Sie sind auf Seite 1von 24

CHAPTER 3

Special Matrices

SPECIAL MATRICES
Banded Matrix:
A banded matrix is a square matrix that has all elements equal to zero,
with the exception of a band centered on the main diagonal.

The dimensions of a banded system can be quantified by two


parameters:
a)

The bandwidth BW, and

b)

The half bandwidth HBW.

These two values are related by

SPECIAL MATRICES
The dimensions of a banded system can be quantified by two
parameters:
a)

The bandwidth BW, and

b)

The half bandwidth HBW.

These two values are related by BW = 2HBW + 1


In general, a banded system is one for which
aij = 0 if |i - j|>HBW

0 0

BW
0

HBW

0 0

HBW + 1

TRIDIAGONAL SYSTEM
A banded matrix with a bandwidth of 3 is called a tridiagonal
system. It can be expressed generally as
f1

e2

g1
f2
e3

g2
f3 g3

en 1

f n 1
en

x1 r1


x
2 r2
x r
3 3


g n 1 xn 1 rn 1
f n xn rn

Notice that we have changed our notation for the


coefficients from a and b to e, f, g and r.

TRIDIAGONAL SYSTEM
By LU decomposition, the tridiagonal system is transformed to
f1 g1

e2 f 2

e3

1
~
e2 1

e~2

g2
f3

g3

en 1 f n 1 g n 1
en
f n
f1 g1

~
f2

e~n 1 1

e~n 1

g2
~
f3

g3

~
f n 1

g n 1
~
fn

THOMAS ALGORITHM
A tridiagonal system can be solved efficiently by using Thomas
algorithm. The algorithm consists of three steps:
1. Decomposition:

e
e~k k
f k 1
where k = 2,3,.,n
~
f k f k e~k g k 1

THOMAS ALGORITHM
2. Forward Substitution:

~
rk rk e~k rk 1
where k = 2,3,.,n
3. Back Substitution:

~
r
xn ~n
fn
~
r g x
xk k ~k k 1
fk

where k = n-1,n-2,..,2,1

EXAMPLE 1
Solve the following tridiagonal system:
0
0 x1 40.8
2.04 1

1
2
.
04

1
0
0
.
8

1 2.04 1 x3
0.8

0
1 2.04 x4 200.8

SOLUTION
LU decomposition

0
0
2.04 1

0
1 2.04 1
0
1 2.04 1

1
2
.
04

0
0
2.04 1

0
1
.
550

1
0

0
1 2.04 1

1
2
.
04

0
0
2.04 1

0
1
.
550

1
0

0
0
1.395 1

0
0
1 2.04

0
0
2.04 1

0
1
.
550

1
0

0
0
1.395 1

0
0
0
1.323

R2 ( 2.041 ) R1

1
R3 ( 1.550
) R2

1
R4 ( 1.395
) R3

SOLUTION
The factors employed to obtain the upper triangular matrix are
~
f 2 2.041 0.49
~
1
f 3 1.550
0.645
~
1
f 4 1.395
0.717
LU decomposition
0
0
1

1
0
0.49
[ A] [ L][U ]
0
0.645
1

0
0
0.717

0 2.04 1
0
0

0 0 1.550 1
0
0 0
0
1.395 1

1 0
0
0
1.323

SOLUTION
Forward decomposition
r2 0.8 (0.49)(40.8) 20.8
r3 0.8 (0.645)(20.8) 14.221
r4 200.8 (0.717)(14.221) 210.996

Thus, the right-hand-side vector has been modified to


40.8
20.8

14.221
210.996

SOLUTION
Back decomposition
x4

r4
f4

x3

r3 g 3 T4
f3

1)(159.480 )
14.221(1.395
124.538

x2

r2 g 2 T3
f2

x1

r1 g1 T2
f1

210.996
1.323

159.480

20. ( 1)(124.538)
1.550

93.778

40.8 ( 1)(93.778)
2.04

65.970

EXAMPLE 2
a) Solve the following tridiagonal system with the Thomas algorithm
0 x1 41
0.8 0.4

0
.
4
0
.
8

0
.
4
x

25

0.4 0.8 x3 105

b) Determine the matrix inverse for [A] based on the LU


decomposition and unit vectors.

