Chapter 2

Introduction to the Stiffness (Displacement) Method

The basic concepts on which the Direct Stiffness Method is founded will be introduced.
The linear spring model will be used because it provides a simple yet generally instructive tool.

2.1 Definition of the Stiffness Matrix:

a) For a single element, a stiffness matrix k is a matrix such that
f k d

where k relates local-coordinate nodal displacements d to the local forces

Note:

(2.1)
of an element.

lacoL
setanidrooc
k

Matrix

Note that

x, y , z will represent local coordinates and

x, y , z will represent global coordinates.

b) For a continuous medium or structure

comprising a series of elements, a stiffness matrix K relates global-coordinate nodal
displacements d to global forces F of the whole structure,
F K d

(2.2)

2.2 Derivation of the Stiffness Matrix for a Spring Element:

For the linear spring shown; the relationship between the

force f applied at the free end and the displacement d is

k
d

f = k d
where k = spring constant or stiffness of spring (N/m)

f
1

Consider the same spring but as a part of a structure, connected to other springs at its ends; both of which
may deflect and may have forces applied to them; so

points 1 & 2 nodes of element

x
local axis, positive direction as indicated

f1x , f 2 x
local nodal forces, positive in x direction
direction.
d1x , d 2 x local nodal displacements (DOF at each node), positive in x

So , we can write the vectors

f1x
f

f2 x

and

d1x
d

d 2 x

Therefore, the stiffness matrix for the spring element shown is of the order (2 x 2) and Eq. (2.1) takes the
form:

f1x k11 k12 d1x

f2x k21 k22 d2x

(2.3)

General Steps Applied to a Spring Structure:

Consider a structure that consists of a series of springs as shown:

Step 1. Discretize and Select Element Type:

Consider the linear spring element shown with undeformed

length of L, with stiffness k and nodes 1 & 2.
Because of the adjacent springs, this element will be subjected
.
to nodal tensile forces T directed along the spring local axis x
So, the element is in equilibrium.

Step 2. Select a Displacement Function:

because a
Here we assume a linear displacement function u
linear function with specified endpoints has a unique path. Therefore,
u a1 a 2 x

(2.4)

In general,

total # of the ai coefficients = total # of DOF associated with the element

Eq. (2.4) in matrix form

u 1

a1

a2

at each node;
Evaluate u

u 0 d1x a1 , and

a2

(2.4a)

u L d 2 x a1 a 2 L

d 2 x d1x
L

and Eq. (2.4) becomes

d 2 x d1x

x d1x

(2.5)

and in matrix form

x d1x

L d 2 x

u 1
L

u N1

or

N1 1

where

(2.5a)

d1x

N2

x
,
L

2x

(2.5b)

and

x
L

N2

(2.6)

are called the shape functions (interpolation functions); noticing

that
N1 = 1 at node 1 and N1 = 0 at node 2
N2 = 0 at node 1 and N2 = 1 at node 2
N1 + N2 = 1

, and

for any x along the element.

Step 3. Define Strain/Displacement & Stress/Strain

Relationships:

For the linear spring, Hooks law states (force/deformation instead of

stress/strain)
T k

T = tensile force,

where

(2.7)
and

= total elongation (deformation) in the spring

for the element we have

( L) u
( 0)
u
using Eq. (2.5), the above equation yields

d 2 x d1 x

(2.8)

, therefore;
Note that d 1 x (see figure on page 2) has a negative value because it in the opposite direction of x

= positive value (negative value). Eqn. (2.7) becomes

T k (d 2 x d1 x )

(2.8a)

Step 4. Derive the Element Stiffness Matrix and Equations:

Recall that we have three alternative methods that can be used to obtain element stiffness matrix, namely;

1.
2.
3.

Direct Equilibrium method.

Work or Energy methods (min. P.E. is the most common one).
Weighted Residual Methods (Galerkins is the most popular one).

