Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

48
UNCERTAINTY MANAGEMENT IN NUCLEAR ENGINEERING HYBRID FRAMEWORK FOR VARIATIONAL AND SAMPLING METHODS Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University SAMSI Program on Uncertainty Quantification: Engineering and Renewable Energy RTP, NC September 20 th , 2011

description

Uncertainty Management In Nuclear Engineering Hybrid Framework for Variational and Sampling Methods. SAMSI Program on Uncertainty Quantification: Engineering and Renewable Energy RTP, NC September 20 th , 2011. Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area - PowerPoint PPT Presentation

Transcript of Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Page 1: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UNCERTAINTY MANAGEMENT IN

NUCLEAR ENGINEERINGHYBRID FRAMEWORK FOR

VARIATIONAL AND SAMPLING METHODS

Hany S. Abdel-Khalik, Assistant ProfessorPI, CASL VUQ Focus Area

North Carolina State University

SAMSI Program on Uncertainty Quantification: Engineering and Renewable Energy

RTP, NCSeptember 20th, 2011

Page 2: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

MOTIVATION: ROLE OF MODELING AND SIMULATION Science-based modeling and simulation is

poised to have great impact on decision making process for the upkeep of existing systems, and optimizing design of future systems

Two main challenges persist: Why should decision makers believe M&S

results? How to be computationally efficient?

Page 3: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

OBJECTIVE: UNCERTAINTY MANAGEMENT Employ UQ to estimate all possible outcomes and

their probabilities

Identify key sources of uncertainty and their contribution to total uncertainty Must be able to calculate the change in response due to

change in sources of uncertainty (sensitivity analysis SA)

Employ measurements to reduce epistemic uncertainties Must be able to correct for epistemic sources of

uncertainties to minimize differences between measurements and predictions (inverse problem, aka data assimilation DA)

Page 4: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

SOURCES OF UNCERTAINTY Input parameters

Parameters input to models are often measured or evaluated by pre-processor models

Measurements and/or pre-process introduce uncertainties Parameters uncertainties are the easiest to propagate

Numerical Discretization Real complex models have no closed form solutions Digitized forms of the continuous equations must be prepared Numerical schemes vary in their stability and convergence properties For well-behaved numerical schemes, numerical errors

can be estimated Model form

Models are approximation to reality The quality of approximation reflects level of insight into physical

phenomena. With more measurements, physicists are often able to

formulate better models Most difficult to evaluate especially with limited measurements

Page 5: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UNCERTAINTY MANAGEMENT

Input Parameters

Outp

ut

Resp

onse

s

Page 6: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UQ APPROACHES APPLIED IN NUCLEAR ENGINEERING COMMUNITY1. Sampling approach

I. Analysis of variance, Scatter plots, Variance based decomposition

II. Efficient sampling strategies2. Surrogate (ROM) approach

I. Response Surface MethodsI. Employing forward model only

Polynomial Chaos Stochastic Collocation MARS

II. Employing forward and adjoint models Gradient Enhanced Polynomial Chaos

II. Variational Methods via adjoint model constructionIII. Hybrid Subspace Methods

a. Response Surface Methods + Variational Methods

Page 7: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Nuclear Engineering Models

Page 8: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Nuclear ReactorDevice that converts nuclear energy into

electricity via a thermodynamic cycle.Nuclear energy is released primarily via

fission of nuclear fuel.Physics governing behavior of nuclear

reactor include:Radiation transportHeat transport through the fuelFluid Dynamics and Thermal analysis

(Thermal-Hydraulics)ChemistryFuel performanceEtc.

Page 9: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Nuclear Reactions Interaction of single nuclear particles

cannot be predicted analytically. However only ensemble average of

interactions of many particles can be statistically estimated.

The constant (cross-section) characterizes probability of interaction between many particles of type A and many particles with type B; and are experimentally evaluated.

A BA B C D R N N

Page 10: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Cross-Section Resonances (Example)

U238 cross-section uncertainty in resonance region leads to 0.15% uncertainty in neutron multiplication ($600K in Fuel Cycle Cost)

21 eV37 eV66 eV

Page 11: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Core Design Heterogeneity

Source: http://www.nei.org

FuelGap

Clad

Uranium is contained in Ceramic fuel pellet

Stack is contained in metal rod

Rods are bundled together in an assembly

Fuel pellets are stacked together

Assemblies are combined to create the reactor core

Page 12: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Physical ModelThe ensemble average of neutron

distribution in a reactor can be described by Boltzmann Equation:

/ / / / / /

4 0

/ / / / / /

4 0

1 ( , ) ( , , , )

( , ) ( , , , )

( ) ( ) ( ) ( , , , )4

( , , , )

t

s

f

r E r E tt

d dE E E r E t

E d dE v E E r E t

s r E t

Page 13: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Nuclear Reactors ModelingWide range of scales:

energy, length, and time, varying by several orders in magnitude

Wide range of physics

Fully resolved description of reactor is not practical

Physical Model Reduction adopted to render calculations in practical run times

FuelGap

Clad

Uranium is contained in

Ceramic fuel pellet

Stack is contained in

metal rod

Rods are bundled together in an

assembly

Fuel pellets are stacked together

Assemblies are combined to create the

reactor coreSpatial Heterogeneity of nuclear reactor core Design

Cross-Sections dependence on neutron energy

Page 14: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

ROM via Multi-Scale ModelingGiven problem complexity, subdivide

problem domain into sub-domains

H

( , , ) ( , )i i i i i i if f f f f f fT x y x

Hi

Sub-domain, generally involving different physics, scale, and mathematical representation, and based on assumed boundary conditions.

Hi

( , ) ( )f f f f fT x y x

Page 15: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

ROM via Multi-Scale Modeling (Cont.)Coarse-scale model describes

macroscopic system behavior

H

( , ) ( )c c c c cT x y x ( , )i i i ic c f fx x y

Hi

Sub-domain solutions are integrated to calculate coarse-scale parameters for the coarse-scale model.

Hi

Page 16: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Uncertainty Management

Page 17: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

MATHEMATICAL DESCRIPTION Most real-world models consist of two stages:

Constraints:

Response:

Example:

, 0x

,R x

( ). aD z z z z S z

and dR z z dz

where a dx z D z z S z

Page 18: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UNCERTAINTY MANAGEMENT REQUIREMENTS To estimate uncertainty and sensitivities to

enable UQ/SA/DA, one must calculate:

R RR xx x

Indirect EffectDirect Effect

d d

d

R z zR xx R z

R z zR x

x R z

Page 19: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

determined by user-defined ranges for possible parameters variations

variation in state due to parameters variations;

requires solution of forward model

describes how responses of interest depend on

the state; easiest to determine for a given response function

only quantity needed by UQ

must be available for SA and DA

RR xx

x

x

R

R

Rx

Page 20: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Sampling approach Sample x and determine and R Perform statistical analysis on R Employ (x, R) samples to estimate sensitivities of R wrt x

Surrogate (ROM) approach Response Surface Methods (RSM)

Use limited samples to find a ROM relating R and x Sample the ROM many more times to get UQ results

Variational Methods Bypass the evaluation of , and directly find a ROM relating

R’s first order variations wrt x. Use deterministic formula to get UQ; no further samples required

Hybrid Subspace Methods Employ variational methods to find first-order ROM Sample ROM to find reduced set of input parameters xr

Use RSM to relate R and xr and get UQ results

RR xx

Page 21: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

RSM VS. VARIATIONAL APPROACH:DEMO TOY PROBLEM Constraint:

Response:

Adjoint Problem: Response: ‘solved once for a given response’ ‘All possible response

variations can be estimated cheaply’

1 2

1 2

2 35 4

73

x xx x

1 27 7x x R

723 4 71

5 1

7

31 1 10R

Page 22: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

VARIATIONAL APPROACH FOR UNCERTAINTY MANAGEMENT Given a well-behaved model, Taylor-series

expand:

Given first-order derivatives evaluated by VA, the surrogate is given by:

Employ the surrogate in place of original model for UQ, SA, and DA

0 01

( ) H.O.T.n

i ii i

yy f x y x xx

0 01

nsurrogate

i ii i

yy y x xx

Page 23: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

VARIATIONAL APPROACH Can be used to estimate first order variations

of a given response with respect to all input parameters using a single adjoint evaluation

For models with m responses, m executions of the adjoint model are required

For linear models (or quasi-linear models), it is the most efficient approach to build the surrogate

For higher order variational estimates (applied to nonlinear models), the number of adjoint evaluations becomes dependent on n. Ex. for quadratic models, n adjoints are needed.

