anal Bound Elem 2013;37:579–84.
[44] Tsiatas GC, Yiotis AJ. Size effect on the static, dynamic and buckling analysis of or-
thotropic Kirchhoff-type skew micro-plates based on a modified couple stress theory:
comparison with the nonlocal elasticity theory. Acta Mech 2015;226:1267–81.
[45] Babouskos NG, Katsikadelis JT. Optimum design of thin plates via frequency opti-
mization using BEM. Arch Appl Mech 2015;85:1175–90.
[46] Katsikadelis JT. Boundary elements: theory and applications. Amsterdam-London:
Elsevier; 2002.
[47] Katsikadelis JT. The boundary element method for plate analysis. Amsterdam–
Boston: Elsevier; 2014.
[48] Katsikadelis JT. The boundary element method for engineers and scientists theory
and applications. 2nd ed. Amsterdam-London: Elsevier; 2016.
[49] Timoshenko S, Woinowsky-Krieger S. Theory of plates and shells. 2nd ed. New York:
McGraw-Hill Book Company; 1959.
[50] Reddy JN. Theory and analysis of elastic plates and shells. 2nd ed. CRC Press, Taylor
& Francis Group; 2007.
[51] Zhang YX, Kim KS. Linear and geometrically nonlinear analysis of plates and shells
by a new refined non-conforming triangular plate/shell element. Comput Mech
2005;36:331–42.
[52] Singh S, Patel BP. Atomistic-continuum coupled model for nonlinear analysis of sin-
gle layer graphene sheets. Int J Non Linear Mech 2015;76:112–19.