Statistical analysis analysis of simulation of simulation - - PowerPoint PPT Presentation

▶
statistical analysis analysis of simulation of simulation
SMART_READER_LITE
LIVE PREVIEW

Statistical analysis analysis of simulation of simulation - - PowerPoint PPT Presentation

Mar 03, 2024 333 likes •689 views

Statistical analysis analysis of simulation of simulation Statistical experiments: : experiments Challenges for industrial industrial applications applications Challenges for Bertrand Iooss Bertrand Iooss

slide-1
SLIDE 1
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

Bertrand Iooss Bertrand Iooss

  • Statistical

Statistical analysis analysis of simulation

  • f simulation

experiments experiments: : Challenges for Challenges for industrial industrial applications applications

slide-2
SLIDE 2
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

Present Present NPP NPP GEN GEN-

  • III

III GEN GEN-

  • IV

IV

Continuity / present NPP Technological rupture

Main industrial stakes for Nuclear Energy

  • !"!

#$

slide-3
SLIDE 3
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

The DEN Simulation Platform for Nuclear Applications % %& %

An Open Source Platform for building multi-physics and multi-scale industrial simulation tools from CAD to post-processing An Open Source Platform for building multi-physics and multi-scale industrial simulation tools from CAD to post-processing

SALOME (open – source):

  • Preprocessing
  • Postprocessing
  • Supervision

!"#$%&!'(#%&)*+ + #* #%&$"*!%,)*+*!"$)+ #'

Fragment de fission Diffusion Remise en solution Création Piégeage Création Diffusion Piégeage Recombinaison Flux au JDG Pore intergranulaire Pore intragranulaire Lacune Interstitiel
  • "&+*)",*"

.%+ "#"/0+) 0#'

Crystal volume Grains Atomic clusters Atoms Dislocations Mechanics

$% "#)%&+*)",*"

Experimental data

*0#"/!'+)*+

Uncertainties (URANIE - ROOT)

& ! '

slide-4
SLIDE 4

1 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

Outline

* () % * +) ,

  • ) &
slide-5
SLIDE 5

2 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

. / & 33 & && , & * & / 01 & 2 / 3

1.0 On uncertainties in simulation experiments

  • 4(

4+ 5

  • 4(

4+ 5 678

9

0 4 3 3 33 43 5

slide-6
SLIDE 6

6 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss Decision criterion Ex: Probability < 10-b Feedback process

Step B: Quantification

  • f uncertainty

sources

Modelisation with probability distributions Direct methods, statistics, expertise

Model (or measurement process)

f(x,d)

Model Model (or (or measurement measurement process process) )

f(x,d)

Input variables

Uncertain : x Fixed : d

Input Input variables variables

Uncertain : x Fixed : d

Variables

  • f

interest

Y = f(x,d)

V Variables ariables

  • f
  • f

interest interest

Y = f(x,d)

Step Step A : Problem specification A : Problem specification

Quantity of interest

Ex: variance, probability ..

Quantity Quantity of

  • f

interest interest

Ex: variance, probability ..

Step C : Propagation of uncertainty sources Step C’ : Sensitivity analysis, Prioritization

Observed variables Yobs Observed Observed v variables ariables Yobs

Step B’: Quantification of sources

Inverse methods, calibration, assimilation

1.1 Bring clear (even schematic) methodology

:;) +<<=>

slide-7
SLIDE 7

7 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

  • (<

(<<<

? (

  • % @
  • %

#&

  • $3

+

%

"3 8* 3

" A! /!$

<

  • ,3

3

  • +

9 3

( 3

;2 ? *3: 3

+

+

  • $;*3

1.2 Give clear classification of most useful methods

slide-8
SLIDE 8

< ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

  • (<

(<<<

? (

  • % @
  • %

#&

  • %

&

+

% " A! /! & A! /!$

<

  • ,3

3

  • 9

3

;2 ? *3: 3

+

(<<

  • !

'

  • !

