ENGR90024
COMPUTATIONAL FLUID DYNAMICS
Lecture O03
Analysis of the Eulers Method!
Higher Order Taylor Method
In general, Eulers method is not very accurate because it is
derived by truncating Taylor series at t2.
Taylor series
(t
l+1
td
) = (t ) +
1! dt
l
tl
Eulers method
(t
l+1
t f (tl , l )
) = (t ) +
1!
l
To get a more accurate formula, we will derive a formula
by truncating Taylor series at t3
Taylor series
td
l+1
l
(t ) = (t ) +
1! dt
t2 d 2
+
2
2!
dt
l
t
tl
Taylor method order 2
2 2
t
t
d
l
l
l+1
l
f
(t
,
)
(t ) = (t ) +
+
1!
2! dt2
tl
t l
l+1
l
(t ) = (t ) +
f (t ,
1!
2 2
t
d
l
)+
2! dt2
t l
l+1
l
(t ) = (t ) +
f (t ,
1!
2
t
d d
l
)+
2! dt dt
t l
l+1
l
(t ) = (t ) +
f (t ,
1!
2
t
d
l
)+
f (t, (t))
2! dt
(t
(t
l+1
l+1
t l
) = (t ) +
f (t ,
1!
l
t l
) = (t ) +
f (t ,
1!
l
t
)+
2!
2
t
)+
2!
@f l
(t ,
@t
@f l
(t ,
@t
tl
tl
Chain rule
tl
@f l
)+
(t ,
@
@f l
)+
(t ,
@
)f (tl ,
d
)
dt
l
(tl+1 ) = (tl ) +
t l
f (t ,
1!
)+
t
2!
@f l
(t ,
@t
)+
@f l
(t ,
@
)f (tl ,
O03.1
Equation (O03.1) is called the Taylor Method order 2. It is!
more accurate than the Eulers method because it has a local!
truncation error O(t3)
Taylor method order 2
Advantage - More accurate
Disadvantage - Need to calculate derivatives of f(t,)
Example O03.1:
!
Using Taylors method order 2, solve
!
!
!
!
!
d
=1
dt
For 0 < t < 8 with (t=0)=0 and
a) t=2
b) t=1
c) t=0.5
d) t=0.1
Compare your solution with Eulers method
Taylors method order 2
(tl+1 ) = (tl ) +
t l
f (t ,
1!
)+
t
2!
@f l
(t ,
@t
)+
@f l
(t ,
@
We need to find all terms in the equation above
For this question
f (t, ) = 1
@f
=0
@t
@f
=
@
)f (tl ,
(tl+1 ) = (tl ) +
t l
f (t ,
1!
)+
t
l+1
l
(t ) = (t ) +
(1
1!
l+1
@f
=
@
@f
=0
@t
f (t, ) = 1
t
l
= +
(1
1!
t
2!
@f l
(t ,
@t
)+
@f l
(t ,
@
)f (tl ,
2
t
l
)+
0 + 1(1
2!
2
t
l
)+
1
2!
Eulers method!
Example O02.1
l+1
t(1
function MPO02p1()
clear all;
close all;
Delta_t=2.0;
t=0:Delta_t:8.0
phi(1)=0.0;
Taylor method order 2!
Example O03.1
l
Only need!
to change one!
line of code!
l+1
t
+
(1
1!
t2
)+
1
2!
function MPO03p1()
clear all;
close all;
Delta_t=0.5;
phi(1)=0.0;
t=0:Delta_t:8.0
for l=1:length(t)-1
phi(l+1)=phi(l)
+Delta_t*(1-phi(l))
end
for l=1:length(t)-1
phi(n+1)=phi(n)+Delta_t*(1phi(l))-(Delta_t^2/2)*(1-phi(l))
end
plot(t,phi,'ko-')
hold on
ezplot(@(t)1-exp(-t),
[0,8,0,2])
xlabel('t');
ylabel('\phi');
legend('Euler','True');
plot(t,phi,'ko-')
hold on
ezplot(@(t)1-exp(-t),[0,8,0,2])
xlabel('t');
ylabel('\phi');
legend('Euler','True');
function MPO03p1()
clear all;
close all;
Delta_t=0.5;
phi(1)=0.0;
t=0:Delta_t:8.0
for l=1:length(t)-1
phi(l+1)=phi(l)+Delta_t*(1-phi(l))-(Delta_t^2/2)*(1-phi(l))
end
plot(t,phi,'ko-')
hold on
ezplot(@(t)1-exp(-t),[0,8,0,2])
xlabel('t');
ylabel('\phi');
legend('Euler','True');
Output
1.6
1.4
1.2
Note: the
Taylor
method for
t=2 predicts
=0 for all
values of t!