For the spring elements, we will use the direct equilibrium approach:
Comparing the element nodal force convention versus the element equilibrium (see figures on p.2 & 3); we can
write
f1x T ,
and
f2 x T
Using equations (2.7) and (2.8a) we get

T f1x k ( d 2 x d1x ) ,

or

f1x k ( d1x d 2 x )

and

T f2 x k ( d 2 x d1 x )

(2.9a)

f2 x k ( d 2 x d1x )

(2.9b)

So the element equations in matrix form are:

f1x k k d1x

f2x k k d2x

(2.10)

and therefore the local stiffness matrix for a spring element is

k
k
k

Note that

k
k

(2.11)

is a symmetric square matrix ( k ij k ji ) .

Step 5. Assemble Element Equations to Obtain Global Equations:

Next we assemble element equations to obtain the total (structure) equations. To accomplish this
we can use either

1. Use nodal force equilibrium, force/deformation, and compatibility equations, or

2. Direct stiffness method.

As a result of either approach the total stiffness matrix and total force matrix of the system
(total structure) will also be obtained:
N

K K k ,
e 1

(e)

(e )

F F f

(e )

(2.12)

e 1

(e)

where k , and f are now element stiffness and force matrices, respectively, in a global
frame of reference. If not in a global coordinate axes, element matrices would have to be
transformed into a global frame then may be assembled.
Note that the sign does not mean a simple summation of element matrices but it means that
these element matrices must be assembled properly.
Step 6. Apply Boundary Conditions and Solve for Nodal Displacements:

Next we impose boundary conditions to the total system F k d ; from which we

obtain a reduced (smaller) system. This reduced system then can be solved for the
unknown ds.
Step 7. Solve for Element Forces:

Back substituting the values of ds into either equation (2.9) or (2.10) will yield the
element forces.
Note:
To illustrate the full details of steps 5-7, we consider solving the spring assemblage as
shown next.

2.3 Assembling the Total Stiffness Matrix Using:

Nodal Force Equilibrium, Force/Deformation, and Compatibility Equations:
Consider the following example (spring assemblage):

Nodes have been numbered 1, 3, and 2 for further generalization because sequential numbering
between elements generally does not occur in large problems.
Node 1 is fixed.
F3x and F2x are applied global forces at nodes 3 and 2, respectively.
x is the global axis of the total structure.
of each element coincides with the global axis of the system, i.e.; di and d i are
Local axis x
the same.

(a) Using Eq. (2.10), force/displacement equation, we write:

For Element (1)

f1(x1) k1 k1 d1(1x)
(1)
(1)
f 3 x k1 k1 d 3 x

(2.13)

For element (2)

f 3(x2) k 2 k 2 d 3( 2x )
( 2)
( 2)
f 2x k2 k2 d 2x

(2.14)

(b) Compatibility (continuity) requirements yield:

d 3(1x) d 3( 2x ) d 3 x

(2.15)

Where,

d 3(1x) = local displacement at node 3 of element (1)

( 2)
d 3x
= local displacement at node 3 of element (2), and

d 3 x = global displacement at node 3 of spring assemblage.

(c) Write nodal equilibrium equations:

Using force convention of elements, we write nodal equilibrium equations at nodes 1, 2, and 3:
Node 1,

F1x f1(x1)

Node 2,

F2 x f 2(x2 )

Node 3,

F3 x f

(1)
3x

(2.16)

( 2)
3x

Where F1x is the unknown reaction; and F2x & F3x are applied external forces.

Substituting equations (2.13), (2.14) & (2.15) into equation (2.16), we get
F1x k1 d1 x k1 d 3 x
F2 x k 2 d 3 x k 2 d 2 x
F3 x k1 d1 x k1 d 3 x k 2 d 3 x k 2 d 2 x

or in matrix form,

F1x k1

F2 x 0
F k
1
3x

k1
k 2
k1 k 2

0
k2
k2

d1 x

d 2x
d
3x

Total or global
Stiffness matrix

Global nodal
force matrix

(2.17)

Global nodal
displacement
matrix

Or in compact form,

F K d
(2.18)
One can see how tedious work would be involved if we had large number of nodes, so we need a more
practical method!

