F.E.M.

- Chapter 3
The direct stiffness method (part II):
pin jointed structures

J.-P. Ponthot
University of Liège - Belgium

February 11, 2021

 Goal of the method

F.E.M. - Chapter 3 - The direct stiffness method (part II) 2 / 62

 Goal of the method

Goal : be able to solve structural problems like ’Truss-structures’

Note : Joint ⇒ Idealized as a ’pin joint’ or hinge

⇒ No moment can be transmitted

F.E.M. - Chapter 3 - The direct stiffness method (part II) 3 / 62

A simple 1D truss/ bar element :
Spring Analogy

F.E.M. - Chapter 3 - The direct stiffness method (part II) 4 / 62

 Spring Analogy (1)

● A truss/ bar element transmits a force between its two end-points.

● The bar can only carry load in the axial direction
● The bar is assumed prismatic (A=cst)
EA EA
Suggests the analogy with the spring k= →N = L (u2 − u1 )
L
u1 u2 u1 u2
EA EA
F1 F2 N N
x L x L

k
● u1 , u2 → Nodal displacements
● F1 , F2 → Nodal loads
● N → Internal force

F.E.M. - Chapter 3 - The direct stiffness method (part II) 5 / 62

 Spring Analogy (2)

u1 u2 u1 u2
EA EA
F1 F2 N N
x L x L

Equilibrium equations 
−N
 
1 −1

u1
 
F1

EA
for the element = L =
N −1 1 u2 F2

K q = Fint

Equilibrium Fint = Fext = g ⇒ Kq=g

F.E.M. - Chapter 3 - The direct stiffness method (part II) 6 / 62

 Alternative derivation (1) : Mechanics of materials

u1 u2 u1 u2
EA EA
F1 F2 N N
x L x L

MoM → we have σ =
E ε ⇒ N = Aσ = E Aε
∆L u2 −u1
ε =
L = L
EA
Internal force N= (u2 − u1 )
L
Nodal equilibrium
EA
● Node 1 : N = −F1 ⇒ F1 = L (u1 − u2 )
EA
● Node 2 : N = F2 ⇒ F2 = L (u2 − u1 )
In matrix form, equilibrium equations for the element
    
EA 1 −1 u1 F1
=
L −1 1 u2 F2

F.E.M. - Chapter 3 - The direct stiffness method (part II) 7 / 62

 Alternative derivation (2) : Elasticity

● Prismatic (means A = cste) bar model (idealization)

1D problem : σij → σ
1D problem : εij → ε
1D problem : xi → x − → σ →
−
n = (−1 0 0) n = (1 0 0)

● equilibrium

∂σij 1D dσ
+ ρb =0 → = 0 → σ = cst.
∂xj |{z}i dx
=0 if no body force
 
σ 0 0
⇒ σ =  0 0 0 with σ = cst and N = σA
0 0 0
(
F1 = t1 A = −σ A = −N
boundary conditions: ti = σij nj = t̄i →
F2 = t2 A = σ A = N

F.E.M. - Chapter 3 - The direct stiffness method (part II) 8 / 62

 Remark

∂σij
● Equilibrium ∂xj + ρ bi = 0 → σ = cst.
 
1 ∂ui ∂uj du
● Compatibility εij = + → ε =
2 ∂xj ∂xi dx
ui = ūi → u(0) = u1 u(L) = u2

● Constitutive relation σij = Hijkl εkl → σ = E ε → N = E A ε

All equations from elasticity are satisfied.

→ exact solution (of the idealized member)

F.E.M. - Chapter 3 - The direct stiffness method (part II) 9 / 62

 Energy considerations

● Strain energy
simple
3D Z tension Z Z L
1 1 EA 2
U = σij εij dV = σ ε dV = ε dx
V 2 V 2 |{z} 2 0
Eε

u2 −u1 EA (u2 −u1 )2

● ε= L ⇒U = 2 L

   
1 T EA 1 −1 u1
or U = q Kq with K= and q=
2 L −1 1 u2

● Since
    
1 EA 1 −1 u1 1 EA u1 − u2 EA (u2 − u1 )2
U= (u1 u2 ) = (u1 u2 ) =
2 L −1 1 u2 2 L u2 − u1 2 L

F.E.M. - Chapter 3 - The direct stiffness method (part II) 10 / 62

 Truss structure

F.E.M. - Chapter 3 - The direct stiffness method (part II) 11 / 62

 Truss structure

Consider the spatial truss structure. (Applied loads and support saved for later)
v2
2 u2

1 4 ● 5 nodes : 1 to 5
3 ● 8 members : 1 to 8
v1 v5
u5 v3 → 10 (15 in 3D) DOF’s : uI and vI
u1 2 5
1 3 u3 for each node i.e. I = 1 to 5.
5
7 6
y 8 qs T = [u1 v1 u2 v2 ... u5 v5 ]
Nodal structural displ. vector
u4 4
x v4
Keypoints: Displacements are continuous:
● along each bar
● AT THE NODES!

