2006-936: SOLVING NONLINEAR GOVERNING EQUATIONS OF MOTION

USING MATLAB AND SIMULINK IN FIRST DYNAMICS COURSE

Ali Mohammadzadeh, Grand Valley State University

ALI R. MOHAMMADZADEH is currently assistant professor of Engineering at School of
Engineering at Grand Valley State University. He received his B.S. in Mechanical Engineering
from Sharif University of Technology And his M.S. and Ph.D. both in Mechanical Engineering
from the University of Michigan at Ann Arbor. His research area of interest is fluid-structure
interaction.

Salim Haidar, Grand Valley State University

SALIM M.HAIDAR is currently associate professor of Mathematics at Grand Valley State
University. He received his B.S. in Mathematics with a Minor in Physics from St. Vincent
College, and his M.S. and Ph.D. in Applied Mathematics from Carnegie-Mellon University. His
research studies are in applied nonlinear analysis: partial differential equations, optimization,
numerical analysis and continuum mechanics

Page 11.1141.1

American Society for Engineering Education, 2006

 Solving Nonlinear Governing Equations of Motion
Using MATLAB and SIMULINK in First Dynamics Course

Overview

Students in first dynamics courses deal with some dynamical problems in which the governing
equations of motion are simultaneous, second order systems of non-linear ordinary differential
equations. There are no known quantitative methods or closed-form solution to these systems of
non-linear differential equations. To teach students analytic methods and solution techniques to
this category of dynamical problems, the authors devised a model example in which the motion
of a particle on a rough cylindrical surface is considered. The authors then formulate qualitative
and quantitative solution methods and employ both MATLAB and SIMULINK to these analytic
methods and solution techniques to arrive at the stable, numerical solution to the proposed
example.

In the model example provided, the governing equations of motion are first obtained in the state
form. Four different approaches then were adopted to arrive at the stable, numerical solution of
the system of first order, simultaneous, non-linear differential equations of motion.

In the first approach, the governing differential equations were converted to appropriate
difference equations, and then a program was written in MATLAB to solve the resulted system
of nonlinear algebraic equations. The second approach employed fourth-order Runge-Kutta
scheme by writing a program in MATLAB to render a stable solution to the system of
differential equations. The third method utilized MATLAB built-in function, ode45, to solve
the governing non-linear system of differential equations. Finally SIMULINK, which is an
extension to MATLAB, was used to provide solutions to the governing differential equations.
The results of these different approaches were then compared with each other.

Although MATLAB is recommended as the programming language for some end-of-the-chapter

problems in the recent, well-known text by Tongue1, none of the popular dynamics texts in the
market1, 2, 3, 4, 5 today (including Tongues) use SIMULINK in any form or way. The authors
believe that SIMULINK, being a graphical user interface program, has a great potential in
promoting better understanding of dynamics subjects, especially when students do not have a
differential equations course in their background. SIMULINK is also an excellent tool to
reinforce the topics students learn in a typical, undergraduate differential equations course. Last
but not least, SIMULINK is a very powerful tool in analysis and design of dynamical systems.
The authors used SIMULINK in analysis and design of an automobile suspension system6 as an
Page 11.1141.2

exemplary model in vibrations class.

This model example, which provided for follow-up homework assignments and a project, helped
students learn about efficient numerical methods, and how to employ technology tools,
MATLAB and SIMULINK, in solving engineering problems, early in the dynamics class. What
students learned here helps them a great deal in the subsequent courses in the curriculum. The
state form of the governing differential equations of motion, introduced to students in the follow-
up homework problems, is certainly subject of further application in studying the dynamical
response of the systems in several subsequent courses, namely dynamic systems, vibrations, and
control. Moreover, this treatment by state form of the governing equations is a novel and
powerful way to treat differential equations in the classroom as all popular undergraduate
differential equations texts in the market today miss out on it.

It should be mentioned that, at our institution, dynamics is taught in the first semester of the
junior year and right after the differential equations class, which is offered in the second semester
of the sophomore year. In other curriculums, where dynamics is taught before students take
differential equations, the method applied here (in the SIMULINK part of this paper) becomes
extremely valuable as it can be introduced and employed in the dynamics course before the
differential equation class (See Appendix D). Appendix D clearly shows that SIMULINK
integration block symbol requires no need for a priori knowledge of differential equations
course. Nonetheless, SIMULINK still can be used in the differential equations course, as we did
in ours.

