Introduction to uncertainty quantification (UQ). With applications.

Size: px
Start display at page:

Download "Introduction to uncertainty quantification (UQ). With applications."

Transcription

1 (UQ). With applications. J. Peter ONERA including contributions of A. Resmini, M. Lazareff, M. Bompard ONERA ANADE Workshop, Cambridge, September 24 th, 2014

2 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

3 INTRODUCTION Example I : drag minimization Deterministic minimization of airplane cruise shape : define shape That minimizes total drag Cd at M=0.82 Satisfying constraints on lift, pitching moment, inner volume... Actually, variations of cruise flight Mach number Waiting for landing slot Speeding up to cope with pilot maximum flight time Variable Mach number described by D(M) Robust definition of airplane cruise shape Minimize Cd(M)D(M)dM, satisfying constraints for all values of M present in D(M)

4 INTRODUCTION Example II : fan design Fan operational conditions subject to changes in wind conditions Manufacturing subject to tolerances Robust design accounts for variability of external parameters tolerances for internal parameters Figure 1: Robust design (from cenaero.be)

5 INTRODUCTION Example III : validation process Unkown data in experiment Upwind Mach number not fully controled in wind tunnels dm = Unknown physical constant needed in numerical model Wall roughness constant (milled, brazed, eroded surface...) Discrepancy in a computational/experimental validation process! Compute the mean and standard deviation of the output of interest due to the uncertain inputs

6 INTRODUCTION Definition of uncertain inputs UNCERTAINTY QUANTIFICATION : describe the stochastic behaviour of OUTPUTS of interest due to uncertain INPUTS Overview of CFD actual uncertain INPUTS : Geometrical (manufacturing tolerance) Operational: flow at boundaries (far field, injection...) Reference: Proceedings of RTO-MP-AVT-147 Evans T.P., Tattersall P. and Doherty J.J.: Identification and quantification of uncertainty sources in aircraft related CFD-computations - An industrial perspective Stochastic behaviour of OUTPUTS. Presently, most often, mean and variance are calculated

7 INTRODUCTION Three issues in (UQ) 1 terminology Lack of agreement on the definition of error, uncertainty... AIAA Guide G Uncertainty is a potential deficiency in any phase are activity of the modeling process that is due to the lack of knowledge. Error is a recognizable deficiency in any phase or activity of the modelling process that is not due to the lack of knowledge ASME Guide V& V 20 (in its simpler version adopted for the Lisbon Workshops on CFD uncertainty). The validation comparison error is defined as the difference between the simulation value and the experimental data value. It is split in numerical, model, input and data errors (assumed to be independant). Numerical (resp. input, model, data) uncertainty is a bound of the absolute value of numerical (resp. input, model, data) error

8 INTRODUCTION Three issues in (UQ) 2 (UQ) validation and verification (UQ) CFD-based exercise leads to a standard deviation on some outputs Compare this standard deviation to the numerical error Richardson method, GCI... Pierce et al. Venditti et al. adjoint based formulas for functional outputs Compare this standard deviation to the modeling error Run several (RANS) models Run better models than (RANS) Numerical (UQ) investigation only make sense if standard deviation due to uncertain inputs not much smaller than modelling or discretization error

9 INTRODUCTION Three issues in (UQ) 3 lack of shared well-defined problems Most often mean and variance of some outputs are computed in (UQ) exercises as this is feasible with usual methods? as this is actually required for applications? from EDF s specialits point of view, three stochastic criteria of interest: central dispersion (mean and variance of output) range (min and max possible values of output) probability of exceeding a threshold Get information from industry in order to define relevant (UQ) exercises Share mathematical industrial test cases and mathematical exercices to split the CFD influence / the one of (UQ) method

10 INTRODUCTION Intrusive vs non-intrusive methods Non-intrusive methods. No change in the analysis code Post-processing of deterministic simulations Intrusive methods. Changes in the analysis code Stochastic expansion of state/primitive variables Galerkin projections Probably not feasible for large industrial codes

