You are on page 1of 74

An efficient computational framework for uncertainty quantification in multiscale systems

Xiang Ma
Presentation for Thesis Defense (B-Exam) Date: Aug 18, 2010
Materials Process Design and Control Laboratory Sibley School of Mechanical and Aerospace Engineering 169 Frank H. T. Rhodes Hall Cornell University Ithaca, NY 14853-3801 Email: xm25@cornell.edu URL: http://mpdc.mae.cornell.edu/ Cornell University
Materials Process Design and Control Laboratory 1

Acknowledgement
SPECIAL COMMITTEE: Prof. Nicholas Zabaras, M & A.E., Cornell University Prof. Subrata Mukherjee, T & A.M., Cornell University Prof. Phaedon-Stelios Koutsourelakis, CEE., Cornell University Prof. Gennady Samorodnitsky, OR & IE., Cornell University

FUNDING SOURCES: Air Force of Scientific Research (AFOSR) National Science Foundation (NSF) Sibley School of Mechanical & Aerospace Engineering

Materials Process Design and Control Laboratory (MPDC)

Cornell University

Materials Process Design and Control Laboratory

Outline of the presentation Motivation: Coupling multiscaling and uncertainty analysis Mathematical representation of uncertainty Problem definition: Stochastic multiscale flow in porous media Adaptive Sparse Grid Collocation (ASGC) Method for the solution of SPDEs ASGC coupled with Adaptive High Dimensional Model Representation technique for high-dimensional SPDEs Mixed Finite Element Heterogeneous Multiscale Method Numerical Examples

Cornell University

Materials Process Design and Control Laboratory

Need for uncertainty analysis All physical systems have inherent associated randomness
SOURCES OF UNCERTAINTIES
Multiscale nature inherently statistical Uncertainties in process conditions Material heterogeneity Model formulation approximations, assumptions

Why uncertainty modeling?


Assess product and process reliability Estimate confidence level in model predictions Identify relative sources of randomness Provide robust design solutions

Cornell University

Materials Process Design and Control Laboratory

Why uncertainty and multiscaling Uncertainties introduced across various length scales have a non-trivial interactions.

Micro

Meso

Macro

Physical properties, structure follow a statistical description

Use micro averaged models for resolving physical scales

Imprecise boundary conditions Initial perturbations

Cornell University

Materials Process Design and Control Laboratory

Outline of the presentation Motivation: Coupling multiscaling and uncertainty analysis Mathematical representation of uncertainty Problem definition: Stochastic multiscale flow in porous media Adaptive Sparse Grid Collocation (ASGC) Method for the solution of SPDEs ASGC coupled with Adaptive High Dimensional Model Representation technique for high-dimensional SPDEs Mixed Finite Element Heterogeneous Multiscale Method Numerical Examples

Cornell University

Materials Process Design and Control Laboratory

Mathematical representation of uncertainty Math: Probability space (, F, P) Sample space Sigma-algebra Probability measure

Stochastic variable can be regarded as a function taking value in the stochastic space.
MAP

S
Sample space of elementary events

Real line

Each outcome is mapped to a corresponding real value

Collection of all possible outcomes

A stochastic process is a random field with variations across space and time.

X : ( x, t , )
Cornell University
Materials Process Design and Control Laboratory 7

Spectral stochastic representation A stochastic process = spatially, temporally varying random function

X : ( x, t , )

CHOOSE APPROPRIATE BASIS FOR THE PROBABILITY SPACE GENERALIZED POLYNOMIAL CHAOS EXPANSION SUPPORT-SPACE REPRESENTATION KARHUNEN-LOVE EXPANSION SPARSE GRID, CUBATURE, LH
8

HYPERGEOMETRIC ASKEY POLYNOMIALS

PIECEWISE POLYNOMIALS (FE TYPE)

SPECTRAL DECOMPOSITION

COLLOCATION, MC (DELTA FUNCTIONS) Cornell University

Materials Process Design and Control Laboratory

Stochastic collocation based framework


Need to represent this function

f:

Sample the function at a finite (i ) M set of points M = {Y }i =1 Use polynomials to get a approximate representation

If (Y ) = f (Y (i ) )ai (Y (i ) )
i =1

Function value at any point is simply If (Y ) Spatial domain is approximated using a FEM discretization. Stochastic function in 2 dimensions Stochastic domain is approximated using multidimensional interpolating functions
Cornell University
Materials Process Design and Control Laboratory 9

Collocation based framework for UQ


Pre A-exam developments:
1) Stochastic natural convection using generalized polynomial chaos expansion. 2) Preliminary studies on the effect of stochastic discontinuities. 3) A framework for solving stochastic partial differential equations based on adaptive sparse grid collocation method (ASGC).

