0% found this document useful (0 votes)
30 views20 pages

Math. Notes

Uploaded by

leo88ji
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)
30 views20 pages

Math. Notes

Uploaded by

leo88ji
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/ 20

a

IEC 9 13 numerical methods to solve problems that can't be solved otherwise


using
.

LEC 1 Mathematical modelling numerical methods &


problem
. ,
soling

Example Bungee sep of varables


Jumper using
:

want v(t)
Analy tical Method :

V Li
e

g
=
-

↑F F =
ma a =

A
=
d my-e"
&
Let a e
m der
=

F =

cht
Fr :

Case

Fu Is = a"-rn fixdx=Etantte
=
-

mg
-

Mee
Jane
& 9 81 M/52

flat
=
.

dr =

= -Let 12 t =
0 ,
v =
0


.

Itunk" It +
1 =
+ d if d -c =

f
·
Diffrential Equation

I tanb =

Cut f =
0

Numerical Method = tank (Catal

: g-GI Em :
:

Enter
v =

tank) It e
ti +
n
-
ti
rct) :
tank (Natt
a g-Ca e

-
=

v Vi (g -GV) (t ti)
I
-
=

+
n -
+1

Vi

+1 i

Vict
-
=

ti t = + n

new value = old value + Slopex Step size Euler's method

Vi +
1
=

Vi +
xxt

.
a Calculate v for first 125 At = 28
M = 68 .
1 kg 2 . Terminal velocity

C =

0 25
.

kg/m & F =
P

rubbweene
2
g =
9 81 Mys
my
=
< dV
n
.

e

v
=

E = 51 6938 .

Ni

Vi +
1
=
gar
-

IF =

my
-

CV =
ma a
A
in
e

g-C
- so ,

(ti
=

w Vi +
1
=
Vi + + 1
-

ti)

jur If 68 Ca
Euler's Method

m
1kg 0
25kgsm At 2 secs
= =

, . . =
Vi =
0

V= =
(0) +
(9 . 81) (2) =
19 62.

= 9.
81.
V3 =
19 62.
+
(8 397)(2) .
=
36 4).

e
V =
36 41 .
+
(4 942)(2) .
=
46 29 .
d =
9 81 -0
.
.

2a
.

V =
46 29
+
(1 . 942)(2) =

50 174
.

=
.

DIFFRENTIAL EQUATIONS REVIEW

SERERABLE
Y' =
f(x) gcy) ·

function of function
a a
multiplied by a of
y

Solving Steps

1 * =

f(x)q(y)

2
=

fix) -
de

3
gydy= /fix) dx

4 Solve for
.
y
LEC 9 15
.

Analytical Solution

, t
e)
= -
kA A =
4 + r3

*
V =

Exes
k =
0 .
08 mm/mm

A Area
-
=

*3
Numerical Solution
f = -
k4x ( )

I
t
fat
A 25 *
1
0

()
= . min

Wi +
1
=
vi +
(at) ar = - k4x
12/3
Va
=

65 45 -
6 283(0 25) .
= 63 87 .
o =
-
k4x(2 5 .

*
. .

V2 =
63 88.
-
6 182(0 25)
.
.
= 62 33 .
=
6 .
283 3V" =
-
k4x() t +

>

d
-
(B
+ (0) 2 5mm
=

= - 2 . 49 .

=
-0 .
129k +
C

"(0) =
65 45 .
=

c
as

5
.

A Entrain Fart-mgocio- matt V ( 129t +

65 45)


=
-
0 .
.

E =

g -
I
v
Fa my
=

Analytical Solution Numerical Solution At =


2

d
e
g
=

Vi
-

At

/gar
k +

fat
1
=

v
g c
=
-
=

19 62
V1 9 81 2 =
0
=
.
+ .
.

du = dV 6 496 2 32 613
V 19 62
=
+
-
.
.
= .

Edr= t
V3 =
32 .
613 +
4 302 .
- 2
=
41 218 .

Ing- ) =
t +
2 VC0) = 0

↓ (n(g) c c 13 5
= = -
- -
.

st x)
In
(g )
+
= -
-

- + +
c)
e
=

-
g
(eact )
x
g)(
+

v
=

(Ing
"e-
jent g)(
-
)
-

v = -

+
(e
-

v =

1) -
1 9
.

- =

Q ,
Qo
Q, 3 Qusint
3 450s ,t
=

- -
~
A = 1250m2
- =
.
-
450
> W
y 0
=
-

↓ >

-> Q0 450mY
=

- day
LE> 4 9 . 20

- +

ERROR

Accuracy - how
closely the computed value
agrees
with the true value

Precision how closely values other measured values


agree with
->

ERROR TRUE VALUE APPROX


magnitude
= -

does not take into account order of


.

