R.

HERMAN

S O LV I N G D I F F E R E N T I A L E Q U AT I O N S
USING SIMULINK

R . L . H E R M A N - V E R S I O N D AT E : J U LY 3 , 2 0 1 6
Copyright 2016 by R. Herman

published by r. l. herman

This text has been reformatted from the original using a modification of the Tufte-book documentclass in LATEX.
See tufte-latex.googlecode.com.

solving differential equations using simulink by Russell Herman is licensed under a Creative Com-
mons Attribution-Noncommercial-Share Alike 3.0 United States License. These notes have resided at
http://people.uncw.edu/hermanr/mat361/simulink since Summer 2015.

First printing, July 2016

Contents

1 Introduction to Simulink 1
1 Solving an ODE . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2 Handling Time in First Order Differential Equations . . . . . 7
3 Working with Simulink Output . . . . . . . . . . . . . . . . . . 11
4 Printing Simulink Scope Images . . . . . . . . . . . . . . . . . 13
5 Scilab and Xcos . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6 First Order ODEs in MATLAB . . . . . . . . . . . . . . . . . . 19
Symbolic Solutions . . . . . . . . . . . . . . . . . . . . . . . . . 20
ODE45 and Other Solvers. . . . . . . . . . . . . . . . . . . . . . 20
Direction Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23

2 First Order Differential Equations 25

1 Exponential Growth and Decay . . . . . . . . . . . . . . . . . . 25
2 Newtons Law of Cooling . . . . . . . . . . . . . . . . . . . . . 27
3 Free Fall with Drag . . . . . . . . . . . . . . . . . . . . . . . . . 31
4 Pursuit Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
5 The Logistic Equation . . . . . . . . . . . . . . . . . . . . . . . 37
6 The Logistic Equation with Delay . . . . . . . . . . . . . . . . . 38
7 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40

3 Second Order Differential Equations 43

1 Constant Coefficient Equations . . . . . . . . . . . . . . . . . . 44
2 Projectile Motion . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3 The Bouncing Ball . . . . . . . . . . . . . . . . . . . . . . . . . . 48
4 Second Order ODEs in MATLAB . . . . . . . . . . . . . . . . . 50
5 Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54

4 Transfer Function and State Space Blocks 57

1 State Space Formulation . . . . . . . . . . . . . . . . . . . . . . 57
2 Transfer Functions . . . . . . . . . . . . . . . . . . . . . . . . . 58

5 Systems of Differential Equations 61

4

1 Linear Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
2 Nonlinear Models . . . . . . . . . . . . . . . . . . . . . . . . . . 63

6 Boundary Value Problems 69

7 Difference Equations 71

8 Partial Differential Equations 73

9 Index 75
 1
Introduction to Simulink

There are several computer packages for finding solutions of dif-

ferential equations, such as Maple, Mathematica, Maxima, MATLAB, etc.
These systems provide both symbolic and numeric approaches to finding
solutions. They often require a bit of coding. However, there are graphical
environments for solving problems, including differential equations. One
such environment is Simulink, which is closely connected to MATLAB. In
these notes we will first lead the reader through examples of solutions of
first and second order differential equations usually encountered in a dif-
ferential equations course using Simulink. We will then look at examples
of more complicated systems.

1.1 Solving an ODE

Simulink is a graphical environment for designing simulations of

systems. When you have access to Simulink and MATLAB, you can start
MATLAB and on the icon bar there is an icon that you can click to launch
Simulink. Alternatively, you can type simulink to bring up the Simulink
Library Browser as shown in Figure 1.1. Next, click the yellow plus to
bring up a new model. We build models by dragging and connecting the
needed components, or blocks, from groups such as the Continuous, Math
Operations, Sinks, or Sources.
As an example, we will use Simulink to solve the first order differential
equation (ODE)
dx
= 2 sin 3t 4x. (1.1)
dt
We will also need an initial condition of the form x (t0 ) = x0 at t = t0 . For
this problem we will let x (0) = 0.
dx
We can solve Equation (1.1) by integrating to formally obtain
dt
Z
x (t) = (2 sin 3t 4x (t)) dt.

We will view this as a system in which the input, x 0 = 2 sin 3t 4x, is fed
into an integrator and the output will be x (t). Generally, we have
Z
x (t) = x 0 (t) dt.

This process is depicted in Figure 1.2.

2 solving differential equations using simulink

Figure 1.1: The Simulink Library

Browser. This is where various blocks
can be found for constructing models.
[As seen in MATLAB 2015a.]

x0 R x Figure 1.2: Schematic for a general

input output system in which the block takes the
input and produces an output.

In order to carry this out, we separately insert the terms 2 sin 3t and
4x into the integration procedure. Since we do not know 4x, we take
the output from the integrator, multiply it by 4, and subtract that from
2 sin 3t. This combined set of terms is then feed back into the integrator.
This is shown schematically in Figure 1.3.

x0 R x Figure 1.3: Schematic for solving

2 sin 3t + output x 0 = 2 sin 3t 4x. The terms 2 sin 3t and
4x are fed into the integrator and x is
output.
4

The simulation in Simulink carries out this procedure and takes the
form shown in Figure 1.4. In the background Simulink uses one of
MATLABs ODE solvers, numerical routines for solving first order dif-
ferential equations, such as ode45. This system uses the Integrator block
1 dx
s to integrate , producing x (t). The Scope is used to plot the out-
Integrator
dt
put of the Integrator block, x (t).
 introduction to simulink 3

The input for the Integrator is the right side of the differential Equation
(1.1), 2 sin 3t 4x. The sine function can be provided by using the Sine
Wave block, whose parameters are set in the component. In order to get
4x, we grab the output of the Integrator (x) and boost it by changing the
Gain value to "4." Then, using the Sum component, these terms are added,
or subtracted, and fed into the integrator. That is the main idea behind
solving this system.

1 Figure 1.4: System for solving first

s order ODE dx dt = 2 sin 3t 4x as a
Sine Wave Integrator Simulink simulation.
Scope
Function
4
Gain

For this example, we implement the following steps:

Drag needed blocks into the model region [Figure 1.5.]:

Integrator block from the Continuous group;

Sum block from the Math Operations group,
Gain block from the Math Operations group,
Sine Wave block from the Math Operations group; and,
Scope block from the Sink group.

1 Figure 1.5: Add needed components to

s the model window.
Sine Wave Integrator Scope

1
Gain

Connect the output of the Sum block to the input of the Integrator
block. [Figure 1.6.]

1 Figure 1.6: Example of connecting

s two components: Align the compo-
Sine Wave Integrator Scope nents, Click on output of one and drag
1 to another. Then, release to finalize
s connection
Sine Wave Integrator Scope

Connect the Integrator to the Scope by clicking on the Integrator out-

put and dragging to the Scope until they are connected.

Right-click the Gain control and choose Flip Block under Rotate &
Flip. Double-click the Gain block and change the Gain block value
from 1 to 4. It should change on the control.
4 solving differential equations using simulink

Figure 1.7: Block Parameters for the

Sum control.

Double-click the Sum control to bring up Block Parameters as shown in

Figure 1.7 and change from |++ to |+- in order to set addition/subtraction
nodes. [Note that the symbol | is a blank node. Also, one can change
the block to rectangular form. This is sometimes useful in displaying an
overall flow direction to the model.]

Double-click the Sine Wave block and change the frequency to 3 rad/s
and the amplitude to 2. [See Figure 1.8] Set the time dropdown menu to
Use Simulation Time.

Figure 1.8: Parameters for the Sine

Wave block. Select the amplitude and
frequency desired.
 introduction to simulink 5

Connect the Gain output to the negative input of Sum and the Sine
Wave output to the positive input on the Sum control. [Note: The Gain
can be set to a negative value and connected to a + node in the Sum
block to obtain the same effect.]

To add a node to route an x value to the Gain, hold the CTRL key and
click on the Output line of the Integrator and drag towards the input
of the Gain. You can also Right-Click the line where you want the node
and drag from there to the Gain block. See Figure 1.9.

1 Figure 1.9: Add a node by right-clicking

s one the line and dragging to the input
Sine Wave Integrator Scope of a block.

1
Gain

The initial value, x (0), of x is inserted by double-clicking the Integrator

and setting the value. For this example we set x (0) = 0.

One can annotate the diagram by clicking near where labels are needed
and typing in the text box. This leads to the model in Figure 1.10.

1 Figure 1.10: Connections for First Order

s ODE model for dxdt = 2 sin 3t 4x.
Sine Wave Integrator Scope
Function
4
Gain

Save the file under a useable file name. This file can be called in
MATLAB, or one can use the run button to run the simulation.

Figure 1.11: Scope plot of the solution

of dx
dt = 2 sin 3t 4x, x (0) = 0, with
Refine Factor= 1.
6 solving differential equations using simulink

Double-click the Scope to see the solution. Figure 1.11 shows the Scope
plot after using the autoscale ( ) feature to rescale the scope view.
A little effort is needed to change the plot attributes and to import the
plots into working documents. This will be discussed in Section 1.4.

Also, one can make further changes to the system by checking the Con-
figuration Parameters under the Simulation menu item. See Figures
1.12-1.13. In particular, changing the Refine Factor can lead to smoother
solutions. The solution shown in Figure 1.11 had a setting of 1 and that
in Figure 1.14 is the result of setting the Refine Factor to 10.

Figure 1.12: System Configuration

Parameters.

Figure 1.13: Configure Data Im-

port/Export Parameters. Changing
As noted in setting the initial value, one can double-click the Integra- theRefine Factor can lead to smoother
tor block and set the initial condition. However, sometimes it is useful to solutions.
 introduction to simulink 7

externally feed the initial condition into the block. Double-click the Inte-
grator block and change the initial condition source from internal to exter-
nal. This adds another input to the block. Drag a Constant block from the
Sources group into the model, connect it to the new input, and change the
constant value to the desired initial value. This results in the simulation
shown in Figure 1.15.

Figure 1.14: Scope plot of the solution

of dx
dt = 2 sin 3t 4x, x (0) = 0, with
Refine Factor= 10.