The authors of this study received positive feedback from students regarding this experience.
Students especially enjoyed using SIMULINK and expressed that this project help them a great
deal in understanding not only dynamics topics but also the subjects covered in their previous
differential equation course.

Problem Statement

In our model example, we propose to evaluate the position, velocity and the time at which the 1
pound block leaves the surface of a cylindrical surface on which it slides. The block is assumed
to have an initial velocity V0 at the top of the cylinder and is subject to a constraint friction force
of kinetic coefficient of friction, k (See Figure 1). To achieve a stable numerical solution, we
assume, without loss of generality, a specific initial speed of 10 ft/s for the block and consider
the coefficient of kinetic friction between the block and surface to be zero in one case and 0.2 in
the other. The radius of cylinder, r = 5 ft.

V0

r
Page 11.1141.3

Figure 1
Formulation

Figure 2 shows the free body and inertia response diagrams of the block, degrees from the top
of the cylinder.

W
t
n mat
man
N F
Figure 2

Adopting the path coordinate and applying Newtons second law of motion, one obtains:
V2
Fn ma n : W cos s / N ? m (1)
r
Ft ma t : W sins F ma t (2)

Where m is the mass, W is the weight, V is the speed of the block degree(s) from the top, r is
the radius of the cylinder, N is the surface normal force, F is the surface friction force and an and
at are the normal and tangential components of accelerations of the block, respectively. However:

F ok N (3)
2
d s
at (4)
dt 2

Where s is the path traversed by the block on the cylindrical surface. Solving for N from
equation (1), one obtains:
V2
N m ( g cos s ) (5)
r

Substituting N from (5) into (3), and the result together with (4) into (2), one gets:

d 2s V2
g (sin s o k cos s ) o k (6)
dt 2 r

The speed of the block can be expressed as:

Page 11.1141.4
 ds
V (7)
dt
ds
Or: V r (8)
dt

Substituting (7) into (6), and rearranging (8), one arrives at the state form of the governing
equations of motion as:

ds
V (9)
dt
dV V2
g (sin s o k cos s ) o k (10)
dt r
ds V
? (11)
dt r

and subject to the initial conditions:

s( 0 ) 0.
V ( 0 ) V0 10 ft / s (12)
s ( 0 ) 0.

The problem at hand is clearly a single degree of freedom autonomous system (since s = r s )
and, therefore, it should be governed by two first order state equations. However, we formulated
the problem in the current manner by using equations 9-11 to obtain numerical solutions
separately for both angular and curvilinear positions of the block. The block leaves the
cylindrical surface when there is no contact with it (N = 0) and, at the same time, when the rate
of change of the normal force with respect to s is negative. When N = 0, equation (5) becomes:

V2
g cos s 0 (13)
r

Equations (9), (10), (11) form a set of nonlinear simultaneous differential equations of motion,
in state form, with initial conditions (12) and subject to condition (13) on angle , at which the
block leaves the surface of the cylinder.

Solution Methods

A closed-form solution to the system of non-linear, time-dependent, ordinary differential

equations (9), (10), (11) subject to initial conditions (12) and constraint equation (13) is not
possible. Three different numerical approaches are employed in this study to calculate the angle,
velocity, position and time at which the block leaves the surface of the cylinder; namely, an
appropriate difference equations scheme, Runge-Kutta method, and a MATLAB built-in function
Page 11.1141.5

ode45. This is the first time that the students try to use numerical methods to solve system of
differential equations in the curriculum. Pedagogically, it is very useful if students try different
solution approaches and compare the results of these different ways. We also introduced students
to SIMULINK to teach them a fourth approach to obtain a stable solution to the problem at hand,
in particular, and to this category of simultaneous non-linear differential equations in general.