F.E.M. - Chapter 3 - The direct stiffness method (part II) 12 / 62

 Problem formulation

Fint
s = Ks qs = gs
|{z} |{z} |{z}
contribution from every element stiffness matrix applied loads

with
qs T = [ u1 v1 u2 v2 ... u5 v5 ]
DOF number 1 2 3 4 ... 9 10

The uxi and uyi are unique at the nodes ⇒ Joint displacement of all members
meeting at the joint will be unique.

⇒ Fundamental property of the structure that we will try to satisfy

F.E.M. - Chapter 3 - The direct stiffness method (part II) 13 / 62

 Disconnection and Localization (1)

1. Disconnection (conceptual, not actually programmed)

v2 2
u2

1
3 4
v1
u1
v5
2 5
1 5u 3
5

v3
y 7 8 u3
6

x 4 u4
v4

F.E.M. - Chapter 3 - The direct stiffness method (part II) 14 / 62

 Disconnection and Localization (2)

2. Localization

● Each element has 4 DOF’s

(
Index vector
● Topology → Define for each element
Freedom pointer

Ex : Element 7 from node 4 to node 1 (arbitrary)

Index(7) = [7, 8, 1, 2] Global DOF (structure numbering)

→ Component = 1, 2, 3, 4 Local DOF (element numbering)

F.E.M. - Chapter 3 - The direct stiffness method (part II) 15 / 62

 Disconnection and Localization (3)

And we have

qe (I) = qs (Index(e, I)) e = element number

| {z } | {z }
Element level Structural level

Some people (mainly mathematicians) also use the ’localization matrix’ such that

qe = Le qs (qie = Leij qjs )

where Le has only binary components (0 or 1) and has a dimension equal to

(
4 × 10 here
ElementDOF × StructureDOF →
4 × 107 if we have a 5 106 node 2D structure

F.E.M. - Chapter 3 - The direct stiffness method (part II) 16 / 62

 Disconnection and Localization (4)

Example of a localization matrix : element 8 from node 4 to node 3

 
0 0 0 0 0 0 1 0 0 0
0 0 0 0 0 0 0 1 0 0
L = 
8 0 0 0 0 1 0 0 0 0 0
0 0 0 0 0 1 0 0 0 0

Note : INDEX(8) = (7, 8, 5, 6) ⇒ Compact storage !

F.E.M. - Chapter 3 - The direct stiffness method (part II) 17 / 62

2D element formulation

F.E.M. - Chapter 3 - The direct stiffness method (part II) 18 / 62

 2D element formulation

x
L 2 (x2,y2)
y
y EA=cst
j
1
(x1,y1)
x

We know equilibrium equations in local (x̄, ȳ) axes which are element dependent
    
EA 1 −1 ū1 F̄1
=
L −1 1 ū2 F̄2

1 DOF/node (local axes)

F.E.M. - Chapter 3 - The direct stiffness method (part II) 19 / 62

 Element formulation (4 DOF’s per element)

x
L 2 (x2,y2)
y
y EA=cst
j
1
(x1,y1)
x

● We want global / structural axes (same ∀ element)

2 DOF per node ⇒ K4×4 e =?
● We know in local axes  
e EA 1 −1
1 DOF per node ⇒ K̄2×2 =
L −1 1

⇒ Transformation of displacement components

F.E.M. - Chapter 3 - The direct stiffness method (part II) 20 / 62

 Element formulation (4 DOF’s per element) (1)