Figure 1.15: Connections for the First

1 Order ODE model for dx
xos dt = 2 sin 3t 4x
Sine Wave showing how to provide an external
Function Integrator Scope initial value.
1
Constant
4
Gain

1.2 Handling Time in First Order Differential Equations

In this section we review the solutions of first order differential equa-

tions, separable first order differential equations and linear first order
differential equations involving explicit time dependence. The time depen-
dent functions are obtained using the Clock block and a Math Function
block. Double-clicking the Math Function block allows for the selection of
a number of common functions.
Example 1.1. Solve the initial value problem

dy 2
= y, where y(1) = 1. (1.2)
dt t
8 solving differential equations using simulink

This is a separable equation. Placing y-variables on the left and

t-variables on the right side, we have
dy 2
Z Z
= dt.
y t
Integrating both sides,

ln |y| = 2 ln |t| + C = ln t2 + C.

Exponentiating, we obtain the general solution,

y(t) = At2 ,

where A = eC .
Using the initial condition, we have the solution, y(t) = t2 .

t 1 1/t 2/t Figure 1.16: First order separable

2 dy/dt y
1 differential equation model.
u s
Clock Math Gain
Function Product Integrator Scope

y' = 2/t y, y(1)=1

Exact solution: y(t) = t2

We can set up the problem in Simulink as shown in Figure 1.16 for the
initial value problem
dy 2
= y,
dt t
where y(1) = 1. Running the simulation, we obtain the solution shown in
Figure 1.17.

Figure 1.17: Scope plot of the solution

dy
of initial value problem (1.2), dt = 2t y,
where y(1) = 1.

The solution looks like y(t) = t2 . We can verify this by plotting t2 along
with the solution t see if they are the same. Another method would be to
 introduction to simulink 9

compute the difference between the numerical and exact solution, y(t) t2 .
In order to do this, we add a Math Function block, selecting the square
function and connect it to the time route and a Sum Block. The solution is
also fed into the latter block and the difference is fed into a second Scope
Block. This is shown in Figure 1.18.

Figure 1.18: First order separable

u2 differential equation model with extra
blocks to plot the difference between
Mat Scope1 the numerical and exact solution,
Function2 y(t) t2 , for Equation (1.2).
t 1 1/t 2/t dy/dt y
2 1
u s
Clock Math Gain
Function Product Integrator Scope

y' = 2/t y, y(1)=1

Exact solution: y(t) = t2

The result of the simulation is shown in Figure 1.19. We note that this is
the numerical error, though the solution is only off by 1.4 105 over the
given interval. Considering that the solution at t = 10 is Y (10) = 100, this
is a relative error of roughly 107 . That seems perfectly acceptable.

Figure 1.19: Scope plot of the difference

between the numerical and exact
solution, y(t) t2 , for Equation (1.2)

It is simple to change the differential equation (1.2) in the previous

example to a linear first order differential equation.

dy 2
= y + t2 .
dt t
10 solving differential equations using simulink

Example 1.2. Solve the linear first order differential equation,

dy 2
= y + t2 , (1.3)
dt t
satisfying y(1) = 1.
We first rewrite Equation (1.3) in standard form,
dy 2
y = t2 . (1.4)
dt t
We can now determine the integrating factor,
 Z t 
2
(t) = exp d

= exp [2 ln t]
= t 2 .

Multiplying Equation (1.4) by the integrating factor, (t), we can

find the solution:
 
2 dy 2
t y = t 2 t 2
dt t
d  2 
t y = 1
dt
t 2 y ( t ) = t + C
y(t) = t3 + Ct2 . (1.5)

Using the initial condition, y(1) = 1, we obtain C = 0. Therefore,

the solution is y(t) = t3 .

t2 y
Figure 1.20: Linear first order differen-
u2 dy/dt 1 tial equation model.
s
t Math Integrator1 Scope1
Function2
Clock 2/t y
1 1/t 2/t 2
2 y' = 2/t y+t , y(1)=1
u
Math Gain1
Product1
Exact solution: y(t) = t3
Function1

The model for this problem is shown in Figure 1.20. Running the sim-
ulation, we obtain the numerical solution, y(t) = t3 , as shown in Figure
1.21. Computing the difference between the numerical and exact solutions
in this case, we find the error is about 6 105 .
Example 1.3. Consider the initial value problem,
dx
= 2 sin 3t 4x, x (0) = 0. (1.6)
dt
This is the example that we first solved using Simulink. It is an-
other linear first order differential equation. The integrating factor is
found to be Z 
(t) = exp 4 dt = e4t .
 introduction to simulink 11

Figure 1.21: Scope plot of the difference

between the numerical and exact
solution, y(t) t2 .

Multiplying Equation (1.6 by the integrating factor, we can obtain the

general solution:
d  4t 
e x = 2e4t sin 3t
dt Z
e4t x = 2 2e4t sin 3t dt + C
2 4t
= e (4 sin 3t 3 cos 3t) + C
25
2
x (t) = (4 sin 3t 3 cos 3t) + Ce4t . (1.7)
25
6
Using the initial condition, x (0) = 0, we find C = 25 . Therefore,
the particular solution is
2 6
x (t) = (4 sin 3t 3 cos 3t) + e4t . (1.8)
25 25
The solution can be found using Simulink. The model for this is
shown in Figure 1.22. The plot on the scope matches that in Figure
1.11.

1.3 Working with Simulink Output

Often we might want to access the solutions in MATLAB. Using

the model in Figure 1.16, add the To Workspace block. Double-click and
rename the variable as y and change the output type to array. Run the
simulation. This will put tout and y data into the MATLAB workspace.
In MATLAB you can plot the data using plot(tout,y). You can add
labels with xlabel(t), ylabel(y), title(y vs t). Adding the command
12 solving differential equations using simulink

Figure 1.22: Model for plotting the

6/25*exp(-4*u(1))
exact solution (1.8 of the initial value
Clock Fcn problem (1.6.

2/25
Sine Wave1 Scope
Gain

Sine Wave

t 1 1/t 2/t y Figure 1.23: Adding To Workspace

2 dy/dt
1 block for sending output to MATLAB.
u s
Clock Math Gain
Function Product Integrator Scope

y' = 2/t y, y(1)=1

Exact solution: y(t) = t2 y
To Workspace

set(gcf,Color,[1,1,1]) makes the plot background white. The result is

shown in Figure 1.24.

y vs t Figure 1.24: Plot of model solution in

120
MATLAB.

100

80

60
y

40

20

0
1 2 3 4 5 6 7 8 9 10
t

Once you have exported your data to the MATLAB workspace and
created a plot, then you can use the menu items under Tools to annotate
the plot. Once you are satisfied with the Figure, go to the Edit menu and
select Copy Figure. Go to your report document and Paste (CTRL-V) the
figure into your document. You can then resize the figure, center it, and
add a numbered Figure caption describing the figure. Other methods for
recording Simulink Scope images and the Simulink model are described
next.
 introduction to simulink 13

1.4 Printing Simulink Scope Images

In this section we discuss different methods for transferring the plots

generated in Simulink models to a document or report. For example, you
might want to copy images produced by the scope or your model into an
MS Word document. There are several ways you can do this. You might be
able to use the Print icon to print to a file or printer, or you can follow one
of the following methods.

Method 1:
Select the Scope figure window in Figure 1.25, then hit ALT+PrintScrn
to copy the figure to a clipboard and paste the figure into your application.

Figure 1.25: Scope plot. Note that the

plots in this section are generated by
the oscillator model in the next chapter.

You might want to change the colors before copying the scope image.
Click the Scope Parameters icon (2nd icon) and go to the Style tab as seen
in Figure 1.26. Change the Figure Color to black, Axes Colors to white
background and black writing, and Line Color to black. The selection of
these parameters is shown in Figure 1.26.
Now the Scope plot looks like Figure 1.27.

Method 2:
Go to Scope Parameters and select the History tab. Check the Save
data to workspace box. Note the variable name. Lets change the name to
MyScopeData for this example. Saving with Structure with time will save
the data as a structure. Run the simulation again.
Now, go into the MATLAB command window. You should see the
MyScopeData data in the variable list. Type

plot(MyScopeData.time, MyScopeData.signals.values)

This gives the MATLAB plot in Figure 1.29 which can be manipulated
14 solving differential equations using simulink

Figure 1.26: Scope color parameters.

Figure 1.27: Scope plot with a white

background.

Figure 1.28: History tab in Scope

parameters.

and saved or copied as an image.

Method 3:
You can save the scope image as a jpg image. Create the MATLAB code
in Table 1.1. Save this code as an m-file with a name like prfig.m. In MAT-
 introduction to simulink 15

Figure 1.29: Plot generated by Method

1
2.

0.8

0.6

0.4

0.2

-0.2
0 1 2 3 4 5 6 7 8 9 10

LAB run prfig (type prfig in the Command Window.) It should produce
the file mypic.jpg in your MATLAB folder. Of course, you can change the
name of the image before running prfig.m.

shh = get(0,ShowHiddenHandles); Table 1.1: MATLAB code for saving the

scope image as a jpg image.
set(0,ShowHiddenHandles,On)
set(gcf,PaperPositionMode,auto)
set(gcf,InvertHardcopy,off)
saveas(gcf,mypic.jpg)
set(0,ShowHiddenHandles,shh)

Now you can Insert the figure into your MS Word document as a Pic-
ture file.

Figure 1.30: Scope plot from Method 3.

Method 4:
You can add a To Workspace block to your simulation. This will auto-
matically place the data in the MATLAB space. Go to the Simulink library
16 solving differential equations using simulink

and add a To Workspace block to your model. Connect this block to the
input to the Scope (right click the input line and drag to connect to the To
Workspace block.) This will give the connection as shown in Figure 1.31.

Figure 1.31: Use of a To Workspace

block.

You can double-click this block and change the variable name that will
be saved. Lets assume it is simout. Then, run the simulation. Go into
MATLAB and type

plot(simout.time,simout.data)

This will give you a plot of the Scope data. Now you can print, save
as an image, or copy (under Edit) to an MS Word document. Below is
what you get using Copy Figure under the Edit menu item in the Figure
window.

Figure 1.32: Scope plot from Method 4.

Printing Models
Once you have made a model, you might want to include it in a report.
It is easy to capture a model, but a complicated model might not print
large enough to see the component annotations.
First open the desired model. Then, in MATLAB you can use the print
command to print the model. For example, typing the following in the
MATLAB command window prints the open model to an encapsulated
postscript file:

print -s -deps -r300 mymodel.eps

 introduction to simulink 17

For jpg files, you can use

print -s -djpeg -r300 mymodel.jpg

For other formats, consult the MATLAB help system.

