Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

International Journal of Non-Linear Mechanics 112 (2019) 117–125

Contents lists available at ScienceDirect

International Journal of Non-Linear Mechanics


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

On a uniformly-valid asymptotic plate theory


Fan-Fan Wang a,b , David J. Steigmann b , Hui-Hui Dai c ,∗
a
Department of Mathematics, East China University of Science and Technology, Shanghai 200237, China
b
Department of Mechanical Engineering, University of California, Berkeley, CA 94720, USA
c
Department of Mathematics and Department of Materials Science and Engineering, City University of Hong Kong, 83 Tat Chee Avenue, Kowloon Tong, Hong Kong

ARTICLE INFO ABSTRACT


Keywords: A uniformly-valid plate theory, independent of the magnitudes of applied loads, is derived based on the two-
Uniformly-valid plate theory dimensional plate theory obtained from series expansions about the bottom surface of a plate. For five different
Asymptotic method magnitudes of surface loads, it is shown by using asymptotic expansions that this unified plate theory recovers
Nonlinear elasticity
five well-known plate models in the literature to leading-order. This demonstrates its uniform validity. More
generally, it provides a uniformly-valid plate model provided that two asymptotic conditions are satisfied,
which can be checked as a posteriori. The weak formulation of the uniformly-valid plate equations is furnished,
which can be used for finite element implementation.

1. Introduction For example, for four different scalings, four different plate models
were derived (see [16]). Naturally, one wonders whether a unified
Two-dimensional plate models have been studied extensively. One plate theory can be derived, which is uniformly valid for all these
approach for deriving plate models from three-dimensional (3D) elas- kinds of scalings. In this paper, we aim to provide such a theory. In
ticity is based on ad hoc hypotheses, and we refer to the book by practice, such a theory is also needed. In a quasi-static loading process,
Ciarlet [1] for a review. On the more mathematical side, two pro- the magnitude of applied loads varies and cannot be of a fixed order
cedures are well-known. One type is based on Gamma convergence, of the thickness. Also, there can be a situation that some part of the
which is concerned with the limiting two-dimensional variational prob- plate undergoes a membrane deformation and some part undergoes a
lem for small thickness [2–5], which gives the leading-order plate bending deformation, and a single plate equation derived by Gamma
theories rigorously. Another type is based on asymptotic methods. One convergence or asymptotic methods based on particular scalings cannot
way is to apply asymptotic expansions to the 3D weak formulation,
cope with this case.
which was developed by Ciarlet and his associates [6–8]. This powerful
Recently, Dai and Song [17] developed an approach for deriving
method can yield convergence results and error estimates (in some
a consistent finite-strain plate theory without any assumption on the
cases) for the derived linear plate theory. For problems with geomet-
magnitudes of applied loads, which stemmed from the method of
rical nonlinearity, in [8], Ciarlet derived formally the von Kármán
coupled series–asymptotic expansions [18–21] for weakly nonlinear
plate equations by this method. Another way is to apply asymptotic
expansions to the 3D differential formulation. This method was ini- theories. More specifically, series expansions about the bottom sur-
tiated in [9] for linearized elasticity (see also [10]). Later on, Erbay face were used for the dimension reduction. Certain algebraic/linear
and Suhubi [11,12] used this approach to derive the von Kármán plate algebraic relations are identified, which lead to recursive relations for
equations for compressible and incompressible isotropic hyperelastic the expansion coefficients. Finally, the top traction condition yields
materials (their zeroth order results) and Erbay [13] derived Föppl the consistent plate equations. This approach was also utilized to
membrane theory. Also by this method, in [14–16], Millet et al. con- deduce finite-strain plate theory for incompressible hyperelastic mate-
sidered four different levels of in-plane and normal surface loads and rials in [22] and dynamic finite-strain plate theory in [23]. Such an
obtained four different plate models, including two membrane models, approach, using series expansions of the local differential equations,
one nonlinear plate model and one linear Kirchhoff–Love model. stands in contrast to that pursued by [24–30], based on expansion of
In the above-mentioned approaches based on Gamma convergence the potential energy for conservative boundary-value problems. In this
and asymptotic methods, some a priori scalings on the applied loads/ paper, based on the plate theory developed in [17], we will propose a
deformations related to the plate thickness need to be imposed. For uniformly-valid plate theory without any assumption on magnitudes of
each kind of imposed scalings, one type of plate model can be derived. surface loads. A large part is to show that, to leading-order, our derived

∗ Corresponding author.
E-mail address: mahhdai@cityu.edu.hk (H.-H. Dai).

https://doi.org/10.1016/j.ijnonlinmec.2019.02.011
Received 19 December 2018; Received in revised form 14 February 2019; Accepted 21 February 2019
Available online 2 March 2019
0020-7462/© 2019 Elsevier Ltd. All rights reserved.
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125

plate theory can recover the plate models in [16] and [13] for five Suppose that the plate is composed of a hyperelastic material with
different levels of in-plane and normal surface loads. Also, our derived the strain energy function 𝛷(𝐅) which is 𝐶 4 in 𝐅. Then the nominal
plate theory is valid for other magnitudes of applied loads as long as stress tensor 𝐒 has the following series expansion:
two asymptotic conditions are satisfied, which can be checked as a 1 2 (2) 1 1 |
𝐒 = 𝐒(0) + 𝑍𝐒(1) + 𝑍 𝐒 + 𝑍 3 𝐒(3) + 𝑍 4 𝐒(4) | , (2.9)
posteriori in practice. In this sense, our plate theory is uniformly-valid. 2! 3! 4! |𝑍=𝑍4
The structure of this paper is as follows. In Section 2, we briefly
where 0 < 𝑍4 < 𝑍 and
recall the derivation of the plate theory in [17] and reformulate the
plate equations into a suitable form for further asymptotic reductions. 𝜕𝛷 (0) 𝜕 2 𝛷 (0) (1)
𝐒(0) = (𝐅 ), 𝐒(1) = (𝐅 )[𝐅 ], (2.10)
In Section 3, to use asymptotic expansions for reductions, we first 𝜕𝐅 𝜕𝐅𝜕𝐅
2
𝜕 𝛷 (0) (2) 3
𝜕 𝛷
nondimensionalize the plate equations. Then, for five different mag- 𝐒(2) = (𝐅 )[𝐅 ] + (𝐅(0) )[𝐅(1) , 𝐅(1) ].
nitudes of surface loads, we carry out the asymptotic analysis to derive 𝜕𝐅𝜕𝐅 𝜕𝐅𝜕𝐅𝜕𝐅
the corresponding plate equations. In Section 4, we conclude that our The explicit expression for 𝐒(3) is not needed and is thus not provided.
plate theory is uniformly valid and the corresponding weak formula- Neglecting the body forces, the field equation is given by
tion together with discussions on its physical meaning are presented.
𝐃𝐢𝐯𝐒 = 𝟎, (2.11)
Finally, we draw some concluding remarks in Section 5.
which leads to the recursive relations for the stress coefficients
2. Two-dimensional plate model
∇ ⋅ 𝐒(0) + (𝐒(1) )T 𝐤 = 𝟎, (2.12)
In this section, we give a brief reaccount of the derivations of the
plate equations in [17]. We consider a hyperelastic plate of uniform ∇ ⋅ 𝐒(1) + (𝐒(2) )T 𝐤 = 𝟎, (2.13)
thickness which occupies the region 𝛺 × [0, 2ℎ] in the reference con-
figuration, where 𝛺 is a plane region with piecewise smooth boundary ∇ ⋅ 𝐒(2) + (𝐒(3) )T 𝐤 = 𝟎. (2.14)
curve 𝜕𝛺 and 2ℎ is the thickness of the plate. It is assumed that the
The linear algebraic equations (2.12) and (2.13) can be used to solve
placement vector 𝐱 = 𝐱(𝑋, 𝑌 , 𝑍) of a material point is a 𝐶 5 function in
𝐱(2) and 𝐱(3) respectively as shown in [17]. The Eq. (2.14) will be used
𝑍, where (𝑋, 𝑌 , 𝑍) is the Cartesian coordinate of a material point in
to eliminate 𝐒(3) later.
the reference configuration. Then we have the Taylor series expansion
The top and the bottom traction conditions are
for 𝐱 based on the bottom surface:
1 2 (2) 1 3 (3) 1 4 (4) 1 5 (5) | 𝐒T 𝐤|𝑍=2ℎ = 𝐪+ , (2.15)
𝐱 = 𝐱(0) + 𝑍𝐱(1) + 𝑍 𝐱 + 𝑍 𝐱 + 𝑍 𝐱 + 𝑍 𝐱 | , (2.1)
2 3! 4! 5! |𝑍=𝑍1
𝜕𝑖 𝐱 | 𝐒T 𝐤|𝑍=0 = −𝐪− , (2.16)
where 𝐱(𝑖) = | for 𝑖 = 0, 1, … , 4 and 0 < 𝑍1 < 𝑍. Corre-
𝜕𝑍 𝑖 |𝑍=0
spondingly, the displacement vector 𝐮 = 𝐱 − 𝐗 has the following series where 𝐪± are the applied tractions on the top and bottom surfaces. By
expansion using (2.9), we have
1 1 1 1 | 1 1
𝐮 = 𝐮(0) +𝑍𝐮(1) + 𝑍 2 𝐮(2) + 𝑍 3 𝐮(3) + 𝑍 4 𝐮(4) + 𝑍 5 𝐮(5) |𝑍 = 𝑍1 . (2.2) (𝐒(0) )T 𝐤 + 2ℎ(𝐒(1) )T 𝐤 + (2ℎ)2 (𝐒(2) )T 𝐤 + (2ℎ)3 (𝐒(3) )T 𝐤 (2.17)
2 3! 4! 5! | 2 3!
1
We note that 𝐮(0) = 𝐱(0) − 𝑋𝐞1 − 𝑌 𝐞2 , 𝐮(1) = 𝐱(1) − 𝐞3 and 𝐮(𝑖) = 𝐱(𝑖) for + (2ℎ)4 (𝐒(4) |𝑍=𝑍5 )T 𝐤 = 𝐪+ ,
4!
𝑖 = 2, 3, 4, 5.
Similarly, we can take Taylor series expansion for the deformation (𝐒(0) )T 𝐤 = −𝐪− , (2.18)
gradient tensor 𝐅 at 𝑍 = 0:
where 0 < 𝑍5 < 2ℎ.
1 2 (2) 1 1 | 𝜕𝛷
𝐅 = 𝐅(0) + 𝑍𝐅(1) + 𝑍 𝐅 + 𝑍 3 𝐅(3) + 𝑍 4 𝐅(4) | , (2.3) Since 𝐒(0) = (∇𝐱(0) + 𝐱(1) ⊗ 𝐤), Eq. (2.18) provides three algebraic
2 3! 4! |𝑍=𝑍2 𝜕𝐅
equations for solving 𝐱(1) in terms of ∇𝐱(0) .
where 𝐅(𝑖) is defined in the same way as 𝐱(𝑖) and 0 < 𝑍2 < 𝑍. It is easy
Subtraction of (2.17) and (2.18), by using (2.12)–(2.14), gives
to obtain that
2 2
𝐅(𝑖) = ∇𝐱(𝑖) + 𝐱(𝑖+1) ⊗ 𝐤, 𝑖 = 0, 1, 2, 3, (2.4) ∇ ⋅ 𝐒(0) + ℎ∇ ⋅ 𝐒(1) + ℎ ∇ ⋅ 𝐒(2) + 𝑂(ℎ3 𝐩) = −𝐪,
̄ (2.19)
3
where 𝐤 = 𝐞3 is the upward unit normal to the reference bottom surface where 𝐪̄ = (𝐪+ + 𝐪− )∕(2ℎ) and 𝐩 = (𝐒(4) |𝑍=𝑍5 )T 𝐤.
𝜕 For the present purpose of asymptotic reductions, we rewrite (2.19)
of the plate and ∇ denotes the in-plane gradient operator ∇ = 𝐞
𝜕𝑋𝛼 𝛼 as
with the summation convention being used. Here and hereafter Greek
indices take values in {1, 2} and Latin indices take values in {1, 2, 3}. ⎧ (0) (1) 2 2 (2) 3
⎪∇ ⋅ 𝐒𝑡 + ℎ∇ ⋅ 𝐒𝑡 + 3 ℎ ∇ ⋅ 𝐒𝑡 + 𝑂(ℎ 𝐩𝑡 ) = −𝐪̄ 𝑡 ,
The series expansion of the Green–Lagrange strain tensor 𝐄 can be ⎨ 2 (2.20)
given by ⎪∇ ⋅ (𝐒(0) 𝐤) + ℎ∇ ⋅ (𝐒(1) 𝐤) + ℎ2 ∇ ⋅ (𝐒(2) 𝐤) + 𝑂(ℎ3 𝑝3 ) = −𝑞̄3 ,
⎩ 3
1( T ) 1 1 1 | where 𝐒(𝑖) (𝑖)
𝐄= 𝐅 𝐅 − 𝐈 = 𝐄(0) + 𝑍𝐄(1) + 𝑍 2 𝐄(2) + 𝑍 3 𝐄(3) + 𝑍 4 𝐄(4) | , 𝑡 = 𝐒𝛼𝛽 𝐞𝛼 ⊗ 𝐞𝛽 (𝑖 = 0, 1, 2), 𝐪𝑡 = 𝑞1 𝐞1 + 𝑞2 𝐞2 , 𝐩𝑡 = 𝑝1 𝐞1 + 𝑝2 𝐞2 ,
2 2 3! 4! |𝑍=𝑍3
𝐪̄ 𝑡 = (𝐪𝑡 + 𝐪𝑡 )∕(2ℎ) and 𝑞̄3 = (𝑞3+ + 𝑞3− )∕(2ℎ). Hereafter the in-plane part
+ −
(2.5) of a vector or a tensor is denoted by a subscript 𝑡. Eq. (2.20) is the plate
equation derived in [17]. The boundary conditions have been discussed
where (⋅)Tdenotes the transpose and 0 < 𝑍3 < 𝑍. By using (2.3) and
extensively in that paper, and here we omit the details.
(2.4), we can show that
Physically, (2.20) (or (2.19)) represents the balance of forces of
1 ( (0) T (0) )
𝐄(0) = (𝐅 ) 𝐅 − 𝐈 , (2.6) a plate element, as it can also be obtained by integrating the field
2
equation (2.11) with respect to 𝑍. On the other hand, from (2.19), upon
1 ( (0) T (1) ) using (2.18), (2.12) and (2.13) (or simply adding (2.17) and (2.18)),
𝐄(1) = (𝐅 ) 𝐅 + (𝐅(1) )T 𝐅(0) , (2.7)
2 one obtains
1 ( (0) T (2) )
𝐄(2) = (𝐅 ) 𝐅 + (𝐅(1) )T 𝐅(1) + (𝐅(2) )T 𝐅(0) . (2.8) (𝐒(0) )T 𝐤 + ℎ(𝐒(1) )T 𝐤 + ℎ2 (𝐒(2) )T 𝐤 + 𝑂(ℎ3 (𝐒(3) )T 𝐤) = 𝐦, (2.21)
2