11 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

12 MONTE-CARLO Basics of probability (1) A classical introduction to probability basics involves a sample space Ω (dice values, interval of Mach number values) events ξ i.e. element of Ω or set of elements of Ω (even dice values, subinterval of Mach number values) event space A, set of subsets of Ω, stable by union, intersection, including null set and Ω (also called σ-algebra) [INPUT] a probability function P on A such that P (Ω) = 1,P ( ) = 0, plus natural properties for complementary parts and union of disjoint parts (then denoted D keeping P for polynomials) [OUPUT] random variables X depending of the event ξ (like CDp or CLp of an airfoil depending on the far-field Mach number through Navier-Stokes equations )

13 MONTE-CARLO Basics of probability (2) Discrete example : regular 6-face Dice thrown once events ξ = 1,2,3,4,5 or 6 sample space Ω ={1,2,3,4,5,6} probability space Ω = null set plus all discrete sets of these numbers {, {1}, {2}, {3}, {4}, {5}, {6}, {1, 2}, {1, 3}, {1, 4}, {1, 5}, {1, 6}, {2, 3}...{1, 2, 3, 4, 5, 6}} probability function P (later denoted D) : P ( ) = 0, P ({1}) = 1./6., P ({2}) = 1./6.,... P ({1, 2}) = 1./3., P ({1, 3}) = 1./3, P ({1, 4}) = 1./3... P ({1, 2, 3, 4, 5, 6}) = 1. random variables X, for example dice value to the power three...

14 MONTE-CARLO Basics of probability (3) Discrete example : Mach number in [0.81,0.85] events ξ = a Mach number value in [0.81,0.85] sample space Ω = [0.81,0.85] probability space Ω = all subparts of [0.81,0.85] probability function P to be defined (later denoted D). For example from a beta distribution with power three : P φ (φ) = (1. φ2 ) 3 φ [ 1, 1] φ = (ξ 0.83)/0.02 P (ξ) = P φ(φ) = (1. (ξ )2 ) 3 random variables X, aerodynamic pitching moment of a wing with variable Mach number M ( event ξ) in the farfield

15 MONTE-CARLO Set of probability density functions of β distributions (on [0,+1] with the α 1 β 1 convention for exponants)

16 MONTE-CARLO Basics of probability (4) Allows to define probablities for both discrete and continuous spaces Quantities of interest = functions of the stochastic variables random variables Example = pitching moment of wing J (random variable) depending via NS equations of uncertain far-field Mach number (uncertain input/event) Mean Variance Z E(J ) = J (ξ)d(ξ)dξ Ω Z V ar(j ) = (J (ξ) E(J )) 2 D(ξ)dξ Ω Covariance of two random variables Cov(J, G) = Z Ω (J (ξ) E(J ))(G(ξ) E(G))D(ξ)dξ

17 MONTE-CARLO Basics Monte-Carlo mimics the law of the event in a series of calculations Reference method for all uncertainty propagation methods Generation of a sampling (ξ 1, ξ 2..., ξ p..., ξ N ) of the p.d.f D(ξ) Example : sampling for normal distribution. a k, b k indep. uniformly distributed in [0.,1.] ξ k = 2 ln(a k )cos(2πb k ) Computation of corresponding flow fields W (ξ p ), p [1, N] Computation of functional outputs J (ξ p ) = J(W (ξ p ), X(ξ p )) Estimation of following statistical quantities : E(J ) = Z J (ξ)d(ξ)dξ J N = 1 N p=n X J (ξ p ) p=1 σ 2 J = E((J E(J ))2 ) σ 2 J N = 1 N 1 Need to quantify accuracy of estimation p=n X p=1 (J (ξ p ) J N ) 2

18 MONTE-CARLO Accuracy of estimation: mean (1) Scalar case. variance σ J is known Reminders about N (0, 1) N J N E(J ) σ J N (0, 1) (Normal distribution) Probability density function (p.d.f.) of N (0, 1)- D N (x) = 1 2Π e x2 2 Symmetric cumulative distribution function - Φ N (x) = 1 2Π x x e t2 2 dt With ɛ confidence : E(J ) [ J uɛ σ N u J ɛ N, JN σ + u J ɛ N ] ɛ = 1 2Π e t 2 2 dt u ɛ ɛ u ɛ With 99% confidence : 2.58 E(J ) [ J N 2.58 σ J N, JN σ J N ] (0.99 = 1 2Π 2.58 e t2 2 dt)