1.5 Scilab and Xcos

There are alternatives to using MATLAB. One example is Xcos.

Xcos is part of Scilab. Scilab is free and open source software for numerical
computation similar to MATLAB. Xcos is a graphical design environment.
The Xcos environment is shown in Figure 1.33.

Figure 1.33: The Xcos workspace.

After downloading and installing Scilab from http://www.scilab.org/,

one can type xcos or click on the icon to launch Xcos. This brings up
the Xcos Palettes browser and Xcos workspace as shown in Figures 1.34
and 1.35. This looks similar to Simulinks Library Browser as shown in
Figure 1.1.
In Figure 1.36 we show the model for solving the first example of this
chapter:
dx
= 2 sin 3t 4x, x (0) = 0.
dt
This is equivalent to the Simulink model in Figure 1.4. We see that this
model is similar to the Simulink construction. However, there are are some
differences. First of all, the block have a different appearance.
Next, there are some differences in setting up the block parameters. The
Sum block is set up by double-clicking the block and entering the signs
and number of input ports as [1;-1]. This indicates that the Sum block has
to inputs, the first is positive and the second is negative.
18 solving differential equations using simulink

Figure 1.34: The Xcos Palette browser.

Figure 1.35: The Xcos workspace.

The scope requires an additional input. Namely the time is entered

using a clock. In Simulink this is automatic, though we had also used the
clock to introduce time as an independent variable when needed.
The initial condition and the sine function parameters are entered by
double-clicking the integrator and sine block, respectively.

Figure 1.36: The Xcos model for solving

the first order ODE dx
dt = 2 sin 3t 4x.

In order to run the simulation, one can click the play icon or select
 introduction to simulink 19

Start under the Simulation menu item. The ODE solver can be changed
through Setup under the Simulation menu. The solution is shown in
Figure 1.37.

Figure 1.37: The Xcos model solution of

dx
dt = 2 sin 3t 4x, x (0) = 0.

The blocks are not labeled like Simulink. One can label the blocks by
right-clicking and selecting Edit under the Format item. There one can
enter text to appear with the block. Annotation of the workspace is done
by selecting a Text_f block and adding text to it and changing the fontsize.
Sample annotations are shown in Figure 1.38.

Figure 1.38: The Xcos model with

annotation added.

We spent time earlier discussing how to capture images of the out-

put and models for reports. In Xcos it is a simple matter to Export the
model or the solutions by selecting Export under the File menu. There
are options for saving these to different formats. The images can also be
modified by changing the axis range, fonts, colors, etc.

1.6 First Order ODEs in MATLAB

One can use MATLAB to obtain solutions and plots of solutions of

differential equations. This can be done either symbolically, using dsolve,
or numerically, using numerical solvers like ode45. In this section we will
provide examples of using these to solve first order differential equations.
We will end with the code for drawing direction fields, which are useful
for looking at the general behavior of solutions of first order equations
without explicitly finding the solutions.
20 solving differential equations using simulink

Symbolic Solutions

The function dsolve obtains the symbolic solution and ezplot

is used to quickly plot the symbolic solution. As an example, we apply
dsolve to solve the main model in this chapter.
At the MATLAB prompt, type the following:

sol = dsolve(Dx=2*sin(t)-4*x,x(0)=0,t);
ezplot(sol,[0 10])
xlabel(t),ylabel(x), grid

The solution is given as

sol =

(2*exp(-4*t))/17 - (2*17^(1/2)*cos(t + atan(4)))/17

Figure 1.39 shows the solution plot.

(2 exp(-4 t))/17 - (2 171/2 cos(t + atan(4)))/17 Figure 1.39: The solution of Equation
(1.1) with x (0) = 0 found using
0.5
MATLABs dsolve command.
0.4

0.3

0.2

0.1

0
x

-0.1

-0.2

-0.3

-0.4

-0.5

0 1 2 3 4 5 6 7 8 9 10
t

ODE45 and Other Solvers.

There are several ODE solvers in MATLAB, implementing Runge-

Kutta and other numerical schemes. Examples of its use are in the differ-
ential equations textbook. For example, one can implement ode45 to solve
the initial value problem
dy yt
= p , y(0) = 1,
dt 2 y2
using the following code:

[t y]=ode45(func,[0 5],1);
plot(t,y)
xlabel(t),ylabel(y)
title(y(t) vs t)
 introduction to simulink 21

One can define the function func in a file func.m such as

function f=func(t,y)
f=-t*y/sqrt(2-y.^2);

Running the above code produces Figure 1.40.

y(t) vs t Figure 1.40: A plot of the solution of

1 dy
dt =
yt 2 , y(0) = 1, found using
0.9 2 y
MATLABs ode45 command.
0.8

0.7

0.6

0.5
y

0.4

0.3

0.2

0.1

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t

One can also use ode45 to solve higher order differential equations. Sec-
ond order differential equations are discussed in Section 4. See MATLAB
help for other examples and other ODE solvers.

Direction Fields

One can produce direction fields in MATLAB. For the differential

equation
dy
= f ( x, y),
dx
we note that f ( x, y) is the slope of the solution curve passing through the
point in the xy=plane. Thus, the direction field is a collection of tangent
vectors at points ( x, y) indication the slope, f ( x, y), at that point.
A sample code for drawing direction fields in MATLAB is given by
[x,y]=meshgrid(0:.1:2,0:.1:1.5);
dy=1-y;
dx=ones(size(dy));
quiver(x,y,dx,dy)
axis([0,2,0,1.5])
xlabel(x)
ylabel(y)

The mesh command sets up the xy-grid. In this case x is in [0, 2] and y
is in [0, 1.5]. In each case the grid spacing is 0.1.
We let dy = 1-y and dx =1. Thus,
dy 1y
= = 1 y.
dx 1
22 solving differential equations using simulink

The quiver command produces a vector (dx,dy) at (x,y). The slope of

each vector isdy/dx. The other commands label the axes and provides a
window with xmin=0, xmax=2, ymin=0, ymax=1.5. The result of using the
above code is shown in Figure 1.41.

Figure 1.41: A direction field produced

1.5
using MATLABs quiver function for
y0 = 1 y.

1
y

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x

One can add solution, or integral, curves to the direction field for dif-
ferent initial conditions to further aid in seeing the connection between
direction fields and integral curves. One needs to add to the direction field
code the following lines:

hold on
[t,y] = ode45(@(t,y) 1-y, [0 2], .5);
plot(t,y,k,LineWidth,2)
[t,y] = ode45(@(t,y) 1-y, [0 2], 1.5);
plot(t,y,k,LineWidth,2)
hold off

Here the function f (t, y) = 1 y is entered this time using MATLABs

anonymous function, @(t,y) 1-y. Before plotting, the hold command is
invoked to allow plotting several plots on the same figure. The result is
shown in Figure 1.42

Figure 1.42: A direction field produced

1.5
using MATLABs quiver function for
y0 = 1 y with solution curves added.

1
y

0.5

0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2
x
 introduction to simulink 23

1.7 Exercises

1. Construct the model in Figure 1.4 for solving the initial value problem
dx
dt = 2 sin 3t 4x, x (0) = 0, and produce a plot of the solution.

dx
2. Modify the model in the Problem 1. to solve = f (t) 2x for a
dt
different function, f (t) and initial condition.

3. Solve the following initial value problems using MATLAB (See Section
1.6) and Simulink. Are the solutions the same? Provide plots of the
solutions.

a. y0 = xy, y(0) = 1.
b. y0 = 2y(3 y), for different initial conditions, y(0) = 4,
y(0) = 2, and y(0) = 1.
c. y0 = 1 + x + y, y(0) = 1.
d. y0 = (y2 4)(y 4) for different initial conditions, y(0) = 5,
y(0) = 3, y(0) = 1, y(0) = 1, and y(0) = 3.

4. Use MATLAB to plot direction fields for the following:

a. y0 = xy.
b. y0 = 2y(3 y).
c. y0 = 1 + x + y.
d. y0 = (y2 4)(y 4).
 2
First Order Differential Equations

We have seen how to solve simple first order differential equations

using Simulink. In particular we have solved initial value problems for the
equations

dy 2
= y, y(1) = 1, (2.1)
dt t
dy 2
= y + t2 , y(1) = 1, (2.2)
dt t
dx
= 2 sin 3t 4x, x (0) = 0. (2.3)
dt
The Simulink models were provided in Figures 1.16, 1.20, and 1.4, respec-
tively.
In this chapter we solve a few more first order equations in the form of
applications. These will include growth and decay, Newtons Law of Cool-
ing, pursuit curves, free fall and terminal velocity, the logistic equation,
and the logistic equation with delay.

2.1 Exponential Growth and Decay

The simplest differential equations are those governing growth

and decay. As an example, we will discuss population models.
Let P(t) be the population at time t. We seek an expression for the rate
dP
of change of the population, . Assuming that there is no migration
dt
of population, the only way the population can change is by adding or
subtracting individuals in the population. The equation would take the
form
dP
= Rate In Rate Out.
dt
The Rate In could be due to the number of births per unit time and the
Rate Out by the number of deaths per unit time. The simplest forms for
these rates would be given by terms proportional to the population:

Rate In = bP and Rate Out = mP.

Here we have denoted the birth rate as b and the mortality rate as m. This
gives the total rate of change of population as

dP
= bP mP kP, (2.4)
dt
26 solving differential equations using simulink

where k = b m.
Equation (2.4) is easily modeled in Simulink. All of the needed blocks
are under the Commonly Used Blocks group. We need an Integrator, Con-
stant, Gain, and a Scope block. The output from the Integrator can be
feed into a Gain control, which represents k, and the output from the
Gain, kP, can then be used as an input to the Integrator. We add the
Scope in order to plot the solution. The model is shown in Figure 2.1.
Note that a Constant block was added to provide an external input of the
initial condition.

Figure 2.1: Simulink model for expo-

-0.5 nential growth and decay. The initial
P' value, P(0) = 10, is set in the Constant
1 P block and k = 0.5 is set in the Gain.
Gain xos
Integrator Scope
10
Constant

The solution for exponential decay with P(0) = 10 and k = 0.5 is

shown in Figure 2.2. The simulation time was set at 10s.