Approach I- Difference Equations Scheme

In this approach we approximate the variables and their derivatives in equations (9), (10), and
(11) as follows:

ds s( t ) s( t Ft )
dt Ft
dV V ( t ) V ( t Ft )
(14)
dt Ft
ds s ( t ) s ( t Ft )
dt Ft

One also approximates the block speed as the average speed value at times t and t + t as:

V(t Ft ) V ( t )
V (15)
2

Substituting (14) and (15) into (9), (10), and (11), one obtains:

s( t ) s( t Ft ) V ( t ) V ( t Ft )
Ft 2
V ( t ) V ( t Ft ) (V ( t F t )) 2
g (sin s ( t F t ) o k cos s ( t F t )) ok (16)
Ft r
s ( t ) s ( t Ft ) V ( t ) V ( t Ft )
Ft 2r

Solving for the variables, velocity, position and angle at time t from equations (16), one arrives
at the governing difference equations:

(V ( t Ft )) 2
V(t ) V(t Ft ) Ft g (sin s ( t Ft ) o k cos s ( t Ft )) o k
r
V(t ) V(t Ft )
s( t ) s( t Ft ) Ft (17)
2
V(t ) V(t Ft )
s ( t ) s ( t Ft ) Ft
Page 11.1141.6

2r
The above system of algebraic equations, subject to initial conditions (12), is then solved by
marching through time from initial t = 0, when the block was launched at the top of the cylinder,
until the time t when the block leaves the cylinder surface. The angle and the velocity V are
substituted into equation (13) at the end of each time step to check whether the block has lost
contact with the surface or not. Without going further into the details, we mentioned to our
students the importance of choosing appropriate time steps to obtain a stable solution. A time
step of Ft = 0.1 ms was adopted in this simulation. Appendix A represent the MATLAB
program used to implement this approach. Although the error of these first order difference
operators is O( t), their numerical stability is much better than the higher order ones.

Approach II- Fourth-Order Runge-Kutta Method

A fourth-Order Runge-Kutta7 solution technique is employed to solve the system of non-linear

time-dependent first-order differential equations (9, 10, 11), subject to the initial conditions (12)
and condition (13) for evaluating the kinematics state of the block at the time of loss of contact
with the cylindrical surface. Appendix B shows the detailed MATLAB program to perform this
integration.

Approach III- MATLAB Built-in Function ode45

The MATLAB built-in function ode458, which is an implicit implementation of fourth-order

Runge-Kutta method, is used to solve the system of non-linear differential equations, (9), (10),
and (11), subject to the initial conditions (12) and condition (13). The detail of calling this
function in MATLAB is also shown in Appendix C.

Approach IV- SIMULINK Solution

SIMULINK, which is an extension to MATLAB, provides its users with a graphical user
interface that is fun to try and much easier to use than MATLAB itself (or Maple) or any other
traditional command-line programs, such as C, FORTRAN or BASIC. It is used to model
complex nonlinear systems, with a relative ease in comparison with these command-line
software programs, and as a result boosts productivity in a significant way.

To employ SIMULINK in solving system of nonlinear simultaneous ordinary differential

equations it is not necessary to convert the system into the state form. Contrary to the above
discussed approaches, we did not use the state form of the equations of motion (Equations 9, 10,
and 11), rather we remodeled our problem using equations (6) (a second order ODE) and (11).

Appendix D shows the detailed SIMULINK model of the problem at hand. It is seen in the
d 2s
SIMULINK model that the left hand side of equation (6), namely, the term ( 2 ), after going
dt
Page 11.1141.7

through two integrator blocks (1/s), renders the position of the block. Further elaboration on this
model is presented in the result and discussion part of this document.
Results and Discussions

To check the result of simulation, we study the exact solution to the governing differential
equations of motion for the case of no friction (k = 0). In that case, governing differential
equations of motion reduce to:

dV
g sin s (18)
dt
ds V
(19)
dt r

Subject to initial conditions (12) and condition (13). These can be solved by elementary
techniques. Using equation (19) and the chain rule, equation (18) is written as:

dV dV ds V dV
g sin s (20)
dt ds dt r ds
Which upon separation of variables and integration we obtain:

V2 V02 2 gr ( 1 cos s ) (21)