2.4 Assembling the Total Stiffness Matrix Using:

Direct Stiffness Method (Superposition)
In this method we use a proper superposition of individual element matrices. For the two-spring
assemblage above, the element stiffness matrices are:
,
(2.19)
where d ix ' s written above indicate the degree of freedom associated with each element. Since the
two matrices are not associated with the same degrees of freedom, they cannot be directly added
together.
To superimpose the element matrices, they must be expanded to the order of the total structure
stiffness matrix, i.e.; each matrix is to be associated with all the degrees of freedom of the structure.
To do that, we simply add rows and columns of zeros for those displacements not associated with
that particular element.
So, equations (2.13) & (2.14) in expanded form are:

d1 x

f1(x1)

f 2(x1)
f (1)
3x

(1 )

k1

k1

d3 x

k1
k1

d3x

d1 x

d3x

( 2)

k2

k2

d2 x

k2
k2

1 0 1 d1(1x)

k1 0 0 0 d 2(1x)
1 0 1 d 3(1x)

d3 x

d 2x

f1(x2)
;

(2.20)

Next, consider force equilibrium at each node (see equation 2.16):

(1)
F1x f1x 0
( 2)

F2 x 0 f 2 x
F f (1) f ( 2)
3x 3x 3x

f 2(x2)
f ( 2)
3x

0 0 0 d1(x2)

k 2 0 1 1 d 2( 2x)
0 1 1 d 3( 2x )

(2.21)

Substituting equations (2.20) into (2.21), yields

(1)
( 2)
F1x
1 0 1 d1x
0 0 0 d 1x

(1)

( 2)
F2 x k1 0 0 0 d 2 x k 2 0 1 1 d 2 x
F
1 0 1 d 3(1x)
0 1 1 d 3( 2x )
3x

(2.22)

(1)
( 2)
(1)
( 2)
Finally, use the compatibility requirements d 3 x d 3 x d 3 x and d1x d1x & d 2 x d 2 x into
equation (2.22), we get

F1x k1

F2 x 0
F k
1
3x

k1
k 2
k1 k 2

0
k2
k2

d1 x

d 2x
d
3x

(2.23)

which is identical to equation (2.17)

We could have directly added the expanded element stiffness matrices (eq. 2.20) to obtain the
total stiffness matrix of the structure and hence the total set of stiffness equations. This method
is called the Direct Stiffness method. It is the most important step in the finite element method.

Now, for problems involving a large number of degrees of freedom, it becomes tedious to expand
each element stiffness matrix to the order of the total stiffness matrix. To avoid such expansions,
we use a direct (shortcut) form of the Direct Stiffness method to obtain the total stiffness matrix:
a) the rows and columns of each element stiffness matrix are labeled according to the degrees of
freedom associated with them as follows:
d1 x
k1
k (1 )
k1

d3 x
k1
k1

d1 x
d3x

d3x
k2
k ( 2)
k2

d2 x
k2
k2

d3 x
d 2x

b) total stiffness matrix K is then constructed by directly adding terms associated with degrees of
freedom in k ( 1) and k ( 2 ) into their corresponding identical degree-of-freedom locations in as
follows:

d1 x
k1
K 0
k1

d2x
0
k2
k2

d3x
k1 d 1 x
k 2 d 2 x
k1 k 2 d 3 x

(2.24)

where elements in K are located on the basis that degrees of freedom are ordered in an increasing node
numerical order for the total structure.