Figure 2.2: Solution for the exponential

decay with P(0) = 10 and k = 0.5.
The simulation time was set at 10.

The exact solution is easily found noting that Equation (2.4) is a separa-
ble equation. Rearranging the equation, its differential form is
dP
= k dt.
P
Integrating, we have
dP
Z Z
= k dt
P
 first order differential equations 27

ln | P| = kt + C. (2.5)

Next, we solve for P(t) through exponentiation,

| P(t)| = ekt+C
P(t) = ekt+C
= Aekt . (2.6)

Here we define the arbitrary constant, A = eC .

If the population at t = 0 is P0 , i.e., P(0) = P0 , then the solution gives
P(0) = Ae0 = A = P0 . So, the solution of the initial value problem is

P(t) = P0 ekt .

In the Simulink model, the initial value was given as P(0) = 10 and k =
0.5. Therefore, the solution in Figure 2.2 is of the function P(t) = 10e0.5t .
Equation (2.4) is the familiar exponential model of population growth: Malthusian population growth.

dP
= kP.
dt
We obtained solutions exhibiting exponential growth (k > 0) or decay
(k < 0). This Malthusian growth model has been named after Thomas
Robert Malthus (1766-1834), a clergyman who used this model to warn
of the impending doom of the human race if its reproductive practices
continued. Later we modify this model to account for competition for
resources, leading to the logistic differential equation.

2.2 Newtons Law of Cooling

If you take your hot cup of tea, and let it sit in a cold room, the tea
will cool off and reach room temperature after a period of time. The law
of cooling is attributed to Isaac Newton (1642-1727) who was probably
the first to state results on how bodies cool.1 The main idea is that a body 1
Newtons 1701 Law of Cooling is
at temperature T (t) is initially at temperature T (0) = T0 . It is placed in an approximation to how bodies cool
for small temperature differences
an environment at an ambient temperature of Ta . The goal is to find the (T Ta  T) and does not take into ac-
temperature at a later time, T (t). count all of the cooling processes. One
account is given by C. T. OSullivan,
We will assume that the rate of change of the temperature of the body Am. J. Phys (1990) p 956-960.
is proportional to the temperature difference between the body and its
surroundings. Thus, we have

dT
T Ta .
dt
The proportionality is removed by introducing a cooling constant,

dT
= k( T Ta ), (2.7)
dt
where k > 0.
28 solving differential equations using simulink

This differential equation can be solved by first rewriting the equations

as
d
( T Ta ) = k( T Ta ).
dt
This now takes the form of exponential decay of the function T (t) Ta .
The solution is easily found as

T (t) Ta = ( T0 Ta )ekt ,

or
T (t) = Ta + ( T0 Ta )ekt .
Example 2.1. A cup of tea at 90o C cools to 85o C in ten minutes. If the
room temperature is 22o C, what is its temperature after 30 minutes?
Using the general solution with T0 = 90o C,

T (t) = 22 + (90 22)ek = 22 + 68ekt ,

we then find k using the given information, T (10) = 85o C. We have

85 = T (10)
= 22 + 68e10k
63= 68e10k
63
e10k = 0.926
68
10k = ln 0.926
ln 0.926
k =
10
0.00764min1 .

This gives the solution for this model as

T (t) = 22 + 68e0.00764t .

Now we can answer the question. What is T (30)?

T (30) = 22 + 68e0.00764(30) = 76o C.

Figure 2.3: Simulation model for New-

0.1 T' tons Law of Cooling, T 0 = k ( T Ta ),
-1
T (0) = T0. Here we set k = 0.1 s1 ,
k 1 T
Product Gain Ta = 20o C, and T0 = 60o C.
s
60 xo Scope

T0
Integrator

20
Newton's Law of Cooling
Ta T' = - k (T-T0)

Next we model Equation (2.7) in Simulink. The input for the integrator
is simply k( T Ta ). We need to define the constants k and Ta . We will
 first order differential equations 29

externally input the initial condition, T (0) = T0 . The simple model is

shown in Figure 2.3. In this case we set k = 0.1 s1 , Ta = 20o C, and
T0 = 60o C. Running the simulation for 100 s, we obtain the solution shown
in Figure 2.4.

Figure 2.4: Solution of Newtons Law of

Cooling example.

How good is the solution? We can solve the problem by hand for this
set of parameters. However, we will take this opportunity to introduce
the idea of a subsystem and set up a model in which we can interactively
modify the constants and get Simulink to automatically provide the exact
solution for comparison.

Figure 2.5: Creating a subsystem for the

1 T' Newtons Law of Cooling model.
k
-1
Product 1 T
Gain 1
s
T(t)
2 xo
T(0)
Integrator

3
Newton's Law of Cooling
T ambient
T' = - k (T-T0)

We begin by replacing the scope with an output block. The Out1 block
can be found in the Sink group. The input to the subsystem will be the Creating a subsystem.
three parameters, k, T0 , and Ta . Each of these constant blocks in Figure 2.3
will be replaced by an In1 block, found in the Sources group. In Figure 2.5
the three inputs and one output are now oval blocks.
Double-click each of the three input blocks, one at a time, and set the
Port Number of k, T0 , and Ta , to 1, 2, and 3, respectively. Finally, rename
each of these controls using the labels that make sense, such as k for k. In
Figure 2.5 we show the subsystem that we have created.
Now highlight the entire subsystem using CTRL-A. In the menu sys-
30 solving differential equations using simulink

Figure 2.6: Subsystem for Newtons

1 k
Law of Cooling, T 0 = k( T Ta ),
k T (0) = T0.
2 T(0) T(t) 1
T(0) T(t)
3 T ambient
T ambient
Subsystem

tem, look for Create Subsystem from Selection. In the 2015 version, this
is under the menu item Diagram and subitem Subsystem & Model Ref-
erence. Rearranging the resulting subsystem, one has something like the
subsystem block in Figure 2.6. This is the equivalent of a black box with
three inputs and one output.
Next, we can make use of the subsystem just created. Replace the three
input ports with constant blocks. Rename the Constant blocks with the
parameter name and fill each block with a value. The output port can be
replaced with a Scope block, or any other form of output desired. This can
be seen in Figure 2.7.
Before finishing with this model, we will build in the exact solution.
Recall that the general solution can be written in terms of the parameters
as
T (t) = Ta + ( T0 Ta )ekt .
So, we can feed the values of the parameters in the model into a Fcn block
and output the exact solution for comparison. We will also need a time
value. So, we will need the Clock block as well.

Newton's Law of Cooling Figure 2.7: Using a user-created subsys-

T' = - k (T-T0)
tem for Newtons Law of Cooling.
0.1 k

60 T(0) T(t)

IC Scope

20 T ambient

Ta
Cooling System

The entire model is shown in Figure 2.8. The subsystem is labeled Cool-
ing System The top portion is a repetition of the Newtons Law of Cooling
model implemented previously.
We have added a Fcn block from the User-Defined Functions group.
The input will be a vector containing all of the variables in the exact solu-
tion. This is accomplished by adding a Mux (or Multiplex) block. Double-
click the Mux block and set the number of inputs to 4.
Now, double-click the Fcn block and enter the exact solution in the form

u(1)+u(2)*exp(-u(3)*u(4))

Here we have assumed that the variables are fed into the Mux block in the
order Ta , T0 Ta , k, and t. In Figure 2.8 one can see how the values are
 first order differential equations 31

Newton's Law of Cooling Figure 2.8: Model of Newtons Law of

T' = - k (T-T0)
Cooling, T 0 = k ( T Ta ), T (0) = T0,
0.1 k using the subsystem feature.
k

60 T(0) T(t)

IC Scope

20 T ambient

Ta
Cooling system

Exact Solution

f(u)
Fcn Scope1

Clock

routed into the Mux block.

The output can be attached to a second scope, as shown, or can be sub-
tracted from the output of the Cooling System block to show the closeness
of the two solutions. One can also send the output to MATLAB using the
To Workspace block.

2.3 Free Fall with Drag

Consider an object falling to the ground with air resistance? Free

fall is the vertical motion of an object solely under the force of gravity. It
has been experimentally determined that an object near the surface of the
Earth falls at a constant acceleration in the absence of other forces, such
as air resistance. This constant acceleration is denoted by g, where g is
called the acceleration due to gravity. The negative sign is an indication
that we have chosen a coordinate system in which up is positive.
We are interested in determining the position, y(t), of a falling body as
a function of time. The differential equation governing free fall is have

y (t) = g. (2.8)

Note that we will occasionally use a dot to indicate time differentiation.

We need to model the air resistance. As an object falls faster and faster,
the resistive force becomes greater. This drag force is a function of the
velocity. The idea is to write Newtons Second Law of Motion F = ma in
the form
my = mg + f (v), (2.9)

where f (v) gives the resistive force and mg is the weight. Note that this
applies to free fall near the Earths surface. Also, for f (v) to be a resis-
tive force, f (v) should oppose the motion. If the body is falling, then f (v)
32 solving differential equations using simulink

should be positive. If the body is rising, then f (v) would have to be nega-
tive to indicate the opposition to the motion.
We will model the drag as quadratic in the velocity, f (v) = bv2 .
Example 2.2. Solve the free fall problem with f (v) = bv2 .
The differential equation that we need to solve is

v = kv2 g, (2.10)

where k = b/m. Note that this is a first order equation for v(t).
Formally, we can separate the variables and integrate over time to
obtain Z v
dz
t+C = . (2.11)
kz2 g
If we can do the integral, then we have a solution for v. One way to
evaluate this integral is to use Partial Fraction Decomposition.
When there are two linear factors in the denominator, the integral
can be rewritten as
Z  
dx 1 1 1
Z
= dx (2.12)
( x a)( x b) ba xa xb
The new integral has two terms which can be readily integrated.
In order to factor the denominator in the current problem, we
first have to rewrite the constants. We let 2 = g/k and write the
integrand as
1 1 1
= . (2.13)
kz2 g k z2 2
Now we use a partial fraction decomposition to obtain
 
1 1 1 1
= . (2.14)
kz2 g 2k z z +
The integrand can be easily integrated giving
 
1 v 
t+C = ln  . (2.15)
2k  v + 
Solving for v, we have

1 Ae2kt
v(t) = , (2.16)
1 + Ae2kt
where A eC . A can be determined using the initial velocity.
There are other forms for the solution in terms of a tanh function, which
the reader can determine as an exercise. One important conclusion is that
forqlarge times, the ratio in the solution approaches 1. Thus, v =
gk as t . This means that the falling object will reach a constant
terminal velocity.
Equation (2.10) can be modeled in Simulink. The model is shown in
Figure 2.9. The solution for k = 0.00159m1 , which is found for the above
sample computation, is shown in Figure 2.10. Weq see that terminal velocity
g
is obtained and matches the predicted value, k = 78 m/s.
 first order differential equations 33

v' Figure 2.9: Model for free fall with drag

9.8 1 v as described by v = kv2 g.
xos
g
0 Integrator

Constant1
Scope
0.00159 u2

k Math Free Fall with Drag

Function v' = kv^2 - g

Figure 2.10: Solution for free fall with

drag with k = 0.00159 starting from
rest.

