Week9 ENG1060 Lecture17

ENG 1060
Computing for Engineers

Lecturer: Dr Christopher Hutchinson
Lecture 17:
MATLAB Solutions to Linear Algebraic Eqns
Last Lecture
• Last lecture
• Today’s Lecture
• Elimination of
• Introduction to Matrices (Matrix
unknowns
format for linear algebraic eqns and
• Gaussian
elimination
basic matrix math)
• Special
matrices
• Lecture
summary
• Using MATLAB to solve sets of
• Next lecture
linear algebraic eqns
2
Last Lecture
• Last lecture
We need to revisit the basics of ‘Matrix Mathematics’
• Elimination of
N linear algebraic equations
unknowns in n unknowns
• Gaussian
a11 x1  a12 x2      a1n xn  b1 This can be written:
elimination
a21 x2  a22 x2      a2 n xn  b2 a11 a12 . . a1n x1  b1 

• Special
    
a21 a22 . . . x 2  b2 
matrices
.......
• Lecture
 . . . . .  .   . 
summary an1 xn  an 2 x2      ann xn  bn     
• Next lecture  . . . . .  .   . 

an1 . . . ann 

x n 
 
bn 

or
Ax  b
3
Last Lecture
• Last lecture We need to revisit the basics of ‘Matrix Mathematics’
Formally, the matrix equation, Ax=b, can be solved by pre-
• Elimination of
unknowns (or left-) multiplying by A-1
• Gaussian
elimination
A-1Ax = A-1 b
• Special
matrices
• Lecture
summary
Remember A-1 A = I
• Next lecture
and Ix=x
so x = A-1 b
4
Last Lecture
• Last lecture
There are two ways that MATLAB can be used to solve linear
• Today’s Lecture algebraic equations expressed in matrix form :
• Elimination of
unknowns
• Gaussian
elimination • Pre-multiplication by the inverse
• Special • >> x=inv(A)*b =A^-1*b
matrices
• Lecture
summary • Left division MATLAB preferred
• Next lecture • >> x=A\b
Ax = b
A\ Ax = A\ b
I x = A\ b
‘dividing from the front’
Of course this is not really division,

since division is not a valid operator
in matrix math. So what is MATLAB doing?
5
Last Lecture
• Last lecture Mixing oil……
• Elimination of
unknowns
• Gaussian Q13
elimination
F Q35
Q55
• Special Q12 Q34
matrices
• Lecture Q44
summary
Q24
• Next lecture
Linear Algebraic Eqns Linear algebraic Eqns expressed in matrix form
(Q12  Q13 )x1  F (Q12  Q13 ) x1  F 
    
Q12 x1  Q24 x 2  0  Q12 Q24 x 2  0 
Q13 x1  (Q34  Q35 )x 3  0  Q13 (Q34  Q35 ) x 3  0 
    
Q24 x 2  Q34 x 3  Q44 x 4  0 Q24 Q34 Q44
 x 4  0 
Q35 x 3  Q55 x 5  0 
 Q35 Q55 

x 5 
 
0 

This is now in matrix form; Ax=B,
where A and B are known 6
Last Lecture
• Last lecture Mixing oil……
(Q12  Q13 ) x1  F 
• Elimination of     
 Q12 Q24 x 2  0 
unknowns
• Gaussian  Q13 (Q34  Q35 ) x 3  0 

elimination     
 Q24 Q34 Q44 x 4  0 
• Special
matrices 
 Q35 Q55 

x 5 
 
0 

• Lecture
summary
• Next lecture
Qx   F 

 Q1Qx  Q1F 
I x  Q1 F 
x  Q1F 
7
Last Lecture
• Last lecture
• Elimination of
unknowns
Mixing oil…… Input starting values
• Gaussian
elimination
• Special
matrices
• Lecture
summary
• Next lecture
Q-1
Note the problem!
This ‘ means transpose.