118
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125

where 𝐦 = (𝐪+ − 𝐪− )∕2. This equation actually represents the balance where 𝛾 = 𝜆∕𝜇. Then, it is easy to obtain
of torques, as it can also be obtained by multiplying the field equation
𝐒(0) = (𝛾Tr𝐄(0) 𝐈 + 2𝐄(0) )(𝐅(0) )T , (3.7)
(2.11) by 𝑍 and then taking an integration. So, one feature of the
present plate model is that the balance of torques is automatically
satisfied when the three equations for balance of forces are satisfied. 𝐒(1) = (𝛾Tr𝐄(0) 𝐈 + 2𝐄(0) )(𝐅(1) )T + (𝛾Tr𝐄(1) 𝐈 + 2𝐄(1) )(𝐅(0) )T , (3.8)
If one does expansions about the middle surface, the resulting plate
model does not enjoy this feature, as shown in Appendix. 𝐒(2) = (𝛾Tr𝐄(0) 𝐈 + 2𝐄(0) )(𝐅(2) )T + 2(𝛾Tr𝐄(1) 𝐈 + 2𝐄(1) )(𝐅(1) )T (3.9)
To put (2.20)2 in a suitable form, we do the manipulation (2.20)2 − (2) (2) (0) T
+(𝛾Tr𝐄 𝐈 + 2𝐄 )(𝐅 ) ,
∇⋅ (2.21), upon using (2.13) and (2.14), to obtain
( ) ( ) where
∇ ⋅ 𝐒(0) 𝐤 − (𝐒(0) )T 𝐤 + ℎ∇ ⋅ 𝐒(1) 𝐤 − (𝐒(1) )T 𝐤 (2.22) (
1
( (2) ) 1 2 𝐄(0) = ∇𝐮(0) + (∇𝐮(0) )T + 𝐮(1) ⊗ 𝐞3 + 𝐞3 ⊗ 𝐮(1) + (∇𝐮(0) )T ∇𝐮(0) (3.10)
2 2 (2) T (1) 2
+ ℎ ∇ ⋅ 𝐒 𝐤 − (𝐒 ) 𝐤 + ℎ ∇ ⋅ ∇ ⋅ 𝐒𝑡 )
3 ( 3
) +(∇𝐮(0) )T (𝐮(1) ⊗ 𝐞3 ) + (𝐞3 ⊗ 𝐮(1) )∇𝐮(0) + (𝐮(1) ⋅ 𝐮(1) )𝐞3 ⊗ 𝐞3
( )
+𝑂 ℎ3 𝑝3 + 𝑂 ℎ3 ∇ ⋅ ∇ ⋅ 𝐒(2) 𝑡 = −𝑞̄3 − ∇ ⋅ 𝐦𝑡 . (
1 (0)
= 𝑢 𝐞 ⊗ 𝐞𝛼 + 𝑢(0) (1) (1)
𝑖,𝛼 𝐞𝛼 ⊗ 𝐞𝑖 + 𝑢𝑖 𝐞𝑖 ⊗ 𝐞3 + 𝑢𝑖 𝐞3 ⊗ 𝐞𝑖
Thus, we obtain the plate equations (2.20)1 and (2.22) for un- 2 𝑖,𝛼 𝑖
knowns 𝐱(0) with 𝐱(1) , 𝐱(2) and 𝐱(3) being obtained from algebraic +𝑢(0) (0) (0) (1) (0) (1)
𝑖,𝛼 𝑢𝑖,𝛽 𝐞𝛼 ⊗ 𝐞𝛽 + 𝑢𝑖,𝛼 𝑢𝑖 𝐞𝛼 ⊗ 𝐞3 + 𝑢𝑖,𝛼 𝑢𝑖 𝐞3 ⊗ 𝐞𝛼
equation (2.18) and linear algebraic equations (2.12) and (2.13), re- )
+𝑢(1) (1)
𝑖 𝑢𝑖 𝐞3 ⊗ 𝐞3 .
spectively. We point out that (2.20)1 and (2.22) are derived without any
assumptions on the magnitudes of 𝐪+ and 𝐪− . In the following section, In particular, the in-plane parts of 𝐒(0) and 𝐄(0) are given by
we shall assume 𝐪+ and 𝐪− are of certain orders of the thickness to do ( )( )
asymptotic reductions and derive different reduced forms of the above 𝐒(0) (0) (0)
𝑡 = 𝐒𝛼𝛽 𝐞𝛼 ⊗ 𝐞𝛽 = 𝛾(Tr𝐄 )𝐈2 + 2𝐄𝑡
(0)
𝐈2 + ∇𝐮(0)
𝑡 (3.11)
plate equations.
and
3. Classification of the plate model 𝐄(0)
𝑡 = E(0) 𝐞 ⊗ 𝐞𝛽
𝛼𝛽 𝛼
(3.12)
( )
1 (0) (0) T (0) T (0) (0) (0)
In [16], directly based on the 3D field equations, four different = ∇𝐮𝑡 + (∇𝐮𝑡 ) + (∇𝐮𝑡 ) ∇𝐮𝑡 + ∇𝑢3 ⊗ ∇𝑢3 .
2
plate equations were derived for four different levels of the in-plane
As in [16], we consider the plate edge to be clamped, i.e. 𝐮 = 𝟎 at
and normal surface loads relative to the thickness of the plate. For the edge. And in the following, we consider five different levels of the
a different level of surface loads, the Föppl membrane theory was in-plane and normal surface loads to obtain five plate models.
derived in [13]. In this section, for these five cases, we will derive
and recover the corresponding five different plate equations presented (Case 1) Non-linear membrane model
in [16] and [13] from the 2D plate theory presented in the previous In this case, we assume that the applied surface tractions satisfy
section, which then demonstrates its uniform validity for all these 𝐪±
𝑡 = 𝑂(𝜀) and 𝑞3± = 𝑂(𝜀). We find the right hand side of (3.2) is 𝑂(1).
levels of loads. For that purpose, we first nondimensionalize the system, For leading-order balance, from (3.7) and (3.10), we find at most
i.e. (2.20)1 and (2.22), through
𝐮(0)
𝑡 = 𝑂(1), 𝑢(1)
3
= 𝑂(1). (3.13)
𝐱 𝐗 𝐮
𝐱̃ = , 𝐱̃ (𝑖) = 𝐱 𝐿 , 𝐗̃ = , 𝐮̃ = , 𝐮̃ (𝑖) = 𝐮(𝑖) 𝐿𝑖−1 ,
(𝑖) 𝑖−1
(3.1) Similarly, since the right hand side of (3.3) is 𝑂(1), we find at most
𝐿 𝐿 𝐿
𝐒 𝐒(𝑖) 𝑖 𝐪+ 𝐪−
𝐒̃ = , 𝐒̃ (𝑖) = 𝐿 , 𝐪̃ + = , 𝐪̃ − = , 𝐮(1)
𝑡 = 𝑂(1), 𝑢(0)
3
= 𝑂(1). (3.14)
𝜇 𝜇 𝜇 𝜇
for 𝑖 = 1, 2, 3, where 𝜇 is the shear modulus for infinitesimal deforma- From (2.12) and (2.13), we find
tion and 𝐿 is the in-plane length scale of the plate. The operators ∇ and 𝐮𝑡(𝑖) = 𝑂(1), 𝑢(𝑖) = 𝑂(1), 𝑖 = 2, 3. (3.15)
̃ = 𝐿∇ and 𝛥̃ = 𝐿2 𝛥. Note that all the 3
𝛥 are nondimensionalized by ∇
tildes will be dropped for brevity hereafter. Thus we assume the following asymptotic expansions
Then the dimensional equations (2.20)1 and (2.22) are transformed
𝑄 = 𝑄0 + 𝜀𝑄1 + 𝜀2 𝑄2 + ⋯ , (3.16)
into the dimensionless system:
2 2 where 𝑄 = 𝑢1(𝑖) , 𝑢2(𝑖) , 𝑢(𝑖)
3
, 𝐱𝑖 , 𝐒(𝑗) , 𝐒(𝑗)
𝑡 , 𝐄(𝑗) , 𝐄(𝑗)
for 𝑖 = 0, 1, 2, 3, 𝑗 = 0, 1, 2.
𝑡
∇ ⋅ 𝐒(0) (1)
𝑡 + 𝜀∇ ⋅ 𝐒𝑡 + 𝜀 ∇ ⋅ 𝐒(2) 3 ̄ 𝑡,
𝑡 + 𝑂(𝜀 𝐩𝑡 ) = −𝐪 (3.2) Substituting (3.16) into (3.2), (3.3) and (2.18), at 𝑂(1), we have
3
( ) ( ) ∇ ⋅ 𝐒(0)
𝑡0
= −𝐪̄ 𝑡 , (3.17)
∇ ⋅ 𝐒(0) 𝐤 − (𝐒(0) )T 𝐤 + 𝜀∇ ⋅ 𝐒(1) 𝐤 − (𝐒(1) )T 𝐤 (3.3)
( ) 1 ( 3 ) ( )
2
+ 𝜀2 ∇ ⋅ 𝐒(2) 𝐤 − (𝐒(2) )T 𝐤 + 𝜀2 ∇ ⋅ ∇ ⋅ 𝐒(1) 𝑡 + 𝑂 𝜀 𝑝3
∇ ⋅ 𝐒(0)
0
𝐤 − (𝐒(0)
0
)T 𝐤 = −𝑞̄3 , (3.18)
3( ) 3
+𝑂 𝜀 3
∇ ⋅ ∇ ⋅ 𝐒(2)
𝑡 = −𝑞̄3 − ∇ ⋅ 𝐦𝑡 , (𝐒(0) )T 𝐤 = 𝟎. (3.19)
0
where 𝜀 = ℎ∕𝐿 and
From (3.7), it can be seen that 𝐒(0)0
depends on 𝐄(0)
0
and ∇𝐮(0)
0
. Since
𝐪+ −
𝑡 + 𝐪𝑡
𝑞3+ + 𝑞3− E(0) and E(0)
depend on 𝐮 (1)
, to deduce equations only containing 𝐮(0) ,
𝐪̄ 𝑡 = , 𝑞̄3 = . (3.4) 033 03𝛼 0 0
2𝜀 2𝜀 we shall use (3.19) to eliminate the former two.
In [16], the plate is assumed to be composed of a Saint-Venant– First we notice that
Kirchhoff material, for which the nominal stress is given by Tr𝐄(0) = Tr𝐄(0) + E(0) , (3.20)
0 0𝑡 033
T
𝐒 = (𝜆(Tr𝐄)𝐈 + 2𝜇𝐄) 𝐅 , (3.5) where
( )
where 𝜆 and 𝜇 are Lamé’s constants. Here we also assume the plate 1
𝐄(0)
0𝑡
= ∇𝐮(0)
0𝑡
+ (∇𝐮(0) T
0𝑡
) + (∇𝐮 (0) T
0𝑡
) ∇𝐮(0)
0𝑡
+ ∇𝑢 (0)
03
⊗ ∇𝑢 (0)
03
. (3.21)
is composed of this material. Then the corresponding dimensionless 2
nominal stress is given by From (3.19) and (3.7), we have
( ) ( )( )
𝐒 = (𝛾(Tr𝐄)𝐈 + 2𝐄) 𝐅T , (3.6) (𝐒(0)
0
)T 𝐤 = 𝛾Tr𝐄(0)0
+ 2E(0)
033
𝐱0(1) + 2 ∇𝐱0(0) E(0) 𝐞 = 𝟎.
03𝛼 𝛼
(3.22)