Displacement components transform as (quite general, not only at nodes, valid for
any point along the bar. N.B. Here we will use ux and uy for displacement
components instead of u and v)
x
L 2 (x2,y2)
y
ux = ūx cos ϕ − ūy sin ϕ y EA=cst

uy = ūx sin ϕ + ūy cos ϕ 1

j

(x1,y1)
x

N.B. ūy = 0 for a bar but we try to be general

    
ux C −S ūx
or =
uy global axes S C ūy local axes

in which C = cos ϕ and S = sin ϕ

F.E.M. - Chapter 3 - The direct stiffness method (part II) 21 / 62

 Element formulation (4 DOF’s per element) (2)

     
ūx C S ux
⇒ =
ūy −S C uy
| {z } | {z }
local axes = q̄ global axes = q
This is valid for any point so in particular, this is valid for nodes, so that for any
node, we can write:
 
C S
q̄ = T q with T2×2 =
−S C

It is easily verified that T is orthogonal i.e. T −1 = T T

Thus q = T T q̄

F.E.M. - Chapter 3 - The direct stiffness method (part II) 22 / 62

 Applying this to a 2 node structure (1)

x
L 2 (x2,y2)
y
y EA=cst
j
1
(x1,y1)
x
     
ū1x C S 0 0 u1x
ū1y  −S C 0 0  u1y 
 =    or q̄ = T q with T 4×4
ū2x   0 0 C S u2x 
can be generalized to any
ū2y 0 0 −S C u2y number of nodes
| {z } | {z }
local axes global axes

T
N.B. T −1
4×4 = T 4×4

F.E.M. - Chapter 3 - The direct stiffness method (part II) 23 / 62

 Applying this to a 2 node structure (2)

And since ū1y = ū2y = 0, one gets

 
    u1x
ū1x C S 0 0  u 1y
 or q̄ = T q with T 2×4
=  
ū2x 0 0 C S u2x  particularized to the bar
u2y

In such a case, it is also verified that one can write

q = T T q̄

F.E.M. - Chapter 3 - The direct stiffness method (part II) 24 / 62

 Stiffness matrix

F.E.M. - Chapter 3 - The direct stiffness method (part II) 25 / 62

 Stiffness matrix (1)

Recall
1 T
U = q̄ K̄ q̄→ local axes
2
with
   
EA 1 −1 u¯1
K̄ = q̄ =
L −1 1 2×2 u¯2 2×1

Strain energy U is a scalar ⇒ Independent of the reference frame

⇒ In global coordinates :
(
1 T K →4×4 Elementary level
U = q K q→ global axes
2 q →4×1

F.E.M. - Chapter 3 - The direct stiffness method (part II) 26 / 62

 Stiffness matrix (2)

● Ā → Local axes
● A → Global axes
Since q̄ = T q, where T 2×4 is the transformation matrix, one gets

q̄T K̄ q̄ = 2U = qT K q = qT T T
| {zK̄T} q
K

e e
K e4×4 = T T4×2 K̄ 2×2 T 2×4 K̄ → Local axes
K e → Global axes

F.E.M. - Chapter 3 - The direct stiffness method (part II) 27 / 62

 Global stiffness matrix of plane bar

 
C2 SC −C 2 −SC (
EA  SC S 2 −SC −S 2  C = cos ϕ
e
K =   with
L  −C 2 −SC C 2 SC  S = sin ϕ
−SC −S 2 SC S2
 
x21 x21 x21 y21 −x21 x21 −x21 y21
EA 
 x21 y21 y21 y21 −x21 y21 −y21 y21 
= 
L3 −x21 x21 −x21 y21 x21 x21 x21 y21 
−x21 y21 −y21 y21 x21 y21 y21 y21

Where xij = xi − xj and yij = yi − yj (form actually implemented)

∆x = x2 − x1 = L cos ϕ
∆y = y2 − y1 = L sin ϕ

F.E.M. - Chapter 3 - The direct stiffness method (part II) 28 / 62

 Test of plane bar element with numerical inputs (2D)

2D Example : Numerical element stiffness matrix (node 1 = (0,0) ; node 2 =

(30,40) ; E = 1000 ; A = 5) :
 
36 48 −36 −48
 48 64 −48 −64 
K 2D4×4 = −36 −48 36

48 
−48 −64 48 64

● Symmetric
● Eigenvalues of K e = {200, 0, 0, 0}
Note : Rigid body mode qj = cste = 1
X
⇒ Kij qj = 0
|{z} ⇒ Kij = 0 ∀i
internal forces generated by RBM j

F.E.M. - Chapter 3 - The direct stiffness method (part II) 29 / 62

 Test of space bar element with numerical inputs (3D)