True Fractional ERROR


Relative Error
Et =

TRUE VALUE treprent = E


x 10

we cannot calculate Error if we don't know the true value

for an iterative method ,

&ELATIVE
Sa Reprox
Previous approx
-
=
ERROR

present approx

Es the iteration
stopping value we stop once
EaLEs
=

Es 102 -u of
significate
sthenumber
0 5

where
x
=

STOPPING CRITERION

eX .
2 =
1 +
x +

2 + ... +
maclaurin series expansion of ex

approximate e. 5 Terms approx


.
et La 1approx-

= =
x 100 %
. 6487
1
True value 1 6487 I 1 %
=

39 3
.

approx : approx : -1
1 33 2 %
-

2 5 9 02 % 2a
=
.

100 %
.
.

- x

approx ·

3 1. 625 1 44 % 7 69 %
If we wanted sol. correct to
1 .
.

2
sig figs W

&

Es =
0 .
5x10-2 %

Es =
0 . 5 % ->
stop when Ea<0 .
5 %
Round-off Errors computer's handle numbers
inability to large
->
, exac t

Truncation Errors ->


caused by approximations

Series Expansion
Taylor

conver t difficult functions into an approximated polynomials



2X-

enx =

Co +21x +
crx +

cyx+ ... +
an infinite series of a
polynomial

may only apply


f(x)
on a

1 Prcx) a polynomial of order m

certain Interval

Cy x Cmx
Pm(X) Co +
+ ... +

x
+
=
,

fix . )
=

Pic x
P1(x) =

P(Xo) +

f(x)(x -

x0)

f(x) fico)(x xo)

I
P1(x) 41(x) =
+ -

i P(Xo) =
f(x)
-
-

No I

P2(x) =

f(x0) +
f(x)(x x0) ,
- +

z(x - x. )

fix :Inx if Xo =
1

enxo + (x-x )
P1(x)
!
=

P1(x) =
x -
1

f(x) =
4nx) at x =
No

f(x) (x-x0) Cz(x-xr)


Exox-x
Co +2 ,
-..
= +

Paxs
=

Co
2<t0)+
=

f(x) Co C , (0)
+ ...
=
+

-
2202 + 6 C > (0) +
=
2C .
POLYNOMIAL of a
C, TAYLOR
f(x)
= +

62s(0) + =
C2
f"(xr) =
22 +

f(x0)
Prix) f(x) <x-xoL
Exocx-x
=

6 C3
+ +

f"(xo) =

c =

f(x0)
I
ex .
Maclavin series expansion of 1Ex

fix) = f(o) +
fico)cx) +

fOLx+ ...

Taylor Series Expansion of f(x) :


cos(X) No cosC)= ?

f(x) =
cos(X.) -simXo(x-) -

co (*-
.

+ *- + -
"

Error Calculation

Terms approx Et Ea

1 0 .
707 41 4 % .

2

Maclaun Series of f(x) =

ex

f(x) = e4 +
exx-x0) +

2 x -x +

Exxo
=
1 +
x +

E +

+ ...

TU 9 21
-

Problems A

- 4t
e 1
y zy y(0)
=
+
= z -

y ext At 1
2
2y
=
0
=
-
- -

1
yo
=

y: + 1
=

yi
+

y =
0 .
1

1) 1 (2-2(1) -
1)0 1 =
0 9
y10
.
=
+
.
.

(2 2)0 9) - i4(0 .

1))0 1
=
0 . 853
y(0 2) a
= . .

0
-
+
.
.

3)
=

0 853 +
(2 -
2(0 853) - =
x(0 .

2))0 . 1
=
0 . 837
y(0
.
.
.
Problem
Exx )
2

fixx= I -
x ·

f(x) =

Inx f(z) =

Xo = 1 Py(x) = (x -
x. )+ ·
*
x -
x - ) +
Ex-x!

Ex- fix-xo
*
+ x
.
" +

(n2 =
0 . 693

P2x) en1 1(x 1) 1 <x-1"-6cx 1)


1x
= +

2
- -
-
- +

3 ! ↓

ORDER f(x) ERROR f(x) =


lax
ERROR
O
W 100 %
fixs E
/ Roxx
= =

44 2 % ==
=

100
1 1 x2 1
.

f"(x)
-
=
= -

2
A

27%
2 0 . 5 fixx) = 2x3 =

E
*
833 %
f4(x) =
-
6 = -

3 0 20 2
.
+
.

0 583 15 9 %
C .
.
LE) 9 22
.

f(x) =
Pm ,
a(X)

f(x) Pm acx) + Rm CX) to the


equal tayher
=

, ,
expansion

Plus the remander

Rm acx)
te
*
(x-al
-

left
,
whatever by truncation
+
...
out

f(x)
R
↑ Rim ,
alx f(a) =
Pcal

4m , a(x)
(f(x) Pmax)) Rma(x)

:
=
-

2

Determine max R(x) value between an interval [a, b]

f(cl =

Pu ,
a
(4)
proof
by a P2 a(x) ,

fical Pi n (a)
fal f(ascx-a)
=

Pr a(x) ficalcx-a)"
, =

+ +

,
11

f"cal =

Pr a(a)
,
2 !