119
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125

1
To proceed further, we follow the argument in [16] by considering where Π(0)
𝑡0
is given by (3.29) with 𝐄(0)
0𝑡
given by 𝐄(0)
0𝑡
= (∇𝐮(0)
0𝑡
+
2
the fact det𝐅 > 0 for orientation preservation. Setting 𝑍 = 0, we have (∇𝐮(0) )T ).
0𝑡
( )
det𝐅(0) = det ∇𝐱(0) + 𝐱(1) ⊗ 𝐞3 > 0. (3.23) Similar to Case (1), it can be shown that 𝐮𝑚
0
= 𝐮(0)
0
, where 𝜀𝐮𝑚 0
is the
first term of the asymptotic expansion of 𝐮𝑚 . Thus 𝐮𝑚 0
satisfies (3.35).
The 𝑂(1) term on the left-side should be larger than zero as 𝜀 can be
The boundary condition is 𝐮𝑚 0
= 𝟎 at 𝜕𝛺. The result agrees with that
arbitrarily small, i.e.
( ) in [16] for this case.
det𝐅(0)
0
= det ∇𝐱0(0) + 𝐱0(1) ⊗ 𝐞3 > 0. (3.24)
(Case 3) The von Kármán plate model
This means that ∇𝐱0(0) and 𝐱0(1) are linearly independent. Then from In this case, we assume that the applied surface tractions satisfy
(3.22), we have 𝐪± 3 ± 3 ± 4 ± 4
𝑡 = 𝜀 𝐪𝑡0 = 𝑂(𝜀 ) and 𝑞3 = 𝜀 𝑞30 = 𝑂(𝜀 ). The right hand side of
(3.2) is 𝑂(𝜀2 ). We find at most
𝛾Tr𝐄(0)
0
+ 2E(0)
033
= 0, E(0)
03𝛼
= E(0)
0𝛼3
= 0. (3.25)
𝐮(0) 2
𝑡 = 𝑂(𝜀 ), 𝑢(1) = 𝑂(𝜀2 ). (3.36)
From (3.20) and (3.25)1 , we obtain 3

(0) 𝛾 Similarly, since the right hand side of (3.3) is 𝑂(𝜀3 ), we find at most
E033 = − Tr𝐄(0)
0𝑡
. (3.26)
𝛾 +2
Through calculations, we have 𝐮(1)
𝑡 = 𝑂(𝜀), 𝑢(0)
3
= 𝑂(𝜀). (3.37)
( )
(0)
𝐒0 𝐤 − (𝐒(0) )T 𝐤 = 𝛾Tr𝐄(0) 𝑢(0) + 2E(0) 𝑢(0) 𝐞𝛼 (3.27) From (2.12) and (2.13), we find at most
0 0 03,𝛼 0𝛼𝛽 03,𝛽
( ) ( )
(0) (0) (1) (0) (1) (0) (0)
− 𝛾Tr𝐄0 + 2E033 𝑢0𝛼 𝐞𝛼 + 2 E0𝛼3 𝑢03 − 𝑢0𝛼,𝛽 E03𝛽 𝐞𝛼 . 𝐮(2) 2
𝑡 = 𝑂(𝜀 ), 𝑢(2)
3
= 𝑂(𝜀), 𝐮(3)
𝑡 = 𝑂(𝜀), 𝑢(3)
3
= 𝑂(𝜀2 ). (3.38)

Upon using (3.20), (3.25) and (3.26), we have Thus we have the following asymptotic expansion

𝐒(0)
0
𝐤 − (𝐒(0)
0
)T 𝐤 = Π(0)
𝑡0
∇𝑢(0)
03
, (3.28) 𝑄 = 𝜀𝑄0 + 𝜀2 𝑄1 + 𝜀3 𝑄2 + ⋯ , (3.39)
where where 𝑄 = 𝑢3(2𝑖) , 𝐮𝑡(2𝑖+1) , 𝐄(𝑗) , 𝐄(1) 𝐒(𝑗) , 𝐒(1)
𝑡 , 𝑡 for 𝑖 = 0, 1, 𝑗 = 0, 1, 2 and
2𝛾
Π(0) = Tr𝐄(0)
0𝑡 2
𝐈 + 2𝐄(0) . (3.29)
𝑡0 𝛾 +2 0𝑡 𝑃 = 𝜀2 𝑃0 + 𝜀3 𝑃1 + 𝜀4 𝑃2 + ⋯ , (3.40)
From (3.11), we obtain
( ) where 𝑃 = 𝐮𝑡(2𝑖) , 𝑢3(2𝑖+1) , 𝐄(2𝑖)
𝑡 , 𝐒𝑡
(2𝑖)
for 𝑖 = 0, 1.
𝐒𝑡0 = Π(0)
(0)
𝑡0
𝐈2 + ∇𝐮(0)
0𝑡
. (3.30) From (3.10) and (3.7), we obtain
( )
1
Then Eqs. (3.17) and (3.18) give the equations for 𝐮(0)
0
: 𝐄(0)
0
= 𝐞3 ⊗ ∇𝑢(0)03
+ ∇𝑢(0)03
⊗ 𝐞3 + 𝐞3 ⊗ 𝐮(1)
0𝑡
+ 𝐮(1)
0𝑡
⊗ 𝐞3 (3.41)
2
[ ( )]
⎧ (0) (0) and
⎪∇ ⋅ Π𝑡0 𝐈2 + ∇𝐮0𝑡 = −𝐪̄ 𝑡 ,
⎨ ( (0) (0) ) (3.31)
⎪∇ ⋅ Π𝑡0 ∇𝑢03 = −𝑞̄3 , 𝐒(0)
0
= 2𝐄(0)
0
. (3.42)

