Lecture 20a. Intro to Matrix Analysis
Lecture 20a. Intro to Matrix Analysis
consider a 2-D structure consisting of four elements, linked at pinned joints with six nodes:
X
Y 4
consider one of the elements, say element # 4
p
LINEAR ELASTIC behaviour and equilibrium =>
4
F = K⋅∆ + fp + fε0 5 6
where F = forces at nodes (any direction =>
two at each node in X and Y) X
fp is the equivalent nodal force resulting in the
same REACTION to the distributed pressure
fε0 is the same for initial strain/stress in the
F = forces at each node
element due to fit, temperature, etc.
for an example of the eqivalent nodal force consider the following uniformly loaded beam:
p*L/2 p*L/2
a force of p*L/2 in the same direction as p will create the same rection force
hence
L L
R1 R2 fp = p⋅
2
1 notes_32_intro_matrix.mcd
henceforth we will assume that the nodal forces already account for the equivalent of distributed forces and that
initial stress/strain = 0 therefore: ...
F = K⋅∆ i.e. The nodal force is linearly proportional to the displacement of the nodes
nod_el := 3
for the fourth element this is expressed as:...
Fe1 1 , 1
Ke Ke
1, 2
Ke
1, 3
∆e1
Fe → Fe2 Ke → Ke2 , 1 Ke2 , 2 Ke2 , 3
∆e → ∆e2
Fe Ke3 , 1 Ke3 , 2 Ke3 , 3 ∆e
3 3
Ke = element stiffness matrix which for now we will assume can be determined by experiment or analysis
similarly a matrix can be found such that:...
σe = Se⋅∆e
Se = element stress matrix
remember that F and ∆ in this two dimensional example each have two components X & Y
FeX1 U
1 U is X displacement
FeY1 V1 V is Y displacement
FeX2 U2 and Ke is a 6 x 6 matrix of coefficients.
Fe := ∆e := we will express the relationship as above until later
FeY2 V2
FeX U
3 3
FeY V
3 3
this examaple is 2-D pinned and involves only X and Y forces and displacements
were this to include clamped joints, there would be a moment and resulting rotation θ for 3 components at each
node Fx, Fy, M and U, V, θ
if this were 3_D, there would be three forces and three moments at each node
later we will also see the concepts of "force" and "displacement" to be generalized and include imposed moments
and resulting rotation θ and termed "degrees of freedom"
2 notes_32_intro_matrix.mcd
above we have expressed the linear elastic behavior of the element in the structure coordinate system
and could do the same for each element. We might have to "pad" some matrices (add some 0 to get the same
number of rows and columns for operations below.
Fe1 1 , 1
Ke Ke
1, 2
Ke
1, 3 ∆1
Fe := Fe2 Ke := Ke2 , 1 Ke2 , 2 Ke2 , 3 ∆e := ∆ 2
∆
Fe Ke3 , 1 Ke3 , 2 Ke3 , 3 3
3
let's now operate in the structure coordinate system and develop some information about the system K (stiffness
matrix) in the relation:
F1 K1 , 1 K
1, 2
K
1, 3
K
1, 4
K
1, 5
K
1, 6
∆1
F2 K2 , 1 K K K K K
2, 6
2, 2 2, 3 2, 4 2, 5 ∆2
∆
F3 K3 , 1 K
3, 2
K
3, 3
K
3, 4
K
3, 5
K
3, 6
∆→
3
F→ K→ ∆4
F4 K4 , 1 K
4, 2
K
4, 3
K
4, 4
K
4, 5
K
4, 6
F K
K K K K K ∆5
5 5, 1 5, 2 5, 3 5, 4 5, 5 5, 6
∆
F K K K K K K 6
6 6, 1 6, 2 6, 3 6, 4 6, 5 6, 6
now to address the structure let's "pad" the element and express the components of nodal force and
displacement in strucuture coordinates: the element node 1 corresponds to structure node 4 etc. so we could
first say: ....