3D problem ⇒ 3 DOF’s per node

Numerical element stiffness matrix (node 1 = (0,0,0) ; node 2 = (2,3,6) ; E =73 =
343 ; A = 10 ; L=7)
 
40 60 120 −40 −60 −120
 60 90 180 −60 −90 −180
 
 120 180 360 −120 −180 −360
K 3D6×6 =
 −40 −60 −120

 40 60 120 

 −60 −90 −180 60 90 180 
−120 −180 −360 120 180 360

● Symmetric
● Eigenvalues
P of K e = {980, 0, 0, 0, 0, 0}
●
j Kij = 0 ∀i

F.E.M. - Chapter 3 - The direct stiffness method (part II) 30 / 62

Work done by applied loads

F.E.M. - Chapter 3 - The direct stiffness method (part II) 31 / 62

 Work done by applied loads

Since q̄ = T q using T4×4

ḡ = T g

One has
local axes
z}|{
P = ḡ1 ū1 + ḡ2 ū2 = ḡT q̄
= gT T T
| {zT} q = gT q
|{z}
identity global axes

Scalar ⇒ Invariant

F.E.M. - Chapter 3 - The direct stiffness method (part II) 32 / 62

Rules that govern assembly

F.E.M. - Chapter 3 - The direct stiffness method (part II) 33 / 62

 Rules that govern assembly (1)

X ∗
z}|{
Fint = Ks q s = gs = Fext
| {z } |{z}
Elements Internal forces Applied load
| {z }
◦

where ∗ : Compatibility : the joint displacements of all members meeting at a joint

will be the same

where ◦ : Equilibrium : the sum of forces exerted by all members that meet at a
joint must balance the external force applied to that joint

F.E.M. - Chapter 3 - The direct stiffness method (part II) 34 / 62

 Rules that govern assembly (2)

At the element level

Fint
e = Ke q e

At the node level

X
Fint = f (1) + f (2) + f (3) (contributions from all elements joining the node)

should be equal to applied load Fext or g.

F.E.M. - Chapter 3 - The direct stiffness method (part II) 35 / 62

Assembly / globalization

F.E.M. - Chapter 3 - The direct stiffness method (part II) 36 / 62

 Assembly / globalization

Strain energy
1 T
● For one element: Ue = qe Ke qe
2
Nelem Nelem
X 1
X
T 1 T
● For the whole structure: Us = Ue = q e Ke q e = q s Ks q s
2 2
e=1 e=1

And since qe = Le qs
One has
1
Ue = qs T LTe Ke Le qs
2
Nelem Nelem Nelem
X X 1 T T 1 T X
⇒ Us = Ue = q s L e Ke L e q s = q s LTe Ke Le qs
2 2
e=1 e=1
|e=1 {z }
Ks
structural stiffness matrix

F.E.M. - Chapter 3 - The direct stiffness method (part II) 37 / 62

 Assembly / globalization (1)

Loop over all elements

● Symbolically
Nelem
X
Ks = Le T Ke Le
|{z}
e=1 elementary stiffness/global axes
Nelem
X
= L e T Te T K̄e Te L e
|{z}
e=1 Global / local axes

N.B. Le = localization matrix : qe = Le qs

F.E.M. - Chapter 3 - The direct stiffness method (part II) 38 / 62

 Assembly / globalization (2)

● Practically

DO N e = 1, Nelem
DO i = 1 , NDOF
DO j = 1 , NDOF
KS (INDEX(N e, i), INDEX(N e, j)) = KS (INDEX(N e, i), INDEX(N e, j))+KN e (i, j)

ENDDO
ENDDO
ENDDO

NDOF = number of DOF per element

Nelem = number of elements in the structure

F.E.M. - Chapter 3 - The direct stiffness method (part II) 39 / 62

 Assembly / globalization (3)

⇒ Fint = Ks qs = gs

Note :
● K is symmetric
● K is sparsely populated (containing relatively few non zero coefficients) in
structures containing a large number of elements. This is because not more
than a few elements are connected to any one node.
● det K = 0 : RBM (Rigid Body Mode)

F.E.M. - Chapter 3 - The direct stiffness method (part II) 40 / 62

 Assembly / globalization (4)

Stiffness matrix is sparse (for this matrix we assume 1 DOF/node)