For (2.18), at 𝑂(𝜀), we have
where Π(0)
𝑡0
is given by (3.29) with 𝐄(0)
0𝑡
given by (3.21).
Suppose that 𝐮𝑚 represents the displacement in the middle surface. (𝐒(0)
0
)T 𝐤 = 2E(0) 𝐞 = 𝟎,
03𝛼 𝛼
(3.43)
From (2.2), we have (for dimensionless quantities)
which leads to
1 2 (2)
𝐮𝑚 = 𝐮(0) + 𝜀𝐮(1) + 𝜀 𝐮 + ⋯. (3.32)
2 E(0) = 0, (3.44)
03𝛼
By substituting (3.16) (𝑄 = 𝐮(𝑖) (𝑖 = 0, 1, 2)) into the above relation, it is
easy to see 𝐮𝑚0
= 𝐮(0)
0
, where 𝐮𝑚0
is the first term of the asymptotic ex- that is (cf. (3.41))
pansion of 𝐮 = 𝐮0 + 𝜀𝐮𝑚
𝑚 𝑚 + ⋯. Thus 𝐮𝑚 satisfies (3.31). For the clamped
1 0 𝐮(1) = −∇𝑢(0) . (3.45)
edge, we have the boundary condition 𝐮𝑚 0
= 𝟎 at 𝜕𝛺. The result agrees 0𝑡 03
with that in [16] for this case, and according to which, this membrane The above relation also implies that 𝐄(0) = 𝟎 and 𝐒(0) = 𝟎. Thus, in fact,
0 0
model can be proved to be identical to the one obtained in [31] based
𝐄(0) and 𝐒(0) are 𝑂(𝜀2 ) and should have the asymptotic expansions as
on asymptotic expansions based on the 3D weak formulation.
(3.40), i.e.
(Case 2) Another membrane model
In this case, we assume that the applied surface tractions satisfy 𝐄(0) = 𝜀2 𝐄(0)
0
+ 𝜀3 𝐄(0)
1
+ ⋯, (3.46)
𝐪± 2 ± 2 ± 3 ± 3
𝑡 = 𝜀 𝐪𝑡0 = 𝑂(𝜀 ) and 𝑞3 = 𝜀 𝑞30 = 𝑂(𝜀 ). With similar analysis as
in Case 1, from (3.2), (3.3), (2.18), (2.12) and (2.13), we can deduce 𝐒(0) = 𝜀2 𝐒(0)
0
+ 𝜀3 𝐒(0)
1
+ ⋯. (3.47)
that
From (3.10), (2.6), (3.7), (3.11) and (3.8), we obtain
𝐮(𝑖)
𝑡 = 𝑂(𝜀), 𝑢(𝑖) = 𝑂(𝜀), 𝑖 = 0, 1, 2, 3 (3.33)
3 (
1
at most. Thus we have the following asymptotic expansions 𝐄(0)
0
= ∇𝐮(0)
0𝑡
+ (∇𝐮(0)
0𝑡
)T + ∇𝑢(0)
03
⊗ ∇𝑢(0)
03
+ ∇(𝑢(0)
13 3
𝐞 ) (3.48)
2 ( ) )
(0) T (1) (1) (1)
𝑄 = 𝜀𝑄0 + 𝜀2 𝑄1 + 𝜀3 𝑄2 + ⋯ , (3.34) +(∇(𝑢13 𝐞3 )) + 2𝑢03 + 𝐮0𝑡 ⋅ 𝐮0𝑡 𝐞3 ⊗ 𝐞3 ,

where 𝑄 = 𝑢(𝑖) , 𝑢(𝑖) , 𝑢(𝑖) , 𝐒(𝑗) , 𝐒(𝑗) 𝐄(𝑗) , 𝐄(𝑗) ( )


𝑡 ,
for 𝑖 = 0, 1, 2, 3, 𝑗 = 0, 1, 2. 1
1 2 3 𝑡
𝐄(0) = ∇𝐮(0) + (∇𝐮(0) T
) + ∇𝑢 (0)
⊗ ∇𝑢 (0)
, (3.49)
Then, by similar calculations as in Case 1, one can deduce the 0𝑡 2 0𝑡 0𝑡 03 03
following equations ( )
1
⎧ 𝐄(1)
0
= ∇𝐮(1)
0𝑡
+ (∇𝐮(1)
0𝑡
)T + 2𝑢(2)
03
𝐞3 ⊗ 𝐞3 , (3.50)
(0) 2
⎪∇ ⋅ Π𝑡0 = −𝐪̄ 𝑡0 ,
⎨ ( ) (3.35) ( )
(0) (0) 1
⎪∇ ⋅ Π𝑡0 ∇𝑢03 = −𝑞̄30 − ∇ ⋅ 𝐦𝑡0 , 𝐄(1) = ∇𝐮(1) + (∇𝐮(1) )T (3.51)
⎩ 0𝑡 2 0𝑡 0𝑡

120
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125
( )
4(𝛾 + 1) (0) 2𝛾 (0) (0)
and = −𝜀3 𝑢 𝑢(0) + 𝑢 𝑢 + 2𝑢(0) 𝑢(0) + 𝑂(𝜀4 ).
𝛾 + 2 03,𝛼𝛽𝛽 03,𝛼 𝛾 + 2 03,𝛼𝛼 03,𝛽𝛽 03,𝛼𝛽 03,𝛼𝛽
𝐒(0)
0
= 𝛾Tr𝐄(0)
0
𝐈 + 2𝐄(0)
0
, (3.52)
For the first term, we have
( )
( )( ) 𝐒(0) 𝐤 − (𝐒(0) )T 𝐤 = 𝜀3 𝐒(0)
1
𝐤 − (𝐒(0)
1
)T 𝐤 + 𝑂(𝜀4 ). (3.69)
𝐒(0)
1
= 𝛾Tr𝐄(0)
1
𝐈 + 2𝐄(0)
1
+ 𝛾Tr𝐄(0)
0
𝐈 + 2𝐄(0)
0
∇𝑢(0)
03
⊗ 𝐞3 + 𝐞3 ⊗ 𝐮(1)
0𝑡
,
Further by using (3.53), (3.59), (3.60) and (3.61), we obtain
(3.53)
𝐒(0)(𝐤 − (𝐒(0) )T 𝐤 ) ( ) (3.70)
𝐒(0) = 𝛾Tr𝐄(0) 𝐈 + 2𝐄(0) , (3.54) = 𝜀3 𝛾Tr𝐄(0) 0
𝐈 + 2𝐄(0)
0
∇𝑢(0)
03
− 𝜀3 𝛾Tr𝐄(0)
0
+ 2E(0)
033
𝐮(1)
0𝑡
+ 𝑂(𝜀4 )
𝑡0 0 2 0𝑡
= 𝜀3 Π(0)
𝑡0
∇𝑢(0)
03
+ 𝑂(𝜀4 ).
𝐒(1)
0
= 𝛾Tr𝐄(1)
0
𝐈 + 2𝐄(1)
0
, (3.55)
For the fourth term, from (3.65), we have
( )( )T
1 2 1
𝐒(1)
1
= 𝛾Tr𝐄(1)
1
𝐈 + 2𝐄(1)
1
+ 𝛾Tr𝐄(1)
0
𝐈 + 2𝐄(1)
0
∇(𝑢(0)
03 3
𝐞 ) + 𝐮(1)
0𝑡
⊗ 𝐞3 , 𝜀 ∇ ⋅ ∇ ⋅ 𝐒(1)
𝑡 = 𝜀3 ∇ ⋅ ∇ ⋅ 𝐒(1)
𝑡0
+ 𝑂(𝜀4 ) (3.71)
3 3
1 4(𝛾 + 1)
(3.56) = − 𝜀3 𝛥2 𝑢(0)
03
+ 𝑂(𝜀4 ).
3 𝛾 +2
𝐒(1)
𝑡0
= 𝛾Tr𝐄(1)
0 2
𝐈 + 2𝐄(1)
0𝑡
. (3.57) Substituting (3.39) and (3.40) into (3.2) and (3.3), upon using
the above relations, at 𝑂(𝜀2 ) and 𝑂(𝜀3 ), respectively, we obtain the
The expression of in 𝐄(0)
1
𝐒(0)
is not needed during calculations and we
1 equations for 𝐮(0) :
0
only need to notice its symmetric property.
Since 𝐮(1) has been given in (3.45), to deduce equations for 𝐮(0) only, ⎧ (0) 4(𝛾 + 1) (0)
0𝑡 0 ⎪∇ ⋅ Π𝑡0 − 𝛾 + 2 ∇(𝛥𝑢03 ) = −𝐪̄ 𝑡0 ,
we want to eliminate 𝑢(1) (occurring in 𝐒(0)
), and 𝑢 (2)
(occurring in 𝐒(1) ⎪
03 0 03 0 ( ) 4(𝛾 + 1)
(1) (1)
and 𝐒1 through 𝐄0 ) by using (2.18) and (2.12). ⎪ 4(𝛾 + 1) 2 (0) (0) (0) (0) (0)
⎨− 3(𝛾 + 2) 𝛥 𝑢03 + ∇ ⋅ Π𝑡0 ∇𝑢03 − 𝛾 + 2 𝑢03,𝛼𝛽𝛽 𝑢03,𝛼 (3.72)
For (2.18), at 𝑂(𝜀2 ), we have ⎪
⎪ 2𝛾 (0) (0)
(𝐒(0) )T 𝐤 = 𝟎, (3.58) ⎪− 𝑢 𝑢 − 2𝑢(0) 𝑢(0) = −𝑞̄30 − ∇ ⋅ 𝐦𝑡0 .
0 ⎩ 𝛾 + 2 03,𝛼𝛼 03,𝛽𝛽 03,𝛼𝛽 03,𝛼𝛽