19 MONTE-CARLO Accuracy of estimation: mean (2) Scalar case: variance σ J is unknown - N J N E(J ) σ JN S(N 1) (Student dist.) With ɛ confidence : E(J ) [ J N u ɛ(n 1) σ JN N, JN + u ɛ(n 1) σ JN N ] u ɛn as function of ɛ and N found in tables Student distribution converges to Normal distribution for large N Tables for u ɛn 1 ɛ N Figure 3: Value of u ɛ(n 1) for Student distribution S(N 1) N 2

20 MONTE-CARLO Accuracy of estimation: mean (3) Scalar case: variance σ J is unknown - N J N E(J ) σ JN With ɛ confidence : E(J ) [ J N u ɛ(n 1) σ JN N, JN + u ɛ(n 1) σ JN N ] u ɛn u ɛn as function of ɛ and N found in tables converges towards u ɛ of normal law for large N S(N 1) (Student dist.) Density function of Student distribution S(N). D S(N) (x) = N+1 Γ( 2 ) N+1 x2 (1 + Γ( N 2 ) ) 2 NΠ N Definition of u ɛn. ɛ = Z uɛn u ɛn D S(N) (t)dt (ɛ ]0, 1.[)

21 MONTE-CARLO Accuracy of estimation: variance (1) (skpd) Scalar case: mean E J is known Notations SJ 2 N = 1 N Chi-square χ 2 N i=n i=1 (J (ξp ) E J ) 2 probability distribution defined on [0, [ with p.d.f. : D χ 2 N (x) = 1 Γ(N/2)2 N/2 x N/2 1 e x/2 Chi-square cumulative d.f. : Φ χ 2 N (x) = x 0 D χ 2 N (t)dt Stochastic variable N S2 J N σ 2 J χ 2 N

22 MONTE-CARLO Chi-square probabilistic density functions D χ 2 N and cumulative density functions Φ χ 2 N (skpd)

23 MONTE-CARLO Accuracy of estimation: variance (2) (skpd) Scalar case: mean E(J ) is known - N S2 J N σ 2 J With ɛ = 1 α confidence : Φ 1 ( α χ 2 2 ) N S2 J N N σj 2 With ɛ = 1 α confidence : σ 2 J [N S2 J N Φ 1 χ 2 N χ 2 N Φ 1 χ 2 N (1. α 2 ) 2 J N Φ 1 χ 2 N (1 α 2 ), N S ( α 2 )] x N Figure 4: Value of Φ 1 χ 2 N (x)

24 MONTE-CARLO Quality of estimation: variance - 99 % confidence Application N = 2 σ 2 J [0.189 S2 J 2, 200 S 2 J 2 ] N = 20 σ 2 J [0.500 S2 J 20, 2.69 S 2 J 20 ] N = 30 σ 2 J [0.558 S2 J 30, 2.17 S 2 J 30 ] N = 100 σ 2 J [0.713 S2 J 100, 1.49 S 2 J 100 ] Convergence speed of bounds towards 1. The cumulative distribution of the Chi-Square law Φ N (x) can be expressed as Φ χ x/2 1 2 N (x) = Γ(N/2) t N/2 e t γ(n/2, x/2) dt = (γ Γ(N/2) lower incomplete Γ function) Check properties of (the inverse of) Φ χ 2 N Check convergence speed of N/Φ 1 (1 α χ 2 2 ) and N/Φ 1 ( α N χ 2 2 ) N 0

25 MONTE-CARLO Accuracy of estimation: variance (3) (skpd) Scalar case: mean E(J ) is unknown - Stochastic variable (N 1) σ2 J N σ 2 J With ɛ = (1 α) confidence : σ 2 J [(N 1) σ 2 J N Φ 1 χ 2 N 1 (1 α ), (N 1) σ 2 Φ 1 2 J N χ 2 N 1 ( α 2 )] χ 2 N 1 x N Figure 5: Value of Φ 1 χ 2 N 1(x)