SOLUTION
a)

0 x1 41
0.8 0.4

0.4 0.8 0.4 x2 25


0
0.4 0.8 x3 105

We have
f1 0.8,

g1 0.4,

e2 0.4, f 2 0.8,

r1 41
g 2 0.4,

e3 0.4, f 3 0.8,

r2 25
r3 105

Decomposition:
For k = 2,

e
0.4
e~2 2
0.5
f1
0.8
~
f 2 f 2 e~2 g1 0.8 (0.5)(0.4) 0.6

SOLUTION
For k = 3,

e
0.4
e~3 ~3
0.66667
0
.
6
f2
~
f 3 f 3 e~3 g 2 0.8 (0.66667)(0.4) 0.53333

Thus, the LU decomposition is


0
0 0.8 0.4
0
1

[ A] [ L][U ] 0.5
1
0 0
0.6
0.4
0
0.66667 1 0
0
0.53333

SOLUTION
0
0 0.8 0.4
0
1

[ A] [ L][U ] 0.5
1
0 0
0.6
0.4
0
0.66667 1 0
0
0.53333

Forward substitution:
For k = 2,
For k = 3,

~
r2 r2 e2 r1 25 (0.5)(41) 45.5
~
r3 r3 e3 ~
r2 105 (0.66667)(45.5) 135.3333

Thus, the right-hand-side becomes

41

45.5
135.3333

SOLUTION
0
0 0.8 0.4
0
1

[ A] [ L][U ] 0.5
1
0 0
0.6
0.4
0
0.66667 1 0
0
0.53333

Back substitution:
For k = 3,

x3

r3 135.3333

253.75
f 3 0.53333

For k = 2,

x2

r2 g 2 x3 [45.5 (0.4)(253.75)]

245
f2
0.6

For k = 1,

x1

r1 g1 x2 [41 (0.4)(245)]

173.75
f1
0.8

Thus, the required solution is


x1 173.75,

x2 245,

x3 253.75

SOLUTION
b) The LU decomposition is
0
0
1

[ L] 0.5
1
0 ,
0
0.66667 1

0
0.8 0.4

[U ] 0
0.6
0.4
0
0
0.53333

To compute the first column of the inverse,


1

[ L]{D} 0 ,
0

d1

where {D} d 2
d
3

0
0 d1 1
1


0.5
1
0 d 2 0
0
0.66667 1 d 3 0

d1 1

0.5d1 d 2 0
0.66667 d 2 d 3 0

d1 1,

d 2 0.5,

d 3 0.333335

SOLUTION
And

[U ]{ A1} {D}
0 A11
1
0 .8 0 .4

0
0 .6
0.4 A21 0.5
0
0
0.53333 A31 0.333335

0.8 A11 0.4 A21 1


0.6 A21 0.4 A31 0.5
0.53333 A31 0.333335

By back substitution, we have

A31 0.625,
The first column of the inverse is

A21 1.25,

1.875

1.25
0.625

A11 1.875

SOLUTION
For the second column,
0

[ L]{D} 1 ,
0

d1

where {D} d 2
d
3

0
0 d1 0
1


0.5
1
0 d 2 1
0
0.66667 1 d 3 0

d1 0

0.5d1 d 2 1
0.66667 d 2 d 3 0

d1 0,

d 2 1,

d 3 0.66667

SOLUTION
And

[U ]{ A1} {D}
0 A12 0
0.8 0.4

0
0.6
0.4 A22 1
0
0
0.53333 A32 0.66667

0.8 A12 0.4 A22 0


0.6 A22 0.4 A32 1
0.53333 A32 0.66667

By back substitution, we have

A32 1.25,
The second column of the inverse is

A22 2.5,
1.25

2 .5
1.25

A12 1.25

SOLUTION
For the third column,

0

[ L]{D} 0 ,
1

0
0 d1 0
1


0.5
1
0 d 2 0
0
0.66667 1 d 3 1

d1 0

0.5d1 d 2 0
0.66667 d 2 d 3 1

d1 0,

d 2 0,

d3 1

SOLUTION
And

[U ]{ A1} {D}
0 A13 0
0.8 0.4


0
0.6
0.4 A23 0
0
0
0.53333 A33 1

0.8 A13 0.4 A23 0


0.6 A23 0.4 A33 0
0.53333 A33 1

By back substitution, we have

A33 1.875,
The third column of the inverse is

A23 1.25,
0.625

1.25
1.875

A13 0.625

SOLUTION
The matrix inverse is
1.875 1.25 0.625

[ A]1 1.25 2.5 1.25


0.625 1.25 1.875

Das könnte Ihnen auch gefallen