Post A-exam developments:


1) An efficient framework for Bayesian inference approach based on ASGC. 2) High Dimensional Stochastic Model Representation (HDMR) technique for the solution of high-dimensional SPDEs. 3) Analysis of stochastic multiscale systems: Uncertainty across length- scales 4) Data-driven approaches to construct stochastic input models from limited information: explore the non-linearity structure of data using Kernel Principal Component Analysis (KPCA).
Cornell University
10

Materials Process Design and Control Laboratory

Outline of the presentation Motivation: Coupling multiscaling and uncertainty analysis Mathematical representation of uncertainty Problem definition: Stochastic multiscale flow in porous media Adaptive Sparse Grid Collocation (ASGC) Method for the solution of SPDEs ASGC coupled with Adaptive High Dimensional Model Representation technique for high-dimensional SPDEs Mixed Finite Element Heterogeneous Multiscale Method Numerical Examples

Cornell University

Materials Process Design and Control Laboratory

11

Flow through heterogeneous media


Necessitates stochastic multiscaling Ground water remediation, nuclear contamination and enhanced oil recovery Availability of enough input data and comparative results Pose and solve questions that provide some insights
Ground water remediation and contamination control

Oil recovery

Cornell University

Materials Process Design and Control Laboratory

12

Problem definition
Basic equations for flow and transport in a domain

with the boundary condition

Length scale of the system,

Length scale of permeability variation, Multiscale paradigm

l << L

But exact permeability unknown. Some statistics or limited data Permeability is a realization from corresponding probability space
Cornell University
Materials Process Design and Control Laboratory

K (, )
13

Karhunen-Love expansion
Karhunen-Lov expansion Based on the spectral decomposition of the covariance kernel of the stochastic process

1. Representation of random process - Karhunen-Loeve, Polynomial Chaos expansions 2. Infinite dimensions to finite dimensions

Random process

Mean Set of random variables to be found

- depends on the covariance

Need to know covariance Converges uniformly to any second order process


Eigenpairs of covariance kernel

Set the number of stochastic dimensions, N Dependence of variables

K ( x, ) = K ( x, Y1 , , YN )
N N

Pose the (N+d) dimensional problem Stochastic space:


Cornell University

i
i =1

PDF:

( Y ) = i (Yi ), Y
i =1

Materials Process Design and Control Laboratory

14

Stochastic variational formulation


Define the deterministic functional spaces:

and duality product Define the stochastic functional spaces:

Tensor product to perform the functional spaces:

Stochastic variational form: Find

Cornell University

Materials Process Design and Control Laboratory

15

Stochastic collocation method


The weak form is equivalent to: for a.e. hold: the following deterministic equations

As before, stochastic variable can be regarded as a function taking value in the stochastic space.
Random variable MAP

S
Sample space of elementary events

Real line

Each outcome is mapped to a corresponding real value

Collection of all possible outcomes

This nature is utilized by the stochastic collocation method to construct the interpolant of the stochastic function. Therefore, we only need the solution to the deterministic problem at the collocation points.
Cornell University
Materials Process Design and Control Laboratory 16

Outline of the presentation Motivation: Coupling multiscaling and uncertainty analysis Mathematical representation of uncertainty Problem definition: Stochastic multiscale flow in porous media Adaptive Sparse Grid Collocation (ASGC) Method for the solution of SPDEs ASGC coupled with Adaptive High Dimensional Model Representation technique for high-dimensional SPDEs Mixed Finite Element Heterogeneous Multiscale Method Numerical Examples

Cornell University

Materials Process Design and Control Laboratory

17

Sparse grid collocation


In higher dimensions, not all the points are equally important. Some points contribute less to the accuracy of the solution (e.g. points where the function is very smooth, regions where the function is linear, etc.). Discard the points that contribute less: SPARSE GRID COLLOCATION1 1D sampling

Tensor product

2D sampling
1 .S. Smolyak, Quadrature and interpolation formulas for tensor product of certain classes of functions, Soviet Math. Dokl, 4 (1963) 240-243.

