Lecture 4

Differential Equations
Lecture 4 Nonhomogeneous and Nonlinear Systems of

First-Order ODEs. Numerical methods for ODEs.
Matylda Jaboska-sabuka
matylda.jablonska-sabuka@lut.fi
Lappeenranta University of Technology
Department of Mathematics and Physics
29 January 2014
Lecture 4 Nonhomogeneous and Nonlinear Systems

Numerical methods for ODEs
Nonhomogeneous and Nonlinear Systems of First-Order ODEs

Method of undetermined coefficients
Method of variation of parameters
Linearization of nonlinear systems

Methods for first-order ODEs
Methods for systems and higher-order ODEs
Matylda Jaboska-Sabuka
Differential Equations 2014

Methods of undetermined coefficients

Nonlinear ODEs and systems of ODEs
Nonhomogeneous linear system of ODEs

y0 = Ay + g
where the vector g(t) is not identically zero. From a general solution
y(h) (t) of the homogeneous system y0 = Ay and a particular solution
y(p) (t) we get a solution of
y = y(h) + y(p) .
y is called a general solution of (1).
(1)


Method of undetermined coefficients

The general rules follow single ODE cases
The modification rule contains vectors of constants, that is for
example if the particular solution is
y(p) = ue t
then u is a vector.
Example 1
Solve
y0 = Ay + g =
3 1
1 3

y+
6
2
e 2t



The general solution of a homogeneous system with two bases can be
written as
"
#

(1)
(2)
y
y
c1
(h)
1
1
y =
= Y(t)c
(1)
(2)
c2
y2
y2
hence, the particular solution is
y(p) = Y(t)u(t).
Substitution into y0 = Ay + g gives
Y0 u + Yu0 = AYu + g
But since Y0 = AY then Yu0 = g and thus u0 = Y1 g.


Example 2
Solve by variation of parameters

3 1
6
0
y = Ay + g =
y+
e 2t
1 3
2



How to analyze the kind and stability properties of critical points of

nonlinear systems?
y0 = f(y), thus
y10 = f1 (y1 , y2 )
y20 = f2 (y1 , y2 )
How to solve nonlinear equations?

00 + k sin = 0


This can be done through linearization.

Taylor series
f (a) =
X
f (n) (a)
n=0
12
n!
(x a)n
function y=exp(x)
1degree Taylor
10
3degree Taylor
5degree Taylor
7degree Taylor
6
4
2
0
2.5
1.5
0.5
0.5
1.5
2.5


Linearization
For nonlinear systems of first-order ODEs, express y0 as
y0 = f(y) = Ay + h(y) and drop h(y).
For nonlinear higher order equations, transform is into a system of
first-order ODEs and proceed as above.


Example 3
y10 = y2
y20 = k sin y1


Consequence of linearization
y1 original
y1 linearized
y2 original
y2 linearized
0.8
0.6
0.4
0.2
0
0.2
0.4
0.6
0.8
1
10
12
14
16
18
20


Methods for Systems and Higher-order ODEs




Step-by-step methods
For initial value problem
y 0 = f (x, y ), y (x0 ) = y0
Method of computing approximate numeric values of the solution y (x) at
the equidistant points on the x-axis
x1 = x0 + h, x2 = x0 + 2h, x3 = x0 + 3h, . . .
where the step size h is a fixed number, for instance, 0.2 or 0.1 or 0.01.
Such formulas are suggested by the Taylor series
y (x + h) = y (x) + hy 0 (x) +
h2 00
y (x) + . . .
2


For a small h the higher powers h2 , h3 , . . . are very small. This suggests
the crude approximation
y (x + h) y (x) + hy 0 (x) = y (x) + hf (x, y ).
Euler (or Euler-Cauchy) method
In an iterative way we get the approximation
y1
y0 + hf (x0 , y0 )
...
yn+1
yn + hf (xn , yn )
Geometrically it is an approximation of the curve y (x) by a polygon,

whose first side is tangent to this curve at x0 .