2.4 Pursuit Curves

Another application that is interesting is to find the path that a

body traces out as it moves towards a fixed point or another moving body.
Such curves are know as pursuit curves. These could model aircraft or
submarines following targets, or predators following prey. For example, a
hawk follows a sparrow, a large fish chases a small fish, or a fox chases a
rabbit.
Example 2.3. A dog at point ( x, y) sees a cat traveling at speed v
along a straight line. The dog runs towards the cat at constant speed
w but always in a direction along line of sight between their posi-
tions. If the dog starts out at the point (0, 0) at t = 0, when the cat is
at ( a, 0), then what is the path the dog needs to follow? Will the dog
catch the cat?
We show the path in Figure 2.12. Let the cats path be along the
34 solving differential equations using simulink

y Figure 2.11: A dog at point ( x, y) sees a

cat at point ( a, vt) and always follows
the straight line between these points.

( a, vt)

( x, y)

x
( a, 0)

line x = a. Therefore, the cat is at position ( a, vt) at time t. The goal is

to find the dogs path, ( x (t), y(t)), or y = y( x ).
First we consider the equation of the line of sight between the
points ( x, y) and ( a, vt). Considering that the slope of this line is the
same as the slope of the tangent to the path, y = y( x ), we have
vt y
y0 = .
ax
The dog is moving at a constant speed, w and the distance the dog
to travels s given by L = wt, where t is the running time from the
origin. The distance the dog travels is also given by the arclength of
the path between (0, 0) and ( x, y) :
Z xq
L= 1 + [y0 ( x )]2 dx.
0
vty
Eliminating the time using y0 = a x , we have
Z xq
w
1 + [y0 ( x )]2 dx = (y + ( a x )y0 ).
0 v
Furthermore, we can differentiate this result with respect to x to get
rid of the integral,
w
q
1 + [y0 ( x )]2 = ( a x )y00 . (2.17)
v
This is the differential equation governing the dogs pursuit. It is
modeled later in Simulink as shown in Figure 2.12.
Even though Equation (2.17)is a second order differential equation
for y( x ), it is a first order separable equation in the speed function
z( x ) = y0 ( x ). Namely,
w vp
( a x )z0 = 1 + z2 .
v x
Separating variables, we find
w dz dx
Z p Z
= ln(z + 1 + z2 ) .
v 1 + z2 ax
 first order differential equations 35

Figure 2.12: Model forp the pursuit

curve, ( a x )y00 = wv 1 + [y0 ( x )]2 ,
sqrt(1+u^2)
y(0) = 0, y0 (0) = 0, for w = 2 and
Fcn v = 1.
1
v

2 1 y'
xos 1
w xos
Integrator
Integrator1 Scope
0 0
Clock
Product y'(0) y(0)
9.000001
Constant

5.997
u(3)*u(2)*u(1)/(u(1)^2-u(2)^2) 6
Display
Fcn1

The integrals can be computed using standard methods from calcu-

lus. We can easily integrate the right hand side,

dx
Z
= ln | a x | + c1 .
ax
The left hand side takes a little extra work, or looking the value up
in Tables or using a CAS package. Recall a trigonometric substitution
is in order. We let z = tan . Then dz = sec2 d. The methods
proceeds as follows:

dz sec2
Z Z
= d
1 + z2
Z
1 + tan2
= sec d
= ln(tan + sec ) + c2
p
= ln(z + 1 + z2 ) + c2 . (2.18)

Putting these together, we have for x > 0,

p v
ln(z + 1 + z2 ) = ln x + C.
w
Using the initial condition z = y0 = 0 and x = a at t = 0,
v
0= ln a + C,
w
or C = wv ln a.
Using this value for c, we find
p v v
ln(z + 1 + z2 ) = ln x ln a
w w
p v x
ln(z + 1 + z2 ) = ln
w a
36 solving differential equations using simulink

p x v
w
ln(z + 1 + z2 ) = ln
a
p x v
w
z+ 1 + z2 = . (2.19)
a
We can solve for z = y0 , to find

1  x  wv  x  wv
 
0
y =
2 a a
Integrating,
"
v
x 1+ w

v
x 1 w
#
a a a
y( x ) = + k.
2 1 + wv 1 wv

The integration constant, k, can be found knowing y( a) = 0. This

gives
 
a 1 1
0 = +k
2 1 + wv 1 wv
 
a 1 1
k =
2 1 wv 1 + wv
avw
= . (2.20)
w2 v2
The full solution for the path is given by
"
1+ v
v#
x w x 1 w
a a avw
y( x ) = a v + 2 .
2 1 + wv 1 w w v2

Can the dog catch the cat? This would happen if there is a time
when y(0) = vt. Inserting x = 0 into the solution, we have y(0) =
avw
w2 v2
= vt. This is possible if w > v.

Figure 2.13: Solution for the pursuit

curve.
 first order differential equations 37

2.5 The Logistic Equation

In this section we will explore a nonlinear population model. Typ-

ically, we want to model the growth of a given population, y(t), and the
differential equation governing the growth behavior of this population is
developed in a manner similar to that done in the section on growth and
decay. Recall that a simple population model

dy
= by my, (2.21)
dt
where we had defined the birth rate as b and the mortality rate as m.
Generally, these rates could depend on the time. In the case that they
are both constant rates, we can define k = b m and obtain the famil-
iar exponential model of population growth. When populations get large
enough, there is competition for resources, such as space and food, which
can lead to a higher mortality rate. Thus, the mortality rate may be a func-
tion of the population size, m = m(y). The simplest model would be a
+ cy. Then, the previous exponential model The logistic model was first published
linear dependence, m = m
in 1838 by Pierre Franois Verhulst
takes the form (1804-1849) in the form
dy
= ky cy2 , (2.22)
 
dN N
dt = rN 1 ,
dt K
where k = b m. This is known as the logistic model of population growth. where N is the population at time t, r is
Typically, c is small and the added nonlinear term does not really kick in the growth rate, and K is what is called
the carrying capacity. Note that in this
until the population gets large enough.
model r = k = Kc.
Example 2.4. Show that Equation (2.22) can be written in the form

z0 = kz(1 z)

which has only one parameter.

We carry this out be rescaling the population, y(t) = z(t), where
is to be determined. Inserting this transformation, we have

y0 = ky cy2
z0 = kz c2 z2 ,

or  c 
z0 = kz 1 z .
k
Thus, we obtain the result, z0 = kz(1 z), if we pick = kc .
The point of this derivation is to show that there is only one free param-
eter, k, and that many combinations of c and k in the original problem lead
to essentially the same solution shape.
We can model the logistic equation, y0 = ry(1 y), with r = 1 and
y(0) = 0.1 in Simulink. The model is shown in Figure 2.14. Running te
model gives the solution in Figure 2.15. It shows the typical sigmoidal
curve bounded by the solutions y = 0 and y = 1.
 38 solving differential equations using simulink

Logistic Equation Figure 2.14: Logistic equation, y0 =

ry(1 y).
y' y
1 1
s
Gain Integrator Scope

Product 1-y
y' = r y (1-y)
1
Constant

Figure 2.15: Solution to the logistic

equation, y0 = ry(1 y), with r = 1 and
y(0) = 0.1.

2.6 The Logistic Equation with Delay

Sometimes the rate of change does not immediately take place when
the system changes. This can be modeled using differential-delay equa-
tions. For example, when the resources are being depleted, the effects
might be delayed. So, a possible model would be the logistic equation with
delay,
y0 = ry(t)(1 y(t )),
where is a fixed delay time.
The problem with trying to solve this model at time t is that we need
to know something about the solution for earlier times, y(t ). One way
to tackle the problem is to specify the solution for times [0, ] and then
to solve the equation with delay using this starting value. So, if y = 2
 first order differential equations 39

initially, we could let y = 2 for [0, ].

2
y' 1 y
Alpha >= 1
s
Integrator Scope
Clock
Switch to enter y=2
for t<=1
0
Solve y'=0
y

y(1-y(t-1)) 1-y(t-1) y(t-1)

Product
Logistic Equation with Delay Transport
Delay
y' = alpha y(t)(1-y(t-1)) 1
Constant
Figure 2.16: Logistic equation with
delay, y0 = ry(t)(1 y(t ))
The Simulink model is shown if Figure 2.16. A Switch block is used to
specify the starting values for times up to = 1. Then, the differential
equation solver takes over with a Delay block used to enter the delay term.
This model produces the solution 2.17.

Figure 2.17: Logistic equation with

delay, y0 = ry(t)(1 y(t ))
 40 solving differential equations using simulink

2.7 Exercises

1. Model the following first order differential equations in Simulink and

find the solutions for different initial conditions.
dy ex
a. = .
dx 2y
dy
b. = y2 (1 + t2 ).
dt p
dy 1 y2
c. = .
dx x
0
d. xy = y(1 2y).
e. y0 (sin x )y = sin x.
f. xy0 2y = x2 .
ds
g. + 2s = st2 .
dt
h. x 0 2x = te2t .
dy
i. + y = sin x,.
dx
dy 3
j. y = x3 .
dx x
2. Consider the case of free fall with a damping force proportional to the
velocity, f D = kv with k = 0.1 kg/s.

a. Using the correct sign, consider a 50 kg mass falling from rest

at a height of 100m. Find the velocity as a function of time.
Does the mass reach terminal velocity?
b. Let the mass be thrown upward from the ground with an
initial speed of 50 m/s. Find the velocity as a function of time
as it travels upward and then falls to the ground. How high
does the mass get? What is its speed when it returns to the
ground?

