Scribd
Upload
181 views

Runge Kutta With MATLAB

The Runge-Kutta method is a numerical method for solving differential equations using approximations of tangent lines to compute intermediate values along segments of the solution curve. The document summarizes the 4th order Runge-Kutta method (RK4) and provides an example of using it to solve the differential equation y' = y - t^2 + 1 with initial value y(0) = 0.5. It compares the numerical solution to the exact solution for different step sizes and finds that smaller step sizes produce more accurate results. Adaptive step size control using the Runge-Kutta-Fehlberg method is introduced to achieve a desired error tolerance.

Uploaded by

Heather Cleveland
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
181 views

Runge Kutta With MATLAB

The Runge-Kutta method is a numerical method for solving differential equations using approximations of tangent lines to compute intermediate values along segments of the solution curve. The document summarizes the 4th order Runge-Kutta method (RK4) and provides an example of using it to solve the differential equation y' = y - t^2 + 1 with initial value y(0) = 0.5. It compares the numerical solution to the exact solution for different step sizes and finds that smaller step sizes produce more accurate results. Adaptive step size control using the Runge-Kutta-Fehlberg method is introduced to achieve a desired error tolerance.

Uploaded by

Heather Cleveland
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Runge-Kutta method

The formula for the fourth order Runge-Kutta method (RK4) is given below. Consider the
problem
(
y 0 = f (t, y)
y(t0 ) =
Define h to be the time step size and ti = t0 + ih. Then the following formula
w0 =
k1 = hf (ti , wi )


k1
h
k2 = hf ti + , wi +
2
2


h
k2
k3 = hf ti + , wi +
2
2
k4 = hf (ti + h, wi + k3 )
1
wi+1 = wi + (k1 + 2k2 + 2k3 + k4 )
6
computes an approximate solution, that is wi y(ti ).
Let us look at an example:
(
y 0 = y t2 + 1
y(0) = 0.5
The exact solution for this problem is y = t2 + 2t + 1 21 et , and we are interested in the value of
y for 0 t 2.
1. We first solve this problem using RK4 with h = 0.5. From t = 0 to t = 2 with step size
h = 0.5, it takes 4 steps: t0 = 0, t1 = 0.5, t2 = 1, t3 = 1.5, t4 = 2.
Step 0 t0 = 0, w0 = 0.5.
Step 1 t1 = 0.5
k1
k2
K3
K4
w1

= hf (t0 , w0 ) = 0.5f (0, 0.5) = 0.75


= hf (t0 + h/2, w0 + k1 /2) = 0.5f (0.25, 0.875) = 0.90625
= hf (t0 + h/2, w0 + k2 /2) = 0.5f (0.25, 0.953125) = 0.9453125
= hf (t0 + h, w0 + K3 ) = 0.5f (0.5, 1.4453125) = 1.09765625
= w0 + (k1 + 2k2 + 2k3 + k4 )/6 = 1.425130208333333

Step 2 t2 = 1
k1
k2
K3
K4
w2

= hf (t1 , w1 ) = 0.5f (0.5, 1.425130208333333) = 1.087565104166667


= hf (t1 + h/2, w1 + k1 /2) = 0.5f (0.75, 1.968912760416667) = 1.203206380208333
= hf (t1 + h/2, w1 + k2 /2) = 0.5f (0.75, 2.0267333984375) = 1.23211669921875
= hf (t1 + h, w1 + K3 ) = 0.5f (1, 2.657246907552083) = 1.328623453776042
= w1 + (k1 + 2k2 + 2k3 + k4 )/6 = 2.639602661132812
1

Step 3 t3 = 1.5
k1
k2
K3
K4
w3

= hf (t2 , w2 ) = 0.5f (1, 2.639602661132812) = 1.319801330566406


= hf (t2 + h/2, w2 + k1 /2) = 0.5f (1.25, 3.299503326416016) = 1.368501663208008
= hf (t2 + h/2, w2 + k2 /2) = 0.5f (1.25, 3.323853492736816) = 1.380676746368408
= hf (t2 + h, w2 + K3 ) = 0.5f (1.5, 4.020279407501221) = 1.385139703750610
= w2 + (k1 + 2k2 + 2k3 + k4 )/6 = 4.006818970044454

