0% found this document useful (0 votes)
139 views36 pages

Yfx - TN: y Flats

The document discusses using finite difference methods to solve systems of second order initial value problems (IVPs). It describes approximating derivatives using Taylor expansions to obtain centered difference formulas. This results in a matrix algorithm to solve the IVPs, where the solution at each interior mesh point depends on the solutions at neighboring points. The method introduces truncation error proportional to the step size, with the error bounded using the intermediate value theorem.

Uploaded by

baba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
139 views36 pages

Yfx - TN: y Flats

The document discusses using finite difference methods to solve systems of second order initial value problems (IVPs). It describes approximating derivatives using Taylor expansions to obtain centered difference formulas. This results in a matrix algorithm to solve the IVPs, where the solution at each interior mesh point depends on the solutions at neighboring points. The method introduces truncation error proportional to the step size, with the error bounded using the intermediate value theorem.

Uploaded by

baba
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 36

8.

31
Today

Given the second order

Dvp
11 if ix
y
Eb
y y1
161
flats 1
Generate a
seeuence
th such that

E.io Cbithl ylb1 p

21 y fcx.y.ly
e b Cal tu
a ex
Hated y
1 has is the
solution
yfx.tn
to the 121
Ivp
with t.tn
We determine Eu
by a root finding
applying
problem
F It
ylb.tl P
Z
options
a secant meth.J
61 The dewton method

deed initialization of
two to t
points
to solve 2

k ta lylbituil rfftu.it
ylb.tn tae
yc6 4
k 2,3
The Newton method
the iteration
step
has
the form

tu ta ylb.tw
d
Cb tu i

Given that we want to

rewrite h and solve


using
two second orede IVP
ul
situate have a solution that
T and X
depends on

we have to replace approximate


d
Cb th i

in a form that we know

how to solve

cx.tl flx.yfx.tk
y cx.tfaEXE6
yCa.t1 a.y'ca.tl t
d
To find bit
tn

IIcx.tl 3tIfx.ycxthyix.tD
the chain rule
Using
3 x
ylx.tl ytx.tl
3Itt3yIlx.ycx.tl.y4x.t1
lx.t1t3yI.lx.ylx.tl y ex.tl 3
Cx.tl
l
X and T are
intenpendent
31 0

Thus

If lx.ylx.tt ilx.tf
3Iy.Cx.ylx.tl.yYx.tH3fIHH
The initial conditions
become

Cattle lattes
The trick is to
take
zcx.tl 3 ix t
and we will assume
that the order of
differentiation between
X and t can be

replaced
we obtain

z fx.tl
of x 2 ix t
y y
Ty
x ZTxt
y y
II

for x Eb

2 fa.tl o Ela.tl L
Newton's method
Using
we need to solve

two initial value


problems
the th ylb.tn
PZC6
tn i

Example
solve the
Dvp
132 125 yy
y
Lex E3 17
414 yep
fcx.y.li f 32t2x
yy
f t
y fyi
8
We solve a

of 2 second
system
order
Ivp
problems
Ysexes f 321 2 X
yy
th
yeah 17 yal
using we obtain
2 fy Z t fy Z

li
nIoY
it

Let U
y Ury
Us Z Uu Z

We solve a
system
of 4 first order
Ivp
problems
We obtain
U Uz
132 1 2X V Uz
Uz
Iii

mutual
U.CI 17
k

Finish at home
Finite Difference Methods

Shootings method are

considered unstable
For linear Brp
Y paly't ytray
Select H2o and
divide tea 6 into Ntl

eeual sub intervals

X Atih is 0,1 Ntl


b
hi
Let Xi be the interior
mesh
points
The differential
Cfaation we
approximate

Y lxil pcxilyyxiltflxilycx.it
rail
Expanding in
Taylor

yfxi.ci yCXith
ycxilthy'Cxi ttiyYxil

thy exile y I
Xi E Ti E Xix

ycxi.it Cxi h1

ycxi hy'CxilthIy exit


HI y cxiltzh y
ll
Xi E Xi
for some
Ey
Adding
ycxi.tl tYCXi i

2yCxilth'yYxi
Ealy nititil

for fxi
Solving y
we have

y t xif hr ylxi.fr lxi

y hitHykid
tycxi
i
1
We use the intermediate
value theorem

Y cxil nt ylxi.it 2Ylxi

ycxi iD
E
Ey Gil
Xi Xi
Eli 11

In similar
way
a

we obtain a centered
difference formula
for
y'lXi
Y'lxiln zn ylxi.ie Yaa
for
HI y Yi
some