Page 24: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

CHALLENGES OF RSM APPROACH Hard to determine quality of predictions at any

points not used to generate the surrogate? Solution: Leave-some-out Approach

Generate the surrogate with a reduced number of points

Use the surrogate to predict the left-out points Determine the surrogate’s functional form

(surface)? How to select the points used to train the

surrogate? Number of points grow exponentially with number of

input parameters Great deal of research goes into reducing number of

training points

Page 25: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Challenges of UQ in Nuclear EngTypical reactor models require long

execution times rendering their repeated execution computationally impractical:◦ Contain millions of inputs and outputs◦ Require repeated forward and/or adjoint

model executions◦ Strongly nonlinear◦ Coupled in sequential and/or circular

manners◦ Based on tightly and/or loosely coupled

physics◦ Employ multi-scale modeling phenomena

Responses’ PDFs deviate from Gaussian shapes, and must be accurately determined for safety analysis

Page 26: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Efficient Subspace Methods - PhilosophyGiven the complexity of physics model, multi-

scale strategies are employed to render practical execution times

Multi-scale strategies are motivated by engineering intuition; designers often interested in capturing macroscopic behavior

Multi-scale strategies involve repeated homogenization/averaging of fine-scale information to generate coarser information

Averaging = Integration = Lost degrees of Freedom

Why not design our solution algorithms to take advantage of lost degrees of freedom?

If codes are already written, why not reduce them first before tightly/loosely coupling them, and to perform UQ/SA/DA

Page 27: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

ESM: Toy ProblemConsider:

i

j

a

b1 2 3x x i x j x k

k

2 41 1 2 2 3 3 1 1 2 2 3 3

2 4T T

y a x a x a x b x b x b x

y a x b x

1 2 3( ) ( , , )y x y x x x

2

R : active subspace

na b A

A

Page 28: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

ESM: Toy Problem

1 11 1 1 1

1 1

x a b a b

x a b a b

A

P AP

2 4T Ty a x b x

2 4

1 11 1

1 1

,T Ty a b y

AP AP

Original Model:

Reduced Model:

Reduction Step:

Page 29: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Efficient Subspace Methodology (ESM)Consider:

Note that:

i

j

a

b1

2

3

x x ix jx k

k

21 1 2 2 3 3

41 1 2 2 3 3

2 4

T T

y a x a x a x

b x b x b x

y a x b x

1 2 3( ) ( , , )y x y x x x

2 2 { , }T Tdy a x a b x b LC a bdx

Page 30: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Tensor-Free Taylor ExpansionIntroduce modified Taylor Series

Expansion:

This expression implies:

0 1 2 21 , 1

3 3 3, , 1

( ) ( ) ( ) ( )

( ) ( ) ( ) ...

n nT T T

i i ij i ji i j

nT T T

ijk i j ii j k

y x y x x x

x x x

1,...,1,...,

iijj n

dy LCdx

Page 31: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Subspace Reduction AlgorithmAssume matrix of influential directions is

known

One can employ a rank revealing decomposition to find the effective range for

Range finding algorithm may be employed:Employ random matrix-vector products of the

form:

Find the effective range:

Check the error:

11 12 ... nn

A

, 1,...,i iq i r A

1 ... n rrq q Q

1,...,

210 max T Tii s

I QQ A I QQ A

A

Page 32: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Subspace Methods - AlgorithmI/O variability can be described by matrix operatorsGiven large dense operator A, find low rank

approximation:

Matrix elements available:A. Frieze, R. Kannan, and S. Vempala, Fast Monte Carlo

algorithms for finding low rank approximations, in Proc. 39th Ann. IEEE Symp. Foundations of Computer Science (FOCS), 1998.

______, Fast Monte Carlo algorithms for finding low-rank approximations, J. Assoc. Comput. Mach., 51 (2004)

Only matrix-(transpose)-vector product available:H. Abdel-Khalik, Adaptive Core Simulation, PhD, NCSU 2004.P.-G. Martinsson, V. Rokhlin, and M. Tygert, A randomized

algorithm for the approximation of matrices, Computer Science Dept. Tech. Report 1361, Yale Univ., New Haven, CT, 2006.

1

rT

i i ii

s u v

A

Page 33: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Singular Values Spectrum

Singular Value Triplet Index

Sing

ular

Va

lue

r

r

u

1s

2s

3s4s

rs

How to determine a cut-off?