Cornell University

Materials Process Design and Control Laboratory

18

Choice of collocation points and nodal basis functions


In the context of incorporating adaptivity, we use the Newton-Cotes grid with equidistant support nodes and the linear hat function as the univariate nodal basis.

In this manner, one ensures a local support and that discontinuities in the stochastic space can be resolved. The piecewise linear basis functions is defined as
1

Y ji 21i
Cornell University

Y ji

Y ji + 21i

Materials Process Design and Control Laboratory

19

Conventional sparse grid collocation (CSGC)


Denote the one dimensional interpolation formula as LET OUR BASIC 1D INTERPOLATION SCHEME BE SUMMARIZED AS In higher dimensions, a simple case is the tensor product formula AS IN MULTIPLE DIMENSIONS, THIS CAN BE WRITTEN

Using the 1D formula, the sparse interpolant Aq, N , where q is the depth of sparse grid
interpolation ( q 0, q Smolyak algorithm as
0

) and N

is the number of stochastic dimensions, is given by the

i = U i U i 1

U0 = 0

Here, we define the hierarchical surplus as:

Cornell University

Materials Process Design and Control Laboratory

20

Nodal basis versus hierarchical basis

Nodal basis

Hierarchical basis

Cornell University

Materials Process Design and Control Laboratory

21

Adaptive sparse grid collocation


1D tree-like data structure:

2 1

3 1

3 w2

2 w2

1 y1

Anisotropic sampling for interpolating functions with steep gradients and other localized phenomena Define a threshold value. If magnitude of the hierarchical surplus is greater than this threshold, refine around this point. Add 2N neighbor points. Scales linearly instead of O(2N) of other stochastic adaptive methods

y12 y13 y14


3 y2

2 y2

4 y2

2D:

Cornell University

Materials Process Design and Control Laboratory

22

Definition of the error indicator


The mean of the random solution can be evaluated as follows:

Denoting

we rewrite the mean as

We now define the error indicator as follows:

In addition to the surpluses, this error indicator incorporates information from the basis functions. This forces the error to decrease to a sufficiently small value for a large interpolation level. This error indicator guarantees that the refinement would stop at a certain interpolation level.
Cornell University
Materials Process Design and Control Laboratory 23

Adaptive sparse grid collocation: Algorithm

Cornell University

Materials Process Design and Control Laboratory

24

Adaptive sparse grid collocation


Ability to detect and reconstruct steep gradients
f ( x, y ) = 1 |103 x 2 y 2 | +103
f ( x, y ) = 1 | 0.3 x 2 y 2 | +10 1

Cornell University

Materials Process Design and Control Laboratory

25

Application to Rayleigh-Bnard instability Cold wall = -0.5 Insulated Insulated


u(x, t , ) = 0
u( x, t , ) + u(x, t , ) u(x, t , ) = p (x, t , ) + Pr 2u(x, t , ) + Pr Ra (x, t , ) t

( x, t , ) + u(x, t , ) (x, t , ) = 2 ( x, t , ) t

hot wall ()
Filled with air (Pr = 0.7), Ra = 2500, which is larger than the critical Rayleigh number, so that convection can be initialized by varying the hot wall temperature. Thus, the hot wall temperature is assumed to be a random variable. h = 0.4 + 0.3Y Y [0,1] Investigation: Stochastic simulation to find the critical wall temperature.
Cornell University
Materials Process Design and Control Laboratory 26

Rayleigh-Bnard instability: stochastic formulation


Max = 0.667891
0.6 0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4

Nonlinear: convection

0.541
Max = 0.436984
0.35 0.25 0.15 0.05 -0.05 -0.15 -0.25 -0.35 -0.45

Linear: conduction

Cornell University

Materials Process Design and Control Laboratory

27

Rayleigh-Bnard instability: Mean


Max = 1.73709
Max = 1.73123
Max = 0.55
1.6 1.2 0.8 0.4 0 -0.4 -0.8 -1.2 -1.6

1.6 1.2 0.8 0.4 0 -0.4 -0.8 -1.2 -1.6

0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4

Max = 1.73715

u
1.6 1.2 0.8 0.4 0 -0.4 -0.8 -1.2 -1.6

Max = 1.73128

v
1.6 1.2 0.8 0.4 0 -0.4 -0.8 -1.2 -1.6

Max = 0.549918

0.5 0.4 0.3 0.2 0.1 0 -0.1 -0.2 -0.3 -0.4

ASGC (top)
Cornell University
Materials Process Design and Control Laboratory

MC (bottom)
28

Rayleigh-Bnard instability: Variance


Max = 3.30617
3 2.6 2.2 1.8 1.4 1 0.6 0.2

Max = 3.28535
3 2.6 2.2 1.8 1.4 1 0.6 0.2

Max = 0.0129218
0.012 0.011 0.01 0.009 0.008 0.007 0.006 0.005 0.004 0.003 0.002 0.001