Change a row into a column
8
Last Lecture
• Last lecture
• Elimination of
unknowns
Mixing oil……
• Gaussian
elimination
• Special
matrices
• Lecture
summary
• Next lecture
Here are the concentrations in each

reactor: x1…..x5
9
Last Lecture
• Last lecture
• Elimination of
unknowns
Mixing oil……
• Gaussian
elimination
• Special
matrices
• Lecture
summary
• Next lecture
Pretty easy!
But how is MATLAB calculating

this?
10
Today’s lecture
• Last lecture
• How does MATLAB solve these
• Elimination of
unknowns
sets of linear algebraic equations?
• Gaussian
elimination
• Special
matrices
• Lecture
summary
• Next lecture
11
MATLAB Solutions
• Last lecture
There are two ways that MATLAB can be used to solve linear
• Today’s Lecture algebraic equations expressed in matrix form :
• Elimination of
unknowns
You should know
• Gaussian this from ENG1091
elimination • Pre-multiplication by the inverse
• Special • >> x=inv(A)*b =A^-1*b
matrices
• Lecture
summary • Left division MATLAB preferred
• Next lecture • >> x=A\b
Ax = b
A\ Ax = A\ b
I x = A\ b
What is really being done here?
12
Elimination of Unknowns
• Last lecture
Consider we had only two equations in two unknowns (x1, x2):
• Elimination of Eq 1. a11  x1  a12  x2  b1

unknowns
• Gaussian Eq 2. a21  x1  a22  x2  b2

elimination
• Special If we multiply Eq. 1 by a21/a11 we obtain:

matrices
a 21 a a a21 a
• Lecture
Eq 1’. a11  x1  21 a12  x 2  21 b1 a21  x1  a12  x2  21 b1
summary a11 a11 a11 a11 a11
• Next lecture
If we then subtract Eq. 1’ from Eq. 2:

Eq. 2 - Eq. 1’
Eq 2. a21  x1  a22  x2  b2
a21 a
a21 a (a22  a12 )  x2  b 2  21 b1
a21  x1  a12  x2  21 b1 a11 a11
Eq 1’. a11 a11
This is only a function of x2
13
Elimination of Unknowns
• Last lecture
Consider we had only two equations in two unknowns (x1, x2):
• Elimination of Eq 1. a11  x1  a12  x2  b1

unknowns
• Gaussian Eq 2. a21  x1  a22  x2  b2

elimination
a21 a
• Special Multiply Eq. 1 by a21/a11 we obtain: Eq 1’. a21  x1  a12  x2  21 b1
matrices a11 a11
• Lecture
summary
Eq. 2 - Eq. 1’
• Next lecture a21 a
(a22  a12 )  x2  b 2  21 b1
a11 a11
Substitute for x2 back
This is only a function of x2
into Eq. 1 or 2 to find x1.
The simultaneous linear algebraic equations can be solved in two steps:

1. Elimination of the unknowns
2. Back substitution
14
Gaussian Elimination
• Last lecture The simultaneous linear algebraic equations can be solved in two steps:
• Today’s Lecture1. Elimination of the unknowns
2. Back substitution
• Elimination of
unknowns
This process can be readily programmed into a computer to solve n
• Gaussian
elimination simultaneous linear algebraic equations.
• Special
matrices It is called Gaussian Elimination (you have seen this in L4 ENG1091)
• Lecture
summary Gaussian elimination is one of the main methods MATLAB uses to solve
sets of equations of the form:
• Next lecture
a11 x1  a12 x2      a1n xn  b1

a21 x2  a22 x2      a2 n xn  b2
.......
an1 xn  an 2 x2      ann xn  bn
15
• Last lecture How do we do it in practice? a11 a12  a1n b1 
 
• Today’s Lecture a11x1  a12 x 2      a1n x n  b1
augmented a21 a22  a2n b2 
• Elimination of a 21x1  a 22 x 2      a 2n x n  b2 matrix       
unknowns  
• Gaussian
an1 an 2  ann bn 
elimination
a n1x1  a n 2 x 2      a nn x n  bn
pivot equation: a21
• Special  multiply row 1 by
matrices a11 the pivot a11
• Lecture element   subtract from row 2
summary