Fe4 Ke1 , 1 Ke1 , 2 Ke1 , 3 ∆4 Ke1 , 1⋅∆ 4 + Ke1 , 2⋅∆ 5 + Ke1 , 3⋅∆ 6
Fe → Fe5 Ke → Ke2 , 1 Ke2 , 2 Ke2 , 3 ∆ → ∆5
Ke⋅∆ → Ke2 , 1⋅∆ 4 + Ke2 , 2⋅∆ 5 + Ke2 , 3⋅∆ 6
∆
Fe6 Ke3 , 1 Ke3 , 2 Ke3 , 3 6
Ke3 , 1⋅∆ 4 + Ke3 , 2⋅∆ 5 + Ke3 , 3⋅∆ 6
or ... with no loss in accuracy padding the nodes not related to the fourth element ...
3 notes_32_intro_matrix.mcd
Fe1
∆1
0 0 0 0 0
0
Fe2
0 0 0 0 0 0 ∆2
Fe3 0 0 0 0 0 0 ∆
Fe → Ke → 0 3
0 0 Ke Ke Ke ∆→
Fe4 1, 1 1, 2 1, 3 ∆4
Fe 0 0 0 Ke Ke Ke
5 2, 1 2, 2 2, 3 ∆5
Fe 0 0 0 Ke Ke Ke ∆
6 3, 1 3, 2 3, 3 6
note that in this expression, we have expanded ("padded") the F and ∆ vectors to include the unrelated nodes with
no loss in accuracy as ...
0
0
0 which compares with the values above
Ke⋅∆ → Ke ⋅∆ 4 + Ke ⋅∆ 5 + Ke ⋅∆ 6
1, 1 1, 2 1, 3
Ke ⋅∆ 4 + Ke ⋅∆ 5 + Ke ⋅∆ 6
2, 1 2, 2 2, 3
Ke ⋅∆ 4 + Ke ⋅∆ 5 + Ke ⋅∆ 6
3, 1 3, 2 3, 3
we're now going to change nomenclature to allow including an additional element say element #3
first so we can keep track we'll rename the previous sttiffness matrix Ke4
Fe41
0 0 0 0 0 0 0
Fe42
0 0 0 0 0 0 0
0 0 0 0 0 0
Fe43 0
Fe4 → Ke4 → 0
Ke4⋅∆ → Ke4 ⋅∆ 4 + Ke4 ⋅∆ 5 + Ke4 ⋅∆ 6
0 0 Ke4 Ke4 Ke4
1, 1 1, 2 1, 3
Fe44 1, 1 1, 2 1, 3
0
Fe4 0 0 Ke4
2, 1
Ke4
2, 2
Ke4
2, 3 Ke4 ⋅∆ 4 + Ke4 ⋅∆ 5 + Ke4 ⋅∆ 6
5 2, 1 2, 2 2, 3
0 0 0 Ke4 Ke4 Ke4
Fe4 3, 1 3, 2 3, 3 Ke4 ⋅∆ 4 + Ke4 ⋅∆ 5 + Ke4 ⋅∆ 6
6 3, 1 3, 2 3, 3
now suppose another element (#3) has nodes 2 and 5 so node 5 is a common node
4 notes_32_intro_matrix.mcd
Fe31 N.B. only
0 0 0 0 0 0 0 components for
Fe32 0 Ke3 0 0 Ke3 0 Ke3 ⋅∆ + Ke3 ⋅∆ F2 and F5 and ..
1, 1 1, 2 1, 1 2 1, 2 5
only two nodes for
Fe33 0 0 0 0 0 0 0 this element
Fe3 → Ke3 → Ke3⋅∆ →
Fe34 0 0 0 0 0 0
0
Fe3 0 Ke32 , 1 0 0 Ke3
2, 2
0 Ke32 , 1⋅∆ 2 + Ke32 , 2⋅∆ 5
5
Fe3 0 0 0 0 0 0 0
6
now we can use equilibrium for forces at the nodes as follows from these two elements: ...
obviously complete equilibrium requires all nodes ...
F := Fe3 + Fe4 this states that the external force at each node is in equilibrium with the
components of that force for each element
∆ is common (compatibility)
F := K⋅∆ F → Ke3
5 2, 1
⋅∆ 2 + Ke4
2, 1
⋅∆ 4 + Ke3 ( 2, 2
+ Ke4
2, 2 )⋅∆ 5 + Ke42 , 3⋅∆ 6
5 notes_32_intro_matrix.mcd
same as above... CONCLUSION
K - the structure stiffness matrix is determined by the sum of element stiffness matrices in structure coordinates
(expanded to include all nodes)
i.e. ...
6 notes_32_intro_matrix.mcd