'

  • 0+

2# 2#

  • @

))

  • ;BBB

))

  • ;BBB

;BC ;BBB DEEEF F ! %"%$ ! E% %"( $ % "$ C

!8 =

1.3 Reserve details for (very) advanced practicioners

slide-9
SLIDE 9

> ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

1.4 Offer strong services

" F & & # /; "G H$ 4 F #& EE"/ $ )? /* I 2 I

slide-10
SLIDE 10

@ ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

Outline

* () % * +) ,

  • ) &
slide-11
SLIDE 11
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

The industrial stakes for nuclear PWR severe accident study

)3 $A /+% 8 B& & "#& && $J .& & J B& & !!J 2 / #"?D;$! & ,& #"K;B$

  • rejets de produits radioactifs

dans l’environnement rejets de produits radioactifs dans l’environnement

  • Fusion du cœur

du réacteur Fusion du cœur du réacteur

  • Conséquences radiologiques sur la

population et l’environnement Conséquences radiologiques sur la population et l’environnement Bâtiments auxiliaires Bâtiment réacteur

  • rejets de produits radioactifs

dans l’environnement rejets de produits radioactifs dans l’environnement

  • Fusion du cœur

du réacteur Fusion du cœur du réacteur

  • Conséquences radiologiques sur la

population et l’environnement Conséquences radiologiques sur la population et l’environnement Bâtiments auxiliaires Bâtiment réacteur

!! !!

slide-12
SLIDE 12
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

, #&& / * I B & I K &!#

Main objectives of these uncertainty studies

+8 +8 " E/ ,E/$

  • )

) 8 8 F/ F

  • 3

3

  • 3

3 4 48 8 . & 3 & ! &

  • &

!* / 2# & !

slide-13
SLIDE 13
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3 use levels – 1) Precise and punctual needs

,*/ !& !! '* &#L&/ 0 & & ; /)B*++ E (! "&#0!$ A8<< "?M$ !! !%E& " #& #L$0" #& #L$A+)8 F&2 &&

./

  • /

/!,

slide-14
SLIDE 14

1 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3 use levels – 2) Detailed analyses for confirmed users

,*/ /!"$ '* #&& /! &

;!! ;!!

  • &#

&# .# .# & &

  • (

) ( )

Z X SRC

i i i 2 2 2 i

) sgn( σ σ β β =

slide-15
SLIDE 15

2 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3 use levels – 3) Help to physical model developers

,*

? ? & & , ,

  • &

& / / ! !

8<<% B&#&& &# &&

slide-16
SLIDE 16

6 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3 use levels – 3) Help to physical model developers

,*

? ? & & , ,

  • &

& / / ! !

8<<% B&#&& &# && !#!

"F&2 )' $

slide-17
SLIDE 17

7 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

Conclusions of this small success story

%!L/ &?D;/* I F! ! 2# & ! / ! EE# 4 *?D; & !" K;B$ ! ) I F! ! !, & #& ! /! ! & / / /&;DDF/ " !K;B$) # # # /0, & ")) ! 3$

slide-18
SLIDE 18

< ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

Outline

* () % * +) ,

  • ) &
slide-19
SLIDE 19

> ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

A lot of scientific challenges

() K " N N$ +) , " $

  • ) , / " $

O) F /!" N N $ 8) %, /0& & " $ P) & " $

slide-20
SLIDE 20

@ ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3.1 Combine many representations of inputs

,) " & &$ ; #2 &/ ', :6) +<<=> %&*! J 3 B ,J ,!! * !! 0N & "N$0B !! "B$ '* %&!&!! # #2 &/! , NB ,

slide-21
SLIDE 21
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3.2 Best initial designs for metamodel fitting

N#& ,J ,*?M ?M"E! 0E!$ #E *3 3 )3 35

?M ?M ?M ?M#& #& # # # #E E

  • 6

6+

+

6 6+

+

  • !E 8E ' :% +<<=>
slide-22
SLIDE 22
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3.2 Best initial designs for metamodel fitting

N#& ,J

  • &&

)3 35 F ,& "?M0&0,0# 03$ ! &% F :) +<<7>

Dimension 20

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.2 0.5 0.8 1.1 1.4 1.7 2 m Aléatoire Strauss Amas Faure Halton Hammersley Niederreiter Sobol Hypercube Latin (HL) HL maximin HL modifié WSP HL_fd_100 HL_fd

