Appunti Statistica per la sperimentazione e le previsioni in ambito tecnologico (Parte 3)

ASCENT-DISCENT

STEEPEST its

statistical computation

model

RS

auber and

Ist flue

i laving dome DoE

=

* (x) estimate the

bo model

bix

+. we

:

= -2)

w)

Xx) :

bo

Q bixi

Xi +, x

=

, -

-

,

..., . . . .

System equations partial deciates

and

of limeau

fin NOT

= a System

indep.

Gegwange multiple

e

deciates the

the system

to

partial

By setting equal equations

of

following +1

teus ,

abtained

is an

: abi k

i

- ...

1

=

G

The to the

by following

ascouting goumula

given

solution is Xi etimated e

reggie

wwoning

&

b

30/extime

00-30

00

30 Stime

, =

2 Stendentite si

,

this

-In case 104 temp the trajectory

0 0 to

40 =

00

, 2 increment

test Imidrange) it

and

abituary moted that

must

estimate be

skin I :

, ,

Fil/Ex trajestay

them /steepest

the

/ the direction

depembs Xi

Xi mange

· on

: :

-

= ,

to

related Sxi

is MIDRANGE DOES

That

Note THE the

to

NOT cor respond

; ARBITRARY INCREMENT

to the to

must

value

compute be computed equal

of

X

· h

: :

,

i

Theregame not must

equations

fo linearly

have independent

system

because of

we we

,

a

,

abituarly teclinical

i me increment restablished

facto

assign for generis , on

an

a

, fluat

considerations) Di so :

, · Montgomery

In ex

.

>

that

So : smin

& Stime

fixed =

we

at

I la

must nevo

gix a

to estimate

A w

one

the estimated obtained

abtained the

consequently I

value

Once values

of a re

w ,

the factors

K-1

remaining

you variables .

PROCEDURE CORNELL

KHURI AND EXAMPLE

AND le

↓

gives

me e n

leve e

to

start

which I

optimize

intersent proded

nidui

Time us

a

stime stime s

1

12 =

a uss -

=

= ,

=

midi trime

S

Gode& calculation multiplier

Igrompo

puoi

12

fixing

By a , D1

Coded

and

calculate w

we ↓

themflue to the

offer to

back

I in go non %

values multiply

vare

coded coded x10

. Sx temp

start center

the

used to

a from

=> the

the direction

and

design along

of move

Xo fount

maximum first DoE

with my

the

I Montgomery

of

from ex.

Istamdaub) usually lave

you center

the

to replicate new 2

there

and a re

lere

stop the

also in

replicates venteres because these is have

- We

wide difference

a => abs

10

between

flese stuials

2 Men

- ugwee

max

replicates entr

center

the

in > & amplie b

Model :

V

otlen

8

7 ↓

, maints

two 6

obtaine

design

the Polynonial model

of S series approx

Taylau

applying a be

model

around simple

Xo , effects

I only include linear

X1 X2

,

e se

Estimated

& me

the Steemest

center

replicates the

I

When do of AsDes

in optimizing

method the 15

/

design expect obtained

to only

close cuter

of

values in

very

the

replicates

therwise ventures

& In

new

: to

stopping

rule I'm the

When

- close

very

:

sofficient between the

and

relationship max/min

angulare considering

=>

twajest that

coefficients

B1 lack

Be fiftess if

and saying

of

any me

,

the pantial acceptable

deciates model

↳ is

a re my

the the

-test lack

of surface of fit

of

tests -test the

of

compresary : curvature

dome

to be

at the

emb considering

design

Add new

a belong to

t thaf

but

m 8

6

. ,

twajesay

the to

neused

a re

le desig e se

complete new

o

=>

also far we

2

lave wen o m

z veut

the .

next

the

at ber

stop ne se

>

center the design

of ,

where do

I amothen Dat the

explating

e se

holints

flua

twajestay and

2VP twajectory abbr just

19 story

twaje 101

points (0

2 ,

doesn't place

It

flue would

I

center lere a s

situation

standard

in a asting

by

replicat

be goes

↳

↓ a naint 2

He build

did not test

design but of

mew bothe

If

a

fast

fuel starte devature

and

e twajectory

22wither design significant

tests

V a re

simple least

at

lave one

I

eache

for 2wo

2 sen auber

of

venteres that

effect is

rignificant the

to

I go

>

- 2nd ades

Fin important

In fluis decomposition

55 ROM is

modee

~ because

S3 to t eurou

fune

leskes, sere

1) quantify

SPEDoE stuictly

of

)

= va r .

& the

fits to

related DoE because

SS black of we

of the

replicates

with

dealing in

a re center design

the

of x pot

the

applied

model I

If

SSOF- correctly

lave

= the

the

at

model end

muder of

contain

it the

should

wos

variability to

related the fast

that I have applied 15t

a

21

instead

auder of a

interactions

lack and

of

>

- quatuativo effect

Esempio

Esempio

RSM PASSAGE THE

1 ST IND

THE

FROM TO

ORDER

: the

As decomposed

the "

well variable

known Smodel

is

se in Stres

of response .

:

, decomposed

The then

point theat Stef Spe

crucial is

is stres

in RSM in +

:

Deviance/

flue related to

silof the Lack-of-Fit

is tor the

/an

where 55 good-of-

e g .

, .

last lot

with

In

model bealing

this

1st

the model

fitness auder awber

of way a re

we

,

. deniance

the

along descent the

ascent

applied Instead

procedure the

steepest is

sige

,

.

the

through besign senter coded

replicates the

calculated Where

in in

is

e Xo

xo g no , ,

. .

values . the

theat to

Please the

note is

while

strictly model

related sipe

the

is

stlof ,

the

to the

stuictly behends the

related replicates

benign spe

and m o reve r .

on no

,

, to tests

for two hypothesis

it the

is

to If auder

In auter verigy 2nd

necessary pass ,

the

be

must test

The

but

carried second curvature

first fest

,

LOF

is named one

one ,

. the

mandately to

is venify passage

.

test lack-of-gif &

model

fitness

Ho-good the

of Geop/agee

fitness model

the

H1 of

bad

= SSpe/Afpe Afpel

If the For ~/

them distribuited

accepted Afeg

statives

Ho is

is , ,

famatheastone

test

Curvature

Bii

Ho Vi

o

=

=

Abito

If Effel

them distribuited

the

accepted statistines For

Ho n

is

is 1

& ,

,

FINAL DECISION tests they

them be jointly to

calculated must evaluated

Once both venify

in auber

a re ,

,

flue

to Fore

setting

the RSM semamide

auter

possible 2nd decove

passage can :

. to

tests

both must

them the

lot significant 2nd dubeu

semauid pass

· .

we

: a re ,

tests the

bothe them

significant the

antimal levels

seting

2nd factor

scemamid of

no ,

a re

· : , ascent the

descent

the steepest

through 1st final

obtained mussede

ader is

,

optimal setting . test

significant this

the not

the

test while

Flo is significant

Foru

3rd is

Scenaus in

:

. , staut

when

several variabilities

surely missed of

one d soluses

we

case the DoE

planning . significant this

the

the test

test significant

while

not

uthe Flot is

Ferar

is

scemavid :

· ;

situation the boutenline

it

If

should namely purbably Fot is

docur decurs ,

.

to .

significant

close be

e very

g .,

.

·

comp