Upon substituting V from condition (13) into (21) we arrive at the angle at which the block
leaves the surface of the cylinder:

V02 2 gr 100 2( 32.2 )( 5 )

s cos 1
cos 1
29.1078 0 (22)
3 gr 3( 32.2 )( 5 )

Which upon substitution of (22) into (13), we obtain:

V gr cos s 32.2 5 cos( 29.1078o ) 11.8603 ft/s

We allotted three 50-minutes class periods to acquaint our students with numerical methods such
as Runge-Kutta, finite difference, and state form of differential equations. Moreover, one of
these class periods was devoted to introducing students to SIMULINK by showing them how to
construct simulation model for the problem at hand. Later on in the semester, students used
MATLAB and SIMULINK to do homework assignments and complete a project that was given
to them to reinforce the ideas and methods presented in this paper. We would also like to point
out that students who go on to take vibrations, system and control, or differential equations
courses will certainly reap great deal of benefit from this first experience in the dynamics class.
Our students who took these subsequent courses further acknowledge this point.

Table 1 next compares all the above cases with the exact solution. As it is evident from the table
Page 11.1141.8

all approaches clearly predict the kinematics values, for the exact solution, with excellent
accuracy, O(10-4).
 TABLE 1

Kinematics Properties of the Block When it leaves the Cylindrical Surface

No Friction Case (k = 0.)
Approach Time Angle Position Velocity
(s) (deg) (ft) (ft/s)
Finite Difference 0.2393 29.1095 2.5403 11.8598
Fourth order Runge-Kutta 0.2393 29.1106 2.5404 11.8606
MATLAB ode45 Function 0.2393 29.1107 2.5404 11.8606
SIMULINK Model 0.2392 29.1069 2.5400 11.8603
Exact Solution 29.1078 2.5401 11.8603

Table 2 shows the results for the friction case, where the kinetic friction coefficient k = 0.4.
Again all the approaches renders basically same values. The differences are mainly due to
numerical truncation error.

Appendix D shows the simulation of the exercise for k = 0.4. The kinematical values that are
observed in Table 2 for the SIMULINK model are values taken from the Display Blocks in
Appendix D. The Stop Block used in the model is to terminate the simulation when condition
(13) is met. The Logic Block seen in the model is used to implement condition (13). The model
uses three integrator blocks to arrive at position, velocity and the angle for the block. The
Clock Block on top of the model keeps track of the simulation time. These visual blocks, taken
from block diagram concept9, aid students a great deal to simulate a dynamical system in a
relatively short period of time in comparison with a typical command-line program. As we
mentioned before, there is also no need to convert the second order differential equations of
motion into first order (state form), to arrive at a stable numerical solution, in a SIMULINK
model.
TABLE 2

Kinematics Properties of the Block When it leaves the Cylindrical Surface

No Friction Case (k = 0.4)
Approach Time Angle Position Velocity
(s) (deg) (ft) (ft/s)
Finite Difference 0.2913 34.1204 2.9776 11.5456
Fourth order Runge-Kutta 0.2912 34.1093 2.9766 11.5453
MATLAB ode45 Function 0.2912 34.1093 2.9766 11.5453
SIMULINK Model 0.2912 34.1105 2.9774 11.5456

As it is seen from the results, friction holds the block to the cylinder for a longer time compared
with the no friction case. This is expected as friction lowers the block velocity, which from
equation (13) predicts a bigger contact angle .
Page 11.1141.9
Conclusion

It is the authors belief that with the availability of powerful programming tools such as
MATLAB and SIMULINK, the students of dynamics and differential equations benefit
tremendously from integrating similar models, as the one in the above, in their course work.
SIMULINKsoftware especially looks very promising in dynamics classes, where students do
not have a differential equation course in their background.

This model lead to homework problems and projects which helped students learn how to solve a
system of non-linear, time dependent, governing differential equations of motion in 4 different
ways. It introduced students to the state form of these differential equations, a topic which they
will deal with in their future course work in vibrations and control systems. It also sharpened our
students math and programming skills by extending their differential equation knowledge to new
dimensions; namely, stable numerical methods for solving nonlinear systems. SIMULINK also
helped students to develop block diagram skills. A skill they use in their control course later.