Example 4
Apply the Euler method to the following initial value problem, choosing
h = 0.2 and computing y1 , . . . , y5 :
y 0 = y + x, y (0) = 0.
Compare the results with functions real values:
y (x) = e x x 1


Euler method
yn+1 = yn + hf (xn , yn )
n
0
1
2
3
4
5
xn
0
0.2
0.4
0.6
0.8
1.0
yn
0
0
0.04
0.128
0.274
0.489
hf (xn + yn )
0
0.2(0.2+0)
0.2(0.4+0.04)
0.2(0.6+0.128)
0.2(0.8+0.274)
ex x 1
0
0.0214
0.0918
0.2221
0.4255
0.7183


Automatic variable step size selection

Modern numeric software automatically selects variable step sizes hn , so
that the error of the solution will not exceed a given maximum size TOL.
Now for the Euler method, when the step size is h = hn , the local error at
xn is about 21 hn2 |y 00 (xn )|. We require that this be equal to a given
tolerance TOL,
s
2TOL
1 2 00
hn |y (xn )| = TOL, thus hn =
.
2
|y 00 (xn )|


Improved Euler Method (Heuns Method)

In the improved Euler method, in each step we compute first the auxiliary
value
yn+1
= yn + hf (xn , yn )
and then the new value
1
yn+1 = yn + h[f (xn , yn ) + f (xn+1 , yn+1

)]
2
It is a predictor-corrector method.


Algorithm for Improved Euler Method

ALGORITHM EULER (f , x0 , y0 , h, n)
For n = 0, 1, . . . , n 1 do
xn+1 = xn + h
k1 = hf (xn , yn )
k2 = hf (xn+1 , yn + k1 )
yn+1 = yn + 21 (k1 + k2 )
OUTPUT: xn+1 , yn+1
End
Stop


Classical Runge-Kutta Method

ALGORITHM RUNGE-KUTTA (f , x0 , y0 , h, n)
For n = 0, 1, . . . , n 1 do
k1 = hf (xn , yn )
k2 = hf (xn + 21 h, yn + 12 k1 )
k3 = hf (xn + 21 h, yn + 12 k2 )
k2 = hf (xn + h, yn + k3 )
xn+1 = xn + h
yn+1 = yn + 61 (k1 + 2k2 + 2k3 + k4 )
OUTPUT: xn+1 , yn+1
End
Stop


Stiff problems
A stiff equation is a differential equation for which certain numerical
methods for solving the equation are numerically unstable, unless the step
size is taken to be extremely small. It has proved difficult to formulate a
precise definition of stiffness, but the main idea is that the equation
includes some terms that can lead to rapid variation in the solution.
Solving stiff equations can be a very long and hard process.
Backward (Implicit) Euler Method
The backward Euler formula for stiff problems
yn+1 = yn + hf (xn+1 , yn+1 )


Multistep Methods
y 0 = f (x, y ), y (x0 ) = y0
Adams-Bashforth Methods
yn+1 = yn +
h
(55fn 59fn1 + 37fn2 9fn3 )
24
Adams-Bashforth Predictor, Adams-Moulton Corrector

h
(55fn 59fn1 + 37fn2 9fn3 )
24
= f (xn+1 , yn+1
)
h
= yn + (9fn+1
+ 19fn 5fn1 + fn2 )
24
yn+1
= yn +
fn+1
yn+1


Euler Method for Systems

yn+1 = yn + hf(xn , yn )
Runge-Kutta Method for Systems
1
yn+1 = yn + (k1 + 2k2 + 2k3 + k4 )
6
Backward (Implicit) Euler Method for stiff problems
yn+1 = yn + hf(xn+1 , yn+1 )


Matlab ODE solvers for nonstiff problems

ode45 is based on an explicit Runge-Kutta formula, the
Dormand-Prince pair. It is a one-step solver in computing
y (xn ), it needs only the solution at the immediately
preceding time point, y (xn1 ). In general, ode45 is the best
function to apply as a first try for most problems.
ode23 is an implementation of an explicit Runge-Kutta, pair of
Bogacki and Shampine. It may be more efficient than
ode45 at crude tolerances and in the presence of moderate
stiffness.
ode113 is a variable order Adams-Bashforth-Moulton PECE solver.
It may be more efficient than ode45 at stringent tolerances
and when the ODE file function is particularly expensive to
evaluate.