2.5 Applying Boundary Conditions & solving for Nodal Displacements:

Without specifying kinematic constraints or support conditions, the structure will be free to move
as a rigid body, and the total stiffness matrix would be singular ( | K | = 0 ) and hence K has no
inverse.
Boundary conditions are of two general types:
a) Homogeneous boundary conditions; nodes are prevented from movement (most common).
b) Non-homogeneous boundary conditions; finite nonzero values of displacement are
specified such as the settlement of a support.

Example for homogeneous B.Cs.: consider that node 1 of the spring assemblage is fixed, i.e.;
d1x = 0, and hence equation (2.23) becomes

 k1
0

k1

k1
k 2
k1 k 2

0
k2
k2

upon expanding we get

0
F1x

d 2 x F2 x
d
F
3x
3x

(2.25)

k1 (0) 0 ( d 2 x ) k1 d 3 x F1x
0 (0) k 2 d 2 x k 2 d 3 x F2 x

(2.26)

k1 (0) k 2 d 2 x (k1 k 2 ) d 3 x F3 x
last two equations can be written in matrix form as:

k 2 d 2 x F2 x

k1 k 2 d 3 x F3 x

k2
k

(2.27)

Instead of expanding equation (2.25) we could, in general, cross out rows and columns
associated with zero displacements; i.e.; we partition the total stiffness equations. So, for the
spring example we partition off 1st row and 1st column of K and 1st row of both d and F,
(Only true for homo. BCs.)
Solving eq.(2.27) for the unknown generalized displacements (ds), we write

d 2x
k2

d 3x
k2

k2
k1 k 2

F2 x

;
F3 x

or

1 1

d 2 x k 2 k1

1
d 3x
k1

1
k1 F2 x

1 F3 x
k1

(2.28)

Next we use the above results into equation (2.25) to solve for the unknown reaction:
F1 x k1 d 3 x
F1 x k1 (

1
1
F2 x
F3 x )
k1
k1

(2.29)

F1 x F2 x F3 x

Example for non-homogeneous B.Cs.: consider that node 1 of the spring assemblage has a
nonzero displacement of d 1 x , equation (2.23) becomes:
k1
0

k1

0
k2
k2

k1
k 2
k1 k 2

F1x

d 2 x F2 x
d
F
3x
3x

(2.30)

upon expanding, we ge
k1 0 ( d 2 x ) k1 d 3 x F1x
0 ( ) k 2 d 2 x k 2 d 3 x F2 x
k1 k 2 d 2 x (k1 k 2 ) d 3 x F3 x
Considering the last two equations, we may write

 k2
k

k2 d 2x
F2 x

k1 k 2 d 3 x
k1 F3 x

(2.31)

Therefore, in general, for non-homogeneous boundary conditions, we must transform the terms
associated with the specified displacements to the right-side force matrix before solving for the
unknown nodal displacements.

Some properties of K:
K is symmetric (as the case in each k)
K is singular (has no inverse) until sufficient boundary conditions are imposes to remove
singularity and prevent rigid-body motion.
Main diagonal elements of K are always positive.

Discuss Examples: 2.1, 2.2, & 2.3 (see Examples 1.1 & 1.2 Liu)

2.6 Potential Energy Approach to Derive Spring Element Equations:

Recall that the alternative methods used to derive the element equations are:
1. Equilibrium method.
2. Energy methods (minimum potential, virtual work, & Castiglianos).
3. Weighted residual methods (Galerkins)
Among the most common energy methods is the minimum potential method. This method are
more adaptable than equilibrium method, specifically, for complicated elements (those with
large DOF) such as plane stress/strain element, axisymmetric stress element, plate bending
element, and 3-D solid stress element. The minimum potential energy method is applicable only
for elastic materials, but virtual work is applicable for any material behavior. These methods
are considered variational methods and can lead to solve nonstructural problems.
The main Idea of the minimum potential energy:
The total potential energy, p , of a structure is expressed in terms of displacements. In finite
element formulation, these are the nodal displacements such that

p p d1 , d 2 , d 3 ,....., d n

(2.32)
The minimization of p with respect to these displacements will yield the equilibrium equations.

