Lecture 2: The Direct Stiffness Method

2.1 Definition of the Stiffness Matrix. First look at structural problems.

The DSM is also known as the Displacement Method We will be treating displacements of key points (node points) as the state variables. State variables are a set of values that define the state of the entire element. If we want a more complex/complete model of the system we tend to have to add state variables.

The stiffness matrix relates forces acting at the nodes to displacements of the nodes.
Generalized force: either of a force or moment. Generalized displacement: either a translation or a rotation.

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

The state of the structural system is defined by a matrix of displacements (generalized displacements), D .
D = {d1x d1 y 1 d 2 x d Ny N }
T

Global, or assembled, displacement matrix

The external factors acting on the system are given by a force (generalized force) matrix, F .
F = { f1x f1 y m1z f2 x f Ny mNz }
T

Global, or assembled, force matrix

The two quantities are related by a stiffness matrix, K .

Global, or assembled, system of equations F = KD k11 k12 k k22 K = 21 k N1 kN 2 k1N k2 N k NN

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Throughout Chapters 2, 3, 4, and 5 we look at deriving the stiffness matrix for line elements. Relates forces at the element nodes to displacements of those nodes. We begin by focusing on line elements.
State of the element is entirely defined by a filament (a curved line) extending between two node points. The actual shape of the line is defined in terms of values that exist at the nodes. Springs (2.), truss/bars (3.), beams(4.), frames (5.), grids (5.).

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Consider a cable (a flexible bar element):
( s)
F

( s+ds)
ds(1 + ( s) )
(s+ds) b( s+ds ) p1

N b( s)

p1(s) n( s)

p2(s)
(h+ w)

p2(s+ds)

n( s+ds ) q( s+ds) , t ( s+ds)

F + dF

s
Centerline of the flexible rod/cable
N + dN

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

The forces and displacements of the nodes are vector quantities. As always, we have to communicate sense of vector entities with a reference frame. For the spring element of 2.: x y z : global ref frame. x y z : local/element ref frame.

d2 x

f1 x

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

2.2 Derivation of the Stiffness Matrix for a Spring Element.
First line element we consider is an ideal linear spring. 2.2 through 2.5 illustrate Logans 8 FEM steps as applied to springs 2.2 deals with the first 4 steps: the formation of the spring element equations.

Step 1: Set the element type.

L = unstretched length. k = spring stiffness (constant) T = tension (compression) The spring can only deform in the x direction.

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Step 2: Set the displacement function.

d1x d2 x

u = a1 + a2 x

We can reformulate the displacement function (the element) to be in terms of State Variables & Shape Functions.

x1

x2

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Step 3: Define the Governing Differential Equations.
For the DSM, Logan shows that we only need to consider the strain/displacement and stress/strain relationships. Remember that we are trying to relate displacements to forces. The way to do this is to;
consider how the assumed displacement field translates into strain, then translate this strain into stress through a material constant (E), Then relate stress to applied load.

For the spring element the only type of stretch we have is axial stretch.
Thus, we only have axial strains to consider, and we only have axial stretch to consider.

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Step 3: (contd)
T = k = (d 2x - d1x ) T = k d 2x - d1x

Step 4: Derive the Element Equations.

Here we have only to relate the nodal loads to the displacements.

f1x = T f2 x = T

f1x = T f 2 x = +T

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Step 4: (contd)
f1x k k d1x = k k d 2 x f2 x
The element equations.

k k k= k k
Element (or local) stiffness matrix.

Any problem dealing specifically with axial stretch of springs can be solved by assembling this spring element. The assembly process requires an understanding of the continuity/compatibility requirement across the element.
kinematic continuity ensures that we have realistic behaviour between elements.

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Step 5: Assemble the Element Equations to Produce the Global Equations & Apply BCs to Reduce the Global System.
D = d
e =1 N (e)

F =f
e =1

(e)

K = k (e)
e =1

F = KD

Step 6: Solve the assembled global system of equations. Step 7: Recovery (Calculate the spring tensions, T). Step 8: Interpret.
MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

2.3 Example of a Spring Assemblage.
The sigma notation used by Logan is deceiving. Assembly is NOT a direct superposition of matrices. It is a simultaneous enforcement of each set of element equations. Each set of element equations is in terms of a different pair of node points. Assembly is superposition with an added step that looks after the continuity/compatibility conditions.

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

2.3 (contd)
2
3 3x

2x

Global x coordinate Element 1 equations

f1x (1) k1 k1 d1x (1) (1) = (1) f 2 x k1 k1 d 2 x
y

Element 2 equations
f 2 x (2) k2 k2 d 2 x (2) = (2) (2) f 3 x k2 k2 d3 x
y

f1x (1)

1
d1x (1)

f 2 x (1)

f2 x(2)

3 f3x(2)
d3x(2)

d2 x (1)

d2 x(2)

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Continuity/compatibility equations need to be enforced at the interelement boundaries. Continuity conditions enforce a continuity of state across the elements. For spring elements state is defined by displacement in the horizontal direction, x . At node 2: u (1) = u (2)
2 2

(1) d 2 x (1) d1x (1) + u = d1x x L x= L (2) d3 x (2) d 2 x (2) (2) u2 = d 2 x + x L x=0 d 2 x (1) = d 2 x (2)
(1) 2

An entirely intuitive result that was guaranteed in our Step 2.

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

2.4 Assembly by superposition. Assembly of the element equations yields:
(1) 0 d1x f1x 0 k1 k1 0 d1x 0 0 k k 0 d + 0 k d = f (1) + f (2) k2 2 x 2 x 2 x 1 2 1 2x d 0 k k2 d3 x 0 f3 x (2) 0 0 3 x 0 2

At each node, the external forces applied must equal the sum of the forces exerted on the spring nodes.
F1x = f1x (1) F2 x = f 2 x (1) + f 2 x (2) F = f (2)
3x 3x

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Final global (assembled) system.
k1 0 d1x F1x k1 k k + k k d = F 2x 2 2 2x 1 1 k2 d3 x F3 x k2 0

2.5 Boundary Conditions. If we look at the global stiffness matrix that resulted in 2.3 det( K ) = 0
0 1 0 k1 k1 k k + k k 1 = 0 2 2 1 1 k2 k2 1 0 0
One of an infinite number of solutions when zero force is applied.

Columns of K are linearly dependent (do not span 3-dimensional space). Some combinations of applied forces can not be sustained.
MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

We must apply boundary conditions in order to solve. The system we analyzed is limited by our Step 1 through Step 4. The choices we made in the element derivation force us to set at least one of the node displacements.

3 3x

2x

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

Problem 2.12: Use a connectivity table to help form the global stiffness matrix in one step. Another way of bookkeeping the assembly process.
y
fix
i

1 2

f jx
d jx

dix

fix kii kij dix = k ji k jj d jx f jx kii = k jj = k1 kij = k ji = k1

Element # 1 2 3

Node i 1 3 2

Node j 3 4 3

MECH 420: Finite Element Applications

Lecture 2: The Direct Stiffness Method

assembling the system:
0 k1 0 d1x F1x k1 F 0 k3 k3 0 d 2 x 2x = F3 x k1 k3 k1 + k3 + k2 k2 d3 x F4 x 0 k2 0 k2 d 4 x
1. Symmetric 2. Positive Definite 3. Singular

Apply the boundary conditions (column manipulation and/or reduction of the matrix equation).
F3 x = 1000 kN d1x = d 2 x = d 4 x = 0

Solution & Recovery

MECH 420: Finite Element Applications