Max = 3.29984

u
3 2.6 2.2 1.8 1.4 1 0.6 0.2

Max = 3.27903

v
3 2.6 2.2 1.8 1.4 1 0.6 0.2

Max = 0.0128783

0.012 0.011 0.01 0.009 0.008 0.007 0.006 0.005 0.004 0.003 0.002 0.001

ASGC (top)
Cornell University
Materials Process Design and Control Laboratory

MC (bottom)
29

Outline of the presentation Motivation: Coupling multiscaling and uncertainty analysis Mathematical representation of uncertainty Problem definition: Stochastic multiscale flow in porous media Adaptive Sparse Grid Collocation (ASGC) Method for the solution of SPDEs ASGC coupled with Adaptive High Dimensional Model Representation technique for high-dimensional SPDEs Numerical Examples

Cornell University

Materials Process Design and Control Laboratory

30

Motivation of HDMR
Conventional and adaptive collocation methods are not suitable for highdimensional problems due to their weakly dependence on the dimensionality (logarithmic) in the error estimate.

Although ASGC can alleviate this problem to some extent, its performance depends on the regularity of the problem and the method is only effective when some random dimensions are more important than others. These modeling issues for high-dimensional stochastic problems motivate the use of the High Dimensional Model Representation (HDMR) technique. Stochastic elliptic: N = 25, Lc = 1/ 16
Cornell University
Materials Process Design and Control Laboratory 31

High dimensional model representation (HDMR)


f ( Y ) = f 0 + fi (Yi ) + f i1i2 Yi1 , Yi2 +
i =1 N i1 < i2 N N

)
N N

i1 < i2 < i3

f i1i2i3 Yi1 , Yi2 , Yi3 +

i1 < < is

f i1

is

( Y , , Y ) +
i1 is

+ f12

(Y1 , , YN )

In this expansion:

f 0 denotes the zeroth-order effect which is a constant. The component function fi (Yi ) gives the effect of the variable Yi acting independently of the other input variables. The component function fi i (Yi , Yi ) describes the interactive effects of the variables Yi and Yi . Higher-order terms reflect the cooperative effects ofincreasing numbers of variables acting together to impact upon f ( Y ) .
12 1 2

The last term f12N (Y1 , , YN ) gives any residual dependence of all the variables locked together in a cooperative way to influence the output f ( Y ) . Several low-dimensional problems
Cornell University
Materials Process Design and Control Laboratory 32

CUT-HDMR
This equation is often written in a more compact notation:

for a given set where denotes the set of coordinate indices and . Here, denotes the - dimensional vector containing those components of whose indices belong to the set , where is the cardinality of the corresponding set , i.e. . For example, if , then and implies

In CUT-HDMR, a reference point is first chosen

where the notation means that the components of other than those indices that belong to the set equal to those of the reference point.
Mean of the random input vector is chosen as the reference point.
Cornell University
Materials Process Design and Control Laboratory 33

CUT-HDMR coupled with ASGC


Therefore, the N -dimensional stochastic problem is transformed to several lower-order -dimensional problems which can easily solved by ASGC:

where indexed by

are the hierarchical surpluses for different sub-problems and is only a function of the coordinates belonging to

Then the mean of the HDMR expansion is simply

In other words, instead of solving the N - dimensional problem directly using ASGC, which is impractical for extremely high dimensional problems, we only need to solve several one- or two- dimensional problems, which can be solved efficiently via ASGC.
Cornell University
Materials Process Design and Control Laboratory 34

Adaptive HDMR1
Identify the important dimensions through the first-order expansion

Then we define the important dimensions as those whose weights are larger than a predefined error threshold . Only higher-order terms which consist of only these important dimensions are considered. We only construct the higher-order terms which is of importance:

Stop the construction if relative error

is small than a threshold

1 .X. Ma, N. Zabaras, An adaptive high-dimensional stochastic model representation technique for the solution of stochastic partial differential equations, JCP, 229 (2010) 3884-3915.