Piacx) fas fas <x-al


:
:
+

fa) =

Pra(al
P2" ,
a(x) =

f"cal

D2 a(x)
=
0

1
,

Rm 14),
:
/ franx) -

Paal *

Rim acx)
,
=
1 final -

Piax)/ P2 a(a)
,
=

f(x) + 0 + 0

:
P2 ,
a(a)
=
fica) + 0

RY n(x) ,
=
/fex) -

Dax) P2 ,
a(a)
=
f"(n)

Racx)=/fcx) -
01
pat = 0 -> a constant
Run (x)
=
If a ce

IR*a (xs) = M M is the maximum value of


for (*)

fRinncx) /Mdx
-> 19R"(x) dx) = /IR x) dx

I /Rnacx) dx) =
/Mdx
23 -
Fal =
0

1RMa(xl) = Mx + > - >


Racas :

/Racx)) <falx -
a) 0 =
Ma
Ma
+ >

CI -

IR* cast =
Ma +
>
C = -Ma ->
max val of 6

/IRmacx1l</MEa
↳ cho -> a = 0

IR* x) =
ME +
s

:
M will
usually either be at

IRmaSxL) ME
a or b ,
=

TAYLOR'S REMAINDER THEOREM


f(x) = cosx No = f() =

fixconor-ancaa co? -Yo"+


xx-xo+ a xx-x. e

Remainder

Because remainder is an infinite series ,


no number but can bound .

/Race in the I
M= max value of

F(x) =
Pcx) +
Raxs
n

f(x) =I

I where
There must be some point

Eis
n+ 1

1 cas1 .
=

=Y


f(x) =
0
=

?
a
=

x
=

0 .
1

f(x) =
P2(x)

P2(X) =
1 +

Ex- *
- 5
↳ =
5x10
P2(0 1)
+

.
= 1 .
04875

Racesl=(Emal) 2 -
10 ,
0 .
1)

=) 10 .

17
f"(x) =(x
42
(R2(x>)
-

=
+
1)

6 25x185
finax (X) at x = 0 -
.

since
*I
:
Euler's

V(t) Vct=)(ti ti) R1


vitz-t e
Vi
-
+
+

+
=

R ,
=

Va
rict :
+
ti +
- t
:

Forward
difference
approximation · A
V
-
plus an error
-
error

therefore

the
is

,
proportional

smaller
smaller

the
the
to

error
step
k

will
size

be
* t

Derivation of Backward Difference Approximation

f(x : )
fix.) fi
-

k
=

+
error &
xi +
1
-
x:

f(x) =
f(xi) - f(x - 1) +
- BDA

begin with TaylorexpX =

Xc-1 a =
X:

f(xi - 1) =
f(x) f(x)(xi-xi) +

fix -
-
x- )

fix)
A fair f?
:
(xi-1- xi) same as FDA

x < 3, xi +
1
DA ficxi) f(x f(x = -1)
:
Centered =

=
+
) -
-
CDA

f(xix
m
+ x2 x)
f(xi 1) f(x f(x)(xi x (x ↳ -
+

1
+ -

i
+ = +

f(x
fixi
-
: - 1)
=

f(x) -
f(x)xi -
+
x- +

-
1
+

xi) - 3
xi + xi)
-

fi) I
2 2

(f (3 )
↓ -
k

f(x) f(x= f(x)(x : 1)


+

1)
=

0 +
+
1 -

xi - + 0 + ,

-
MVT
by A "c cTal

faifixieties of" sh
fIC3)
fixit
:

+
2

erro a
xi +
b -
xi - 1
LEC 10 4 .

Forward Difference Approximation

fam-fix)
h
fix = , envorom order of h

·
divide h , divede Error
by h

Backward Difference Approximation

fix: faxis-f(x)
L
+
Ei o(h)

Centre Difference Approximation

fax :
ficx. =

fat -
Be OCh

· drude ↳ divide er ror

f(x)
*
ex .
f(x) = -0 . 1x -0 . 15x3-0 5x . - 0 25+
.
1 .
2 at x =
0 . 5

xi 1
x= 0 5 x = -1 0
f'(0
=
=
1 True value 5)
=

9125
. +
, = -0 .
.

f(x : 1) =
1 2
f(xi 1) f(0 25)
.
-

1 1035
=
=
.
.
-

f(x : ) 0 925
f(x)
=

f(0
.

. 5) =
0 . 925

f(xi + 1) =

0 . 2
f(x: 1) =

f(0 >5) .
=
0 636
+ .