26 MONTE-CARLO Restriction for Monte-Carlo method. Meta-models Convergence speed of Monte-Carlo for mean value estimation is 1 N Increasing precision of Monte-Carlo estimation by a factor of 10 requires multiplying the number of evaluations by a factor of 100 Not usable in the context of complex aeronautics simulation Cost of one evaluation very high (requires resolution of (RANS) equations) Aerodynamic function relatively smooth replace flow field / aerodynamic function by a model built with a database of flow fields / function values

27 Figure 6: Monte-Carlo method with meta-models Introduction to uncertainty quantification METAMODEL BASED MONTE-CARLO Use meta-models with Monte-Carlo method Monte Carlo (large) stochastic sampling Sampling and m.m. outputs Meta model (small) Sampling Sampling and outputs Build meta model (optimize inner parameters..) Meta model CFD code Compute flow Compute functional outputs Evaluate meta model based mean, variance,...

28 METAMODEL BASED MONTE-CARLO Meta-models Most often, approximation of a function of interest Meta-models used at ONERA for CFD: Kriging, Radial Basis Function, Support Vector Regression Other meta-models of specific interest for UQ: Adjoint based linear or quadratic Taylor expansion Discussed in more details in the extension to multi-dimension section Influence of meta-model accuracy on mean and variance accuracy?

29 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Presentation Search confidence interval on aerodynamic function C L with uncertainty on AoA Nominal configuration: NACA0012, M = 0.73, Re = 6M, AoA = 3 elsa (a) (RANS+(k-w) Wilcox turbulence model) solver (Roe flux+van Albada lim.) Figure 7: Mesh (a) The elsa CFD software: input from research and feedback from industry Mechanics and Industry 14(3) L. Cambier, S. Heib, S. Plot

30 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Distribution of uncertainty Beta distribution (parameters (2.,2.)) [pdfb(ξ) = (1 ξ)2 (1 + ξ) 2 ] over [-1,1] p.d.f of angle of attack AoA [pdfa(α) = pdfb(10.(α 3.))] over [2.9,3.1] Figure 8: Beta distribution of AoA

31 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Monte-Carlo method: application on C L (1) Estimated mean value 90 percent 95 percent 99 percent Best MC with MM value Figure 9: Mean of C L coefficient and confidence interval

32 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Monte-Carlo method: application on C L (2) Estimated variance value 90 percent 95 percent 99 percent Best MC with MM value e-05 6e-05 4e-05 2e Figure 10: Variance of C L coefficient and confidence interval

33 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Monte-Carlo method with meta-models: learning sample Use learning sample based on Tchebyshev polynomials Figure 11: Tchebychev distribution (11 points)

34 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Monte-Carlo method with meta-models: reconstruction of C L sample11.dat outputok.dat outputrbf.dat outputsvr.dat Figure 12: C L

35 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Monte-Carlo method with meta-models: application on C L (1) Estimated mean value 90 percent 95 percent 99 percent Best MC with MM value e+06 1e+07 Figure 13: Mean of C L coefficient and confidence interval

36 APPLICATION OF (METAMODEL BASED) MONTE-CARLO Monte-Carlo method with meta-models: application on C L (2) Estimated variance value 90 percent 95 percent 99 percent Best MC with MM value e-05 6e-05 4e-05 2e e+06 1e+07 Figure 14: Variance of C L coefficient and confidence interval

37 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

38 ORTHOGONAL POLYNOMIALS Orthogonal/orthonormal polynomials Polynomial/spectral expansion and stochastic post-processing Polynomials orthogonal for the product defined by p.d.f D(ξ) : < P l, P m >= P l (ξ)p m (ξ)d(ξ)dξ = δ lm Basic references Orthogonal polynomials Abramowitz and Stegun: Handbook of Mathematical functions. (1972). Chapter 22 Spectral expansions J. P. Boyd: Chebyshef and Fourier spectral methods (2001) Expand fields on this basis In non-intrusive PC method fields of interest In intrusive PC method state/primitive variables