which yields For the middle-surface displacement 𝐮𝑚 , it is easy to deduce that


𝛾Tr𝐄(0)
0
+ 2E(0)
033
= 0, (3.59) 𝐮𝑚 = 𝐮𝑚 + 𝑢𝑚 𝐞 = 𝐮(0) − ∇𝑢(0) , (3.73)
0 0𝑡 03 3 0 03
1
where =E(0)
033
𝑢(1)
03
+ 𝑢(0) 𝑢(0) . where 𝜀2 𝐮𝑚 and 𝜀𝑢𝑚 denote the first terms of the asymptotic expansions
2 03,𝛼 03,𝛼 0𝑡 03
Noticing that of 𝐮𝑚
𝑡 and 𝑢 𝑚 , respectively. Substituting the above equation into (3.72),
3
we obtain
Tr𝐄(0)
0
= Tr𝐄(0)
0𝑡
+ E(0)
033
, (3.60)
⎧∇ ⋅ Π𝑚 = −𝐪̄ ,
Eq. (3.59) gives ⎪ 𝑡0 𝑡0
⎨ 4(𝛾 + 1) 2 𝑚 ( ) (3.74)
E(0) =−
𝛾
Tr𝐄(0) , (3.61) ⎪− 𝛥 𝑢03 + ∇ ⋅ Π𝑚
𝑡0
∇𝑢𝑚
03
= −𝑞̄30 − ∇ ⋅ 𝐦𝑡0 ,
033 𝛾 +2 0𝑡 ⎩ 3(𝛾 + 2)

from which 𝑢(1) can be expressed in terms of 𝐮(0) . where Π𝑚 𝑡0


is given by (3.29) with E(0) 0𝑡
given by (3.49) with the
03 0
To eliminate 𝑢(2) , we consider (2.12). At 𝑂(𝜀), we have replacement of 𝐮(0)
0
by 𝐮 𝑚 . For the clamped edge, we have the boundary
0
03
conditions 𝐮𝑚0
= 𝟎 and 𝐮(0)
0𝑡
⋅ 𝝂 = 0, where 𝝂 is the normal vector of 𝜕𝛺.
(𝐒(1)
0
)T 𝐤 = 𝟎, (3.62) The latter also leads to ∇𝑢𝑚 ⋅ 𝝂 = 0. The result agrees with the result
03
which, upon using (3.45), gives in [16] for this case. This von Kármán plate model was also derived
in [11] (to leading-order) for a general isotropic hyperelastic material.
𝛾
𝑢(2)
03
= 𝑢(0) . (3.63)
𝛾 + 2 03,𝛼𝛼 (Case 4) The linear Kirchhoff–Love plate model
Now we turn our attention to the plate equations (3.2) and (3.3). In this case, we assume that the applied surface tractions satisfy
For the first term on the right-hand side of (3.2), from (3.54) (upon 𝐪± 𝛼 ± 𝛼 ±
𝑡 = 𝜀 𝐪0𝑡 = 𝑂(𝜀 ) and 𝑞3 = 𝜀
𝛼+1 𝑞 ± = 𝑂(𝜀𝛼+1 ), where 𝛼 > 3. The
03
𝛼−1
right hand side of (3.2) is 𝑂(𝜀 ). We find at most
using (3.60) and (3.61)), we obtain
2𝛾 𝐮(0) 𝛼−1
𝑢(1) = 𝑂(𝜀𝛼−1 ).
𝐒(0)
𝑡0
= Tr𝐄(0)
0𝑡 2
𝐈 + 2𝐄(0)
0𝑡
, (3.64) 𝑡 = 𝑂(𝜀 ), 3
(3.75)
𝛾 +2
Similarly, since the right hand side of (3.3) is 𝑂(𝜀𝛼 ), we find at most
which only depends on 𝐮(0)0
. We also note 𝐒(0)
𝑡0
= Π(0)
𝑡0
, as defined in
(0)
(3.29) with 𝐄0𝑡 given by (3.49). For the second term in (3.2), from 𝐮(1)
𝑡 = 𝑂(𝜀
𝛼−2
), 𝑢(0)
3
= 𝑂(𝜀𝛼−2 ). (3.76)
(3.57) and (3.48), and by using (3.45) and (3.63), we have
From (2.12) and (2.13), we find at most
4(𝛾 + 1)
∇ ⋅ 𝐒(1) =− ∇(𝛥𝑢(0) ). (3.65)
𝑡0 𝛾 +2 03 𝑢(2)
𝛼 = 𝑂(𝜀
𝛼−1
), 𝑢(2)
3
= 𝑂(𝜀𝛼−2 ), 𝑢(3)
𝛼 = 𝑂(𝜀
𝛼−2
), 𝑢(3)
3
= 𝑂(𝜀𝛼−1 ).
For the third term in (3.2), we have (3.77)
2 2
𝜀 ∇ ⋅ 𝐒(2) 4
𝑡 = 𝑂(𝜀 ). (3.66) Thus we have the following asymptotic expansions
3
For (3.3), for the third term, it can be shown from (3.9) that 𝑄 = 𝜀𝛼−2 𝑄0 + 𝜀𝛼−1 𝑄1 + 𝜀𝛼 𝑄2 + ⋯ , (3.78)
2 2 ( )
𝜀 ∇ ⋅ 𝐒(2) 𝐤 − (𝐒(2) )T 𝐤 = 𝑂(𝜀4 ). (3.67) where 𝑄 = 𝑢3(2𝑖) , 𝐮𝑡(2𝑖+1) , 𝐄(𝑗) , 𝐄(1)
𝑡 , 𝐒(𝑗) , 𝐒(1)
𝑡 for 𝑖 = 0, 1, 𝑗 = 0, 1, 2 and
3
For the second term, by using (3.56), (3.45) and (3.63), we have 𝑃 = 𝜀𝛼−1 𝑃0 + 𝜀𝛼 𝑃1 + 𝜀𝛼+1 𝑃2 + ⋯ , (3.79)
( ) ( )
𝜀∇ ⋅ 𝐒(1) 𝐤 − (𝐒(1) )T 𝐤 = 𝜀3 ∇ ⋅ 𝐒(1)
1
𝐤 − (𝐒(1)
1
)T 𝐤 + 𝑂(𝜀4 ) (3.68) where 𝑃 = 𝐮𝑡(2𝑖) , 𝑢3(2𝑖+1) , 𝐄(2𝑖)
𝑡 , 𝐒(2𝑖)
𝑡 for 𝑖 = 0, 1.

121
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125

Similar with Case (3), we can show that Then by similar analysis as in Case (3), we obtain the equations for
𝐮(0) :
𝐮(1)
0𝑡
= −∇𝑢(0)
03
, 𝐄(0)
0
= 𝟎, 𝐒(0)
0
= 𝟎. (3.80) 0

⎧ (0)
Thus 𝐄(0) and 𝐒(0) should have the following asymptotic expansions ⎪∇ ⋅ Π𝑡0 = −𝐪̄ 𝑡0 ,
⎨ ( ) (3.94)
(0) (0)
𝐄(0) = 𝜀𝛼−1 𝐄(0) + 𝜀𝛼 𝐄(0) + ⋯, (3.81) ⎪∇ ⋅ Π𝑡0 ∇𝑢03 = −𝑞̄30 ,
0 1 ⎩

where Π0𝑡0 is given by (3.29) with 𝐄(0) 0𝑡


given by (3.49). Similar to
𝐒 (0)
= 𝜀𝛼−1 𝐒(0)
0
+ 𝜀𝛼 𝐒(0)
1
+ ⋯. (3.82) the former cases, it can be shown that 𝐮𝑚 = 𝐮(0) , where 𝜀2𝛼∕3 𝐮𝑚 and
0 0 0𝑡
𝜀 𝑢03 denote the first terms of the asymptotic expansions of 𝐮𝑚
𝛼∕3 𝑚
𝑡 and
Then, by similar calculations as in Case (3), one can obtain the
𝑢𝑚 , respectively. Thus 𝐮𝑚 satisfies (3.94). The boundary condition is
following equations for 𝐮(0) : 3 0
0 𝐮𝑚
0
= 𝟎 at 𝜕𝛺. Specially, when 𝛼 = 3∕2, this Föppl membrane model
⎧ (0) 4(𝛾 + 1) (0) agrees with that in [13] (to leading order).
⎪∇ ⋅ Π𝑡0 − 𝛾 + 2 𝛥(∇𝑢03 ) = −𝐪̄ 𝑡0 ,
⎨ 4(𝛾 + 1) (3.83)
⎪− 𝛥2 𝑢(0) = −𝑞̄30 − ∇ ⋅ 𝐦𝑡0 , Remark. The non-linear membrane model and Föppl membrane model
⎩ 3(𝛾 + 2) 03
can also be derived by 𝛤 -convergence (see [5]), which have additional
1 effect of relaxation. While, in the current series expansion and formal
where Π0𝑡0 is given by (3.29) with 𝐄(0)
0𝑡
given by 𝐄(0)
0𝑡
= (∇𝐮(0)
0𝑡
+
2 asymptotic procedure based on the 3D equilibrium equations such
(∇𝐮(0)
0𝑡
)T ).
effect is not present.
For the middle-surface displacement 𝐮𝑚 , we have

𝐮𝑚 = 𝐮𝑚 + 𝑢𝑚 𝐞 = 𝐮(0)
03 3
− ∇𝑢(0) , (3.84)
0 0𝑡 0 03 4. A uniformly-valid plate theory and its weak formulation
where 𝜀𝛼−1 𝐮𝑚
0𝑡
and 𝜀𝛼−2 𝑢𝑚03
denote the first terms of the asymptotic
expansions of 𝐮𝑚 𝑚
𝑡 and 𝑢3 , respectively. Substituting the above equation All the five types of plate models derived in the previous section
into (3.83), we obtain come from the single plate model (2.20)1 and (2.22) with (2.18),
⎧∇ ⋅ Π𝑚 = −𝐪̄ , (2.12) and (2.13). Also, we observe from the derivations that the terms
⎪ 2 2 2 2 ( (2) )
𝑡0 𝑡0
(3.85) ℎ ∇⋅𝐒(2) (2) T
𝑡 in (2.20)1 and 3 ℎ ∇⋅ 𝐒 𝐤 − (𝐒 ) 𝐤 in (2.22) do not come
⎨ 4(𝛾 + 1) 2 𝑚 3
⎪− 𝛥 𝑢30 = −𝑞̄30 − ∇ ⋅ 𝐦𝑡0 , into play for the leading-order correct results in all five cases. Based on
⎩ 3(𝛾 + 2) this observation, one may drop them to obtain
1( 𝑚
where Π𝑚 is given by (3.29) with E(0) given by 𝐄(0) = ∇𝐮0𝑡 + ∇ ⋅ 𝐒(0) (1)
(4.1)
) 𝑡0 0𝑡 0𝑡 2 𝑡 + ℎ∇ ⋅ 𝐒𝑡 = −𝐪
̄ 𝑡,
(∇𝐮𝑚
0𝑡
)T . Similar to the previous case, for the clamped edge, the