:C) +<<7>

slide-23
SLIDE 23
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

(<</&, &

*#&&& C D@1° ° ° °*

?!2

  • 3.3 Estimation of extremal events

!* / # !! && %& * D;%0D;% % / % %, %: )+<<=> ' :D2 +<<O> ,' , /& 3

slide-24
SLIDE 24

1 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3.4 Functional data in computer experiments

# QA"B$ #& Q /! B / /! !43: 3

%#& # , Q"$A"4"$$ % * Q"/$A"4("($ 3 4"$ /$ / "! $ ,)/* ", 1$ F 3 ,)Q4* Q"/$A 4"$A# 4"$Q"/$!

slide-25
SLIDE 25

2 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

(P /!" R 3$ ! & ! ε", $ #? 3 3

50 100 150 200 250 300 50 100 150 200 50 100 150 200 250 300 50 100 150 200

+! &!

3.4 An example : hydrogeological model

"7<$"# $ B D

  • *
slide-26
SLIDE 26

6 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

%, & J I ?M' :! SM/2 T (777> I 1 :&S <U> / I ?!1 &#& "$* & 1 #& ! :?! SF +<<=> I & ! /!:5!1 +<<<> & I I 3

3.4 Open issues for functional inputs

slide-27
SLIDE 27

7 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

/

:!) +<<P ) +<<P M ) +<<= N ) +<<U % +<<=>

0/ %/ D1 & 'D %!L/ 1

3.4 Open issues for functional (and multidimensionnal) outputs

slide-28
SLIDE 28

< ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3.5 Mixing variational and stochastic methods

B0&0&03 &/ ! /! L&

?! %D " % &D$

F&2 )N "KT$

slide-29
SLIDE 29

> ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3.5 Mixing variational and stochastic methods

, V8<#& && V(<

. 4 4 E @@5

D ! ,E ,*// "L& $

:B2 +<<< %)(77->

% % , ,#& #& , , , ,#& #&

  • #&

#&

  • @

@// //

slide-30
SLIDE 30

@ ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

ν !G ! H /!ε, #& B , * ! &)) / " 3$ & " L 3$ #2 &/! & & / "L$

:C S%(77< & (7=7 5!1E%1& +<<< BS;!+<<7 N2/ +<<7>

3.6 Stochastic computer codes

& **ℜ ℜ B QA&"B$@ν"ε B*ε$ "QWB$@σ"QWB$ "QWB$ B Q

slide-31
SLIDE 31
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

3.6 Stochastic computer codes

/ Q"B$A"QWB $Q "B$AC"QWB $ # &/ C:Q"B ε$ >AC:Q"B$ >@:Q "B$ > / B * / ε * JJ/ &!# ε 4JJ * 0; B: ; ε ! % & !"QWB$ + 33 :;&) +<<7> ) ( Var )] Var[E( m Y X Y S

i

i X =

) ( Var ) ( Y Y E S

d

T = ε

slide-32
SLIDE 32
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

Conclusion: The main challenge: Place of « DACE » teams

/38 !43 33 33 3 33 3 F& ! ! X

: !"$ F& F& "8<$ +<<=>

B J & /# B J ! L B J ?2 & B J ?/# & +8 4 3 3 *3()*"3 & # /* !%K%"KR$ % "K$ F#$%+*0 ;,$-8444G;33G 3

slide-33
SLIDE 33
  • ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss
slide-34
SLIDE 34

1 ENBIS-EMSE 2009 Conference – Saint Etienne – July 2009 – Iooss

References

N ) F& +<<U N2/ S F& ; K/ +<<7 ) +<<= ;) K . +<<= ) , & +<<P ) &)B)?!) +<<7 M ) T +<<= BS;! ;! S +<<7 B2 ) ;2 +<<< ?! SF B)T'& B +<<7 % F&YBF +<<= %) F& (77- D2 T; +<<O ! SM/2 F& (777 6) F& +<<= ;&) F& ! &S ' +<<U & T; N (7=7 C) T +<<7 C S% T6 F& (77< 5!1E%1& F&YK/ +<<<