Plate (Shell) Elements

The Finite Element Library of the MIDAS Family Programs includes 3-node triangular and 4-node
quadrilateral Plate/Shell elements.

The MIDAS shell elements are flat shell elements and are formulated as a combination of a plane stress
and plate-bending elements. The stiffness terms associated with two in-plane translational degrees of
freedom (DOF) are formulated in the same way as for plane stress element. The stiffness terms related to
one out-of-plane translational DOF and two out-of-plane rotational DOF are formulated by the same
procedure used for the plate-bending element.

Shell Plane stress Plate-bending

Figure 1 Formulation of a Shell Element

For details on formulation of a plane stress element, refer to “Plane Stress & Plane Strain Element”.

A plate-bending element is used to model a shell subjected to out-of-plane bending. This element is
classified into a thin plate element, which ignores shear deformations and a thick plate element, which
accounts for the effects of shear deformations.

The formulation of a thin plate element is based on the concept of Discrete Kirchhoff Triangular (DKT)
element. As shown in Fig. 2., this element has one translational DOF and two rotational DOF in the local
coordinates. A four node quadrilateral element is constructed using this DKT element. As shown below, a
quadrilateral is divided into two triangles, and the two triangular elements are superimposed. Since there
are two directions of diagonals, the four triangular elements are superimposed, and the resultant stiffness
is obtained by dividing the sum by 2 (see Fig. 2).
 Figure 2 Formulation of a Plate Element based on the DKT Element
The plate finite element formulation based on the concept of the DKT element is described below. This
element is derived from the assumptions ignoring shear deformations (Kirchhoff thin plate hypothesis).
The plate theory accounting for transverse shear deformations (Mindlin-Reissner plate theory) is based on
the following assumption: “The particles located normal to the neutral surface of an undeformed plate,
after deformation, remain on a straight line but this straight line is not necessarily normal to the deformed
neutral surface.” Using this assumption, the displacement components at an arbitrary point in the element
are,

u = z β x ( x, y ) , v = z β y ( x, y ) , w = w ( x, y )

where, β x and β y are the rotation angles of the lines normal to the undeformed neutral surface in the
x-z and y-z planes, respectively. According to Kirchhoff plate theory, which ignores shear deformations,
∂w ∂w
βx = − and β y =
∂x ∂y

The equation for linear flexural strains is,

 ∂β x 
 
 ∂x 
 ∂β y 
ε b = zκ = z  
 ∂y 
 ∂β ∂β y

 x+ 
 ∂y ∂x 

where, z is the distance from the neutral surface and κ is the vector of curvatures.
And the equation for shear strains is,

 ∂w 
 ∂x + β x 
γ = 
 ∂w + β 
 ∂y y

The state of the stresses is assumed equal to that of a plane stress element. Therefore, for an isotropic
element, the flexural stresses of the element are,

σ x  1 ν 0 
  E  
σ b = σ y  = zDκ = z ν 1 0 κ
1 −ν 2
τ xy 
 
 0 0 (1 −ν ) 2 
and the shear strains are,

τ xz  E 1 0 
σ s =   = Eγ = γ
τ yz  2 (1 + ν ) 0 1 

Now, the strain energy is obtained as follows:

1 1
U = Ub + U s =
2 ∫ A
κ T Dbκ dA + ∫ γ T Dsγ dA
2 A
where,
Db = Stress-strain (material) matrix for bending
Ds = Stress-Strain matrix for shear
For an isotropic material, Db and Ds take the following form:

1 ν 0 
h2 Eh3  
Db = ∫ Dz dz =
2
ν 1 0
−h 2
12 (1 − ν ) 2  
 0 0 (1 −ν ) 2 
h2 kEh 1 0 
Ds = k ∫ Edz =
−h 2 2 (1 + ν ) 0 1 

in which, k is the correction factor for considering the non-uniform characteristic of shear strains, and is
usually assumed as 5/6.
Then, by definition, flexural moments M and shear forces Q are obtained through integration of stresses
over the thickness as follows:

 Mx 
  h2
M =  M y  = ∫ σ zdz = Dbκ
−h 2
 M xy 
 
 Qx  h2
Q =   = k ∫ σ s dz = Dsγ
Qy 
−h 2

The boundary condition of this element is C0 continuation condition, which includes only the
compatibility of displacements and rotations. Since it does not involve differential terms, it is an easy task
to satisfy the condition.
The above equations represent the formulation of a thick plate element, which includes shear
deformations. Now, we proceed to formulation of a DKT element without shear deformations. The DKT
element and node numbering convention is illustrated in Fig. 3. We first ignore the Us term and assume
the following:

(1) The element has only three DOF, i.e., displacement w and two rotations θ x and θ y at each node.