Principle of minimum potential energy states:

Of all displacements that satisfy the given boundary conditions of a structure, those that satisfy
the equations of equilibrium are distinguishable by a stationary value of the potential energy. If
the stationary value is a minimum, the equilibrium state is stable.

To explain this principle, we introduce the following concepts:

1) Total potential energy is defined as the sum of the internal strain energy U and the potential
energy of the external forces , that is
p U
(2.33)
Internal strain energy, U, is the work done by internal forces (or stresses) through deformations
(strains) in the structure.
Potential energy of the external forces, , is the potential energy which is lost when work is done
by external forces (body forces, surface traction forces, and applied nodal forces), i.e.;
W
(2.34)
As an example, consider the linear spring shown:
The differential internal strain energy,
k
dU = internal force * change in displacement through which the force moves
or

F
x

dU = F dx and since F = k x
then

dU = k x dx
k

therefore, the total strain energy is

U

U 12 ( kx) x

1
2

Fx

k x dx

1
2

k x

1
(2.35)

or area under the force-deformation curve.

dx

Also, the potential energy of the external force can be written as:
W F x

(2.36)

Substituting equations (2.35) and (2.36) into equation (2.33), the total potential energy will be:

p 12 k x 2 F x

(2.37)

10

2) The stationary value of a function G(x) can be a maximum, a minimum, or a neutral point on
G(x). From differential calculus, any stationary value x of G(x) must satisfy
dG
G
0.
Maximum
dx
Neutral
Minimum

x
A similar process can be used with p but here

we use variational calculus; where the first variation of p could be minimized (avoiding
details):
p p d1 , d 2 , d 3 ,....., d n

If

Then the first variation of p has the general form:

p
d1

d1

p
d2

d2

p
dn

dn

(2.38)

The principle of minimum potential energy states that equilibrium exists when the di define a
structure state such that p 0 for arbitrary admissible variations d i from the
equilibrium state. An admissible variation is one in which the displacement function still
satisfies the boundary conditions and inter-element continuity.
To satisfy p 0 , all coefficients associated with the d i must be zero independently. Thus,

p
di

(i 1,2,, n)

or

(2.39)

where n equations must be solved for the n values of dIi that define the static equilibrium
state of the structure. The above equation shows that we can interpret the variation of p as
a compact notation equivalent to differentiation of p with respect to the unknown nodal
displacements for which p is expressed.
It can be shown that, for a linear-elastic spring subjected to an axial force, the minimum
potential energy also corresponds to the equilibrium position of the spring (see example 2.4).

Next we derive the spring element equations using the above principle:
Consider the linear spring element shown (repeated);

The total potential energy is

p

1
2

k ( d2 x d1 x ) 2 f1 x d1 x f2 x d2 x

(2.40)

11

where d2 x d1x deformation of the spring x.

Upon expanding, we get

p 12 k (d22x 2 d1x d2 x d12x ) f1x d1x f2 x d2 x
Minimizing p with respect to each nodal displacement, i.e.;
p
0
i 1, 2
(assuming that spring is in equilibrium); then,
di

1
2

k ( 2 d2 x 2 d1 x ) f1x 0

1
2

k ( 2 d2 x 2 d1x ) f2 x 0

1x

2x

Simplifying, we get
k ( d1 x d2 x ) f1 x
k ( d d ) f
1x

2x

2x

In matrix form, we write the element equations as

k k d1x f1x
k k
d2x f2x

(2.41)

therefore, the element stiffness matrix for a spring element is

k
k
k

k
k

(2.42)

which is identical to equation (2.11).

It can be shown that the total potential energy of an entire structure can be minimized with
respect to each nodal degree of freedom and this minimization will result in the same total
equations of the structure as those obtained by the direct stiffness method.
Note that the total potential energy for the structure is:
NE

p (pe )

(2.43)

e 1

(see example 2.5 Logan)

12