1 2 3 4 5 6 7 8 9
1
3 2
3
1 2 4 5
4
5
6
6 7 8 9 7
8
9

N.B. Boundary conditions are not taken into account of this matrix

F.E.M. - Chapter 3 - The direct stiffness method (part II) 41 / 62

 Assembly / globalization (5)

Typical stiffness matrix sparsity pattern (2 DOF/node)

y NF = 50
1 6 11 16 21

(1) (5) (9) (13)

2 7 12 17 22

(2) (6) (10) (14)

3 8 13 18 23 K=
x
(3) (7) (11) (15)
4 9 14 19 24

(4) (8) (12) (16)

5 10 15 20 25

16 elements
25 nodes
50 DOF’s

F.E.M. - Chapter 3 - The direct stiffness method (part II) 42 / 62

Stress/Force computation

F.E.M. - Chapter 3 - The direct stiffness method (part II) 43 / 62

 Stress computation

● From Ks qs = gs get qs = Ks −1 gs

● For each element, the internal force is given by

Global axes : qe = Le qs (qe (I) = qs (INDEX(I))
Local axes : q̄e = Te qe = Te Le qs
Internal force (local axes) :

Ee Ae e
F int
= K̄e q̄e = K̄e Te Le qs ⇒ Ne = F̄int
1 = −F̄int
2 = (ūx2 − ūex1 )
Le
Stresses (local axes) :

Ne (ūex2 − ūex1 )
σe = = Ee
Ae Le

F.E.M. - Chapter 3 - The direct stiffness method (part II) 44 / 62

 Assembly : examples

F.E.M. - Chapter 3 - The direct stiffness method (part II) 45 / 62

 Simplified assembler example: 2-bar problem

1(1,2) (1) 2(3,4)

E = 3000 and y
A = 2 for all bars
1 (1) 2 (2)
3(5,6)
3 (2) P assembly
x
3 1(1,2)
4 1 2(3,4)

P
• node number (bold)
3(5,6)
• element number (parentheses)
• 3 nodes ⇒ 6 DOF’s Global DOF numbers
for the structure written in parenthesis
after node number

F.E.M. - Chapter 3 - The direct stiffness method (part II) 46 / 62

 2-bar problem (cont’d)

Start by cleaning the structural stiffness matrix K

 
0 0 0 0 0 0 1
 0 0 0 0 0 0  2 Global DOF
 
 0 0 0 0 0 0  3 numbers (aka
K= 
 0
 0 0 0 0 0   4 global equation
 0 0 0 0 0 0  5 numbers)
0 0 0 0 0 0 6

and merge
Form stiffness of bar (1)  
1500 0 −1500 0 0 0 1
 
1500 0 −1500 0 1 Index   0 0 0 0 0 0 
 2
 0 0 0 0  2  −1500 0 1500 0 0 0  3
K (1) =  K =  
 −1500 0 1500 0  3 
 0 0 0 0 0 0 
 4
0 0 0 0 4  0 0 0 0 0 0 
0 0 0 0 0 0

F.E.M. - Chapter 3 - The direct stiffness method (part II) 47 / 62

 2-bar problem (cont’d)

and merge (K)

Form stiffness of bar (2) K (2)  
1500 0 −1500 0 0 0
  0 0 0 0 0 0 
768 576 −768 −576 3 
 
576 432 −576 −432   −1500 0 2268 576 −768 −576  3
4
  

0 0 576 4
432 −576 −432 

 −768 −576 768 576  5 
−576 −432 576 432 6
 0 0 −768 −576 768 576  5
0 0 −576 −432 576 432 6

Rem : For bar (2) (and other bars in general), choosing the Index vector to be
[3 4 5 6] (node 2 to node 3) or [5 6 3 4] (node 3 to node 2) will lead to the same
results due to the structure of K.

