Numerical Integration of Ordinary Differential Equations: Mathematical Methods and Modeling Laboratory Class

Mathematical Methods and Modeling
Laboratory class
Numerical Integration of
Ordinary Differential Equations
Exact Solutions of ODEs
• Cauchy’s Initial Value Problem in normal form:
• Recall:
• if f is locally Lipschitz-continuous
 there is an unique local solution
• if f is (uniformly) Lipschitz-continuous on all I
 there is an unique solution in I, i.e.
global solution
Numerical Analysis
• If solution can hardly be explicited  numerical
• Numerical Analysis is the branch of mathematics

studying approximation methods for solving
equations  applications on calculators
Explicit Euler Method
• For finding an approximate (or numerical)

solution of an ODE in normal form, first discretize
the domain I into n subintervals of width h:
s.t.
• Discretize the derivative y’(x) with the (forward)
finite difference:
• Denote and
• suppose to be able to calculate
Explicit Euler Method
• The ODE can then be approximated as
• Approximate solution:
• This formula for calculating the approximate

solution is called Explicit/Forward Euler Method
• Necessary:
• Initially:
• At each step:
• Decide discretization step h
• It can be shown consistent:
Implicit Euler Method
• Conversely, if we take the backward finite

difference
we have the approximation:
• This formula is called Implicit/Backward Euler

Method
• However, in this method we have to solve the
equation (with f appearing) in the unknown
• In practice, when f(x,y) behaves badly, Implicit
method is preferred if computationally feasible
Theta-Method
• Instead, fix in [0,1] and take the intermediate

value of f:
• Then it is easy to derive a generalization of EMs:
• This formula is called the -method

• Clearly, with it is Explicit/Implicit EM
• Is it explicit or implicit?
• It has better convergence properties
Example
• Find the numerical solution to Cauchy’s IVP
• Compare it with the exact solution
1. Define the M-file for function f(t,y) = 3t-ty

2. Fix h>0 and find numerical solution u by
EEM
3. Draw y, u and the error y-u
Example
• Let's try to numerically solve previous IVP with

IEM. Unluckly, this time it is not possible to use
MATLAB for inverting f(t,y) w.r.t. y. Hence, invert it
with paper & pencil
• That yields
• Compare IEM's to EEM's and exact solution y,

finding maximum absolute error and sum-of-
squares error
MATLAB ODE Solver
• Luckly, MATLAB already has algorithms for ODEs

• ode45 function finds approximate solutions for
most of simple non-stiff problems. Basic syntax:
[T,Y] = ode45(@fun,[x0 xF],initvals);
• @fun is handle to funcion fun defining f(x,y)
• x0,xF are initial and final values of interval I
• vector initvals contains Cauchy’s initial value(s)
of the solution(s): y(x0)
• odeset function can set some options of ODE
Solver specified as additional argument. Eg.:
odeopt = odeset('RelTol',1e-4,'AbsTol',…
[1e-4 1e-4 1e-5]);
Example
1. Define the function f(t,y) = 3t-ty

2. Call ode45 function
Example
1. Define the function f(t,y)

2. Call ode45 function
Higher order ODE
• A n-th order ODE y(n)=f(t,y,y',...,y(n-1)) can be

transformed into a system of n ODEs with n
variables
• Example: equation of Van der Pol Oscillator
after assigning y=x', it can be rewritten as system
• Try solving through MATLAB with
in the interval [0, 20]

Higher order ODE
• Create the M-file vaderpol.m defining the vector

function f(t,y) needed by ode45:
function out = vanderpol(t,x)
%as stated in the documentation this function has
%to take as arguments: time t and state-variable x.
mu = 2; %parameter of the Van der Pol oscillator
out = [0; 0];

out(1) = x(2);
out(2) = mu*(1-x(1)^2)*x(2) - x(1);
end
Higher order ODE
•Then in the command window run the instruction

[T,Y] = ode45(@vanderpol, [0 20], [2 0]);
• and draw the trajectory:
plot3(T,Y(:,1),Y(:,2));
xlabel('t'); ylabel('x'), zlabel('y');

Numerical Integration of Ordinary Differential Equations: Mathematical Methods and Modeling Laboratory Class

Uploaded by

Copyright:

Available Formats

Numerical Integration of Ordinary Differential Equations: Mathematical Methods and Modeling Laboratory Class

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Numerical Integration of Ordinary Differential Equations: Mathematical Methods and Modeling Laboratory Class

Uploaded by

Copyright:

Available Formats

Mathematical Methods and Modeling

• Cauchy’s Initial Value Problem in normal form:

• If solution can hardly be explicited  numerical

• Numerical Analysis is the branch of mathematics

• For finding an approximate (or numerical)

• The ODE can then be approximated as

• This formula for calculating the approximate

• Conversely, if we take the backward finite

we have the approximation:

• This formula is called Implicit/Backward Euler

• Instead, fix in [0,1] and take the intermediate

• Then it is easy to derive a generalization of EMs:

• This formula is called the -method

• Find the numerical solution to Cauchy’s IVP

• Compare it with the exact solution

1. Define the M-file for function f(t,y) = 3t-ty

• Let's try to numerically solve previous IVP with

• Compare IEM's to EEM's and exact solution y,

• Luckly, MATLAB already has algorithms for ODEs

• Find the numerical solution to Cauchy’s IVP

• Compare it with the exact solution

1. Define the function f(t,y) = 3t-ty

• Find the numerical solution to Cauchy’s IVP

• Compare it with the exact solution

1. Define the function f(t,y)

• A n-th order ODE y(n)=f(t,y,y',...,y(n-1)) can be

after assigning y=x', it can be rewritten as system

• Try solving through MATLAB with

in the interval [0, 20]

• Create the M-file vaderpol.m defining the vector

out = [0; 0];

•Then in the command window run the instruction

You might also like