Well-Posed

Ill-PosedIll-

Conditioned

Page 34: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Subspace-Based HybridizationApproach #1Methods Hybridization inside each

componentsReduce subspace first, then employ

forward method to sample the reduced subspace

nx my

Random Sampling of

1st Local Derivatives

Find Reduced

Input Parameters

( )r rx Original

ModelMapping

Page 35: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Subspace-Based HybridizationApproach #2Hybridization across componentsEmploy different method(s) for each

components, and perform subspace reduction across components interface

nx k my

Find Reduced

Parameters

( )r r

Mapping

Page 36: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Implementation – Subspace MethodsGiven a chain of codes, one attempts to

reduce dimensionality at each I/O hand-shake

nx k my

nx ( )r r my

ESM Reduction

Page 37: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

BWR REACTOR CORE CALCULATIONSHYBRID SUBSPACE SAMPLING APPROACH, W/ LINEAR APPROXIMATIONBASED ON WORK BY MATTHEW JESSEE, HANY ABDEL-KHALIK, AND PAUL TURINSKY

MG XS

Lattice Calcs

FG XS

Core Calcs

keff, power, flux, margins,

etc.

ENDF MG Gen Codes

6# of Data > 10 410

610

510

Runtime ~ mins

hrs

mins

Page 38: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UQ AND SA RESULTS CORE K-EFFECTIVE & AXIAL POWER DISTRIBUTION

0 2 4 6 8 10 12 14 16 18 200

1

2

3

4

5

6

7x 10

-3

Exposure (GWD/MTU)

Rela

tive

Stan

dard

Dev

iatio

n in

k-e

ffect

ive

U-238Pu-Am-CmGdU-235Total

0 5 10 15 20 250

0.005

0.01

0.015

0.02

0.025

Stan

dard

Dev

iatio

n in

Nod

al P

ower

0 5 10 15 20 250

0.5

1

1.5

2

2.5

Rela

tive

Noda

l Pow

er

Axial Position Bottom -> Top

U-238Pu-Am-CmGdU-235Total

Page 39: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

DA RESULTS W/ VIRTUAL PLANT DATAPOWER DISTRIBUTION

Page 40: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

DA RESULTS W/ REAL PLANT DATACORE REACTIVITY

Page 41: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

SINGULAR VALUES FOR TYPICAL REACTOR MODELS

Page 42: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UQ STATE-OF-THE-ART:WHAT WE CAN DO! Linear or quasi-linear models with:

few inputs and many outputs: smpl many inputs and few outputs: var many inputs and many outputs: hbrd var-smpl-sub

Nonlinear smooth models with: few inputs and many outputs: smpl, rsm

smpl: sampling methods rsm: response surface methods hbrd: hyhrid var: variational sub: subspace

Page 43: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UQ ONGOING R&D Nonlinear smooth models:

with many inputs and few/many responses: hbrd var-smpl-sub

Linear models coupled sequentially: Possible to reduce dimensionality of data

streams at each code-to-code interface: hbrd-var-sub

Nonlinear models coupled sequentially: Possible: perform reduction at each code-to-code

interface using a hbrd var-smpl-sub

Page 44: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UQ CHALLENGES: CURRENTLY NOT ADDRESSED Nonlinear non-smooth models

(e.g. bifurcated models and discrete type events) Nonlinear models coupled with feedback How to estimate uncertainties for low-probability

events, e.g. tails of probability distributions? How to evaluate uncertainties on a routine basis for

multi-physics multi-scale models? How to efficiently aggregate all sources of

uncertainties, including parameters, numerical, and model form errors?

How to identify validation domain beyond the available experimental data?

How to design experiments that are most sensitive to key sources of uncertainties?

Page 45: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

CONCLUDING REMARKS Most complex models can be ROM’ed. This is

not coincidental due to the multi-scale strategy often employed.

Recent research in engineering and applied mathematics communities has shown that: It is possible to find ROM efficiently One can preserve accuracy of original complex

model

Hybrid algorithms appear to have the highest potential of leveraging the benefits of various ROM techniques

Page 46: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

UQ EDUCATION Very little focus is given to UQ in

undergraduate and graduate education Future workforce, expected to rely more on

modeling and simulation, should be conversant in UQ methods

Ongoing educational efforts: Validation of Computer Models, Francois Hemez, LANL SA and UQ Methods, Michael Eldred, Sandia V&V & UQ, Ralph Smith, NCSU V&V&UQ in Nuclear Eng, Hany Abdel-Khalik, NCSU

Page 47: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Thank you for your attention

[email protected]

Page 48: Hany S. Abdel-Khalik, Assistant Professor PI, CASL VUQ Focus Area North Carolina State University

Tensor-Free Generalized ExpansionIntroduce modified Taylor Series

Expansion:

This expression implies:

0 1 2 21 , 1

3 3 3, , 1

( ) ( ) ( ) ( )

( ) ( ) ( ) ...

n nT T T

i i ij i ji i j

nT T T

ijk i j ii j k

y x y x x x

x x x

1,...,1,...,

iijj n

dy LCdx