F.E.M. - Chapter 3 - The direct stiffness method (part II) 48 / 62

 2-bar problem (cont’d)

The structural stiffness matrix is thus :

 
1500 0 −1500 0 0 0

 0 0 0 0 0 0 

 −1500 0 2268 576 −768 −576 
K= 

 0 0 576 432 −576 −432 

 0 0 −768 −576 768 576 
0 0 −576 −432 576 432

● Eigenvalue check shows 4 zeros : Eigenvalues are [0 0 0 0 1.59 103 3.81 103 ]
● It is a little bit surprising to have 4 zeros ! Beyond the 3 rigid body modes, there is also
a mechanism ≡ deformation mode with zero associated strain energy and that is not a
rigid body mode !

u
Mechanism

F.E.M. - Chapter 3 - The direct stiffness method (part II) 49 / 62

 2-bar problem (cont’d)

Consider the following loading and boundary conditions :

1(1,2)
2(3,4)

3(5,6)

Kq=g→
1 2 3 4 5 6
    
1500 0 −1500 0 0 0 u1 R1

 0 0 0 0 0 0 
 u2   R2 
   

 −1500 0 2268  u3  =  0 
576 −768 −576     

 0 0 576  u4  −P 
432 −576 −432     
 0 0 −768 −576 768 576  u5   R5 
0 0 −576 −432 576 432 u6 R6

→ Displacement imposed to zero (support)

F.E.M. - Chapter 3 - The direct stiffness method (part II) 50 / 62

 2-bar problem (cont’d)

After taking into account boundary conditions, we have Kq = g :

    
2268 576 u3 0
=
576 432 u4 −P

● Symmetric.
● det K 6= 0 ! ⇒ q = K −1 g.
      
u3 0.007 −0.009 0 0.009 P
= =
u4 −0.009 0.0035 −P −0.0035 P

→ u3 and u4 are known as a function of the load P !

F.E.M. - Chapter 3 - The direct stiffness method (part II) 51 / 62

More complex example

F.E.M. - Chapter 3 - The direct stiffness method (part II) 52 / 62

 More complex example: plane truss structure

1 (1,2) (1) 2(3,4) (2) 3(5,6)

E = 3000 and y
A = 2 for all bars (3) (4) (5)
1 (1) 2 (2) 3
4(7,8)
(4)
3 (3) (5) assembly
x
4 2(3,4)
4 4 1 (1,2) 3(5,6)

• node number (bold)

4(7,8)
• element number (parentheses)
• 4 nodes ⇒ 8 DOF’s Global DOF numbers
for the structure written in parenthesis
after node number

F.E.M. - Chapter 3 - The direct stiffness method (part II) 53 / 62

 Plane truss assembly process

Start by clearing the master

Start by cleaning the stiffness
structuralarray K matrix K
stiffness
0 0 0 0 0 0 0 0
 
1
0 0 0 0 0 0 0 0 2
 
0
 0 0 0 0 0 0 0 3 Global DOF
0
K= 
0 0 0 0 0 0 0 4 numbers (aka
0 0 0 0 0 0 0 0

0
 5 global equation
 0 0 0 0 0 0 0 6 numbers)
0 0 0 0 0 0 0 0 7
0 0 0 0 0 0 0 0 8

Form stiffness of bar (1) and merge

1500 0 −1500 0 0 0 0 0
 
1

1500 0 −1500 0

1 INDEX  0 0 0 0 0 0 0 0 2
or 0 1500 0 0 0 0 0 
 
 0 0 0 0 2  −1500 3

−1500 0 1500 0

3 Element  0

0 0 0 0 0 0 0

4
Freedom  0 0 0 0 0 0 0 0
 
0 0 0 0 4
Table  0 0 0 0 0 0 0 0
 

(EFT) 0 0 0 0 0 0 0 0
 
0 0 0 0 0 0 0 0

F.E.M. - Chapter 3 - The direct stiffness method (part II) 54 / 62

 Plane truss assembly process (cont’d)

Form stiffness of bar (2) and merge

1500 0 −1500 0 0 0 0 0
 
 1500 0 −1500 0  3  0 0 0 0 0 0 0 0
 
 0 0 0 0 4  −1500
 0 3000 0 −1500 0 0 03
 0 0 0 0 0 0 0 0
0 1500 0 4
 
−1500 5 
 0 0 −1500 0 1500 0 0 0
0 0 0 0 6  5
 0 0 0 0 0 0 0 0
 6
 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0
Form stiffness of bar (3) and merge
2268 −576 −1500 0 0 0 −768 576
 
1
 −576 432 0 0 0 0 576 −432  2
768 −576 −768 576
 
1
0 3000 0 −1500 0 0 0 
 
 −1500
 −576 432 576 −432  2
 0 0 0 0 0 0 0 0 
 
−768 576 768 −576
 
7
 0 0 −1500 0 1500 0 0 0 
 
576 −432 −576 432 8
 0 0 0 0 0 0 0 0 
 
576 0 0 0 0 768
 
−768 −576 7
576 −432 0 0 0 0 −576 432 8

F.E.M. - Chapter 3 - The direct stiffness method (part II) 55 / 62

 Plane truss assembly process (cont’d)