39 ORTHOGONAL POLYNOMIALS Orthonormal polynomials (1/3) Families of polynomials Normal distribution D n (ξ) = 1 2Π e ξ2 2 on R Hermitte polynomials Gamma distribution D g (ξ) = exp( ξ) on R + Laguerre polynomials Uniform distribution D u (ξ) = 0.5 on [ 1, 1] Legendre polynomials Chebyshev distribution D cf (ξ) = 1/Π/ 1 ξ 2 on [ 1, 1] Chebyshev (first-kind) polynomials Chebyshev distribution D cs (ξ) = 1 ξ 2 on [ 1, 1] Chebyshev (second-kind) polynomials Beta distribut. D β (ξ) = (1 ξ) α (1 + ξ) β / 1 1 (1 u)α (1 + u) β du α > 1., β > 1. on [ 1, +1] Jacobi polynomials (incl. Chebyshev polynomials) Non-usual probabilistic density functions, D l (ξ) computed by Gram- Schmidt orthogonalisation process.

40 ORTHOGONAL POLYNOMIALS Orthonormal polynomials (2/3) Example : 1D problem. Hermite polynomials for normal law Probability density function (p.d.f.) of D n (ξ) = 1 2Π e ξ2 2 Family of orthonormal polynomials for < f, g >= R + f(ξ)g(ξ)d n(ξ)dξ P H 0 (ξ) = 1 P H 1 (ξ) = ξ P H 2 (ξ) = ξ 2 1 P H 3 (ξ) = ξ 3 3ξ P H 4 (ξ) = ξ 4 6ξ Normalization of Hermite polynomials P H j (ξ) = 1 j!(2π) (1/4) P H j (ξ) Orthonormality relation for P H j : < P H j, P H k >= + P H j(ξ)p H k (ξ)d n (ξ)dξ = δ jk

41 ORTHOGONAL POLYNOMIALS Orthonormal polynomials (3/3) Example : 1D problem on [-1,1]. First-kind Chebyshef polynomials 1 1 ξ 2 Probability density function (p.d.f.) of D cf (ξ) = 1 Π Family of orthonormal polynomials for < f, g >= R 1 1 f(t)g(t)d cf (t)dt T 0 (ξ) = 1 T 1 (ξ) = ξ T 2 (ξ) = 2ξ 2 1 T 3 (ξ) = 4ξ 3 3ξ T 4 (ξ) = 8ξ 4 8ξ Normalization of Chebyshev polynomials T 0 = T 0... T n = 2 T n (n 1) Orthonormality relation for T j : < T j, T k >= + T j(ξ)t k (ξ)d cf (ξ)dξ = δ jk specific property T n (cos(θ)) = cos(nθ) (hence T n 1. )

42 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

43 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Basics Polynomial/spectral expansion and stochastic post-processing Polynomials orthogonal for the product defined by p.d.f D(ξ) : < P l, P m >= P l (ξ)p m (ξ)d(ξ)dξ = δ lm Expand field of in interest (entire flow field, flow field at the wall...). i is the discrete space variable (mesh index) Spectral expansion : W (i, ξ) P CW (i, ξ) = l=m 1 l=0 C l (i)p l (ξ) Stochastic post-processing for P CW instead of W References [Wiener 1938][Ghanem et al. 1991]

44 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Stochastic post-processing Expansion of flow field on structured mesh depending on stochastic variable ξ W (i, ξ) P CW (i, ξ) = l=m 1 l=0 C l (i)p l (ξ) Stochastic post-processing (mean and variance) done for the expansion P CW instead of W straightforward evaluation of mean value E(P CW (i, ξ)) = l=m 1 l=0 C l (i)p l (ξ)d(ξ)dξ = C 0 (i) E(P CW (i, ξ)) =< P CW (i, ξ), 1 >= C 0 (i) straightforward evaluation of variance E((P CW (i, ξ) C 0 (i)) 2 ) = ( l=m 1 l=1 C l (i)p l (ξ)) 2 D(ξ)dξ E((P CW (i, ξ) C 0 (i)) 2 ) = l=m 1 l=1 C l (i) 2

45 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Accuracy of truncated spectral extension (1/3) General idea : the highest the regularity of the function of interest the fastest the convergence of the truncated spectral expansion Discussed for Chebyshef first-kind polynomials Classical expension of a continuous fonction over [-1,1] f(ξ) = n=0 C nt n (ξ) Definition of coefficients C 0 = Π 1 f(ξ)t 1 ξ 2 o(ξ)dξ C n = 2 +1 Π 1 1 f(ξ)t 1 ξ 2 n(ξ)dξ (n > 0) Convergence speed of truncated series f(ξ) = P N n=0 C nt n (ξ)?