Cornell University

Materials Process Design and Control Laboratory

35

Adaptive HDMR algorithm

Cornell University

Materials Process Design and Control Laboratory

36

Stochastic Porous Media Flow


production well

.u = f u = K p

in in

D D

Grid: 30 30

1 1

injection well
x 10
-3

L = 0.25, 2 = 2.0, N = 500 :


7.5 7 Standard Deviation 6.5 6
PDF 70

Yi [1,1]
MC 60 50 40 30 20 10 0 -0.02
= 2.0
2

MC: 1,000,000 HDMR: = 10 , 1 = 2 10


-6 -4

1 = 10 1 = 10 = 10
-6

-2 -3 -4

1 = 2 10

5.5 5 4.5 4

= 2.0

0.2

0.4 x

0.6

0.8

-0.01

0.01

0.02 v

0.03

0.04

0.05

0.06

Standard deviation of velocity at y = 0.5

PDF of velocity at point ( 0, 0.5)

Convergence of the normalized errors of the standard deviation

Highest stochastic dimension problem reported based on non-MC method


Cornell University
Materials Process Design and Control Laboratory 37

Stochastic Natural Convection


Pr = 1.0, Ra = 1000

Insulated Cold wall = -0.5

hot wall ()

Grid: 30 30

N = 10, Yi [1,1]
3.5 3 2.5 2 1.5 1 0.5 0 -0.5 -1 -1.5 -2 -2.5 -3 -3.5
0.45 0.4 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 -0.05 -0.1 -0.15 -0.2 -0.25 -0.3 -0.35 -0.4 -0.45

Insulated
3.5 3 2.5 2 1.5 1 0.5 0 -0.5 -1 -1.5 -2 -2.5 -3 -3.5

Mean

Std

0.075 0.07 0.065 0.06 0.055 0.05 0.045 0.04 0.035 0.03 0.025 0.02 0.015 0.01 0.005

1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

L = 0.1, = 0.05

L = 0.1, = 1.0

L = 0.1, = 2.0

38

Cornell University

Materials Process Design and Control Laboratory

Outline of the presentation Motivation: Coupling multiscaling and uncertainty analysis Mathematical representation of uncertainty Problem definition: Stochastic multiscale flow in porous media Adaptive Sparse Grid Collocation (ASGC) Method for the solution of SPDEs ASGC coupled with Adaptive High Dimensional Model Representation technique for high-dimensional SPDEs Mixed Finite Element Heterogeneous Multiscale Method Numerical Examples

Cornell University

Materials Process Design and Control Laboratory

39

Spatial finite element discretization


Permeability is defined as a cell-wise constant on a fine scale grid

element face

A coarse-scale grid is also defined, where we are seeking the coarse-scale solution. This grid is assumed to be conforming to the fine-scale grid where the permeability is defined.

element face

Cornell University

Materials Process Design and Control Laboratory

40

Mixed finite element method


The mixed finite element method for the elliptic problem on the coarse grid Find
c 4

(x
where the lowest-order Raviart-Thomas space for velocity

L 1

R , x2 )

(x
E

R 1

R , x2 )

1c

3c

(x

L 1

L , x2 )

c 2

(x

R 1

L , x2 )

is the value of the coarse-scale flux at the middle point of the side and piecewise constant approximation for pressure E

Cornell University

Materials Process Design and Control Laboratory

41

Multiscale bilinear form


All the fine-scale information of the permeability is incorporated in the bilinear form, i.e. . Define the global matrix for the bilinear form: , where

2x2: Numerical quadrature

where and and weights. are the quadrature points



4

It is clear that the realization of the permeability field at the quadrature point is not able to capture the full information at the subgrid scale in the coarse element.
Cornell University

Materials Process Design and Control Laboratory

42

Modified multiscale bilinear form


Therefore, we need to modify the bilinear form at the quadrature points following the framework of heterogeneous multiscale method1:

where sampling domain

is the solution to the following local subgrid problem in the

with appropriate boundary condition. pressure.

can be considered as the subgrid

1 . W. E, B. Engquist, The heterogeneous multi-scale methods, Comm. Math. Sci 1 (2002) 87-132.

Cornell University

Materials Process Design and Control Laboratory

43

Choice of the sampling domain


We would like to take the sampling domain the same as the coarse element, in order to capture all the fine-scale information of the permeability within the current coarse element.