3. A paratrooper, 322 lbs including munitions, jumps from 10,000 ft.

Model this free fall with air resistance f (v) = 15v2 in Simulink. First,
write down the free fall equation. Use the model to solve for v(t). Is
there a terminal velocity? Find the time to land and the impact velocity.

4. Model the following problem in Simulink: The temperature inside your

house is 70o F and it is 30o F outside. At 1:00 A.M. the furnace breaks
down. At 3:00 A.M. the temperature in the house has dropped to 50o F.
Assuming the outside temperature is constant and that Newtons Law
of Cooling applies, determine when the temperature inside your house
reaches 40o F.

5. Model the following problem in Simulink: A body is discovered during

a murder investigation at 8:00 P.M. and the temperature of the body is
70o F. Two hours later the body temperature has dropped to 60o F in a
room that is at 50o F. Assuming that Newtons Law of Cooling applies
 first order differential equations 41

and the body temperature of the person was 98.6o F at the time of death,
determine when the murder occurred.
 3
Second Order Differential Equations

We now turn to second order differential equations. Such

equations involve the second derivative, y00 ( x ). Lets assume that we can
write the equation as

y00 ( x ) = F ( x, y( x ), y0 ( x )).

We would like to solve this equation using Simulink. This is accomplished

using two integrators in order to output y0 ( x ) and y( x ).

y0 R y Figure 3.1: Basic schemes for using

(a) input output Integrator blocks for solving second
order differential equations.
y00 R y0
(b) input output

y00 R y0 R y
(c) input output

As shown in Figure 3.1(b), sending y00 ( x ) into the Integrator block, we

get out y0 ( x ). This is similar to using y0 ( x ) to get y( x ) in Figure 3.1(a). As
shown in Figure 3.1(c), combining two Integrator blocks, we can input
y00 ( x ) = F ( x, y, y0 ) and get out y and y0 . Feeding this output into F ( x, y, y0 ),
we then obtain a model for solving the second order differential equation.
The general schematic for solving an initial value problem of the form
y00 = F ( x, y, y0 ), y(0) = y0 , y0 (0) = v0 , is shown in Figure 3.2.

y00 R y0 R y Figure 3.2: This is a general schematic

output for solving an initial value problem of
the form y00 = F ( x, y, y0 ), y(0) = y0 ,
y 0 (0) = v0 .
y 0 (0) y (0)

F(x, y, y0 )

In this chapter we will demonstrate the modeling of second order con-

stant coefficient differential equations and show some simple applications.
 44 solving differential equations using simulink

3.1 Constant Coefficient Equations

We can solve second order constant coefficient differential

equations using a pair of integrators. An example is displayed in Figure
3.3. Here we solve the constant coefficient differential equation

ay00 + by0 + cy = 0

by first rewriting the equation as

b c
y00 = F (y, y0 ) = y0 y.
a a
Example 3.1. Model the initial value problem

y00 + 5y0 + 6y = 0, y(0) = 0, y0 (0) = 1,

in Simulink.
The simulation in Figure 3.3 solves the equation

y00 + 5y0 + 6y = 0

with appropriate initial conditions. There are two integrators. One

integrates the first input, y00 , and the other integrates the output of
the first integrator, y0 , giving an output of y. Each Integrator block
needs an initial condition. The first takes y0 (0) = 1 and the second
needs y(0) = 0.

Second Order Constant Coefficient ODE Figure 3.3: Second Order Constant
Coefficient ODE.
y'' 1 y' 1 y
s s
Integrator Integrator1 Scope

b/a y'
5
b/a

c/a y
6
c/a

The outputs, y and y0 are multiplied by the appropriate constants

using a Gain block. They are then combined to form the input,
F (y, y0 ) = 5y0 6y, to the integrators. Running the simulation
for 5 units of time, the Scope gives the solution shown in Figure 3.4.
Example 3.2. Model the forced, damped harmonic oscillator.
A simple application is the forced, damped harmonic oscillator.
Recall that this is modeled using a second order, constant coefficient
equation,
mx 00 + cx 0 + kx = F (t)
 second order differential equations 45

y(x) vs x Figure 3.4: Solution plot for y00 + 5y0 +

0.15
6y = 0, y(0) = 0, y0 (0) = 1.

0.1
y

0.05

0
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
x

for some driving force F (t). Rewriting the equation, we have

1 c k
x 00 = F (t) x 0 x.
m m m

Forced, damped oscillator Figure 3.5: Forced, damped oscillator.

Scope
1
Step 1/m
x'' 1 x' 1 x
s s
Integrator Integrator1 Scope1

0.5
c/m

XY Graph
2
k/m

This suggests a model like that shown in Figure 3.5. In this exam-
ple the forcing term was taken as a step function.
(
0, t < 1,
F (t) =
1, t 1.

The step function parameters are set to start at F = 0 and is increased

to a constant value of F = 1 after t = 1. The constants are given as
m = 1.0 kg, c = 0.5 kg/s, and k = 2.0 N/m.
In Figure 3.6 is shown the solution plot for the forced, damped,
harmonic oscillator model with initial values of x (0) = 1 and x 0 (0) =
0. In this model there is also an XY Graph block. The position and
velocity data is fed into this block and the output is a plot of the
solution in the phase plane. This is shown in Figure 3.7.
 46 solving differential equations using simulink

Figure 3.6: Scope plot of the solution of

the forced, damped, harmonic oscillator
model.

Figure 3.7: XY Graph output for the so-

lution of the forced, damped, harmonic
oscillator model.

3.2 Projectile Motion

Another example is that of projectile motion. This is a system of

equations or a single equation for a vector function. Let the position vector
for the projectile be given by r = [ x, y]. Then, the projectile satisfies the
second order equation r00 = g We can solve this using two integrators
and setting up the system with a two component vector.
To make things more interesting, we can add a drag force. Thus, we
solve the system
r00 = g kvv.
The magnitude of the drag is proportional to v2 . If the projectile is moving
 second order differential equations 47

directly upward, the drag is negative, opposing the motion. The model
will need functions to compute the speed, v, and will need two integrators
with appropriate initial position and velocity. The gravitational force will
also be provided with a constant block. This model is shown in Figure 3.8
The model is done in British units (foot-pound-second). The initial
position is [0, 4] ft and the initial velocity is [80, 80] ft/s. The gravitational
constant is g = [0, 32] ft/s2 . The value of the drag coefficient does not
show in the figure. It can be made to show if the Gain block is resized.
The position and speed vs time plots are shown in Figure 3.9. Note that
changing the simulation time is one way to only display the time that the
mass is above y = 0. Also, the plot of speed shows that the speed is always
positive.
Also shown in this model is the use of the XY Graph block. This takes
two inputs in order to plot the path y vs x. XY Graphs automatically plot
when the simulation is run, as opposed to the Scope plots, which need to
be double-clicked to show the plots. One needs to double-click the block to
change the scale shown. For this model the output is shown in Figure 3.10.
This plot is useful for determining the maximum height and range of the
projectile.

Projectile Motion Figure 3.8: Projectile motion.

[x'',y'']
1 [x',y']
xos 1 [x,y]
xos
Integrate x''
Integrate x'
XY Graph
[80,80] [0,4]
Initial Velocity Initial Position
y vs t

u
Product Dot Product Sqrt v vs t

gravitational
[0,-32] acceleration -K-

Drag Coefficient

3.3 The Bouncing Ball

As seen in the projectile motion model output in Figure 3.9, the

projectile may not stop when it reaches the ground. One needs a way
to determine when this has happened and reverse the direction of the
motion. In this section we will look at a simpler model in which a ball
goes through free fall and bounces when it reaches the ground.
48 solving differential equations using simulink

Figure 3.9: Output of the Scope Blocks

for the projectile motion model.

Figure 3.10: Output of the XY Graph

block for the projectile motion model.

The ball satisfies the second order equation x 00 = g. Noting that the
velocity is v = x 0 , this can be written as two first order equations,

x 0 = v,
v0 = g. (3.1)

This system of equations can be then be put into matrix form,

" # " #" # " #
d x 0 1 x 0
= +
dt y 0 0 y g

Free Fall Figure 3.11: Free fall model.

x'=v [0;-9.8]
v'=-g
Acceleration [x,v]' x
[0 1;0 0]* u 1
[x,v]
[3;0] xos Scope
v
Gain Constant Integrator
Terminator
 second order differential equations 49

This system can be used to produce the Simulink model in Figure 3.11,
where we have introduced initial conditions x (0) = 3 and v(0) = 0. Here
the 2 2 matrix is entered in the gain and the acceleration term is added
separately. In order to plot the position vs time, we put a Demux block to
separate out the components of the state vector and added a Terminator
block to terminate one of the branches.

Figure 3.12: Output from the free fall

model.

The output of the simulation, which was run for a time of 1 second is
shown in Figure 3.12. Note that the ball has fallen below ground level.
We wish to allow for the ball to bounce from the ground. We will need to
test to see when x 0 and v 0. This is accomplished by adding some
conditions to the Integrator block.
Double-click the Integrator block and set the External reset to rising.
This will add a third input as shown in Figure 3.13. Then, replace the
initial condition Constant block with an IC block. This is found in the
Signal Attributes group. It looks like the IC block in Figure 3.13.

Figure 3.13: Modified Integrator and IC

blocks.

Next, we need to enter the conditions determining when the block hits
the ground and change the block velocity. The input to the condition con-
sist of the Boolean condition, (u[1]<=0)&&(u[2]<0), and the new position
and velocity. Here u[1] and u[2] are the position and velocity components.
We set the position as u[1] and the velocity as -0.8*u[2]. These expressions
are entered using Fcn blocks from the User-Defined Function group. This
 50 solving differential equations using simulink

model is shown in Figure 3.14 with the needed connections to the Fcn
blocks and the Integrator block. This output is shown in Figure 3.15.