Taylor 2nd Order
True
1.8
1
0.8
0.6
0.4
0.2
0
1.8
t=2.0
1.6
1.6
1.2
1.2
1.4
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
Taylor 2nd Order
True
t=1.0
1.8
1.4
4
t
1.6
1.2
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
4
t
4
t
t=0.1
Taylor 2nd Order
True
1.6
1.2
1.8
1.4
1.4
Taylor 2nd Order
True
t=0.5
1.8
Taylor 2nd Order
True
4
t
Taylor 2nd order
Euler
1exp(t)
2
Euler
True
1.8
1.4
1.4
t=1.0
1.2
1.2
1
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
0
0
4
t
4
t
1.4
1.6
1.4
1.2
1.2
0.8
0.8
0.6
0.6
0.4
0.4
0.2
0.2
4
t
Taylor 2nd Order
True
1.8
t=0.5
1.6
2
Euler
True
1.8
t=1.0
1.6
1.6
Taylor 2nd Order
True
1.8
t=0.5
4
t
For a given t, the Taylor 2nd order method is more accurate than
Euler
End of Example O03.1
STABILITY OF THE EXPLICIT
EULER METHOD
(SEE PAGE 8 PRINTED LECTURE NOTES)
In Lecture O02, we looked at the accuracy of the explicit Euler
method. Lets now look at the stability of the Euler method.!
We know from previous lecture, the smaller t, the more accurate
the solution, i.e. the truncation error go to zero. This feature is
known as consistency.!
When we make t big enough, the solution will blow up. Can we
predict the value of t where the solution will blow up? For a
numerical method to be useful, we need it to be stable.
Stability and consistency are two quite different things (see later). It is
possible for a method can be consistent but not stable.
If a numerical method is both consistent and stable, it is convergent.
To carry out stability analysis, consider the model problem
d
=
dt
(O03.2)
where is a constant and is a complex number
Re
+i
Im
In many engineering problems, Re is zero or negative. For
stability analysis, you will always assume that Re is zero or
negative.
Find analytical solution!
at time level lt
Strategy
d
=
dt
Compare
Find approximated Eulers!
solution at time level lt
The analytical solution for Eq. (O03.2) can be written as
(t)
1 (
Re t
ei
Re t
(cos(
e
e
t
Re +i Im )t
Im t
Im t)
where 1 is the initial value of
+ i sin(
Im t))
(t)
Re t
(cos(
Im t)
+ i sin(
(t)
Im
(t)
(t)
Im
(t)
(t)
Re
Re
Analytical solution will decay with time if lRe<0
Re
Im t))
(t)
Re t
(cos(
Im t)
+ i sin(
Im t))
Find analytical solution!
at time level lt
d
=
dt
Compare
Find approximated Eulers!
solution at time level lt
Applying Eulers formula for the model problem (Eq. (O03.2)) gives
l+1
= (1 +
=
t)
where = (1+t) is the amplification factor
(O03.3)
l+1
Thus, running a program using Eulers method will give you
1
2
3
4
..
.
l+1
=
=
=
=
=
=
Given
1
2
3
..
.
=
l
2 1
=
=
=
3 1
..
.
l 1
(O03.4)
For solution to be
stable!| | 1
| |2
1+
(1 +
Re
(1 +
Re ) + ( t
+i t
2
Im )
2
Im ) 1
Im
-2
-1
Re
Example O03.2:
!
Write a Matlab program that uses Eulers method to solve
!
!
!
!
!
d
=
dt
for 0 < t < 2 with (t=0)=1. Using the analysis from the previous slides, what
are the values of t for the numerical solution to be stable?
We have
d
= 8
dt
So for this problem =-8. So Re=-8 and Im=0.0. In order for the
numerical solution to be stable, t need to be within the stable
region.!