boundary conditions for (3.85) are 𝐮𝑚 = 𝟎 and ∇𝑢𝑚 ⋅ 𝝂 = 0 at 𝜕𝛺. ( ) ( ) 1


0 03 ∇ ⋅ 𝐒(0) 𝐤 − (𝐒(0) )T 𝐤 + ℎ∇ ⋅ 𝐒(1) 𝐤 − (𝐒(1) )T 𝐤 + ℎ2 ∇ ⋅ ∇ ⋅ 𝐒(1)
𝑡 = −𝑞̄3 − ∇ ⋅ 𝐦𝑡
The result agrees with that in [16] for this case. 3
(4.2)
(Case 5) The Föppl membrane model
In this case, we assume that the applied surface tractions satisfy with (2.18) and (2.12) for solving 𝐱(1) and 𝐱(2) in terms of 𝐱(0) , respec-
𝐪±
𝑡 = 𝜀
2𝛼∕3+1 𝐪± = 𝑂(𝜀2𝛼∕3+1 ) and 𝑞 ± = 𝜀𝛼+1 𝑞 ± = 𝑂(𝜀𝛼+1 ), where
𝑡0 3 30 tively. The constitutive relations are provided by (3.7) and (3.8) or
0 < 𝛼 < 3. The right hand side of (3.2) is 𝑂(𝜀2𝛼∕3 ). We find at most (2.10)1 and (2.10)2 for a general strain energy function. This system
is valid for all the five different magnitudes of 𝐪± ±
𝑡 and 𝑞3 and is a
𝐮(0)
𝑡 = 𝑂(𝜀
2𝛼∕3
), 𝑢(1)
3
= 𝑂(𝜀2𝛼∕3 ). (3.86) uniformly-valid plate model good for these magnitudes. In fact no
matter what the orders of 𝐪± ±
𝑡 and 𝑞3 are, as long as
Similarly, since the right hand side of (3.3) is 𝑂(𝜀𝛼 ), we find at most ( ) ( )
2 2 (2)
ℎ 𝐒𝑡 = 𝑜 𝐒(0) or 𝑜 ℎ𝐒(1) (4.3)
𝐮(1)
𝑡 = 𝑂(𝜀
𝛼∕3
), 𝑢(0)
3
= 𝑂(𝜀𝛼∕3 ). (3.87) 3 𝑡 𝑡

From (2.12) and (2.13), we find at most and


2 2 ( (2) )
𝐮(2) 2𝛼∕3
𝑢(2) = 𝑂(𝜀𝛼∕3 ), 𝐮(3) 𝛼∕3
𝑢(3) = 𝑂(𝜀2𝛼∕3 ). ℎ 𝐒 𝐤 − (𝐒(2) )T 𝐤 (4.4)
𝑡 = 𝑂(𝜀 ), 𝑡 = 𝑂(𝜀 ), 3 ( )
3 3 ( (0) (0) T
) ( (1) (1) T
) 2 (1)
(3.88) = 𝑜 (𝐒 𝐤 − (𝐒 ) 𝐤) or 𝑜 ℎ(𝐒 𝐤 − (𝐒 ) 𝐤) or 𝑜 ℎ ∇ ⋅ 𝐒𝑡 ,

Thus we have the following asymptotic expansion for leading-order correct results, one can drop them respectively in
𝛼∕3 2𝛼∕3 𝛼 (2.20)1 and (2.22). So the validity of (4.1) and (4.2) is not limited to
𝑄=𝜀 𝑄0 + 𝜀 𝑄1 + 𝜀 𝑄2 + ⋯ , (3.89)
the five cases considered in Section 3.
where 𝑄 = 𝑢(2𝑖)
3
, 𝐮(2𝑖+1)
𝑡 , 𝐄(𝑗) , 𝐄(1)
𝑡 , 𝐒(𝑗) , 𝐒(1)
𝑡 for 𝑖 = 0, 1, 𝑗 = 0, 1, 2 and
Remarks. 1. In practice, one can compute the solution according to
𝑃 = 𝜀2𝛼∕3 𝑃0 + 𝜀4𝛼∕3 𝑃1 + 𝜀2𝛼 𝑃2 + ⋯ , (3.90) (4.1) and (4.2) first, and as a posteriori, check the validity of (4.3) and
where 𝑃 = 𝐮(2𝑖) (2𝑖+1)
, 𝐄(2𝑖) (2𝑖) (4.4). If it turns out that they do not hold, then one needs to re-do
𝑡 , 𝑢3 𝑡 , 𝐒𝑡 for 𝑖 = 0, 1.
Similar with Case (3), we can show that the computation by (2.19) (which is more complicated than (4.1) and
(4.2) as 𝐒(2) is involved). Of course, when surface loads belong to the
𝐮(1)
0𝑡
= −∇𝑢(0)
03
, 𝐄(0)
0
= 𝟎, 𝐒(0)
0
= 𝟎. (3.91) five cases considered in Section 3, no a posteriori checking is needed.
2. In a quasi-static loading process, the magnitudes of 𝐪± 𝑡 and 𝑞3
±
Thus 𝐄(0) and 𝐒(0) should have the following asymptotic expansions
may vary, and it is difficult and not convenient to apply a hierarchy of
𝐄 (0)
= 𝜀2𝛼∕3 𝐄(0)
0
+ 𝜀 𝐄(0)
1
𝛼
+ ⋯, (3.92) plate models, while the present uniformly-valid plate model can be used
to handle this case without any difficulty. Also, if in different domains
of 𝛺, surface loads have different magnitudes, and the hierarchy of
𝐒(0) = 𝜀2𝛼∕3 𝐒(0)
0
+ 𝜀𝛼 𝐒(0)
1
+ ⋯. (3.93) plate models does not apply, whereas the present model is applicable.

