You are on page 1of 7

CHAPTER 1

Fundamentals of the finite


element method

Everybody nowadays knows what finite elements are: they are the methods for solving bound-
ary value problems in which one divides the domain of the problem into little pieces over
which the solution is approximated using polynomials. The little pieces are finite elements,
and the polynomials are called shape functions—J. T. Oden (1990).
… I wanted to have a better name for the method because ‘direct stiffness’ did not seem to
convey the essence of the idea. That name is suitable to describe the procedure for assembly
of discrete elements of any type, but it does not imply anything about stress analysis. So
thinking of the work we had done at Boeing in comparing discrete element assembly results
with finite different results, the name ‘finite element method’ seemed to be appropriate … it
seems to have been a good choice because it has stood the test of time—R. W. Clough (1994).
The limitations of the human mind are such that it cannot grasp the behavior of its complex
surroundings and creations in one operation. Thus the process of subdividing all systems into
their individual components or ‘elements’, whose behavior is readily understood, and then
rebuilding the original system for such components to study its behavior is a natural way in
which the engineer, the scientist, or even the economist proceeds—O. C. Zienkiewicz and
R. L. Taylor (1989).
The finite element method (FEM) is one of the most powerful tools in structural
analysis. It can be successfully applied to various field problems in solid mechanics,
fluid mechanics, and electrical engineering. Concentrating on the displacement and
stress analysis, this chapter reviews the finite element formulation for beam bending
problems and plane stress/strain elasto-problems.

1.1 Finite element formulation for beam bending analysis


A structural problem is usually governed by three different kinds of equations,
namely the kinematic relationship, the constitutional law, and the equilibrium
equation, along with attendant boundary conditions. For beam bending problems,
these three equations can be expressed as follows:
kinematic relationship:
κ(x) = −d2 w(x)/dx 2 ; (1.1)
constitutional law:
M(x) = D(x)κ(x); (1.2)
equilibrium equation:
d2 M(x)/dx 2 = −q(x). (1.3)

“chap01” — 2003/6/2 — page 9 — #1


10 Fundamentals of the finite element method

In eqns (1.1)–(1.3), x is the coordinate in the direction of the beam length, w(x) is
the deflection of the beam, κ(x) is the generalized strain of the beam or the curvature
of the deformation, and M(x) is the generalized stress or the bending moment in the
beam; D(x) = E(x)I (x) is the bending stiffness of the beam, where E(x) is Young’s
modulus, I (x) is the moment of inertia of the cross-section of the beam, and q(x)
is the distributed load. Here D(x) is assumed to be the function of the coordinate x.
In other words, the material property of the beam can be inhomogeneous and the
cross-section of the beam can vary.
Combining eqns (1.1)–(1.3) by substituting eqn (1.1) into eqn (1.2) and then
into eqn (1.3), we get the displacement-based equilibrium equation for the bending
beam:
 
d2 d2 w
D(x) 2 = q(x). (1.4)
dx 2 dx
The boundary conditions for the bending beam are mainly divided into the following
three different kinds:
(a) simply supported end:

w = 0, M = M̄; (1.5)

(b) fixed end:

w = 0, θ = dw/dx = 0; (1.6)

(c) free end:

M = M̄, Q = Q̄. (1.7)

Here

θ = dw/dx (1.8)

is the slope,

Q = dM/dx (1.9)

is the shear force, and M̄ and Q̄ are the prescribed moment and the shear force,
respectively.
The fundamental idea of the finite element method is to discretize the continuous
structures into separate ‘elements’. Assume that the beam is divided into a number
of elements and that the ith element is defined by its two nodes xi and xi+1 . The
displacement (deflection) in the ith element is assumed to be a cubic polynomial:

w(ξ ) = a0 + a1 ξ + a2 ξ 2 + ξ 3 , (1.10)

where ξ = 2x/(xi+1 − xi ) − (xi+1 + xi )/(xi+1 − xi ) is the local coordinate of the


ith element. ξ = −1 and 1 at the left and right nodes of the element, respectively.