(2) Since the shear deformations are ignored, the nodal rotations must satisfy the Kirchhoff boundary
∂w ∂w
conditions, θx = and θ y = − .
∂y ∂x
(3) The assumptions of the Kirchhoff plate theory can be constrained at some discrete points.

(4) The compatibility condition of rotations, β x and β y , must be satisfied.

Also, we assume the following to formulate a DKT element:

(1) The rotations β x and β y are imposed in a quadratic variation in the element, i.e.,
6 6
β x = ∑ fi β xi and β y = ∑ f i β yi .
i =1 i =1

where, β xi and β yi represent the values of rotations at each node and mid-edge point, fi

represents the shape functions associated with i-th node and defined as follows:

 f1 = ξ1 ( 2ξ1 − 1)

 f 2 = ξ 2 ( 2ξ 2 − 1)
 f3 = ξ 3 ( 2ξ 3 − 1)
 ξ1 + ξ 2 + ξ 3 = 1
 f 4 = 4ξ 2ξ 3
 f 5 = 4ξ 3ξ1

 f 6 = 4ξ1ξ 2

(2) The Kirchhoff hypothesis is constrained at each node and mid-node. That is, the following
conditions must be satisfied:

 ∂w 
 β +
∂x 
x
γ =  = 0 (at nodes 1, 2 & 3)
 β + ∂w 
 y ∂y 

∂w
βs + =0 (at mid-edge points)
∂s
where, s is the direction along the edge.
 (3) The variation of w along the edge is cubic, i.e.,
 ∂w  3 1  ∂w  3 1  ∂w 
  =− wi −   + wj −  
 ∂s  k 2lij 4  ∂s i 2lij 4  ∂s  j
where, k denotes the mid-node of the side ij, and lij represents the length of the side ij.
(4) The rotational angle βn about the tangential direction to the side is assumed to vary linearly
along the side, i.e.,
1
β nk =
2
( β ni + β nj )

Now, the nodal degrees of freedom are,

T
u =  w1 θ x1 θ y1 w2 θ x 2 θ y 2 w3 θ x 3 θ y 3 

The following geometrical relations are required on each side:

 β x  c − s   β n 
β  =   
 y  s c   βs 
 ∂w 
 ∂s  c s  θ x 
 =  
 ∂w   s −c  θ y 
 ∂n 
where, c = cos(x, nij) and s = sin(x, nij)

Figure 3 Node numbering convention for DKT element

β x and β y can be obtained by the assumed shape functions as follows:

1.5 ( a6 f 6 − a5 f 5 ) 
T
 w1 
  θ 
 b5 f5 + b6 f 6   x1 
 f1 − c5 f5 − c6 f 6  θ y1 
   
1.5 ( a4 f 4 − a6 f 6 )   w2 
β x = HTx (ξ 2 , ξ 3 ) U =  b6 f 6 + b4 f 4  θ x 
   2
 f1 − c6 f 6 − c4 f 4  θ y2 
1.5 ( a f − a f )  w 
 5 5 4 4
  3
 b f
4 4 + b f
5 5  θ x3 
 f −c f −c f  θ 
 1 4 4 5 5   y3 

1.5 ( d 6 f 6 − d5 f 5 ) 
T
 w1 
  θ 
 − f1 + e5 f 5 + e6 f 6   x1 
 − ( b5 f 5 + b6 f 6 )  θ y1 
   
1.5 ( d 4 f 4 − d 6 f 6 )   w2 
β y = HTy (ξ 2 , ξ 3 ) U =  − f1 + e6 f 6 + e4 f 4  θ x 
   2
 − ( b6 f 6 + b4 f 4 )  θ y2 
1.5 ( d f − d f )  w 
 5 5 4 4
  3
 − f1 + e4 f 4 + e5 f5  θ x3 
   