Step 4 t4 = 2
k1
k2
K3
K4
w4

= hf (t3 , w3 ) = 0.5f (1.5, 4.006818970044454) = 1.378409485022227


= hf (t3 + h/2, w3 + k1 /2) = 0.5f (1.75, 4.696023712555567) = 1.316761856277783
= hf (t3 + h/2, w3 + k2 /2) = 0.5f (1.75, 4.665199898183346) = 1.301349949091673
= hf (t3 + h, w3 + K3 ) = 0.5f (2, 5.308168919136127) = 1.154084459568063
= w3 + (k1 + 2k2 + 2k3 + k4 )/6 = 5.301605229265987

Now lets compare what we got with the exact solution


ti
0.0
0.5
1.0
1.5
2.0

Exact solution y(ti )


0.5
1.425639364649936
2.640859085770477
4.009155464830968
5.305471950534675

Numerical solution wi
0.5
1.425130208333333
2.639602661132812
4.006818970044454
5.301605229265987

Error |wi y(ti )|


0
0.000509156316603
0.001256424637665
0.002336494786515
0.003866721268688

All this can be done by using Matlab:


function rungekutta
h = 0.5;
t = 0;
w = 0.5;
fprintf(Step 0: t = %12.8f, w = %12.8f\n, t, w);
for i=1:4
k1 = h*f(t,w);
k2 = h*f(t+h/2, w+k1/2);
k3 = h*f(t+h/2, w+k2/2);
k4 = h*f(t+h, w+k3);
w = w + (k1+2*k2+2*k3+k4)/6;
t = t + h;
fprintf(Step %d: t = %6.4f, w = %18.15f\n, i, t, w);
end
%%%%%%%%%%%%%%%%%%
function v = f(t,y)
v = y-t2+1;
2

2. Solve the problem using RK4 with h = 0.2. All you need to do is to replace h = 0.5; and
for i=1:4 in the above Matlab program into h = 0.2 and for i=1:10. Then we have
ti
0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0

Exact solution y(ti )


0.5
0.829298620919915
1.214087651179365
1.648940599804746
2.127229535753766
2.640859085770477
3.179941538631726
3.732400016577663
4.283483787802442
4.815176267793527
5.305471950534674

Numerical solution wi
0.5
0.829293333333333
1.214076210666667
1.648922017041600
2.127202684947944
2.640822692728752
3.179894170232231
3.732340072854980
4.283409498318406
4.815085694579435
5.305363000692655

Error |wi y(ti )|


0
0.000005287586582
0.000011440512698
0.000018582763146
0.000026850805823
0.000036393041726
0.000047368399496
0.000059943722683
0.000074289484036
0.000090573214092
0.000108949842019

3. Solve the problem using RK4 with h = 0.05. Again, all need to do is to set h = 0.05 and
for i=1:40. Then we have
ti
0.0
0.05
0.10

1.80
1.85
1.90
1.95
2.00

Exact solution y(ti ) Numerical solution wi


0.5
0.5
0.576864451811988 0.576864446614583
0.657414540962176 0.657414530368210

4.815176267793529 4.815175898599096
4.942590238699086 4.942589852008494
5.067052778860367 5.067052374183828
5.188156209705356 5.188155786548850
5.305471950534677 5.305471508400809

Error |wi y(ti )|


0
0.000000005197405
0.000000010593966

0.000000369194433
0.000000386690592
0.000000404676539
0.000000423156505
0.000000442133868

4. Comparing the results By comparing the above results, we can see that
for h = 0.5,
for h = 0.2,
for h = 0.05,

the error at t = 2 is 0.003866721268688, using 4 steps


the error at t = 2 is 0.000108949842019, using 10 steps
the error at t = 2 is 0.000000442133868, using 40 steps