E k = E

In other words, the subgrid problem is solved on the same coarse element for each quadrature point. Since the governing equations and the solution domain are the same for all the subgrid problem, the only difference between them is the boundary conditions applied on the coarse element, which is in general plays a significant role in the accuracy of the multiscale method.
Cornell University
Materials Process Design and Control Laboratory 44

Choice of boundary condition


Due the theory of HMM, the Neumann boundary condition is
constant

where denotes the value of the ith coarse-scale RT0 basis function at the kth quadrature point.

c 1,2 ( k )

c 1,2 ( k )

In another view, the boundary condition is equivalent to apply the total flux applied along the coarse element boundary , which is

Transmissibility is a measure of the ability of the interface to transport the flow across the boundary of the element.

ei

Ki

ej K j x j
1

xi

Hence we modify the boundary condition to

Transmissibility in the x-direction:

x x j Tva = 2 | a | i + K K j i
Cornell University
Materials Process Design and Control Laboratory

45

Choice of boundary condition


The modified boundary condition is
Q11

as example
Q11T1 / Ti | i |
i =1 4

1 2 3 4

Q11T5 / Ti | i | Q11T6 / Ti | i | Q11T7 / Ti | i | Q11T8 / Ti | i |


i =5 i =5 8 i =5 8 i =5 8

Q11T2 / Ti | i | Q11T3 / Ti | i |
i =1 4 i =1 4

6 7 8

Q11T4 / Ti | i |
i =1

The sum of the flux applied on the fine-scale element is equal to the total flux applied on the same coarse element boundary

We just redistribute the total flux on the coarse-scale element boundary according to the ability to transport the flow at the interface of each fine-scale element . This is clearly a better choice for boundary condition since it determines the flow conditions across the inter-block boundaries.
Cornell University
Materials Process Design and Control Laboratory 46

Subgrid problem in the coarse element


Therefore, our subgrid problem in one coarse-element E is defined as follows: For each quadrature point we seek the solution to the following subgrid problem for each coarse-scale RT0 basis function

subject to the following Neumann boundary condition

uik n = 0

E
uik n = 0

Cornell University

Materials Process Design and Control Laboratory

47

The coarse-scale problem


We will define the modified bilinear form as: for any

where

is defined through the subgrid problem.

The MxHMM version of the Darcys equation on the coarse scale reads: Fine the coarse scale such that

with the boundary condition The only difference is the modified bilinear form, which needs the solution of the local subgrid problem. It is through these subgrid problems and the mixed formulation that the effect of the heterogeneity on coarse-scale solutions can be correctly captured.
Cornell University
Materials Process Design and Control Laboratory 48

Reconstructing the fine-scale velocity


The idea is to solve Darcys equations within each coarse element using Neumann boundary condition given by the coarse-scale flux :

In order to make the problem well-posed, pressure is fixed in one fine-scale element with the coarse-scale pressure.

Again, using mixed finite element, the velocity field is conservative on both fine and coarse scales.
Cornell University
Materials Process Design and Control Laboratory

p = pc

49

Solution of the transport equation


The weak formulation is find such that

Let be the time step and denote by the approximation of the water saturation in element , then the discrete equation is

where

Degree of freedom in RT0 basis, can directly obtained from the solution of multiscale problem.

Cornell University

Materials Process Design and Control Laboratory

50

Quality assessment
To assess the quality of our multiscale approach, we will use the so called water cut curve , which defines the fraction of water in the produced fluid, i.e. where is the flow rate of produced oil at the outlet boundary and rate of the produced water. The water cut curve can be calculated as is the flow

where

refers to the outflow boundary condition.

The dimensionless time is measure in pore volume injected (PVI):

where is the total pore volume of the system, which is equal to the area of the domain here and is the total flow rate.