“chap01” — 2003/6/2 — page 10 — #2


1.1 Finite element formulation for beam bending analysis 11

The displacement in eqn (1.8) and its first derivative (slope) are required to coincide
with the displacement and slope degrees of freedom, respectively, at two nodes of
the element, namely

w|ξ =−1 = a0 − a1 + a2 − a3 = wi ,

dw 
= [a1 − 2a2 + 3a3 ] = lθi ,
dx ξ =−1
(1.11)
w|ξ =1 = a0 + a1 + a2 + a3 = wi+1 ,

dw 
= [a1 + 2a2 + 3a3 ] = lθi+1 ,
dx ξ =1

where l = (xi+1 − xi )/2 is half the length of the element. Solving the con-
stant a0 , a1 , a2 , and a3 from eqn (1.11) and then substituting the results back into
eqn (1.10), we obtain the interpolation form of the displacement as follows:

w(ξ ) = N1 (ξ )wi + N2 (ξ )θi + N3 (ξ )wi+1 + N4 (ξ )θi+1 , (1.12)

where

N1 (ξ ) = 41 (ξ − 1)2 (ξ + 2),
N2 (ξ ) = 41 (ξ − 1)2 (ξ + 1),
(1.13)
N3 (ξ ) = 41 (ξ + 1)2 (−ξ + 2),
N4 (ξ ) = 41 (ξ + 1)2 (ξ − 1),

are the shape functions associated with each node. In matrix form, eqn (1.12) can
be written as

w(ξ ) = N (ξ )δ, (1.14)

where

δ = [wi θi wi+1 θi+1 ]T


(1.15)
N (ξ ) = [N1 (ξ ) N2 (ξ ) N3 (ξ ) N4 (ξ )].

Substituting eqn (1.14) into eqn (1.1) gives the generalized strain, or the curvature

d2 N (ξ ) d2 N (ξ )
κ(ξ ) = − δ = − 2 2 δ. (1.16)
dx 2 l dξ

The generalized stress, or the bending moment, M(x), is then

d2 N (ξ )
M(x) = −D(x) δ. (1.17)
dx 2

“chap01” — 2003/6/2 — page 11 — #3


12 Fundamentals of the finite element method

The principle of virtual work applying to the ith element has the form
 xi+1  xi+1
M(x)κ(x) dx = q(x)w(x) dx + (t + p i )δ. (1.18)
xi xi

The left side of eqn (1.18) is the virtual work done by the generalized stress on
the virtual generalized strain, and the right side of eqn (1.18) is the virtual work
done by the forces acting on the ith element. It is noted that the concentrated forces
on the two nodes have been divided into two parts, namely the externally applied
force t and the inner force p i due to the traction between the adjacent elements.
Substituting eqns (1.14), (1.16), and (1.17) into eqn (1.18), we get

δ T kδ = (p0 + p i )T δ (1.19)

where k and p i are the element stiffness matrix and the equivalent nodal force vector,
respectively:
 xi+1  T
d2 N(ξ ) d2 N (ξ )
k= D(x) dx,
xi dx 2 dx 2
 (1.20)
xi+1
p = 0
N (ξ )q(x) dx + t.
T
xi

Owing to the arbitrariness of the virtual displacement, eqn (1.19) gives rise to the
following element-level finite element equilibrium equation:

kδ = p0 + p i . (1.21)

Substituting eqn (1.13) into eqn (1.20) and assuming that the bending stiffness
D(x) = Di is a constant in the element, we can write the explicit expression for the
element stiffness matrix:
 
12 6l −12 6l
Di  6l 4l 2 −61 2l 2 
k= 3  
. (1.22)
l −12 −6l 12 −6l 
61 2l 2 −61 4l 2

The global finite element equilibrium equation is obtained by assembling the element
equilibrium equations for all elements. It reads

KU = F , (1.23)

where k is the global stiffness matrix, U is the global nodal displacement vector,
and F is the global equivalent nodal external force vector. Note that the equivalent
nodal inner forces are eliminated after assembling the elements due to the principle
of action and reaction.

“chap01” — 2003/6/2 — page 12 — #4


1.2 Finite element formulation for plane stress/strain analysis 13

1.2 Finite element formulation for plane


stress/strain analysis
The three sets of equations for the plane stress and strain elasticity are, respectively,
kinematic relationship:
   
εx   ∂u/∂x 
ε = εy = ∂v/∂y ; (1.24)
   
εz ∂u/∂y + ∂y/∂x

constitutional law:
    
 σx  E −µE 0  εx 
σ = σy = −µE E 0  εy = Dε; (1.25)
   
τxy 0 0 1
2 (1 + µ)E ε z

equilibrium equation:

∂σx /∂x + ∂τxy /∂y + qx = 0,


(1.26)
∂τxy /∂x + ∂σy /∂y + qy = 0,

where u is the displacement in the x-direction and v is the displacement in the


y-direction, qx and qy are distributed body forces in the x and y directions, respect-
ively; εx and εy are extensional strains in the x and y directions and γxy is the shear
strain; σx and σy are extensional stresses in the x and y directions and τxy is the shear
stress; E is Young’s modulus, and µ is the Poisson ratio. The attendant boundary
conditions can be described as follows.
On the displacement prescribed boundary:

u = ū, v = v̄. (1.27)

On the force prescribed boundary:

tn = t¯n , ts = t¯s , (1.28)

where tn and ts are tractions in the normal and tangential directions of the boundary,
respectively, and an overbar represents the prescribed value.
Here we will only derive a four-node quadrilateral element for plane stress or
strain analysis. The geometrical domain of the element is defined in terms of shape
functions and its nodal coordinates, as follows:
  4  
x x
x= = Ni (ξ, η) i = N (ξ, η)X, (1.29)
y yi
i=1

where x = (x, y)T is the coordinate system, (xi , yi ) (i = 1, 2, 3, 4) is the


coordinate of the ith node of the element, X = [x1 , y1 , x2 , y2 , x3 , y3 , x4 , y4 ]T

“chap01” — 2003/6/2 — page 13 — #5


14 Fundamentals of the finite element method

is the vector of element nodal coordinates, N (ξ, η) is the shape function matrix
defined as
 
N1 (ξ, η) 0 N2 (ξ, η) 0 N3 (ξ, η) 0 N4 (ξ, η) 0
N (ξ, η) = ,
0 N1 (ξ, η) 0 N2 (ξ, η) 0 N3 (ξ, η) 0 N4 (ξ, η)

Ni (ξ, η) = 41 (1 + ξ ξi )(1 + ηηi ), i = 1, 2, 3, 4, (1.30)

where ξ and η consist of the local coordinate system associated with the ele-
ment, and (ξi , ηi ) is the local coordinate of the ith node of the element. The
displacements in the element are interpolated by the same expressions as for the
coordinates:
  4  
u u
u= = Ni (ξ, η) i = N (ξ, η)δ, (1.31)
v vi
i=1

where (ui , vi ) (i = 1, 2, 3, 4) is the displacement of the ith node of the element,


and

δ = [u1 , v1 , u2 , v2 , u3 , v3 , u4 , v4 ]T (1.32)

is the vector of the element nodal degrees of freedom. The element with the same
interpolation expressions for the displacements and the coordinates is called the
‘isoparametric element’.
Substituting eqn (1.31) into eqn (1.24) gives the strain vector in terms of the shape
functions and the nodal degrees of freedom:

ε = Bδ, (1.33)

where
 
∂Ni /∂x 0
B = [B1 , B2 , B3 , B4 ], Bi =  0 ∂Ni /∂y  , i = 1, 2, 3, 4.
∂Ni /∂y ∂Ni /∂x
(1.34)

B is called the strain matrix. The principle of minimum potential energy applied to
the element requires
  
π= 2 ε Dε dx dy −
1 T
q u dx dy −
T
(t 0 + t i )T u ds = min, (1.35)
A A ∂A

where A is the area of the element, ∂A is the boundary of A, q = (qx , qy )T is the vec-
tor of body forces, and t 0 and t i are boundary tractions due to the externally applied
load and the inner forces between adjacent elements. Substituting eqns (1.29),
and (1.33) into eqn (1.35) gives

π = 21 δ T kδδ − (p 0 + p i )T δ, (1.36)

“chap01” — 2003/6/2 — page 14 — #6


1.2 Finite element formulation for plane stress/strain analysis 15

where k is the element stiffness matrix, and p 0 and p i are equivalent nodal
forces of externally applied load and boundary tractions between adjacent elements,
respectively. They read

k= B T DB dx dy,
A
 
p0 = N T q dx dy + N T t 0 ds, (1.37)
A ∂A

pi = N T t i ds.
∂A

Minimizing eqn (1.36) results in the following element-level finite element equation:

kδ = p0 + p i . (1.38)

Again, after assembling all element equilibrium equations, the inner force terms
are eliminated from each other due to the priniciple of action and reaction between
adjacent elements. The global finite element equilibrium equation is then written as

KU = F , (1.39)

where K is the global stiffness matrix, U is the global vector of all nodal degrees
of freedom, and F is the global vector of all equivalent nodal applied forces.

“chap01” — 2003/6/2 — page 15 — #7

You might also like