This brings out a question. How to find the most appropriate step size, if we want to make sure the
error is less than a given , for example, = 0.00001?
5. Adaptive step size control and the Runge-Kutta-Fehlberg method The answer is, we will
use adaptive step size control during the computation. The idea is to start with a moderate step
size. When we detect the expected error is larger than , reduce the step size and recalculate the
current step. When we detect the expected error is less than , keep the current step and slightly
enlarge the step size in the next step. This requires us to have a good estimation of the expected
error.

Heres the formula for the Runge-Kutta-Fehlberg method (RK45).


w0 =
k1 = hf (ti , wi )


k1
h
k2 = hf ti + , wi +
4
4


3h
3
9
k3 = hf ti + , wi + k1 + k2
8
32
32


12h
1932
7200
7296
k4 = hf ti +
, wi +
k1
k2 +
k3
13
2197
2197
2197


439
3680
845
k5 = hf ti + h, wi +
k1 8k2 +
k3
k4
216
513
4104


8
3544
1859
11
h
k3 +
k4 k5
k6 = hf ti + , wi k1 + 2k2
2
27
2565
4104
40
25
1408
2197
1
k1 +
k3 +
k4 k5
216
2565
4104
5
16
6656
28561
9
2
wi+1 = wi +
k1 +
k3 +
k4 k5 + k6
135
12825
56430
50
55
1
R = |wi+1 wi+1 |
h
 1/4
= 0.84
R
wi+1 = wi +

if R

keep w as the current step solution


and move to the next step with step size h

if R >

recalculate the current step with step size h

The Matlab program is given below:


function rk45
epsilon = 0.00001;
h = 0.2;
t = 0;
w = 0.5;
i = 0;
fprintf(Step %d: t = %6.4f, w = %18.15f\n, i, t, w);
while t<2
h = min(h, 2-t);
k1 = h*f(t,w);
4

k2 = h*f(t+h/4, w+k1/4);
k3 = h*f(t+3*h/8, w+3*k1/32+9*k2/32);
k4 = h*f(t+12*h/13, w+1932*k1/2197-7200*k2/2197+7296*k3/2197);
k5 = h*f(t+h, w+439*k1/216-8*k2+3680*k3/513-845*k4/4104);
k6 = h*f(t+h/2, w-8*k1/27+2*k2-3544*k3/2565+1859*k4/4104-11*k5/40);
w1 = w + 25*k1/216+1408*k3/2565+2197*k4/4104-k5/5;
w2 = w + 16*k1/135+6656*k3/12825+28561*k4/56430-9*k5/50+2*k6/55;
R = abs(w1-w2)/h;
delta = 0.84*(epsilon/R)(1/4);
if R<=epsilon
t = t+h;
w = w1;
i = i+1;
fprintf(Step %d: t = %6.4f, w = %18.15f\n, i, t, w);
h = delta*h;
else
h = delta*h;
end
end
%%%%%%%%%%%%%%%%%%
function v = f(t,y)
v = y-t2+1;
Run this program, we see that starting with h = 0.2, the program outputs
Step
Step
Step
Step
Step
Step
Step
Step
Step

0:
1:
2:
3:
4:
5:
6:
7:
8:

t
t
t
t
t
t
t
t
t

=
=
=
=
=
=
=
=
=

0.0000,
0.2000,
0.4353,
0.6766,
0.9264,
1.1902,
1.4806,
1.8537,
2.0000,

w
w
w
w
w
w
w
w
w

=
=
=
=
=
=
=
=
=

0.500000000000000
0.829299076923077
1.287432405787216
1.827289794651997
2.448301479233138
3.153049280338359
3.955581050460808
4.952039512278185
5.305486816572746

Notice the error is


|y(2) w8 | = 0.00001486603807077103

You might also like