ODE45 results
Difference between constant and variable step size when solving
y 0 = x + y , y (0) = 0.
h = 0.1
hn variable
CPU time
0.0312
0.1872
MSE
9.8416e-018
2.6404e-018
After changing the initial condition to y (0) = 5 we get

h = 0.1
hn variable
CPU time
0.0312
0.1872
MSE
3.5430e-016
9.5056e-017


Matlab ODE solvers for stiff problems

ode15s is a variable order solver based on the numerical
differentiation formulas (NDFs). Optionally, it uses the
backward differentiation formulas (BDFs, also known as
Gears method) that are usually less efficient. Like ode113,
ode15s is a multistep solver. Try ode15s when ode45 fails,
or is very inefficient, and you suspect that the problem is
stiff, or when solving a differential-algebraic problem.
ode23s is based on a modified Rosenbrock formula of order 2.
Because it is a one-step solver, it may be more efficient than
ode15s at crude tolerances. It can solve some kinds of stiff
problems for which ode15s is not effective.


Matlab ODE solvers

Solver
ode45
ode23
ode113
ode15s
ode23s
Problem Type
Nonstiff
Nonstiff
Nonstiff
Stiff
Stiff
Order of Accuracy
Medium
Low
Low to high
Low to medium
Low


Consequences of solver choice nonstiff problems

y10 = y2 y3 ,
y1 (0) = 0
y20
y30
y2 (0) = 1
= y1 y3 ;
= 0.51y1 y2 ; y3 (0) = 1
t = [0, 12]
With ode45 the solution takes about 0.38sec.
With ode15s the solution takes about 0.59sec.
The stiff solver can treat a nonstiff problem efficiently with minor solution
difference.


Consequences of solver choice stiff problems

An example of a stiff system is provided by the van der Pol equations in
relaxation oscillation. The limit cycle has portions where the solution
components change slowly and the problem is quite stiff, alternating with
regions of very sharp change where it is not stiff.
y10 = y2 ,
y20
= 1000(1
y1 (0) = 0
y12 )y2
y1 ; y2 (0) = 1
t = [0, 3000]
With ode15s the solution takes about 1sec.
The nonstiff solver (ode45) cannot handle the problem in reasonable
time, and significantly influences the solution.

Lecture 4

Caricato da

Informazioni sul documento

Titolo originale

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Lecture 4

Caricato da

Copyright:

Formati disponibili

Differential Equations

Lecture 4 Nonhomogeneous and Nonlinear Systems of

Lecture 4 Nonhomogeneous and Nonlinear Systems

Nonhomogeneous and Nonlinear Systems of First-Order ODEs

Numerical methods for ODEs

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods of undetermined coefficients

Nonhomogeneous linear system of ODEs

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods of undetermined coefficients

Method of undetermined coefficients

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods of undetermined coefficients

Method of variation of parameters

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods of undetermined coefficients

Differential Equations 2014

Methods of undetermined coefficients

Lecture 4 Nonhomogeneous and Nonlinear Systems

Nonlinear ODEs and systems of ODEs

How to analyze the kind and stability properties of critical points of

How to solve nonlinear equations?

Differential Equations 2014

Methods of undetermined coefficients

Lecture 4 Nonhomogeneous and Nonlinear Systems

This can be done through linearization.

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods of undetermined coefficients

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods of undetermined coefficients

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods of undetermined coefficients

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods for first-order ODEs

Lecture 4 Nonhomogeneous and Nonlinear Systems

Numerical methods for ODEs

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods for first-order ODEs

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods for first-order ODEs

Geometrically it is an approximation of the curve y (x) by a polygon,

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods for first-order ODEs

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods for first-order ODEs

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems

Methods for first-order ODEs

Automatic variable step size selection

Differential Equations 2014

Lecture 4 Nonhomogeneous and Nonlinear Systems