UiECXi Xiii
The calculations
will
previous
result in a

matrix algorithm
Tale and
nd use them in

Ax 6

ycxitf
2ylx.lt

fxi i

plxil nycxi.it
ycxitilz

1 fail ya thx hI zpcxi


Y Kil y ll
A finite difference method
truncation error

fifty focal
with boundary
conditions 1 61 2

Nets
Wed Wat P
and

Witi 2W twin

tplxilfwitiz nwi 1
fcxilwii rcx.fr
for each
i l N
we can write as

wi
it
Epcxi I

2th'ecxif wi

Ct E pH wi.tt

h r Cx

A U b
A will be
composed
of the coefficients
as above

Aw 6
A we b

th h th pKil
I o
z
Ix
I I pay 2
that It pcx1 o.ro
Ail

l
L o it t

an

In
I

h r Cx f I pay u
Theorem suppose
that

17 I are continuous on ca.ly

Then the tri diagonal


Aw b has
system a

lenieue solution provided That


I
h L
L

L mail
GEXE 6
pull
Note that the
a
theorem guarantees
Unida solution to the
but not
system the local truncation
about
error
Proof main idea
Need to show
that
1PCxh 121 implies

I I hzplx.it tHthzpcxil
2
Observe the matrix
N

A airs I

show that for A


he have

I Aii l Z lai.it t aiijti


for it 2,3 d l
a an
17 I am a
il
this will
imply A
has solution
a
anifle

Finite difference methods


Lineart
forhon
Given the non linear DVP
fix
y yy
yal 1 seat
years
similar to the linear
case we divide Ea 6

into Ntl Equally spaced


sub intervals and use
finite difference formulas for
we have
Y Y
y Kit I 2ycx.tt Cxi

ffxi.ycxi1 Ylxitilz
Hi
hyca
HE
y nil
We solve the system of
N equation with Wo L

Was
p
Witi 2W 1W wi
so
1 f
Xi wi jh
0

2W Wath fi w Wz
L to

Wi it 2W wit thtffxi.wi.wiInwi 1 i ro

Wa 12W th fcxn.wn.Pzwna 1p.io


i
We
resolve that f
and its derivatives will
be continuous and that
fylx y y't S on
a Exe 6
D laxity'll p X S
Fy fy are bounded
in D
Then we can use

Theorem ll l in the
book to ensure lenigueness

can be cast
as a roof finding
problem
Fix 1 0

Iterative methods
Linear
for solving
non

Dvp
Newtons meth.J
If intial estimation
is close ekoadh
then the method provides
fuadratic convergence
Fix o

x'mix IJfx THE


where J is the Jacobian
matrix of F

F Ix't f.ci f.cxT fnEIj

a
Calculating J can

take a lot of effort

JNeu

i
Quasi Newton method
avoids
J
explicitly computing

zf fjlktseatfcx.se
In
n
X
Ts

if he i
u i
otherwise

makes the method easier

to
implement

ok but we still
need many evaluations
of FCI

r
The number of
evaluation increases
as Ocn as the
number of etuations
grows
T
ydenImet
Generalization of the
secant method

fix
effxtf

X X
Writing 42 in the multiple
variable cage

I t
A f HE t Fai t
A is to
suppose
replace JCI
we need to make
sure A is unite
We need that
it and X will
be known
To find a matrix A
which is
knife
Recall that any
Zero vector
non
yea
can be written as

a sum of IX x
and the
orthogonal
of X
complement
E I 2

A 2 JCx Z
when
12
Ic
01 0

Combine 130 1190


to make sure A
is
knifed
X
updating using
Jfk will not
influence vectors orthogonal
to X I

kH tJlx Yix ix
H
A ytIFH a X Hi

A instead J CX

Ai will be the

approximation for Jai t


method has
Broy dens
the
following form
i
Updating
sit
Ai Ai it Ai.is
Tsi
it i

Ici
It FCI t F
I
Cii Updating
join I Ajax t

I
0cal
Improved Broydens
methot

Sherman Morrison
formula
that A
suppose
matrix
is singular
non

satisfy that
yell
A I't l Then
f't
I'ta
ixytj.at
CA A x

t
Ttx
f transpose
inverse
ro o f
p t
A I A
y
I
j't Iff
ix

FA t

I y.it A.Y
A A
I
f't t't x'ly ta i'I yti
Tix
I t ni
if ta t
x'y
I na
A faint

i
f

taking
it Yi
Ai Si

Tsi
j Si

Applying the Sherman


Morrison formula
sittin
A
A titis
sit itA
titis
Ai A
yi s sit A

iAy it
A y
Csi A yilsitA
Titty
1 T
Ai
so we can find
A i from A
without the need to
invert the matrix
at each iteration
Which will be much were

effective
Instead of 0cm
we have 0 NZ

You might also like

pFad - Phonifier reborn

Pfad - The Proxy pFad of © 2024 Garber Painting. All rights reserved.

Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.


Alternative Proxies:

Alternative Proxy

pFad Proxy

pFad v3 Proxy

pFad v4 Proxy