 − ( b4 f 4 + b5 f 5 )  θ y3 
where
3 1 1 
ak = − xij lij2 , bk = xij yij lij2 , ck =  xij2 − yij2  lij2
4 4 2 
1 1 
d k = − yij lij2 , ek =  yij2 − xij2  lij2 , lij2 = xij2 + yij2
4 2 
in which for ij = 23, 31 & 12, k = 4, 5 & 6, respectively, and xij = xi-xj, yij = yi-yj
Then, the vector of curvatures κ is obtained as follows:

κ = Bu
 y31HTx ,ξ2 + y12 HTx ,ξ3 
1  
B (ξ 2 , ξ 3 ) =  − x31HTy ,ξ2 − x12 HTy ,ξ3 
2A
 − x31HTx ,ξ − x12 HTx ,ξ + y31HTy ,ξ + y12 HTy ,ξ 
 2 3 2 3 

where, A is the area of the element

Accordingly, the stiffness matrix of the DKT element becomes,
1 1−ξ3
K DKT = 2 A∫ ∫ BT Db Bdξ 2 dξ 3
0 0

The stiffness matrix in the above equation is calculated by numerical integration using the 3-point
Gaussian quadrature.

Also, the bending moments M at any point in the element are obtained as follows:

M ( x , y ) = Db B ( x , y ) U
 x = x1 + ξ 2 x21 + ξ 3 x31

 y = y1 + ξ 2 y21 + ξ 3 y31

The thick plate element included in MIDAS Finite Element Library is a modified Discrete Kirchhoff-
Mindlin Triangular/Quadrilateral (DKMT/DKMQ) element.

The formulation procedure for the DKMT element is quite similar to that used for DKT element. Except,
the terms related to the shear stiffness are included in the formulation. Also, for the sake of convenience
of formulation, some parts are expressed in somewhat different forms.

First, the equation for the curvatures and the stress-strain relation are identical to those used in the DKT
element.
The DKMT element and node numbering convention is illustrated in Fig. 4. The shape functions are
expressed by dividing the node terms and mid-node terms as follows:
 3 6

 x ∑ i xi ∑ Pk ck ∆β sk
β = f β +
 i =1 k =4
 3 6
β =
 y ∑ ∑
f β
i yi + Pk sk ∆β sk
i =1 k =4
 Figure 4 Node numbering convention for DKMT element
where,

fi = ξ i , P4 = 4ξ1ξ 2 , P5 = 4ξ 2ξ 3 , P6 = 4ξ 3ξ1 (ξ1 + ξ 2 + ξ 3 = 1) (i =1, 2, 3)

Then, the curvature vector is calculated by,

κ = Bbβ u + Bb∆β ∆β

where:

0 − y32 0 0 − y13 0 0 − y21 0 

1 
B bβ = 0 0 x32 0 0 x13 0 0 x21 
2A 
0 x32 − y32 0 x13 − y13 0 x21 − y21 
  ∂P ∂P  
 −  k y13 + k y21  ck 
  ∂ξ 2 ∂ξ 3  
  ∂Pk  
1  ∂Pk  ( k = 4,5, 6 )
B b∆ β = "  x + x s
21  k "
 ∂ξ 2 ∂ξ 3
13
2A   
 
  ∂Pk ∂Pk   ∂Pk ∂Pk  
  x13 + x21  ck −  y13 + y21  sk 
  ∂ξ 2 ∂ξ 3   ∂ξ 2 ∂ξ 3  
∆β = [ ∆β s 4 ∆β s 6 ]
T
∆β s 5

Also, the shear strains are calculated by,

γ xz 
γ =   = B s∆β ∆β
γ yz 
where
  s5 s6   s6 s4   s4 s5  
 ξ 2 − ξ1  φ4  ξ3 − ξ 2  φ5  ξ1 − ξ3  φ6 
2  A2 A1   A3 A2   A1 A3  
B s∆β =  
3  c6  c4 c6   c5 c4  
 ξ1 − c5 ξ 2  φ4  ξ 2 − ξ3  φ5  ξ − ξ φ
1 6
 A1 A2   A2 A3   A3
3
A1  

in which

2  h2 
φk =   , A1 = c4 s6 − c6 s4 , A2 = c5 s4 − c4 s5 , A3 = c6 s5 − c5 s6
k (1 − ν )  lij2 