Im t
-8t, t large
-2
-1
Re
t0
8 t0
t 1/4
So for solution to !
be stable t must!
be between 0 and
0.25
Im
-2
-1
Re
function MPO03p2()
clear all;
close all;
Delta_t=0.01;
t=0:Delta_t:2.0;
%Preallocating memory
phi=zeros(size(t));
phi(1)=1.0;
for l=1:length(t)-1
phi(l+1)=phi(l)+Delta_t*(-8*phi(l));
end
plot(t,phi,'ko-')
hold on
ezplot(@(t)exp(-8*t),[0,2,-2,2])
xlabel('t');
ylabel('\phi');
legend('Euler','True');
100
Euler
True
80
t=0.5
60
0.5
40
Euler
True
t=0.25
1.5
20
0.5
20
40
1.5
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
0.2
1.5
0.8
1
t
1.2
1.4
1.6
1.8
0.5
0.5
0.5
0.5
1.5
1.5
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
Euler
True
t=0.01
1.5
t=0.1
0.6
Euler
True
0.4
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
100
Euler
True
80
t=0.5
60
0.5
40
Euler
True
t=0.25
1.5
20
0.5
20
40
1.5
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
0.2
1.5
0.8
1
t
1.2
1.4
1.6
1.8
0.5
0.5
0.5
0.5
1.5
1.5
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
Euler
True
t=0.01
1.5
t=0.1
0.6
Euler
True
0.4
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
Euler solution is unstable for t=0.5
Euler solution is neutrally stable for t=0.25
Euler solution is stable for t=0.1 & t=0.01
Im
-2
-4
-1
Re
100
Euler
True
80
t=0.5
60
40
20
t=
20
40
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
8 0.5 =
Im
-2
-1
t=0.25
1.5
Re
Euler
True
0.5
0.5
t=
1.5
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
8 0.25 =
Im
-2
-1
Re
2
Euler
True
t=0.1
1.5
0.5
0.5
t=
1.5
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
8 0.1 =
0.8
Im
-2
-1
Re
2
Euler
True
t=0.01
1.5
0.5
0.5
t=
1.5
0.2
0.4
0.6
0.8
1
t
1.2
1.4
1.6
1.8
8 0.01 =
0.08
End of Example O03.2
Consider now the case where is purely imaginary,
Re
+i
Im
d
=
dt
d
=i
dt
Im
For this case, iImt, is always on the vertical axis of the
stability diagram. It is not within the stability region. Hence,
in this case, Eulers method is always unstable.
Example O03.3:
!
Write a Matlab program that uses Eulers method to solve
!
!
!
!
!
d
= 2i
dt
for 0 < t < 10 with (t=0)=1. Using the analysis from the previous slides, what
are the values of t for the numerical solution to be stable?
d
= 2i
dt
0 t 10
(t = 0) = 1
The analytical solution for this problem can be written as
= Ae2it
= (ARe + iAIm )e2it
= (ARe + iAIm )(cos(2t) + i sin(2t))
= (ARe cos(2t)
AIm sin(2t)) + i(ARe sin(2t) + AIm cos(2t))
But we are given that =1 for t=0
1 + 0i = ARe + iAIm
Hence
ARe = 1
AIm = 0
d
= 2i
dt
0 t 10
= (ARe + iAIm )e2it
(t = 0) = 1
ARe = 1
AIm = 0
So the analytical solution for this equation is
(t)
e2it
cos(2t) + i sin(2t)
function MPO03p3()!
clear all;!
close all;!
Delta_t=0.02;!
!
!
t=0:Delta_t:10.0;!
%Preallocate Memory!
phi=zeros(size(t));!
phi(1)=1.0;!
!
!
for n=1:length(t)-1!
phi(n+1)=phi(n)+Delta_t*(i*2*phi(n));!
end!
!
plot(t,imag(phi),'ko-')!
xlabel('t');!
ylabel('\phi');!
hold on!
ezplot(@(t)sin(2*t),[0,10,-5,5])!
legend('Euler','True');
Im()
sin(2 t)
5
4
t=0.5
Euler
True
t=0.2
4
3
sin(2 t)
5
Euler
True
1
2
2
3
3
4
4
5
5
t
10
sin(2 t)
Euler
True
0
1
5
t
5
t
10
10
Euler
True
t=0.02
2
1
sin(2 t)
t=0.1
5
t
10
Im
t = 2i 0.5 = i
-1
Re
sin(2 t)
Euler
True
t=0.5
3
2
1
-2
0
1
2
3
4
5
10
Im
0.4i
t = 2i 0.2 = 0.4i
-1
Re
sin(2 t)
Euler
True
t=0.2
4
3
2
1
-2
0
1
2
3
4
5
10
Im
t = 2i 0.1 = 0.2i
0.2i
-1
Re
sin(2 t)
Euler
True
t=0.1
4
3
2
1
-2
0
1
2
3
4
5
5
t
10
Im
t = 2i 0.02 = 0.04i
0.04i
-1
Re
t
sin(2 t)
Euler
True
t=0.02
4
3
2
1
-2
0
1
2
3
4
5
10
For all values of t, t is outside the stable region. Hence
the numerical solution obtained using Eulers method will
never be stable for any value of t.
End of Example O03.3
