Beruflich Dokumente
Kultur Dokumente
Zeyun Wu and Hany Abdel-Khalik Department of Nuclear Engineering North Carolina State University November 12th, 2012
Page 11 of 18 Page of 21
Introduction
Sensitivity Analysis (SA) determines the most influential components in the input parameters corresponding to the output responses, meanwhile it can provide valuable information to facilitate Uncertainty Quantification (UQ) procedure. Perturbation theory (PT) and Generalized perturbation theory (GPT) provides an efficient tools to estimate response sensitivities in nuclear reactor calculations for decades. Recently a forward based GPT-Free method is introduced and intended to perform response sensitivity analysis without the need to solve adjoint equations associated with GPT applications. GPT-Free method has been successfully implemented to both deterministic and Monte Carlo modeling problems and results shown in previous ANS meetings verifies its feasibility and efficiency.
Page 2 of 21
Sensitivity:
dk d
Sensitivity:
dR d
Page 3 of 21
Limitations of GPT
In applications where both the number of input parameters and output responses are significantly large, GPT can become computationally intractable due to the considerable number of adjoint calculations needed. For those engineering systems that are modeled stochastically, e.g., the Monte Carlo particle transport model commonly used in reactor analysis benchmark calculations, there currently exists no general extension of GPT theory.
GPT-free method is developed to address these limitations.
Page 4 of 21
GPT-Free: Theory
1. The system multiplication, k, can be written as an unknown function of the state-space (neutron flux), f k = f ( ) Consider a response functional that is an inner product of some cross-sections, , and the flux R = ,
2.
3.
The multiplication may be implicitly related to all generalized responses of interest, described mathematically as: k = f ( R1 Rm ) 4. Differentiate with respect to cross-sections m dk f dRi General Response Fundamental = Sensitivity Profiles Sensitivity Profile d i =1 Ri d
ANS Winter Meeting, Nov. 11-15, 2012, San Diego, CA Page 5 of 21
Subspace:
dk span d
dR2 d
dk d
dk d
dR1 d
4
dk d
dk d
Page 6 of 21
K&
= ( I Qr QT r )
& = ( QQT )
K &
K K K K f ( 0 + ) = f ( 0 + & ) !!!
ANS Winter Meeting, Nov. 11-15, 2012, San Diego, CA Page 7 of 21
K&
Then,
f ( 0 + ) f ( 0 +
K&
i =1
i =1
r r K K if linear K K K K f q f = f 0 + i qi = f ( 0 ) + i + ( ) ( 0 0 ) i i =1 i =1
r << n !!!
ANS Winter Meeting, Nov. 11-15, 2012, San Diego, CA Page 8 of 21
Page 9 of 21
Epistemic vs. Aleatoric uncertainty Two uncertainties are reduced by distinct strategies Independence of the two uncertainties
a e
X
Independnce of a and e
ANS Winter Meeting, Nov. 11-15, 2012, San Diego, CA Page 11 of 21
3. 4. 5.
Simplified!
Page 12 of 21
Case Study
Page 13 of 21
Page 14 of 21
n = 68306
Reference eigenvalue:
k =1.0723 0.0001
ANS Winter Meeting, Nov. 11-15, 2012, San Diego, CA Page 15 of 21
dk d
3.
QR = dk d
4.
n r \ ... dk d r
Evaluate the -metric; increase r until the error of the metric is below user-defined tolerance
= kpert kapp
Where,
200
150
kpert = k ( 0 + )
kapp = k ( 0 + ) K K T = k 0 + ( Qr Qr ) K K&
100
50
100
200
300
600
700
800
900
= ( I Qr QT r )
& = ( Q r QT r )
k eff
Page 18 of 21
th
10
15 Case number
20
25
30
Page 19 of 21
Averaged Computing 140.40 18.00 Time (min) Total Computing 86910.50 11143.70 Time (min)
ANS Winter Meeting, Nov. 11-15, 2012, San Diego, CA Page 20 of 21