You are on page 1of 26

4.12.

21
with.at tours off error

Wit wit hf ft wi
w L

Uit Ui th f Iti Ui 1 Siti


U ht So

we assume that

IS I E S for all i
d w

l Uil
y Iti E
Lai a

El Hee L
D
Cti a

1 IS e

for all i o
N
E 1h1 Iz E
h

µ
hi JI14
If he find then

E Chico Echl is

decreasing
If h Ff E ht o

Echl is
increasing
Theorem

If Taylor's method of
order h
is used to approximate
the solution

y Ctl fit YAH aetes


1 191 2

with step size h and if


C t a s then the
ye
local truncation error

is 01h
yr we can re
arrave
the Taylor expansion
Using
around
yit
Yin Yi hfyl.fi 1 hIf'cti.y
II f Ct yi if i Hi
for some

Ii C ti Titi

Based on we have

Fitial yiYI
h TYti.fi

ah f hi Mi D

for each i o d I

1
ti.fi fCfi
yi thf'cti
hn
f ft yi
since c c a S3
ye
yl f f it
yity
bounded on a S

TitichlE cha h
since I y ff Iffy
D8

Limitation of
Taylor
methods

For more accurate


we
approximations
need high orders
which
require explicit
calculation at each

for the derivatives


step
of f It yall
Runge Kutta methods

21
Taylor expansion
fit y f fo
y
It.to E ly yoPE
It t.TII.tlt t.ny.yi.IE
HII It
it s
citation
reminder term

4,11 left tl
rely
It
yl I
f 7 2
fit Putt ll t Ralf
Recall
y ll
the Taylor
method of order 2

witi withfffti.li thztafftiivil


With 1 t.hr
T It yl
T t.nl Taylor
method of order 2
We would like to
CH ft y
approximate c

with error that is

no
bigger
than 06h4
TYt.y flt.gl
t El t
Iff
Main idea
approximate
T'Ct yl using
f t 12 which
yep
will be expanded in

series
Taylor
fft.iq tpI
flt.y
t2IItpE kiEEIfI
igs3IIy
E III

n
to x
comparing

t2IftpIfy
fct.lt

fCt.ylth
fIItEyfI

Comparing coefficients
we have

t.kz
p HI
T ft y E

f t 1
ythflt.it
R
hIEftaIfIt.ylaEI
hfffit.yl 12,11
that the
Assuming partial
derivatives are bounded then
the error is 01h
The midpoint method

it swi hfftityh.withffti.IM
p p
This method resembles
the Euler however it is

different

ear

The method approximates


the at the center
slop
of each interval
We can derive many
different methods which
are och

T It.ly flt.y1thfCt.y
th f t.tl
A
flt.y taufftta ytp.flf.yY
Therefore we have

fct.ylth FH.ly t

tho 3 17 f

tt
iEI
3ft
i
we need to find
an 7 ans
fx
oh

fct.yltaeffttdr.ytfhfft. i
a.ta
1flt.y
ta.de
ttaupefIT
a
take y
a
f If
pi

y
H

i
wait h't teach
Tunisian
have
We
flexibility
the
to choose parameters
a a I
4
2
12
Then we have the
method
following

wit With tfftith


WithfHiiv W
t.it
L

E
The one Sified Eater method
12h4 The most common
12h method

K h.fcti.lv
kzihfftit.kz with
k hfftith
yh.it
Kui hff.IE ti Wit Ks

Wit Wit f k 12ha 12k 1ha

The method has a

local truncation error

of Och
numerical
Comparing
methods there are

several consideration

accuracy
Amount of computation
effort

Should we use a

method with
simple a

Shaker site
step
or
complicated
a more

method h
with larger
Eater 1 evaluation per
step
cc cc cc
12h4 4
p p.li atioi
ig
t
i

Error control
12nF

Ahh with step


site in

Nhl with site


Error control
n

01h nigh
run

We want to nah
sure that
ti
lyttit wit ee

we would like to

ensure that the error


in the
approximation
It
is less than E

Given of
if
Imagine that at each

time we have nth


step a

corter method

ycti.it yctilth4Hiyltil.h
Y0fhht
which probe
wit Witholfti Wish
Suppose that we use another
method with local truncation
error of hut
ylti.li ycfilthoTHi yltilih

tOChnt Weyfti1n
wiEwi

Final ylti.it yCt 0ftiiHtihh


h
wi
YHiti loltiwiihl
g
Wi
ylti.it WitIt iwihf

lti I Wit
T
In a similar way we

get wit
Ei.nu Yltiti
T T
corresponding
to 0T

1 we
Combining
obtain

C
itichl Eit.ch tFi itl
since the left hand side

is och we must

have that we must have

that w it Witi
is
a

the dominant term


since by our
assumption

Iit Ch is ochh
we are to use
going
the or for method
higher
to the error
approximate
in the last step of
our lower error methot
We define
12 th twit Witil
If change
the step
site to then
eh
E tilth 5k 41

The h Kh
e
Eckh I a R
R
To keep the local
truncation error bellow
E we regain
9 REE
T
ee E a

L
eecwf.tw
h 7 eh
At each step size

we calculate 12 and

and based on the

quantity computed the


step site is
accepter
and the algorithm
proceetes or
rejected
and step site

is modified

I E i E

etually spaced
a'Tweeting
mm
TTY.in

You might also like