• Next lecture
 a11 a12  a1n b1 
 a 21 
 21 a11 a11
a  a22  aa1121 a12  a  a 21
a b  a 21
 2n a11 1n 2 a11 1 b
      
 
 an1 an 2  ann bn 
0 a22 a2n b2

16
• Last lecture a11 a12  a1n b1 
 
 a22  a2n b2
• Elimination of       
unknowns
  repeat for
• Gaussian an1 an 2  ann bn  remaining
elimination
rows
• Special
matrices
a11 a12  a1n b1 
 

• Lecture
 a22  a b 
2n 2 
summary (n-1) rows
      {<=> (n-1)
• Next lecture
  unknowns}
 an 2  a b 
nn n 
a11 a12  a1n b1 

 
  a22  a2n b2
get rid of x2 systematically
     (n-2) rows
 
  ann bn
 {<=> (n-2) unknowns}
17
a11 a11 x1  a12 x 2                  a1(n1) x n1  a1n x n  b1
• Last lecture a12  a1n b1 
  a22 x 2                  a2(n1) x n1  a2n x n  b2
• Today’s Lecture  a22  a2n b2

• Elimination of     
unknowns   (n1) x n1  a(n1)n
a(n1)(n1) (n1) x n  bn1
(n1)
 annn bnn 
• Gaussian annn x n  bnn
elimination
ELIMINATION COMPLETED:
• Special
matrices upper triangular matrix of aij’s
 
• Lecture
summary
bn
• Next lecture xn  n
annn
(n1)  a(n1)n
(bn1 (n1) x n )
x n1  BACK
(n1)
a(n1)(n1)
SUBSTITUTE
  
  
x1 
b1  a12 x 2                  a1(n1) x n1  a1n x n 
a11
18
• Last lecture
If only 3 variables:
• Elimination of
unknowns
• Gaussian
elimination
• Special
matrices
• Lecture
summary
• Next lecture
19
M-file: Gaussian Elimination
function x = GaussNaive(A,b)
% GaussNaive(A,b):
% Gauss elimination without pivoting
% input:
% A: coefficient matrix
% b: right hand side vector
% Output
% x: solution vector
Check on squareness of A
[m,n] = size(A);
if m~=n, error('Matrix A must be square'); end
nb = n+1;
Aug = [A b]; Form augmented matrix
% forward elimination
for k = 1: n-1
for i = k+1:n
factor = Aug(i,k)/Aug(k,k); Elimination steps
Aug(i,k:nb) = Aug(i,k:nb)-factor*Aug(k,k:nb);
end
end
% back substitution
c=zeros(n,1);
c(n) = Aug(n,nb)/Aug(n,n);
for i = n-1:-1:1 Back substitution steps
c(i) = (Aug(i,nb)-Aug(i,i+1:n)*c(i+1:n))/Aug(i,i);
end
x=c;
20
% GaussNaive(A,b):
% Gauss elimination without pivoting
% input:
% A: coefficient matrix
% Output
[m,n] = size(A);
nb = n+1;
Aug = [A b];
for k = 1: n-1
for i = k+1:n
factor = Aug(i,k)/Aug(k,k);
end
end
% back substitution
c=zeros(n,1);
for i = n-1:-1:1
end
x=c;
21
• Last lecture
• Today’s Lecture nb = n+1; a11 a12 . .  b1 

Aug = [A b];    
• Elimination of
a21 . . . 
unknowns A b   
• Gaussian
 . . . .   
elimination    
 . . . ann  b n 
• Special
matrices
• Lecture
summary
• Next lecture a11 a12 . . b1 

  
a21
 . . .  n
Aug 
 . . . . 
 