Form stiffness of bar (4) and merge

2268 0 0 0 −768 576
 
−576 −1500

0 0 0 0

3  −576 432 0 0 0 0 576 −432 
0 3000 0 −1500 0 0 0 
 
3
 0 2000 0 −2000  4  −1500
 0 0 0 2000 0 0 0
 

0 0 0 0

7 −2000  4
 0 0 −1500 0 1500 0 0 0 
 
0 −2000 0 2000 8
 0 0 0 0 0 0 0 0 
 
576 0 0 0 0 768
 
−768 −576 7
576 −432 0 −2000 0 0 −576 2432 8

Form stiffness of bar (5) and merge

2268 −576 −1500 0 0 0 576
 
−768
 −576 432 0 0 0 0 576 −432 
768 576 −768 −576
 
5
0 3000 0 −1500 0 0 0 
 
−1500
 576 432 −576 −432  6
 0

0 0 2000 0 0 0 −2000

−768 −576 768 576
 
7
 0 0 −1500 0 2268 576
 
−768 −576  5
−576 −432 576 432 8
 0

0 0 0 576 432 −576

−432  6
576 0 0 1536 0
 
−768 −768 −576 7
576 −432 0 −2000 −576 −432 0 2864 8

F.E.M. - Chapter 3 - The direct stiffness method (part II) 56 / 62

 Plane truss assembly process (cont’d)

When all elements have been processed

 
2268 −576 −1500 0 0 0 −768 576
 −576 432 0 0 0 0 576 −432 
 
 −1500 0 3000 0 −1500 0 0 0 
 
 0 0 0 2000 0 0 0 −2000 
K= 

 0 0 −1500 0 2268 576 −768 −576 

 0 0 0 0 576 432 −576 −432 
 −768 576 0 0 −768 −576 1536 0 
576 −432 0 −2000 −576 −432 0 2864

is the structural stiffness matrix.

Eigenvalue check shows 3 zeros.

F.E.M. - Chapter 3 - The direct stiffness method (part II) 57 / 62

 Application

Consider the following loading and boundary conditions :

F
2 (3,4)
1(1,2) 3(5,6)

4(7,8)

P
N.B. Statically indeterminate problem !

F.E.M. - Chapter 3 - The direct stiffness method (part II) 58 / 62

 Application (2)

Kq=g→

    
2268 −576 −1500 0 0 0 −768 576 u1 R1

 −576 432 0 0 0 0 576 −432  u 2   R2 
   

 −1500 0 3000 0 −1500 0 0 0 
 u 3   R3 
   

 0 0 0 2000 0 0 0 −2000 
 u 4  =  F 
   

 0 0 −1500 0 2268 576 −768 −576  u 5   R5 
   

 0 0 0 0 576 432 −576 −432  u 6   R6 
   
 −768 576 0 0 −768 −576 1536 0   u7   R7 
576 −432 0 −2000 −576 −432 0 2864 u8 −P

→ Displacement imposed to zero (support)

→ Displacement equal to zero (by symmetry)

F.E.M. - Chapter 3 - The direct stiffness method (part II) 59 / 62

 Application (3)

After taking into account Boundary Conditions and symmetries :

Kq=g
         
2000 −2000 u4 F u4 0.0017 0.0012 F
= ⇒ =
−2000 2864 u8 −P u8 0.0012 0.0012 −P

● SYMMETRIC
● det K 6= 0 !

F.E.M. - Chapter 3 - The direct stiffness method (part II) 60 / 62

 Summary

F.E.M. - Chapter 3 - The direct stiffness method (part II) 61 / 62

 Software organisation

Input nodes - elements - properties

e
compute K
loop
over Graphics
elements Assemble Ke in KS preprocessing
(model and data handling)
Apply Bc’s q,g

Solve KSqS=gS

Compute for each element

loop
More complex F.E. codes
Fint=Keqe
over
elements quite similar, but K e , g, · · ·
stresses are different

File
listing
Graphics
(postprecessing: results)

F.E.M. - Chapter 3 - The direct stiffness method (part II) 62 / 62