46 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Accuracy of truncated spectral extension (2/3) Cos Fourier series of g periodic over [ Π, +Π] : g(θ) = P n=0 A ncos(nθ) R A 0 = 1 +Π f(θ)dθ A R 2Π Π n = 1 +1 g(θ)cos(nθ)dθ Π 1 Integration by part of A n for a C k function A n = 1 nπ A n = 1 n 2 Π... A n = 1 n k Π +1 1 g (θ)sin(nθ)dθ +1 1 g(2) (θ)cos(nθ)dθ +1 1 g(k) (θ)cos(nθ + kπ/2)dθ Upper bound of the coefficients for a C k function A n 1 n k Π +1 1 g(k) (θ) dθ 2 n k g (k) Upper bound of the residual g(θ) N n=0 A ncos(nθ) N+1 A n g (k) 1 du 2 g(k) N u k (k 1)N k 1

47 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Accuracy of truncated spectral extension (3/3) Chebyshev expansion of a C k function f(ξ) over [-1,1] f(ξ) = k=0 A kt k (ξ) Change variable: ξ = cos(θ), g = f cos is C k over [ Π, Π] g(θ) = f(cosθ) = k=0 A kt k (cos(θ)) g(θ) = k=0 A kcos(k θ) is a Fourier series!!! All results concerning Fourier series have a counterpart for Tchebyshev expansions Express g (k) from f (k)... f Express the upper bound of A k, use T k = 1. Find the upper bound of truncation error of spectral expansion of f Extension to f(i, ξ) = P k=0 A k(i)t k (ξ)

48 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Coefficient computations (1/4) Expansion of flow field on structured mesh depending on stochastic variable ξ W (i, ξ) P CW (i, ξ) = l=m 1 l=0 C l (i)p l (ξ) Straightforward stochastic post-processing for mean and variance (presented before) Accuracy of ideal P CW depending on degree and regularity - theory of spectral expansions (just discussed) Computation of C l coefficients? Accuracy of actual P CW expansion? (articles on this topic?)

49 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Coefficients computation - Gaussian quadrature (2/4) Expansion of flow field on structured mesh depending on stochastic variable ξ W (i, ξ) P CW (i, ξ) = l=m 1 l=0 C l (i)p l (ξ) From orthonormality property C l (i) =< P CW, P l > Under regularity assumptions C l (i) =< W, P l > Gaussian quadrature C l (i) =< P CW, P l >=< W, P l >= W (i, ξ)p l (ξ)d(ξ)dξ Computation by Gaussian quadrature associated to D with G points. f(ξ)d(ξ)dξ k=g k=1 w kf(ξ k ) (w k,ξ k ) depend on D(ξ) Exact quadrature if f(ξ) is a polynomial of degree (2G 1). Polynomial chaos : C l (i) exact if W (i, ξ)p l (ξ) polynomial of ξ of degree lower than/equal to (2G 1).

50 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Coefficients computation - collocation (3/4) Other way : collocation - least-square collocation Identify W (i, ξ l ) and P CW (i, ξ l ) for M values of ξ. Collocation W (i, ξ k ) = l=m 1 l=0 C l (i)p l (ξ k ) k {1, M} solved for C l (i) Solve least-square problem for more values of W (i, ξ l ) than coefficients in the expansion Less general accuracy results than for Gauss quadrature

51 GENERALIZED POLYNOMIAL CHAOS METHOD (N.-I.) Coefficients computation - collocation (4/4) Theorical result for Chebyshev series (see Boyd pages 95-96). Expansion of a C k function f(ξ) over [-1,1] f(ξ) = P k=0 A kt k (ξ) Collocation at x k = cos( kπ M 1 ) k [0, M 1] (Chebyshev extrema) obtain P M 1 (ξ) = M 1 k=0 Ā k T k (ξ) Collocation at x k = cos( (2k+1)Π 2M ) k [0, M 1] (Chebyshev roots) compute Q M 1 (ξ) = M 1 k=0 Ã k T k (ξ) Upper bounds of truncation error f(ξ) N n=0 ĀnT n (ξ) 2 N+1 A n... f(ξ) N n=0 ÃnT n (ξ) 2 N+1 A n... Factor TWO w.r.t. truncation of exact series