Students feedback regarding this model and its follow-up homework assignment and a project,
was very positive. They indicated that the model, homework and project combination helped
them a great deal to further understand the topics learned in the dynamics and differential
equation classes. They enjoyed, in particular, programming with SIMULINK because of its user
graphical interface character and relative ease to use.

Bibliography

1. Tongue, B.H., and S.D., Sheppard: Dynamics: Analysis and Design of Systems in Motion, John Wiley, 2004.
2. Beer, F. P., and J. E. Russell: Vector Mechanics for Engineers- Dynamics, 6th. Edition, McGraw Hill, 2004.
3. Meriam, J.L., and L.G. Kraige: Engineering Mechanics- Dynamics, 5th. Edition, John Wiley, 2001.
4. Riley, W.F., and L.D. Sturges: Engineering Mechanics- Dynamics, 2nd Edition John Wiley, 1995.
5. Hibbeler, R.C.: Dynamics, 10th Edition, Prentice Hall, 2004.
6. Mohammadzadeh, A.R., and S. Haidar: Analysis and Design of Automotive Suspension System, ASEE Annual
Proceedings, 2006.
7. Ayyub, B.M., and R.H. McCuen: Numerical Methods for Engineers, Prentice Hall, 1996.
8. Chapra, S.C.: Applied Numerical Methods with MATLAB for Engineers and Scientists, McGraw Hill, 2005.
9. Dabney J.B., and T.L. Harman: Mastering SIMULINK, Prentice Hall, 2004.

Page 11.1141.10
 APPENDIX A

MATLAB File FiniteDifferenceFric.m for Finite Difference Approach

% Finite Difference Approach To solve ODE System of Equations

i=1;
% Specifying Constants
r=5; muk=0.4;dt = 0.0001;g=32.2;
% Creating Vectors
Theta= []; V = []; t= [];
% Initializing Vectors
V(i)=10; Theta(i)=0; t(i)=0; s(i)=0;
u=t(1);
% Testing Whether the Block Leaves the Surface and Evaluating
% The New Positions, Angle, and Velocity of the Block
while cos(Theta(i))-(V(i))^2/(r*g)>eps
i= i+1;
V(i) = V(i-1)+dt*(g*(sin(Theta(i-1))-muk*cos(Theta(i-1)))+muk*(V(i-
1))^2/r);
Theta(i)= Theta(i-1)+dt/(2*r)*(V(i)+V(i-1));
s(i)=s(i-1)+ dt*(V(i)+V(i-1))/2;
t(i)= u+(i-1)*dt;
end
% Evaluating the angle, Position, and velocity at which the block leaves
% the surface
Theta(i)=((Theta(i)+Theta(i-1))/2);
s2 = r*Theta(i);
Theta(i) = Theta(i)*180/pi;
V(i) = (V(i)+V(i-1))/2;
s(i) = (s(i)+s(i-1))/2;
% Printing the Results
fprintf('Time = %5.4f sec\n',t(i))
fprintf('Theta = %5.4f deg\n',Theta(i))
fprintf('Position= %5.4f ft\n',s(i))
fprintf('Position using r*Theta = %5.4f ft\n',s2)
fprintf('Velocity = %5.4f ft/s\n',V(i))

Running FiniteDifference.m for the case k = 0.

Time = 0.2393 sec

Theta = 29.1095 deg
Position= 2.5403 ft
Position using r*Theta = 2.5403 ft
Velocity = 11.8598 ft/s

Running FiniteDifference.m for the case k = 0.4

Time = 0.2913 sec

Theta = 34.1204 deg
Position= 2.9776 ft
Page 11.1141.11

Position using r*Theta = 2.9776 ft

Velocity = 11.5456 ft/s
 APPENDIX B

MATLAB File FrictionIntegRun to call on Runge-Kutta Integration Method

%%Provide the Initial Conditions

y0 =[0;0;10]; r=5; g=32.2;

%Call on Runge-Kutta Function to Perform Integration

[t,y]=rkgen('fric',[0 1],y0,0.0001);