122
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125

1 2ℎ 1 2ℎ
In the following discussion, we will give the weak formulation of where 𝐒̄ 𝑡 = ∫ 𝐒 d𝑍 and 𝐮̄ 𝑡 = ∫ 𝐮 d𝑍. By setting 𝝓𝑡 = 𝐮(0) 𝑡 +
2ℎ 0 𝑡 2ℎ 0 𝑡
the uniformly-valid plate model (4.1) and (4.2), which may be used (1)
ℎ𝐮𝑡 (= 𝐮𝑚 +𝑂(ℎ2 )), it is seen that the right hand side of (4.11) represents
𝑡
for finite element computation. First, multiplying both sides of (4.1) by the virtual work done by the applied force 𝐪𝑡 (cf. (4.8)).
the virtual displacement 𝛿𝝓𝑡 and integrating by parts, we have 2. The virtual work due to the bending moment at the plate edge
{ } about the middle surface is
(𝐒(0) (1)
𝑡 +ℎ𝐒𝑡 ) ∶ ∇𝛿𝝓𝑡 d𝐴−𝐪 ̄ 𝑡 ⋅𝛿𝝓𝑡 d𝐴 = (𝐒(0) +ℎ𝐒(1) T
𝑡 ) 𝝂⋅𝛿𝝓𝑡 d𝑠, (4.5)
∫𝛺 ∫𝜕𝛺 𝑡 2ℎ
− [ 𝐪𝑡 ⋅ 𝝂(𝑍 − ℎ)d𝑍]𝛿𝑢𝑚
3,𝝂
d𝑠 (4.14)
where 𝝂 is the outward unit normal on 𝜕𝛺. ∫𝜕𝛺𝑞 ∫0
2ℎ
Similarly, multiplying both sides of (4.2) by 𝛿𝜙3 and integrating by
= − [ 𝐒T𝑡 𝝂 ⋅ 𝝂(𝑍 − ℎ)d𝑍]𝛿𝑢𝑚
3,𝝂
d𝑠
parts, we have ∫𝜕𝛺𝑞 ∫0
{[ ] 2ℎ ( )T
(𝐒(0) + ℎ𝐒(1) )𝐤 − (𝐒(0) + ℎ𝐒(1) )T 𝐤 + 𝐦𝑡 ⋅ ∇𝛿𝜙3 (4.6) = − [ 𝐒(0) (1) (1)
𝑡 + ℎ𝐒𝑡 + (𝑍 − ℎ)𝐒𝑡
∫𝛺 ∫𝜕𝛺𝑞 ∫0
}
1 2 (1) × [𝝂, 𝝂](𝑍 − ℎ)d𝑍]𝛿𝑢𝑚 d𝑠 + 𝑂(ℎ4 )
− ℎ 𝐒𝑡 ∶ ∇∇𝛿𝜙3 − 𝑞̄3 𝛿𝜙3 d𝐴 3,𝝂
{[ 3
] 1 ( ) 2ℎ3
= (𝐒 + ℎ𝐒 )𝐤 − (𝐒 + ℎ𝐒(1) )T 𝐤 ⋅ 𝝂 + ℎ2 (𝐒(1)
(0) (1) (0) T = − (𝐒(1) )T [𝝂, 𝝂]𝛿𝑢𝑚 d𝑠 + 𝑂(ℎ4 ).
∫𝜕𝛺 3 𝑡 ) [𝝂, 𝝉] ,𝑠 3 ∫𝜕𝛺𝑞 𝑡 3,𝝂
}
1 1 2 (1) T
+ ℎ2 (∇ ⋅ 𝐒(1)𝑡 ) ⋅ 𝝂 + 𝐦𝑡 ⋅ 𝝂 𝛿𝜙3 d𝑠 − ℎ (𝐒𝑡 ) [𝝂, 𝝂]𝛿𝜙3,𝝂 𝑑𝑠, By setting 𝜙3 = 𝑢𝑚 = 𝑢(0) + ℎ𝑢(1) + 𝑂(ℎ2 ), then the term ∫𝜕𝛺 𝑚̂ 3 𝛿𝜙3,𝜈 d𝑠 in
3 ∫𝜕𝛺 3 3 3 3 𝑞
(4.12) represents the virtual work done by the applied bending moment
where 𝝉 is the unit tangent vector of 𝜕𝛺 such that (𝝉, 𝝂, 𝐤) forms a right- (cf. (4.10)).
hand triple, and we define 𝐀[𝐚, 𝐛] = 𝐀𝐚 ⋅ 𝐛 for a second-order tensor 𝐀 3. The twisting moment at the edge due to the applied force 𝐪 ⋅ 𝝉 is
and vectors 𝐚, 𝐛. 2ℎ 2ℎ
Suitable boundary conditions may be imposed accordingly. On the 𝑀 = 𝐪𝑡 ⋅ 𝝉(𝑍 − ℎ)d𝑍 = 𝐒T𝑡 𝝂 ⋅ 𝝉(𝑍 − ℎ)d𝑍 (4.15)
∫0 ∫0
displacement edge 𝜕𝛺0 , from (4.5) and (4.6), we may impose 2ℎ ( )T
= 𝐒(0) (1)
𝑡 + ℎ𝐒𝑡 + (𝑍 − ℎ)𝐒𝑡
(1)
𝝂 ⋅ 𝝉(𝑍 − ℎ)d𝑍 + 𝑂(ℎ4 )
𝝓𝑡 = 𝐮̂ 𝑡 , 𝜙3 = 𝑢̂ 3 , 𝜙3,𝝂 = 𝛼,
̂ (4.7) ∫0
2ℎ3 (1) T
where 𝐮̂ and 𝛼̂ are respectively prescribed displacement and bending = (𝐒𝑡 ) [𝝂, 𝝉] + 𝑂(ℎ4 ).
3
angle. On the traction edge 𝜕𝛺𝑞 , we may impose It was shown in [32] (Section 2.5; the authors attribute this to Kirch-
( )T hoff) that the twisting moment 𝑀 is statically equivalent to a dis-
2ℎ 𝐒(0)
𝑡 + ℎ𝐒𝑡
(1)
𝝂 = 𝐪̂ 𝑡 , (4.8) 2ℎ3 ( (1) T
tributed shear force (along the thickness direction) 𝑀,𝑠 = (𝐒𝑡 )
) 3
{ [𝝂, 𝝉] ,𝑠 . Thus the total effective shear force is
1 ( )
2ℎ [(𝐒(0) + ℎ𝐒(1) )𝐤 − (𝐒(0) + ℎ𝐒(1) )T 𝐤] ⋅ 𝝂 + ℎ2 (𝐒(1) T
𝑡 ) [𝝂, 𝝉] ,𝑠 (4.9) 2ℎ ( )
3 2ℎ3
1 } (𝐒(1)
𝐒T 𝝂 ⋅ 𝐤d𝑍 + )T [𝝂, 𝝉] (4.16)
+ ℎ2 (∇ ⋅ 𝐒(1)
𝑡 ) ⋅ 𝝂 + 𝐦𝑡 ⋅ 𝝂 = 𝑞̂3 , ∫0 3 𝑡 ,𝑠
3 ( )T ( )
2 2ℎ3
2 3 (1) T = 2ℎ 𝐒(0) + ℎ𝐒(1) + ℎ2 𝐒(2) 𝝂 ⋅ 𝐤 + (𝐒(1) T 4
𝑡 ) [𝝂, 𝝉] ,𝑠 + 𝑂(ℎ ).
− ℎ (𝐒𝑡 ) [𝝂, 𝝂] = 𝑚̂ 3 , (4.10) 3 3
3
By using (2.21), the above equation can be rewritten as
where 𝐪̂ 𝑡 and 𝑞̂3 are respectively the applied in-plane and shear forces 2ℎ ( )
2ℎ3
(per unit arc length of 𝜕𝛺), and 𝑚̂ 3 is the applied bending moment 𝐒T 𝝂 ⋅ 𝐤d𝑍 + (𝐒(1) T
𝑡 ) [𝝂, 𝝉] ,𝑠 (4.17)
∫0 3
about the middle surface. (
2
Thus the weak form of the uniformly-valid plate system is = 2ℎ [(𝐒(0) + ℎ𝐒(1) )𝐤 − (𝐒(0) + ℎ𝐒(1) )T 𝐤] ⋅ 𝝂 + ℎ2 (𝐒(2) 𝐤 − (𝐒(2) )T 𝐤) ⋅ 𝝂
) ( 3 )
( ) 1 2ℎ 3
2ℎ (𝐒(0) (1)
̄ 𝑡 ⋅ 𝛿𝝓𝑡 d𝐴 = 𝐪̂ ⋅ 𝛿𝝓𝑡 d𝑠, (4.11) +𝐦𝑡 ⋅ 𝝂 − ℎ2 (𝐒(2) )T 𝐤 ⋅ 𝝂 + (𝐒(1) T 4
𝑡 ) [𝝂, 𝝉] ,𝑠 + 𝑂(ℎ ).
∫𝛺 𝑡 + ℎ𝐒𝑡 ) ∶ ∇𝛿𝝓𝑡 − 𝐪 ∫𝜕𝛺𝑞 𝑡 3 3
Upon using (2.13), we have
{ 2ℎ ( )
[(𝐒(0) + ℎ𝐒(1) )𝐤 − (𝐒(0) + ℎ𝐒(1) )T 𝐤 + 𝐦𝑡 ] ⋅ ∇𝛿𝜙3 2ℎ3
2ℎ
∫𝛺
(4.12) 𝐒T 𝝂 ⋅ 𝐤d𝑍 + (𝐒(1) T
𝑡 ) [𝝂, 𝝉] ,𝑠 (4.18)
} ∫0 3
(
1
− ℎ2 𝐒(1)
𝑡 ∶ ∇∇𝛿𝜙3 − 𝑞̄3 𝛿𝜙3 d𝐴 = 𝑞̂ (𝑠)𝛿𝜙3 d𝑠 + 𝑚̂ (𝑠)𝛿𝜙3,𝝂 d𝑠. = 2ℎ [(𝐒(0) + ℎ𝐒(1) )𝐤 − (𝐒(0) + ℎ𝐒(1) )T 𝐤] ⋅ 𝝂 + 𝐦𝑡 ⋅ 𝝂
3 ∫𝜕𝛺𝑞 3 ∫𝜕𝛺𝑞 3
) ( )
1 2ℎ3
+ ℎ2 (∇ ⋅ (𝐒(1)
𝑡 )) ⋅ 𝝂 + (𝐒(1) T 4
𝑡 ) [𝝂, 𝝉] ,𝑠 + 𝑂(ℎ ),
3 3
Remarks. 1. Suppose that 𝐪 is the applied force (per unit area) at the
2 2 (2)
plate edge in the 3D case and its in-plane part is 𝐪𝑡 . The corresponding where ℎ (𝐒 𝐤 − (𝐒(2) )T 𝐤) has been dropped as in obtaining (4.2) from
3
virtual work by 𝐪𝑡 is
(2.22) (also see (4.4)).
2ℎ 2ℎ
Thus, the term ∫𝜕𝛺 𝑞̂3 𝛿𝜙3 d𝑠 in (4.12) (for 𝜙3 = 𝑢(0) + ℎ𝑢(1) + 𝑂(ℎ2 ) =
𝐪𝑡 ⋅ 𝛿𝐮𝑡 d𝑠d𝑍 = 𝐒T𝑡 𝝂 ⋅ 𝛿𝐮𝑡 d𝑍d𝑠 (4.13) 𝑞 3 3
∫0 ∫𝜕𝛺𝑞 ∫𝜕𝛺𝑞 ∫0 𝑢̄ 3 + 𝑂(ℎ2 ) = 𝑢𝑚 + 𝑂(ℎ2 )) is the virtual work done by the applied 3D
3
2ℎ force due to the virtual displacement 𝛿𝜙3 (cf. (4.9)).
= (𝐒𝑡 − 𝐒̄ 𝑡 + 𝐒̄ 𝑡 )T 𝝂 ⋅ 𝛿(𝐮𝑡 − 𝐮̄ 𝑡 + 𝐮̄ 𝑡 )d𝑍d𝑠 4. Based on the above three remarks, we conclude that physically
∫𝜕𝛺𝑞 ∫0
2ℎ the weak form (4.11) and (4.12) represents the virtual work principle
= 2ℎ 𝐒̄ T𝑡 𝝂 ⋅ 𝛿 𝐮̄ 𝑡 d𝑍d𝑠 + (𝐒𝑡 − 𝐒̄ 𝑡 )T 𝝂 ⋅ 𝛿(𝐮𝑡 − 𝐮̄ 𝑡 )d𝑍d𝑠 of the 2D plate system, which can be used to implement a finite element
∫𝜕𝛺𝑞 ∫𝜕𝛺𝑞 ∫0
scheme.
2ℎ 2ℎ
+ (𝐒𝑡 − 𝐒̄ 𝑡 )T 𝝂 ⋅ 𝛿 𝐮̄ 𝑡 d𝑍d𝑠 + 𝐒̄ T𝑡 𝝂 ⋅ 𝛿(𝐮𝑡 − 𝐮̄ 𝑡 )d𝑍d𝑠
∫𝜕𝛺𝑞 ∫0 ∫𝜕𝛺𝑞 ∫0 5. Concluding remarks
= 2ℎ 𝐒̄ T𝑡 𝝂 ⋅ 𝛿 𝐮̄ 𝑡 d𝑍d𝑠 + 𝑂(ℎ3 )
∫𝜕𝛺𝑞 In this paper, based on series expansions about the bottom surface, a
( )T uniformly-valid plate theory without any assumption on magnitudes of
= 2ℎ 𝐒(0)
𝑡 + ℎ𝐒(1)
𝑡 𝝂 ⋅ 𝛿(𝐮(0)
𝑡 + ℎ𝐮(1) 3
𝑡 )d𝑍d𝑠 + 𝑂(ℎ ),
∫𝜕𝛺𝑞 applied loads is presented. It is shown that for a Saint-Venant–Kirchhoff