A1 = c4 s6 − c6 s4 , A2 = c5 s4 − c4 s5 , A3 = c6 s5 − c5 s6

Accordingly, the strain-displacement matrix B is given by,

κ = Bbu = ( Bbβ + Bb∆β A n ) u

γ = B s u = B s ∆β A n u

where, An is the transformation matrix, which relates ∆β = A n u and is given by,

A n = A −∆1β A w
2 
 3 l12 (1 + φ4 ) 0 0 
 
 2 
A ∆β = 0 l23 (1 + φ5 ) 0
 3 
 
2
 0 0 l31 (1 + φ6 ) 
 3 
 x21 y21 x21 y21 
 1 − 2 − 2 −1 − 2 − 2 0 0 0 
 
x y x y32 
Aw =  0 0 0 1 − 32 − 32 −1 − 32 −
 2 2 2 2 
 y 
 −1 − x13 − y13 0 0 0
x
1 − 13 − 13 
 2 2 2 2 

Therefore, the final stiffness matrix is given by,

K = Kb + K s
K b = ∫ BTb Db Bb dA, K s = ∫ BTs Ds B s dA
A A

Also, the bending moments and shear forces at any point in the element are defined as,
M = Db B b u
T = Ds B su
The formulation procedure of the DKMQ element is quite similar to that used for the DKMT element.
Except, some differences exist due to the differences in the number of nodes and shape functions. The
matrices for the DKMQ element corresponding to those for the DKMT element are as follows:

 1 1
 f1 = 4 (1 − ξ )(1 − η ) P5 =
2
(1 − ξ 2 ) (1 − η )

 f = 1 (1 + ξ )(1 − η ) 1
P6 = (1 + ξ ) (1 − η 2 )
 2 4 2

 f = 1 (1 + ξ )(1 + η ) 1
P7 = (1 − ξ 2 ) (1 + η )
 3 4 2
 1 1
 f 4 = (1 − ξ )(1 + η ) P8 = (1 − ξ ) (1 − η 2 )
 4 2

 ∂fi ∂fi
 0 ai  ai = j11 ∂ξ + j12 ∂η
0 
 
B bβ = " 0 0 bi " ,
 , ( i = 1, 2,3, 4 )
 0 bi b = j ∂f i + j ∂fi
ai 
 i 21
∂ξ
22
∂η
  ∂Pk ∂P  
  j11 + k j12  ck 
  ∂ξ ∂η  
  ∂Pk ∂Pk  
B b∆ β = "  j + j s
22  k "  ( k = 5, 6, 7,8)
 ∂ξ ∂η
21
  
 
  ∂Pk ∂Pk   ∂Pk ∂Pk  
 j21 + j22  ck −  j11 + j12  sk
  ∂ξ ∂η   ∂ξ ∂η  

∆β = [ ∆β s 5 ∆β s 8 ]
T
∆β s 6 ∆β s 7

1  − j11 (1 − η ) l12φ5 − j12 (1 + ξ ) l23φ6 j11 (1 + η ) l34φ7 j12 (1 − ξ ) l41φ8 

B s∆β =  
6  − j21 (1 − η ) l12φ5 − j22 (1 + ξ ) l23φ6 j21 (1 + η ) l34φ7 j22 (1 − ξ ) l41φ8 

2  h2 
φk =   ( k = 5, 6, 7,8)
k (1 − ν )  lij2 

where, j11, j12, j21 and j22 are the components of the inverse Jacobian matrix.

The transformation matrix An takes the following form:

A n = A −∆1β A w
2 
 3 l12 (1 + φ5 ) 0 0 0 
 
 2 
0 l23 (1 + φ6 ) 0 0
 3 
A ∆β =  
2
 0 0 l34 (1 + φ7 ) 0 
 3 
 2 
 0 0 0 l41 (1 + φ8 ) 
 3 
 x21 y21 x21 y21 
 1 − 2 − 2 −1 − 2 − 2 0 0 0 0 0 0 
 
0 x y x32 y32
0 0 1 − 32 − 32 −1 − − 0 0 0 
 2 2 2 2 
Aw = 
0 x y x43 y 
0 0 0 0 0 1 − 43 − 43 −1 − − 43 
 2 2 2 2 
 x y x y 
 −1 − 14 − 14 0 0 0 0 0 0 1 − 14 − 14 
 2 2 2 2 