Bouncing Ball
x'=v
v'=-g [0;-9.8]
Integrator
Acceleration
[x,v]'
[0 1;0 0]* u
x
1 [x,v] Scope
(u[1]<=0)&&(u[2]<0) v
Gain s
Fcn
[3;0] Terminator
xo
u[1]
Fcn1
IC
-0.8*u[2]
Fcn2

Figure 3.14: Bouncing ball.

Figure 3.15: Output from thebouncing

ball model.

3.4 Second Order ODEs in MATLAB

We can also use ode45 to solve second and higher order differential
equations. The key is to rewrite the single differential equation as a system
of first order equations. Consider the simple harmonic oscillator equation,
x + 2 x = 0. Defining y1 = x and y2 = x,
and noting that

x + 2 x = y 2 + 2 y1 ,
 second order differential equations 51

we have

y 1 = y2 ,
y 2 = 2 y1 .

Furthermore, we can view this system in the form y = y. In particular,

we have " # " #
d y1 y1
=
dt y2 2 y2
Now, we can use ode45. We modify the code slightly from Chapter 1.

[t y]=ode45(func,[0 5],[1 0]);

Here [0 5] gives the time interval and [1 0] gives the initial conditions

y1 (0) = x (0) = 1, y2 (0) = x (0) = 1.

The function func is a set of commands saved to the file func.m for
computing the righthand side of the system of differential equations. For
the simple harmonic oscillator, we enter the function as

function dy=func(t,y)
omega=1.0;
dy(1,1) = y(2);
dy(2,1) = -omega^2*y(1);

There are a variety of ways to introduce the parameter . Here we simply

defined it within the function. Furthermore, the output dy should be a
column vector.
After running the solver, we then need to display the solution. The
output should be a column vector with the position as the first element
and the velocity as the second element. So, in order to plot the solution as
a function of time, we can plot the first column of the solution, y(:,1), vs t:

plot(t,y(:,1))
xlabel(t),ylabel(y)
title(y(t) vs t)

The resulting solution is shown in Figure 3.16.

We can also do a phase plot of velocity vs position. In this case, one can
plot the second column, y(:,2), vs the first column, y(:,1):

plot(y(:,1),y(:,2))
xlabel(y),ylabel(v)
title(v(t) vs y(t))

The resulting solution is shown in Figure 3.17.

Finally, we can plot a direction field using a quiver plot and add solu-
tion curves using ode45. The direction field is given for = 1 by dx=y
and dy=-x.
52 solving differential equations using simulink

y(t) vs t Figure 3.16: Solution plot for the simple

1
harmonic oscillator.
0.8

0.6

0.4

0.2

0
y

-0.2

-0.4

-0.6

-0.8

-1
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
t

v(t) vs x(t) Figure 3.17: Phase plot for the simple

1
harmonic oscillator.
0.8

0.6

0.4

0.2

0
v

-0.2

-0.4

-0.6

-0.8

-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
x

clear
[x,y]=meshgrid(-2:.2:2,-2:.2:2);
dx=y;
dy=-x;
quiver(x,y,dx,dy)
axis([-2,2,-2,2])
xlabel(x)
ylabel(y)

hold on
[t y]=ode45(func,[0 6.28],[1 0]);
plot(y(:,1),y(:,2))
hold off

The resulting plot is given in Figure 3.18.

 second order differential equations 53

Figure 3.18: Phase plot for the simple

2
harmonic oscillator.
1.5

0.5

0
y

-0.5

-1

-1.5

-2
-2 -1.5 -1 -0.5 0 0.5 1 1.5 2
x

3.5 Exercises

1. Model the following initial value problems in Simulink and compare

solutions to those using ode45.

a. y00 9y0 + 20y = 0, , y(0) = 0, y0 (0) = 1.

b. y00 3y0 + 4y = 0, y(0) = 0, y0 (0) = 1.
c. 8y00 + 4y0 + y = 0, y(0) = 1, y0 (0) = 0.
d. x 00 x 0 6x = 0 for x = x (t), x (0) = 0, x 0 (0) = 1.

2. Model the given equation in Simulink for an appropriate initial condi-

tion and plot the solution. Analytically determine and plot the solution
and compare to the model solution.

a. y00 3y0 + 2y = 10.

b. y00 + 2y0 + y = 5 + 10 sin 2x.
c. y00 5y0 + 6y = 3e x .
d. y00 + 5y0 6y = 3e x .
e. y00 + y = sec3 x.
f. y00 + y0 = 3x2 .
g. y00 y = e x + 1.

3. Consider the model in Figure 3.19. Fill in the question marks with the
correct expression at that point in the computation. What differential
equation is solved by this simulation?

4. Model the given equation in Simulink for an appropriate initial condi-

tion and plot the solution. Analytically determine and plot the solution
and compare to the model solution.

a. x2 y00 + 3xy0 + 2y = 0.
b. x2 y00 3xy0 + 3y = 0, y(1) = 1, y0 (1) = 0.
c. x2 y00 + 5xy0 + 4y = 0.
54 solving differential equations using simulink

What does this model solve? Figure 3.19: Mystery model for Problem
4.
?
1 ?
0 xos 1 ?
1 xos
y'(0) Integrator
y(0) Integrator1 Scope

3 ?

b Divide Clock
?
u2
10
Math
c Divide1 Function

d. x2 y00 2xy0 + 3y = 0, y(1) = 3, y0 (1) = 0.

e. x2 y00 + 3xy0 3y = x2 .
f. x2 y00 + 3xy0 3y = x2 .
g. 2x2 y00 + 5xy0 + y = x2 + x.
h. x2 y00 + 5xy0 + 4y = 0.
i. x2 y00 2xy0 + 3y = 0.

5. Consider an LRC circuit with L = 1.00 H, R = 1.00 102 , C =

1.00 104 f, and V = 1.00 103 V. Suppose that no charge is present
and no current is flowing at time t = 0 when a battery of voltage V is
inserted. Use a Simulink model to find the current and the charge on
the capacitor as functions of time. Describe how the system behaves
over time.

6. A certain model of the motion light plastic ball tossed into the air is
given by
mx 00 + cx 0 + mg = 0, x (0) = 0, x 0 (0) = v0 .
Here m is the mass of the ball, g=9.8 m/s2 is the acceleration due to
gravity and c is a measure of the damping. Since there is no x term, we
can write this as a first order equation for the velocity v(t) = x 0 (t) :

mv0 + cv + mg = 0.

a. Model this problem using Simulink.

b. Determine how long it takes for the ball to reach its maxi-
mum height?
c. Assume that c/m = 5 s1 . For v0 = 5, 10, 15, 20 m/s, plot the
solution, x (t), versus the time.
d. From your plots and the expression in part b., determine the
rise time. Do these answers agree?
e. What can you say about the time it takes for the ball to fall as
compared to the rise time?
 4
Transfer Function and State Space
Blocks

4.1 State Space Formulation

There are other more elegant approaches to solving a differential

equation in Simullink. Take for example the differential equation for a
forced, damped harmonic oscillator,

mx 00 + bx 0 + kx = u(t). (4.1)

Note that we changed the driving force to u(t).

Defining x1 = x 0 and x2 = x 0 , this second order differential equation can
be written as a system of two first order differential equations,

b k 1
x10 = x x2 + u ( t )
m 1 m m
x20 = x1 (4.2)

Note that x200 = x10 gives the second order equation with x = x2 . Also, this
is not the typical order of equations usually encountered when studying
systems of differential equations. This order is chosen to be consistent with
the State Space Block which we will use later.
This system can be written in matrix form: x0 = Ax + Bu, where
" #
x1
x= ,
x2

b k

A= m m ,
1 0

1
B = m .
0

We now think of x0 = Ax + Bu, as a system whose input is given by the

forcing term u(t) and we need to integrate the right hand side for a given
input function. The output of this system is the solution vector, x. Also,
we might want to output a plot of the forcing function. Thus, the complete
output from the system can be written as

y = Cx + Du,
 56 solving differential equations using simulink

where C is a row vector and D = 0 or 1. In particular, we might only want

to output the solution component x2 . So, we let C = [0, 1] and D = 0. The
block diagram for this process is shown in Figure 4.1.

D Figure 4.1: State space representation of

the system x0 = Ax + Bu, y = Cx + Du,
x x y
u B ++ 1
s C ++
input output
A

The whole process is captured in the State Space Block. This block is
found in the Continuous group. The implementation of this system with
a sinusoidal forcing term is depicted in Figure 4.2. This shows the pair of
equations

x0 = Ax + Bu,
y = Cx + Du. (4.3)

Figure 4.2: The use of theState Space

u x' = Ax+Bu y Block displaying a Sine Wave input
y = Cx+Du and output to a Scope.
Sine Wave Scope
Function State-Space

As an example, we consider the case where m = 2 kg, b = 0.2 kg/s, and

k = 1.0 N/m. Then, we have
" #
0.1 0.5
A= ,
1 0
" #
0.5
B= ,
0

C = [0, 1], and D = 0. These values are put into the block by going into the
Function Block Parameters dialog box for the State Space block as shown
in Figure 4.3. Note that there is a place to enter the initial condition, such
as [ x1 (0), x2 (0)] T = [0, 1] T . In this case one would type [0; 1]. A compari-
son of outputs from using this initial condition to zero initial conditions is
shown in Figure 4.4. Figure 4.5 shows the system needed to produce this
plot.

4.2 Transfer Functions

Another method for solving the differential equation com-

pactly is to use the Transfer Fcn block. This is shown in Figure 4.6. One
needs to enter the transfer function numerator and denominator in the
Function Block Parameter box, shown in Figure 4.7. Essentially, this is
 transfer function and state space blocks 57

Figure 4.3: State space block parame-

ters.

Figure 4.4: Solution to the forced,

damped harmonic oscillator problem
with initial conditions set to 0 or [0;1].

Figure 4.5: The use of theState Space

x' = Ax+Bu y Block dispaying a Sine Wave input
y = Cx+Du and output to a Scope. The Mux block
(from Signal Routing) is used to feed
u State-Space solutions from two systems using
different initial conditions.