123
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125

material it can recover the plate models in [16] and [13] for five In the special case of
different cases of magnitudes of applied loads. We believe that such
ℎ2 ∇ ⋅ 𝐒(1) = 𝑜((𝐒(0) )T 𝐤), (A.4)
a plate theory is needed, as in all cases there is no need to worry about
its applicability for different levels of applied loads, an advantage not to leading-order, (A.2) gives an algebraic equation
enjoyed by a single plate model derived by Gamma convergence or
asymptotic methods. (𝐒(0) )T 𝐤 = 𝐦 (A.5)
Here, we do not make direct comparisons with the hierarchy of
for solving 𝐱(1) .
Then, the complexity of these two sets of plate systems
plate models derived by Gamma convergence in [5], which considers
for middle-surface and bottom-surface expansions respectively is com-
body forces and zero surface tractions (although in their work surface
parable. The five cases in Section 3 belong to this special case and the
tractions can be added with appropriate scaling; and they did not do
asymptotic reductions based on (A.3) and (A.5) yield the same results,
that for simplicity of exposition). Nevertheless, when the right-hand 1 ℎ2 ( (2) )
sides of (2.20)1 and (2.22) are regarded as the body forces, the cases while ℎ2 ∇ ⋅ 𝐒(2) (2) T
𝑡 in (A.3)1 and 6 ∇ ⋅ 𝐒 𝐤 − (𝐒 ) 𝐤 in (A.3)2 do not
6
(1, 3–5) considered in Section 3 conform with the four plate models make contributions for leading-order correct results. So, one may drop
derived in [5] for a clamped edge. That is, the membrane model, them to obtain
von Kármán plate model, Kirchhoff–Love plate model (linearized von ⎧ (0)
Kármán plate model) and Föppl membrane model in [5] come out ⎪∇ ⋅ 𝐒𝑡 = −𝐪̄ 𝑡 ,
⎨ ( (0) (0) T
) ℎ2 (1)
(A.6)
as special cases of our uniformly-valid plate theory. The physically ⎪∇ ⋅ 𝐒 𝐤 − (𝐒 ) 𝐤 + 3 ∇ ⋅ ∇ ⋅ 𝐒𝑡 = −𝑞̄3 − ∇ ⋅ 𝐦𝑡 ,
meaningful weak formulation of this uniformly-valid plate model is ⎩
presented, which can be used to implement a finite element scheme with (A.5) and (2.12). Comparing with (4.1) and (4.2), (A.6) is not valid
for specific applications. if ℎ2 ∇ ⋅ 𝐒(1) is not smaller than (𝐒(0) )T 𝐤 (e.g. one or whole part of 𝛺
We note that the purpose of this paper is to show the uniform undergoes inextensional rotation, i.e. 𝐅(0) = 𝐑 ∈ 𝐒𝐎(3)). In our view,
validness of the plate model in [17], in the sense that membrane, (4.1) and (4.2) are preferable.
von Kármán, Föppl membrane and Kirchhoff–Love plate models arise
as special cases when different specific scalings are imposed. Some References
examples have been given in [17] and related references (see [22,23]),
which show that plate models obtained by the adopted procedure [1] P.G. Ciarlet, Mathematical Elasticity, Volume II: Theory of Plates, in: Studies in
give asymptotically-correct results in comparison with the available Mathematics and its Applications, vol. 27, North-Holland, Amsterdam, 1997.
higher-dimensional exact solutions. [2] H.L. Dret, A. Raoult, The nonlinear membrane model as variational limit of
nonlinear three-dimensional elasticity, J. Math. Pures Appl. (9) 74 (1995)
Acknowledgments 549–578.
[3] G. Friesecke, R.D. James, S. Müller, A theorem on geometric rigidity and the
derivation of nonlinear plate theory from three dimensional elasticity, Comm.
The work is supported by GRF grants from the Research Grants Pure Appl. Math. 55 (2002) 1461–1506.
Council of Hong Kong SAR, China (Project No. CityU 11303015), a SRG [4] G. Friesecke, S. Müller, R.D. James, Rigorous derivation of nonlinear plate theory
grant from City University of Hong Kong (Project No. 7004670), the and geometric rigidity, C. R. Math. Acad. Sci. Paris 334 (2002) 173–178.
National Natural Science Foundation of China (Project No. 11102068) [5] G. Friesecke, R.D. James, S. Müller, A hierarchy of plate models derived from
nonlinear elasticity by gamma-convergence, Arch. Ration. Mech. Anal. 180
and the US National Science Foundation (Grant No. CMMI-1538228).
(2006) 183–236.
[6] P.G. Ciarlet, P. Destuynder, A justification of a nonlinear model in plate theory,
Appendix. Asymptotic plate model based on the middle-surface
Comput. Methods Appl. Mech. Engrg. 17 (1979) 227–258.
expansions [7] P.G. Ciarlet, P. Destuynder, Justification of the 2-dimensional linear plate model,
J. Mec. 18 (1979) 315–344.
Consider a plate occupying the region 𝛺 × [−ℎ, ℎ]. We assume the [8] P.G. Ciarlet, A justification of the von Kármán equations, Arch. Ration. Mech.
sufficient smoothness for the unknown position vector and the strain Anal. 73 (1980) 349–389.
[9] A.L. Goldenveizer, The principles of reducing three-dimensional problems of
energy function. If one does expansions about 𝑍 = 0 (middle surface),
elasticity to two-dimensional problems of the theory of plates and shells, in:
by almost same procedure in Section 2 the following plate system can H. Gortler (Ed.), Applied Mechanics, Proceedings of the Eleventh International
be derived Congress of Applied Mechanics, Springer, Berlin, 1966, pp. 306–311.
1 [10] A. Goldenveizer, J. Kaplunov, E. Nolde, On Timoshenko-Reissner type theories
∇ ⋅ 𝐒(0) + ℎ2 ∇ ⋅ 𝐒(2) + 𝑂(ℎ4 𝐩+ ) = −𝐪, ̄ (A.1) of plates and shells, Int. J. Solids Struct. 30 (1993) 675–694.
6
[11] H.A. Erbay, E.S. Suhubi, An asymptotic theory of thin hyperelastic plates: Part
1
(𝐒(0) )T 𝐤 − ℎ2 ∇ ⋅ 𝐒(1) + 𝑂(ℎ4 (𝐒(4) )T 𝐤) + 𝑂(ℎ5 𝐩− ) = 𝐦, (A.2) I. General Theory, Int. J. Eng. Sci. 29 (1991) 447–466.
2 [12] H.A. Erbay, E.S. Suhubi, An asymptotic theory of thin hyperelastic plates: Part
with (2.12) and (2.13), where 𝐩± = (𝐒(4) |𝑍=𝑍1 )T 𝐤 ± (𝐒(4) |𝑍=𝑍2 )T 𝐤 for II. Incompressible solids and an application of the general theory, Int. J. Eng.
Sci. 29 (1991) 467–480.
𝑍1 and 𝑍2 between 0 and 𝑍 and 𝐒(𝑖) (𝑖 = 0, 1, 2) are given by (2.10).
[13] H.A. Erbay, On the asymptotic membrane theory of thin hyperelastic plates, Int.
(A.1) and (A.2) are independent, representing the balances of forces J. Eng. Sci. 35 (1997) 151–170.
and torques of a plate element respectively. The system consists of six [14] O. Millet, A. Hamdouni, A. Cimetiere, Dimensional analysis and asymptotic
PDEs and two sets linear algebraic equations for solving 𝐱(2) and 𝐱(3) , expansions of the equilibrium equations in nonlinear elasticity, Part I: the
while the plate system obtained from bottom surface expansions only membrane model, Arch. Mech. 50 (1998) 953–973.
consists of 3 PDEs (2.19), one set of algebraic equations (2.18) and two [15] O. Millet, A. Hamdouni, A. Cimetiere, Dimensional analysis and asymptotic
expansions of the equilibrium equations in nonlinear elasticity, Part II: the
sets of linear algebraic equations (2.12) and (2.13). This is the reason two-dimensional Von Kármán model, Arch. Mech. 50 (1998) 873–899.
why the latter is preferred. [16] O. Millet, A. Hamdouni, A. Cimetiere, A classification of thin plate models by
Upon using (A.2), it is easy to show that (A.1) can be put in the asymptotic expansion of non-linear three-dimensional equilibrium equations, Int.
form J. Non-Linear Mech. 36 (2001) 165–186.
[17] H.-H. Dai, Z. Song, On a consistent finite-strain plate theory based on three-
⎧ (0) 1 2 (2) 4 + dimensional energy principle, Proc. R. Soc. Lond. Ser. A Math. Phys. Eng. Sci.
⎪∇ ⋅ 𝐒𝑡 + 6 ℎ ∇ ⋅ 𝐒𝑡 + 𝑂(ℎ 𝐩𝑡 ) = −𝐪̄ 𝑡 , 470 (2014) 2014.0494.
⎪ ( ) 2 ( ) 2
⎨∇ ⋅ 𝐒(0) 𝐤 − (𝐒(0) )T 𝐤 + ℎ ∇ ⋅ 𝐒(2) 𝐤 − (𝐒(2) )T 𝐤 + ℎ ∇ ⋅ ∇ ⋅ 𝐒(1)𝑡
[18] H.-H. Dai, Z.X. Cai, Phase transitions in a slender cylinder composed of an
⎪ 6 3 incompressible elastic material. i. asymptotic model equation, Proc. R. Soc. Lond.
⎪+𝑂(ℎ4 𝑝+ ) + 𝑂(ℎ4 ∇ ⋅ (𝐒(4) )T 𝐤) + 𝑂(ℎ5 ∇ ⋅ 𝐩− ) = −𝑞̄3 − ∇ ⋅ 𝐦𝑡 . Ser. A Math. Phys. Eng. Sci. 462 (2006) 75–95.
⎩ 3
[19] H.-H. Dai, F.-F. Wang, Analytical solutions for the post-buckling states of an
(A.3) incompressible hyperelastic layer, Anal. Appl. 10 (2012) 21–46.

124
F.-F. Wang, D.J. Steigmann and H.-H. Dai International Journal of Non-Linear Mechanics 112 (2019) 117–125

[20] H.-H. Dai, Y. Wang, F.-F. Wang, Primary and secondary bifurcations of a [26] D.J. Steigmann, Koiter’s shell theory from the perspective of three-dimensional
compressible hyperelastic layer asymptotic model equations and solutions, Int. nonlinear elasticity, J. Elasticity 111 (2012) 91–107.
J. Non-Linear Mech. 52 (2013) 58–72. [27] D.J. Steigmann, A well-posed finite-strain model for thin elastic sheets with bond
[21] H.-H. Dai, F.-F. Wang, J. Wang, J. Xu, Pitchfork and octopus bifurcations in a bending stiffness, Math. Mech. Solids 13 (2013) 103–112.
hyperelastic tube subjected to compression: Analytical post-bifurcation solutions [28] D.J. Steigmann, R.W. Ogden, Classical plate buckling theory as the small-
and imperfection sensitivity, Math. Mech. Solids 20 (2015) 25–52. thickness limit of three-dimensional incremental elasticity, J. Appl. Math. Mech.
[22] J. Wang, Z. Song, H.-H. Dai, On a consistent finite-strain plate theory 94 (2014) 7–20.
for incompressible hyperelastic materials, Int. J. Solids Struct. 78–79 (2016) [29] D.J. Steigmann, Mechanics of materially-uniform thin films, Math. Mech. Solids
101–109. 20 (2015) 309–326.
[23] Z. Song, H.-H. Dai, On a consistent dynamic finite-strain plate theory and its [30] M. Taylor, K. Bertoldi, D.J. Steigmann, Spatial resolution of wrinkle patterns in
linearization, J. Elasticity 125 (2016) 1–35. thin elastic sheets at finite strain, J. Mech. Phys. Solids 62 (2014) 163–180.
[24] D.J. Steigmann, Thin-plate theory for large elastic deformations, Int. J. [31] D. Fox, A. Raoult, J.C. Simo, A justification of nonlinear properly invariant plate
Non-Linear Mech. 42 (2007) 233–240. theories, Arch. Ration. Mech. Anal. 124 (1993) 157–199.
[25] D.J. Steigmann, Asymptotic finite-strain thin-plate theory for elastic solids, [32] E. Ventsel, T. Krauthammer, Thin Plates and Shells: Theory, Analysis, and
Comput. Math. Appl. 53 (2007) 287–295. Applications, Marcel Dekker, New York, 2001.

125

You might also like