0 925
-
f'(0 5)
=

Forward . = 1 .
45
h =
0 - 25

h = 0 5
.
0 . 5

f(0 5) : -2925= -1 155

a5
.

-0
.
.

/
25
x 100 % 58 % 0
Ex
.
=
=

Et 26 5%
2
=

Backward f 'co 5) .
:

a = -0 .
5
f'(0 . 5) = 0 . 925-1 1025
-
.

=
- 0 .
714

=1-05-05)
0 . 25

x 100% =

39 7
Ex
-

Eq = 21 7 .
% ~Ed

f'(0 5) -
-1. 1035
Centre 934
=

0
.

= -

b
.

not
f(c0 -
5) : = -
1 0 . 5

Ex = 2 . 4 %N Ex

21 /
+
1)x 100% %
-

= :

9 6 .
Total Numerical Errors ->
Includes roundoff er ror

At-faxi- F
CDF
fix- :

hi

f(x - +
1) =
xi +
1) +
2: + 1 roundoff error

rounded value

f(x= 1) -
=
Ax1) : -
+
ei +
1 round off error

fix- eits----ei-1h"
3!
2 ho

-x xin
einen As
-

2h
-
men -
Approx .
roundoff er ror truncation error

(i+1 has value of decrual


assume :
:-1 a maximum
->
points of computer

Roundoff
Max
error
-

- Max f =

M
Tot Error
.

Ech]
Total

↓ Error = +
he -> error a h kP error

" x ht error
error

k · need to find an optimal value for h to halance

·
Eck) 0

A
=

Echi= +

Ech)
Me hopt=
= -

function
+

->
M must be maximum value of entire

OPTIMAL H VALVE
=- Mt
BERIVING CENTERED DERIVATIVE APPROXIMATION for SECOND DERIVATIVE

Forward

f(x, +
1) =
f(xi) +
f(x)(xi
-
+ 1 -xi) +

fxi)(xi + 1
-
xi) +

+xi)(xi
! 3
-
+ 1 -x) +
En I +

u
h

Backward
Xi -1 -
X=

f(xi) f(xi)(x=
f(xi 1) xi C
-
=
-
-
-
- -
1
-
3
h

F +
W
q(b) -

f(xix) f(x-- + =
2 f(x) + 0 +
ficmh -
o +

fish". que
q(ni -

O ⑧

*
- I (f"(3! =

f
+

<32)] a
b

f(x , + , +
f(x = -) =

2f(x) +
f"cxih - n
MVT
=-
2 f"(3 There exists a C where

f"(xi)
2x) fx+ fx f
=

g(aq(b)
-
+ +

q()
-
=

f4f 3
*
+
f ( =
3
-

APPROX
TRUNCATION ERROR

f"cx fixe
x-1-2f(x )
Aysh"
= :
-

Li

differenc of first derivatives

fir . F i
Atf(x .
: -
h
+ ERROR

E
x 1

f(x) =

cos(x)
f "(0 . 8) =
? f"c0 . 8) : -cos(0 8) .

= - 0 . 6967067093

Xi -1 0 79 xi 8 xi 0 81
=

.
=
0 .
= .

k
-8-0
.

1f"(0
0
=

00) = -0 690000 000580409


Et
= .
,
=
0 .

b =
0 .
01
-[10 29)
f"cx - 118 .

=-0 .
6969000 Ex =
0 .
00019329)
0 . 012 e
smallest er ror
h = 0 . 001

1991
f"0 - 200 -
000706709
.

. 00
=
= -
0 .
696/26300 =
0 .

0 . 0072
E E

f(x i fi 1) 2f(xi)
fix =
-
1
+
+
-
+

Roundoff
A +
Total
=
ERROR Error

Ech) = *
M
E:
=
1 -

+
M
=

Kopt
*
2x10-9
Gopt
: 48 . 0 . =
0 . 01244
ERROR

TRUE VALUE =
APPROX+ ERROR

in an iterative process we
may
E ->
TRUE ERROR
want to know when to stop
Et TRUE VAL-APPROX
=

Kal <Es
Et - TRUE PERCENT RELATIVE ERROR

TRUE VAL APPROX


STOPPING CRITERION E ,

t
-

S =
-
t x 100 % =

TRUE VAL "


5x18
-

Es =
0 .
%
-
STOPRING
Eat APPROXIMATE PERCENT RELATIVE ERROR
CRITERION E ,

APProx :
-
approx : -1
En %
=

x 100
approx :

TAYLOR SERIES

Any function be approximated


smooth can as a
polynomial

f(x)
Prcx)=Efx-xo

Pr(x) f(x)

!
=

f(x)(x
facx -x
+

x0
xd
+
-

- ... +
-

TRUNCATION ERRORS

·
occur
by voing an approximation instead of an
analytical procedure

You might also like