Sine Wave Scope

Function1 x' = Ax+Bu y
y = Cx+Du

State-Space1
58 solving differential equations using simulink

Figure 4.6: The use of the Transfer

1 Fcn Block with a Sine Wave input and
2
2s +.2s+1 output to a Scope.
Sine Wave Transfer Fcn Scope
Function

recognized as the Laplace transform of the differential equation with zero

initial conditions. In this case, one enters the coefficients of the second or-
der differential equation into the denominator as [2 .2 1]. Unfortunately,
one can only solve problems with zero initial conditions.
If one knows the transfer function, then one can use it to create an
equivalent State Space block. This is done using the MATLAB function
tf2ss(1,[2,.2,1]). We note that this produces the parameters A, B, C, D, but
what it gives is " #
1
B= ,
0

and C = [0, 0.5], This differs from what we derived above. The difference
lies in the fact that we can multiply B by any constant and divide C by
that constant and not affect the solution of the problem. In this case, the
constant in question is the mass, m = 2.0.

Figure 4.7: Block parameter display for

the Transfer Fcn block.
 5
Systems of Differential Equations

5.1 Linear Systems

We consider the linear system

x0 = ax + by
0
y = cx + dy. (5.1)

This can be modeled using two integrators, one for each equation. Due to
the coupling, we have to connect the outputs from the integrators to the
inputs.
As an example, we show in Figure 5.1 the case a = 0, b = 1, c = 1,
d = 0. This is the linear system of first order equations for x 00 + x = 0, and
y = x 0 . We also insert the initial conditions x (0) = 1, y(0) = 2. Running the
model, results in the plots in Figures 5.2 and 5.3.

y Figure 5.1: Linear system using two

1 integrators.

b
x
0
a

1 x
1 xo s
Integrator
x(0)

Scope XY Graph
1 y
2 xo s
Integrator1
y(0) Linear System of Differential Equations
y x'=ax+by
0 y'=cx+dy
d
x
-1
c
60 solving differential equations using simulink

Figure 5.2: Linear system using two

integrators.

Figure 5.3: Linear system using two

integrators.

This system can by put in matrix form,

" #0 " #" #
x 0 1 x
=
y 1 0 y

This can be modeled by introducing matrix multiplication in a gain block

as shown in Figure 5.4. The input and output to the Integrator block are
vectors. The output is split using a Demux block to plot x and y sepa-
rately. The Scope block plots the two signals separately as functions of t.
The XY Graph block is used to plat the phase plane, y vs x..
We can also use a State Space block to solve this system. This is shown
in Figure 5.5. We set the input as u = 0. In order to output both x and y,
we set A = [01; 0 1], B = [0; 0], C = [10; 01], and D = [0; 0]. We also set
 systems of differential equations 61

[x';y'] Figure 5.4: Linear system using matrix

operation.
1 [x;y]
xos
[1;2]
Integrator
IC x

y
Scope XY Graph

0 1
-1 0 * u
Linear System of Differential Equations

Gain

the initial conditions to [1; 2]. The solution plots are the same as shown in
Figures 5.2 and 5.3.

Figure 5.5: Linear system using matrix

x operation.
x' = Ax+Bu [x;y]
0 y = Cx+Du y
Constant State-Space Scope XY Graph

5.2 Nonlinear Models

Here are simulations of some nonlinear systems. The jerk model and Chua
circuit are typical models displaying complex behavior.
62 solving differential equations using simulink

Nonlinear Pendulum Figure 5.6: Linear and nonlinear pendu-

x''+k sin x=0 lum.

sin(u)
-1 1
Fcn xos 1
Product Gain xos
Integrator
Integrator1
4
Constant
0 1
Constant1 Constant2

Linear Pendulum
x''+k x=0
Sine Wave Scope

-1 1
xos 1
4 Product1 Gain1 xos
Integrator2
Constant3 Integrator3

0 1
Constant4 Constant5

Van der Pol Equation Figure 5.7: van der Pol equation.
x''=mu (1-x^2)x'-x

x'' x' x
1-u^2 1 1
1 s s
Fcn Integrator Integrator1
Product Gain
Scope

Reset to fixed-step, Runge-Kutta, dt = 0.1

XY Graph
systems of differential equations 63

Figure 5.8: Solution plot for the van der

Pol equation.

Figure 5.9: Phase plane plot for the van

der Pol equation.
64 solving differential equations using simulink

Figure 5.10: Predator-Prey model.

x' Predator-Prey Model
0.02 1 x
xos
a 50 x' = x - axy
Rabbits y' = - y + bxy
IC

xy y
Product Scope

0.01 1
xos
y' y
b Foxes
XY Graph
10
IC1

Figure 5.11: SIR epidemic model.

.5 -1 1
s
beta Gain
Product Integrator S

SIR Model Divide N

1
s
Integrator1 I

1
.33 s
Product1 Integrator2 R
gamma
 systems of differential equations 65

Nonlinear Jerk Equation Figure 5.12: Nonlinear jerk model.

x''' + cx'' + bx' + ax + x2 = 0

XY Graph
x''' 1 x'' 1 x' 1 x
s s s x2
Integrator Integrator1 Integrator2
Product

1
a

1.1
b

0.42
c

1 |u| 5/7 x
9 x' 1 x
Constant Abs Product s
b
alpha Integrator

8/7 .5 XY Graph

1 |u| a Gain
y
y' 1
Constant1 Abs1 s
Integrator1 y
Chua's Circuit
Product1
x' = alpha [y - x + bx + 0.5(a-b)(|x+1|-|x-1|)]
z' 1 z
y' = x - y + z -100/7
s
z' = - beta y Integrator2 z
- beta

Figure 5.13: Nonlinear Chua model.

 6
Boundary Value Problems

x Figure 6.1: Shooting method applied

f(u) to the boundary value problem, y00 =
4(y x ), y(0) = 0, y(1) = 2.
Fcn Exact Solution
4 Clock

Gain y''
1 y' 1 y
yp xos s
Scope
Integrator Integrator1
Constant
y
y'' = 4(y-x), y(0) = 0, y(1) = 2.
To Workspace
2 4 2x -2x
y(x) = e (e - 1) (e - e ) + x

clear all;
N=20; % Number of iterations
beta=2; % Target value
yp=1; % y(0) guess
alpha=.1; % Fraction of difference
sim(BVPmodel)
for i=1:N
yp=yp-alpha*(y.signals.values(end,1)-beta);
sim(BVPmodel)
plot(tout,y.signals.values,b)
hold on
end
hold off
axis square
xlabel(x)
ylabel(y)
title(y vs x)
68 solving differential equations using simulink

Figure 6.2: Plot of solution using

y vs x the shooting method applied to the
2 boundary value problem, y00 = 4(y x ),
y(0) = 0, y(1) = 2.

1.5

1
y

0.5

0
0 0.2 0.4 0.6 0.8 1
x
 7
Difference Equations

Difference equations provide other interesting models.

Figure 7.1: Fibonacci sequence, xn+2 =

simout xn+1 + xn , x1 = 1, x2 = 1.
Fibonnaci Sequence Iteration
To Workspace

x(n+2) x(n+1) x(n)

1 1
z z
Unit Delay Unit Delay1 Display

Scope

Figure 7.2: Logistic map, xn0 +1 =

1 rxn (1 xn ).
1 - x(n)
One
Need to set solver
to Fixed Step
x(n+1) and Discrete
3.1 x(n)
Z-1
Product r
Delay Scope

Logistic Map 0.5

x(n+1) = r x(n)(1-x(n))
IC

XY Graph
70 solving differential equations using simulink

Figure 7.3: Henon map, xn+1 = 1

axn2 + yn , yn+1 = bxn

y(n+1)
0.3
Z-1 y(n)
1.4
b
a Delay1
Math
u2 Function
1 0.5
Constant
y(0)
x(n+1)
XY Graph
Z-1 x(n)

Delay Scope
Henon Map
0.5 x(n+1) = 1-a x(n)^2 + y(n)
y(n+1) = b x(n)
x(0)

yn 1 yn-1 Figure 7.4: Eulers Method, yn+1 =

z yn + x f ( xn , yn ).
Unit Delay Scope

yn-1+ h(xn-1+ yn-1)

y_n

n-1 xn-1
.125 x_n
Clock1 Gain1
xn-1+ yn-1
simout
To Workspace
0.125
h(xn-1+ yn-1)
Gain
8
Partial Differential Equations

N 19 Figure 8.1: Model of the 1D heat equa-

N - is odd
tion, ut = ku xx .
Constant Display
1D Heat Equation
du/dt = k(Au+c) dx
u2 10
Math Product1 b-a
Function

Product

1
-C- xos K*u 1
Integrator Gain k
ICs

-C-
BCs

Scope
 9
Index

acceleration due to gravity, 31 free fall, 31

air resistance, 31
annotate, 5 history, 13
anonymous function, 22 hold command, 22

blocks, 1 initial condition

Clock, 7 external, 7
Constant, 7 initial value, 5
Continuous, 1 integrating factor, 10
Delay, 39
Demux, 49 Law of cooling, 27
Fcn, 49 Library Browser, 1
Gain, 3 logistic equation, 37
IC, 49
Inegrator, 2 Malthus, Thomas Robert, 27
Math Function, 7 MATLAB, 1
Math Operations, 1 matrix form, 48
Parameters, 4 MS documents, 13
Scope, 2 MyScopeData, 13
Sine Wave, 3
Sinks, 1 Newtons Law of Cooling, 27
Sources, 1 Newton, Isaac, 27
Sum, 3 numerical error, 9
Switch, 39
ODE, 1
Terminator, 49
ode45, 2, 20
Boolean condition, 49
second order, 51
bouncing ball, 48

partial fraction decomposition, 32

carrying capacity, 37
phase plot, 51
Configuration Parameters, 6
plot, 11
cooling, 27
plotting images, 13
population models, 25, 37
delay model, 38 logistic, 37
direction field, 52 Malthusian, 27
direction fields, 21 prfig, 14
drag, 31 printing models, 16
dsolve, 20 projectile motion, 46
pursuit curves, 33
ezplot, 20
quiver, 22, 52
first order
linear, 7 refine Factor, 6
separable, 7
forced, damped harmonic oscillator, 44 Scilab, 17
74 solving differential equations using simulink

Scope parameters, 13 subsystem, 29

second order, 43
constant coefficients, 44 terminal velocity, 31
simout, 16
simple harmonic oscillator, 50 Verhulst, Pierre Franois, 37
Simulink, 1
step function, 45 Xcos, 17