 . . . ann bn 
nb=n+1
 22
• Last lecture for k = 1:n-1
for i = k+1:n Two
• Elimination of
Aug(i,k:nb) = Aug(i,k:nb) - factor* Aug(k,k:nb);
nested
unknowns
end loops
• Gaussian
end
elimination
first time
• Special
matrices
• Lecture
k 1 i2 factor  a21 a
summary 11
• Next lecture
  a2(1:n 1)  a2(1:n 1)  factor * a1(1:n 1)


    
  This subtracts factor*row 1 from
   
Aug   row 2 right across the augmented
     matrix, getting rid of the first
  element of row 2
    
23
for i = k+1:n Two
• Elimination of
nested
unknowns
end loops
• Gaussian
end
elimination
next
• Special
matrices
k 1 i3 a31
factor 
• Lecture
summary a11
• Next lecture
 a3(1:n 1)  a3(1:n 1)  factor * a1(1:n 1)
    
     This subtracts factor*row 1 from
Aug   row 3 right across the augmented
     matrix, getting rid of the first
  element of row 3
    
24
for i = k+1:n Two
• Elimination of
nested
unknowns
end loops
• Gaussian
end
elimination
until
• Special
matrices
k 1 in an1
factor 
• Lecture
summary a11
• Next lecture
 an (1:n 1)  an (1:n 1)  factor * a1(1:n 1)
    
     This subtracts factor*row 1 from
Aug   row n right across the augmented
     matrix, getting rid of the first
  element of row n
    
25
for i = k+1:n Two
• Elimination of
nested
unknowns
end loops
• Gaussian
elimination
end then
• Special
matrices
k 2 i3 a32
factor 
• Lecture
summary a22
• Next lecture
a3( 2:n 1)  a3( 2:n 1)  factor * a2 ( 2:n 1)
    
     This subtracts factor*row 2 from
Aug    row 3 right across the augmented
    matrix, getting rid of the first
     26
for i = k+1:n Two
• Elimination of
nested
unknowns
end loops
• Gaussian
end
elimination
next
• Special
matrices
k 2 i4 factor  a42
• Lecture
summary a22
• Next lecture
a4 ( 2:n 1)  a4 ( 2:n 1)  factor * a2( 2:n 1)
    
    27
for i = k+1:n Two
• Elimination of
nested
unknowns
end loops
• Gaussian
end
elimination
until
• Special
matrices
k  n 1 an ( n 1)
in factor 
• Lecture
summary
a( n 1)( n 1)
• Next lecture
an ( n 1)  an ( n 1:n 1)  factor * a( n 1)( n 1:n 1)
    
    
Aug   
   
 
   28
Called naïve because in
% GaussNaive(A,b):
calculating the factor it is possible
% Gauss elimination without pivoting we will divide by zero and get an
% input:
% A: coefficient matrix error
% Output
[m,n] = size(A);
nb = n+1;
Aug = [A b];
for k = 1: n-1
for i = k+1:n
factor = Aug(i,k)/Aug(k,k); Elimination steps
end
end
% back substitution
c=zeros(n,1);
end
x=c;
29
for i = k+1:n Two
• Elimination of
nested
unknowns
end loops
• Gaussian
end
elimination
next
• Special
matrices
k 2 i4 factor  a42
• Lecture
summary a22
• Next lecture
a4 ( 2:n 1)  a4 ( 2:n 1)  factor * a2( 2:n 1)
    
    30
Gaussian Elimination: Pivoting
function x = GaussPivot(A,b) To avoid the potential problem of dividing
% GaussPivot(A,b):
% Gauss elimination with pivoting by zero when evaluating the factor, we
% input:
% A: coefficient matrix can check if the pivot element is zero.
% Output
% x: solution vector If so we just swap that row
with a row beneath it!
[m,n] = size(A);
nb = n+1;
The order of the rows does
Aug = [A b]; not matter. This is called
for k = 1: n-1 ‘partial pivoting’.
% Partial pivoting
[big,i] = max(abs(Aug(k:n,1)));
ipr = i+k-1; Search for biggest
if ipr~=k
%pivot the rows pivot element and
Aug([k,ipr],:)=Aug([ipr,k],:);
end use it in each case
for i = k+1:n
Elimination steps
end
end
% back substitution
c=zeros(n,1);
end
x=c;
31
Special Matrices
• Last lecture
e.g. Tri-diagonal matrix: particular case of sparse banded matrix with
bandwidth of 3
• Elimination of
f1 g1  x1   r1 
unknowns     
e2 f2 g2  x 2  r2 
• Gaussian  e3 f3 g3  x 3  r3 
elimination
    