Cornell University

Materials Process Design and Control Laboratory

51

Solution Methodology
Generate the permeability sample given the collocation point, set coarse discretization
Generate collocation point

Compute the stiffness matrix for each coarse element

Solve the subgrid problems for each basis function at quadrature points

Solve stochastic multiscale problem with HDMR

Compute the stochastic coarse-scale fluxes

Return function value at collocation point

Reconstruct the fine-scale velocity

Solve the subgrid problems with coarsescale flux

POSTPROCESSING: Compute the statistics of the solution Cornell University

Solve the transport problem

Materials Process Design and Control Laboratory

52

Outline of the presentation Motivation: Coupling multiscaling and uncertainty analysis Mathematical representation of uncertainty Problem definition: Stochastic multiscale flow in porous media Adaptive Sparse Grid Collocation (ASGC) Method for the solution of SPDEs ASGC coupled with Adaptive High Dimensional Model Representation technique for high-dimensional SPDEs Mixed Finite Element Heterogeneous Multiscale Method Numerical Examples

Cornell University

Materials Process Design and Control Laboratory

53

Example 1: Realistic permeability


In order to verify the accuracy of the deterministic multiscale code, in this example, we consider a deterministic permeability field from the top layer of the SPE10 comparative project. p =0 The logarithm of the fine-scale permeability is defined on 60x220 grid as shown on the right. We assume the flow is from bottom to top with the boundary conditions:

For saturation equation, we assume zero initial conditions and the following boundary condition:

p =100

Cornell University

Materials Process Design and Control Laboratory

54

Contour of x-velocity for various coarse grids

(a ) Fine Scale

(b) 30 110

(c) 15 55

(d ) 10 44

(e) 6 22

Cornell University

Materials Process Design and Control Laboratory

55

Contour of y-velocity for various coarse grids

(a ) Fine Scale

(b) 30 110

(c) 15 55

(d ) 10 44

(e) 6 22

Cornell University

Materials Process Design and Control Laboratory

56

Saturation contour for various coarse grids at 0.4 PVI

(a ) Fine Scale

(b) 30 110

(c) 15 55

(d ) 10 44

(e) 6 22

Cornell University

Materials Process Design and Control Laboratory

57

Saturation Movies

(a ) Fine Scale

(b) 15 55

Cornell University

Materials Process Design and Control Laboratory

58

Water cut curves for various coarse grids


0.8 0.7 0.6 0.5 Reference 30 x 110 15 x 55 10 x 44 6 x 22

F(t)

0.4 0.3 0.2 0.1 0

0.2

0.4

0.6

0.8

PVI

Cornell University

Materials Process Design and Control Laboratory

59

Relative errors for various coarse grids

For velocity: For saturation: For water cut: The reference solution is taken from the solution of the fine-scale problem using mixed finite element method directly.

Cornell University

Materials Process Design and Control Laboratory

60

Example 2: stochastic simulation


= [ 0,1]
2

u n = 0

Fine-scale grid: 64 64 Coarse-scale grid: 8 8

p =1
u n = 0

p=0

where Y is a zero mean Gaussian random field with covariance function. is the correlation length and is the standard deviation

Here

are assumed i.i.d uniform random variables on [-1,1].

For comparison, the reference solution is taken from 106 MC samples, where each direct problem is solved using the fine-scale solver. The stochastic problem is solved using HDMR, where the solution of each deterministic problem at the collocation point is from the multiscale solver. We fix and investigate the effect of the anisotropy of the random field.
61

Cornell University

Materials Process Design and Control Laboratory

Example 2: Isotropic random field


In this problem, we take Due to the slow decay of the eigenvalue, the KLE is truncated after 100 terms. Therefore, the stochastic dimension is 100. The problem is solved with HDMR where each sub-problem is solved through ASGC. We take First, we show the results at 0.2 PVI.

It is interesting to note that the mean is nearly the same as the homogeneous solution with the mean permeability. This is called heterogeneity-induced dispersion, where the heterogeneity smoothes the water saturation profile in the ensemble sense, although individual realization shows heterogeneity. Cornell University
Materials Process Design and Control Laboratory 62

Isotropic field: std of saturation at 0.2 PVI

The result indicates that higher water saturation variance are concentrated near displacement fronts, which are areas of steep saturation gradients.

Cornell University

Materials Process Design and Control Laboratory

63

Isotropic filed: Convergence of HDMR

N i Number of important dimensions

N c Total number of expansion component functions

The normalized error is defined the same as before. From the table, it is seen that the results are indeed quite accurate despite the fact that 64-fold upscaling is used to solve the deterministic problem and adaptive methods are used to solve the stochastic problem.

Cornell University

Materials Process Design and Control Laboratory

64

Isotropic filed: Interpolation properties

We randomly generate one input vector and reconstruct the saturation from HDMR. At the same time, we run a deterministic problem with the fine-scale model and the same realization of the random input vector. The two results are exactly the same.

