You are on page 1of 5

APPENDIX A

Matrices
Introduction

FEA solves structural problems by defining its behavior in terms of differential


equations. These equations are converted into a set of linear algebraic equations,
which are represented in the form of matrix equations.
Matrix equations are easy to solve with computers, and therefore its important to
understand the matrix algebra.
We must always remember that when we solve matrix equations, we are, in fact,
solving the simultaneous algebraic equations (and the differential equations) of the
structure.
Since all the FEA calculations are based on matrix algebra, its important to
understand the matrix operations: addition, subtraction, multiplication, inversion,
transpose, etc.

Relationship between Algebraic and Matrix equations


a11x11 + a12x2 + a13x3 = b1
a21x11 + a22x2 + a23x3 = b2
a31x11 + a32x2 + a33x3 = b3
a11 a12 a13
a21 a22 a23
a31 a32 a33

x1
b1
x2 = b2
x3
b3

(A)

or [A] {x} = {b}

(B)

Equations (A) & (B) represents the same system of equations.

Transpose Matrix
Let [A] =

3 8
5 2

then

6 3

[A]T = 3 5 6
8 2 3

The rows & column position of elements is interchanged.

Also, [AB]T = [B]T [A]T

Orthogonal Matrix
Orthogonal matrix is a square matrix. For an orthogonal matrix, the inverse of the matrix
is equal to its transpose, thus,
[A]-1 = [A]T
Also [A][A]-1 = [I]
[AT][A] = [I]
[A][AT] = [I] Thus,
a11 a12 a13
a21 a22 a23
a31 a32 a33

a11 a21 a31


a12 a22 a32
a13 a23 a33

= [I]

Matrix partitioning
Matrix partitioning is the sub division of a matrix into several smaller matrices, called
submatrices.
In FEA, a matrix is usually partitioned into two or four submatrices.
A matrix is partitioned so that the number of columns to the left of the vertical
partition in the coefficient matrix equals the number of rows above the horizontal
partition.
The submatrices can be manipulated with matrix operations in the same manner as
the original matrix.
EX:
a11 a12 a13
a21 a22 a23
a31 a32 a33
A11 A12
A21 A22

x1
x2
x3
X1
X2

b1
b2
b3

B1
B2

[A11] {X1}+ [A12] {X2} = [B1]


[A21] {X1} + [A22] {X2}= [B2]

The advantage of partitioning a matrix is that the resulting sub matrices can be
manipulated with matrix operations in the same way as the original.
The ability to deal with a system of equations, either as one matrix equation or two
matrix equations provides a convenient means for solving an equation system when
some of the unknowns are in the force vector [B] and the remainder are in the
displacement vector X.
This is frequently the case when applying the FEA method to structures, where the
number of algebraic equations is very large and contains unknown forces.

Inverse Matrix
[A] [A-1] = [I]
[A-1] is the inverse of matrix [A].
The equation [A][X] = [B] is solved by pre-multiplying the L.H.S with A-1, thus
[A-1][A][X] = [A-1][B]
or [X] = [A-1][B]
* [A] must be a square matrix. Non-square matrix cannot be inverted.

Orthogonal Matrix
If A-1 = AT, the matrix [A] is called an orthogonal matrix.

Determinants
The determinant of a square matrix is a single number, a scalar quantity. Its symbol is
det[A] or A
Example:
det[A] = a11 a12
a21 a22

= (a11)(a22)-(a12)(a21)

Sub-matrix solution procedure:

In FEA, generally, some of the displacements and most of the forces are known and
the structural matrix equation is best solved by partitioning it.
The rows and columns of the stiffness matrix are arranged so that the known forces
and known displacements are placed in submatrices.

Example:
K11
K21
K31
K41
K51

K12
K22
K32
K42
K52

K13
K23
K33
K43
K53

K14
K24
K34
K44
K54

K15
K25
K35
K45
K55

q1
q02
q3
q04
q5

f01
f2
f03
f4
f05

The displacements and the forces with subscript 0 are known. [q02, q04 & f01, f03, f05]

Before partitioning, the rows of matrix K are rearranged so that the known forces are
grouped in the first three rows, as shown below.
K11
K31
K51
K21
K41

K12
K32
K52
K22
K42

K13
K33
K53
K23
K43

K14
K34
K54
K24
K44

K15
K35
K55
K25
K45

q1
q02
q3
q04
q5

f01
f03
= f05
f2
f4

Next, the columns of the K matrix are rearranged so that the unknown displacements are
grouped on the top.

Note that, the validity of the algebraic equations represented by the matrix equations
is maintained only if the change in position of q values is compensated by swapping
the corresponding columns in the K matrix.
We will change the order of displacements to:
q1
q3
q5
q02
q04

Therefore, we must swap the columns in K matrix as:


Column 1: unchanged
Column 2: Replace by column 3
Column 3: Replace by column 5
Column 4: Replace by column 2
Column 5: Replace by column 4
4

The final matrix eqn. is


K11
K31
K51
K21
K41

K13
K33
K53
K23
K43

K15
K35
K55
K25
K45

K12
K32
K52
K22
K42

K14
K34
K54
K24
K44

q1
f01
q3
f03
q5 = f05
q02
f2
q04
f4

Now we can partition the matrix as shown. The submatrices can be written as
Kff Kfs
Ksf Kss

qf
qo =

Fo
Fs

[Kff] {qf} + [Kfs] {qo} = Fo


[Ksf] {qf} + [Kss] {qo} = Fs

(1)
(2)

The equations can be solved for qf & Fs as


{qf} = [Kff-1] ({Fo} [Kfs] {qo})
{Fs} = [Ksf]{qf} + [Kss]{qo}

You might also like