%Test Wether the Block Has left the Surface

for i=1:10001
s= cos(y(2,i))-((y(3,i))^2)/(r*g);
if s<eps
y(:,i)=(y(:,i)+y(:,i-1))/2;
y(2,i)=y(2,i)*180/pi;

%Print the Result

fprintf('Time = %5.4f sec\n',t(i))
fprintf('Theta = %5.4f deg\n',y(2,i))
fprintf('Position = %5.4f ft\n',y(1,i))
fprintf('Velocity = %5.4f ft/s\n',y(3,i))
break
end
end

MATLAB Function rkgen to Implement Runge-Kutta Integration Method

% Fourth Order Runge Kutta Method for Solving Simultaneous first order
% Differential Equations

function[tvals,yvals]= rkgen(f,tspan,startval,step)

% Creating Coefficient Vectors

b=[];d=[];
b=[1/6 1/3 1/3 1/6]; d =[0 0.5 0.5 1];

% Indicating the Number of Time Steps and Initial values

steps = (tspan(2) -tspan(1))/step +1;
y=startval; t=tspan(1);
yvals=startval; tvals=tspan(1);

% Calculating k1, k2, k3, and k4

for j=2:steps
k(1,:) = step*feval(f,t,y);
for i=2:4
if (i==2 | i==3)
cc=0.5;
else
cc=1;
end
k(i,:)= step*feval(f, t+step*d(i),y+(cc*k(i-1,:))');
Page 11.1141.12

end

% Updating Function Values and Time

 y1 = y+(b*k)';
t1=t +step;
tvals=[tvals, t1]; yvals = [yvals, y1];
t = t1; y =y1;
end

MATLAB Function fric.m to provide derivatives to Runge-Kutta Integration Method

function ydot = fric(t,y)

muk=0.4; r=5; g=32.2;

% Provide the derivatives

ydot=[ y(3); y(3)/r; g*(sin(y(2))-muk*cos(y(2)))+muk*((y(3))^2)/r];

Running Runge-Kutta Approach for the case k = 0.

Time = 0.2393 sec

Theta = 29.1106 deg
Position = 2.5404 ft
Velocity = 11.8606 ft/s

Running Runge-Kutta Approach for the case k = 0.4

Time = 0.2912 sec

Theta = 34.1093 deg
Position = 2.9766 ft
Velocity = 11.5453 ft/s

Page 11.1141.13
 APPENDIX C

MATLAB File FrictionInteg.m to call on ode45 Integration Method

%Provide Time Span of Integration

for j=1:10000
tspan(j)= 0+(j-1)*0.0001;
end

%Provide the Initial Conditions

y0 =[0;0;10]; r=5; g=32.2;

%Call on ode45 to Perform Integration

[t,y]=ode45(@fric,tspan,y0);

%Test Whether the Block Has left the Surface

for i=1:10000
s= cos(y(i,2))-((y(i,3))^2)/(r*g);
if s<eps
y(i,:)=(y(i,:)+y(i-1,:))/2;
y(i,2)=y(i,2)*180/pi;

%Print the Result

fprintf('Time = %5.4f sec\n',t(i))
fprintf('Theta = %5.4f deg\n',y(i,2))
fprintf('Position = %5.4f ft\n',y(i,1))
fprintf('Velocity = %5.4f ft/s\n',y(i,3))
break
end
end

MATLAB Function fric.m to provide derivatives to ode45 Integration Method

function ydot = fric(t,y)

muk=0.4; r=5; g=32.2;

% Provide the derivatives

ydot=[ y(3); y(3)/r; g*(sin(y(2))-muk*cos(y(2)))+muk*((y(3))^2)/r];

Running MATLAB built in Function ode45 Approach for the case k = 0.

Time = 0.2393 sec

Theta = 29.1107 deg
Position = 2.5404 ft
Velocity = 11.8606 ft/s

Running MATLAB built in Function ode45 Approach for the case k = 0.4

Time = 0.2912 sec

Theta = 34.1093 deg
Page 11.1141.14

Position = 2.9766 ft
Velocity = 11.5454 ft/s
 APPENDIX D

SIMULINK Model for k = 0.4

Page 11.1141.15