Cornell University

Materials Process Design and Control Laboratory

65

Isotropic field: PDF and CDF

PDF and CDF are plotted at the point (0.2, 0), where the highest standard deviation of saturation occurs. The results are obtained from the saturation realizations through HDMR.

Cornell University

Materials Process Design and Control Laboratory

66

Example 2: Anisotropic random field


In this problem, we take Due to the increasing of correlation length in x dimension, the KLE is truncated after 50 terms. Therefore, the stochastic dimension is 50. We first solve this problem again at time 0.2 PVI using HDMR with ASGC. We take the relevant parameters as

We get the same results as in the isotropic case.

Cornell University

Materials Process Design and Control Laboratory

67

Anisotropic field: std of saturation at 0.2 PVI

It is interesting to note that the shape of contour is nearly the same as that of isotropic random field. Only the value of standard deviation are different. The introduction of anisotropy has the effect of increasing the output uncertainty.

Cornell University

Materials Process Design and Control Laboratory

68

Anisotropic field: Convergence of HDMR

According to our previous numerical results, larger uncertainty requires more expansion terms. Here, indeed more expansion terms and collocation points are needed compared with that of isotropic case. In addition, the highest HDMR expansion order is 3. There are 3 third-order component functions, which indicating the existence of higher-order cooperative effects among the inputs.

Cornell University

Materials Process Design and Control Laboratory

69

Anisotropic field: Results at 0.2 PVI

Cornell University

Materials Process Design and Control Laboratory

70

Conclusions
An efficient computational framework developed for analysis of complex multiscale systems. The key aspects of these developments is to utilize the adaptive HDMR coupled with ASGC for solving stochastic PDEs involved. The most important rational that has made this technique important is its nonintrusive character, where only repetitive function calls are required at a much less number of sampling points than that of MC method. We have developed a black box stochastic toolkit that can seamlessly link with any deterministic simulator to facilitate stochastic analysis.

Cornell University

Materials Process Design and Control Laboratory

71

Uncertainty Quantification Across Length Scales


Uncertainty in initial and/or boundary (operating) conditions at the macro scale + Topological uncertainty at lower scales How do mechanical properties and damage evolve? Will the device fail? With what probability? - Simple one way coupling: Upscale the property statistics to the macro scale and use to compute statistics - Two way coupling: For tightly coupled problems, solve micromacro stochastic PDEs simultaneously. Need information theoretic models to propagate uncertainty across scales
72

Uncertain heat flux on the nozzle flap

Limited information about the microstructure of the material

Cornell University

Materials Process Design and Control Laboratory

Continuum-Discrete Stochastic Coupling


Exact microstructural features unknown. Experiments Atomistic model only provide limited statistical information - Where does fracture occur? Depends on the local microstructure => need a stochastic Averaged framework for analysis. properties - Stochastic homogenization can only be applied hierarchically/adaptively (e.g. near or far from crack tips) - Assumes significance in the robust design of materials that withstand failure Continuum model Microstructure Is it possible to couple discrete simulators (e.g. MD, MC) with a continuum stochastic (higher scale) simulator? - Since lower scales posses higher information content, estimate statistics and utilize it to run macro simulator

- Issues with computational efficiency (model reduction)


Cornell University
Materials Process Design and Control Laboratory 73

Publications
Xiang Ma and N. Zabaras, A stabilized stochastic finite element second-order projection method for modeling natural convection in random porous media", J. Computational Physics, Vol. 227, pp. 8448-8471, 2008. Xiang Ma and N. Zabaras, An adaptive hierarchical sparse grid collocation algorithm for the solution of stochastic differential equations", J. Computational Physics, Vol. 208, pp. 3084-3113, 2009. Xiang Ma and N. Zabaras, An efficient Bayesian inference approach to inverse problems based on an adaptive sparse grid collocation method", Inverse Problems, Vol. 25, 035013 (27pp) , 2009. (selected as the highlights of 2009 in Inverse Problems) Xiang Ma and N. Zabaras, An adaptive high-dimensional stochastic model representation technique for the solution of stochastic partial differential equations", J. Computational Physics, Vol. 229, pp. 3884 -3915, 2010. Xiang Ma and N. Zabaras, A stochastic mixed finite element heterogeneous multiscale method for flow in porous media", J. Computational Physics, submitted. Xiang Ma and N. Zabaras, Kernel principal component analysis for stochastic input model generation", J. Computational Physics, submitted. Cornell University
74

Materials Process Design and Control Laboratory

You might also like