         
• Special
matrices          
    
• Lecture en1 f n1 gn1 x n1 r
summary    n1

 en f n 

 x n 
 
rn 

• Next lecture
 f1     g1 
     
f
 2  e
 2  g2 
 There are special modifications of the  f 3  e3  g3 
traditional approaches such as Gaussian      
elimination that can be used to solve f     e     g    
these special sets of equations         
     
f
 n1 e
 n1 gn1

 f n 
 
en 
 
32
MATLAB Solutions
• Last lecture
MATLAB preferred
• Left division
• Elimination of
• >> x=A\b Ax = b
unknowns
A\ Ax = A\ b
• Gaussian I x = A\ b
elimination
• Special
matrices
What is really being done here?
• Lecture
summary
• Next lecture MATLAB uses a complex algorithm to:
1. Check the nature of the matrices (e.g. sparsely distributed,

tridiagonal, etc.)
2. If MATLAB can use a fast technique because the matrix has
special properties it will.
3. Otherwise MATLAB uses Gaussian Elimination
33
Lecture Summary
• Last lecture
• Elimination of
• We have looked at what MATLAB
unknowns
is doing when it solves matrix
• Gaussian
elimination
equations.
• Special
matrices
• Lecture
summary
• Considered coding a ‘Gaussian
• Next lecture
elimination’ routine to solve sets of
linear algebraic equations
34
Next lecture
• Last lecture
What about the iterative root finding

• Elimination of
unknowns
techniques we learnt in lecture 13?
• Gaussian
elimination
• Special
(e.g. Newton-Raphson)
matrices
These techniques can also be modified

• Lecture
summary
• Next lecture
to solve sets of equations:
These ‘Iterative methods’ are the topic

of L18.
35

Week9 ENG1060 Lecture17

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Week9 ENG1060 Lecture17

Uploaded by

Copyright:

Available Formats

ENG 1060

Computing for Engineers

a21 x2  a22 x2      a2 n xn  b2 a11 a12 . . a1n x1  b1 

Of course this is not really division,

• Gaussian  Q13 (Q34  Q35 ) x 3  0 

Note the problem!

This ‘ means transpose.

Here are the concentrations in each

But how is MATLAB calculating

What is really being done here?

• Elimination of Eq 1. a11  x1  a12  x2  b1

• Gaussian Eq 2. a21  x1  a22  x2  b2

• Special If we multiply Eq. 1 by a21/a11 we obtain:

If we then subtract Eq. 1’ from Eq. 2:

• Elimination of Eq 1. a11  x1  a12  x2  b1

• Gaussian Eq 2. a21  x1  a22  x2  b2

The simultaneous linear algebraic equations can be solved in two steps:

a11 x1  a12 x2      a1n xn  b1

0 a22 a2n b2

a11 a12  a1n b1 

• Today’s Lecture nb = n+1; a11 a12 . .  b1 

• Next lecture a11 a12 . . b1 

  a2(1:n 1)  a2(1:n 1)  factor * a1(1:n 1)

 a3(1:n 1)  a3(1:n 1)  factor * a1(1:n 1)

 an (1:n 1)  an (1:n 1)  factor * a1(1:n 1)

a3( 2:n 1)  a3( 2:n 1)  factor * a2 ( 2:n 1)

a4 ( 2:n 1)  a4 ( 2:n 1)  factor * a2( 2:n 1)

an ( n 1)  an ( n 1:n 1)  factor * a( n 1)( n 1:n 1)

a4 ( 2:n 1)  a4 ( 2:n 1)  factor * a2( 2:n 1)

• Next lecture MATLAB uses a complex algorithm to:

1. Check the nature of the matrices (e.g. sparsely distributed,

What about the iterative root finding

These techniques can also be modified

These ‘Iterative methods’ are the topic

You might also like