52 PROBABILISTIC COLLOCATION (N.-I.) Basics (1/2) Another approach for non-intrusive polynomial chaos based on Lagrangian polynomial expansion. [Tatang 1995] [Xiu et al. 2005] [Loeven et al. 2007] for compressible CFD Presented for 1D problem ( i is the discrete space variable mesh index) W (i, ξ) SCW (i, ξ) = l=n l=1 W l(i)h l (ξ) H l (ξ) = Π m=n (ξ ξ m ) m=1,m<>l ( polynomial of degree (N 1) ) (ξ l ξ m ) Note that actually SCW (i, ξ l ) = W l (i). no coefficient computation, define W l (i) = W (i, ξ l ) Definition of (ξ 1, ξ 2,..., ξ N ) (most often, not absolutely necessary) N points of the N-point quadrature associated to D(ξ) m=n f(ξ)d(ξ)dξ m=1 ω mf(ξ m )

53 PROBABILISTIC COLLOCATION (N.-I.) Basics (2/2) Expansion of flow field depending on stochastic variable ξ Stochastic post-processing (mean and variance) done for the expansion SCW instead of W straightforward evaluation of mean value E(SCW (i, ξ)) =< SCW (i, ξ), 1 >= m=n m=1 ω mw m (i) straightforward evaluation of variance E((SCW (i, ξ) E(SCW (i))) 2 ) = m=n m=1 ω mw m (i) 2 ( m=n m=1 ω mw m (i)) 2 Both exact for SCW as N -point Gaussian quadrature is exact for polynomial of degree up to 2N 1

54 PROBABILISTIC COLLOCATION (N.-I.) Link between polynomial chaos and probabilistic collocation Two polynomial expansions Case when the methods merge same degree (M 1) for both expansions coefficents of chaos expansion computed by collocation same set of M collocation points polynomials of degree (M 1) having same values in M distinct abscissas SCW = P CW (expressed in two different basis) exact evaluation of mean and variance for and SCW and P CW same evaluation of mean and variance by the two methods

55 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

56 INTRUSIVE POLYNOMIAL CHAOS References A stochastic projection method for fluid flow. I Basic formulation. Le Maître OP, Knio OM, Najm, HN, Ghanem RG. JCP 173 (2001) A stochastic projection method for fluid flow. II Random process. Le Maître OP, Reagan MT, Najm HN, Ghanem RG, Knio OM. JCP 181 (2002) Intrusive polynomial chaos method for the compressible Navier Stokes equations. Chris Lacor. Slides presented at ONERA Scientific Day on error approximation and uncertainty quantification. 3/12/2010 (available on ONERA s web site) Comparison and intrusive and non-intrusive polynomial chaos methods for CFD applications in aeronautics. Onorato G, Loeven GJA, Ghobarnias G, Bijl H, Lacor C. Proceedings of ECCOMAS CFD 2010 [Application] Assessment of intrusive and non-intrusive non-deterministic CFD methodologies based on polynomial chaos expansions. Dinsecu C, Smirnov S, Hirsch C, Lacor C. IJESMS 2 (2010)

57 INTRUSIVE POLYNOMIAL CHAOS Principle Expand primitive variables (not state-variables! unless you get ratio of polynomial expansions) on PC basis U(x, t, ξ) = l=m 1 l=0 U l (x, t)p l (ξ) Input expansions in stochastic partial differential equations Do a Galerkin projection for all P l (l [0, M 1]) : Multiply all equations by P l (ξ) Apply the mean (in other words, multiply all equations by P l (ξ) D(ξ) sum over ξ domain) Discretize resulting system of coupled partial differential equations Stochastic post-processing

58 INTRUSIVE POLYNOMIAL CHAOS Example 1 (1/3) Steady incompressible flow. Continuity equation One uncertain parameter (angle of attack or...) following D cf (ξ) = 1 Π Order 2 expansion of primitive variables U(x, ξ) = l=2 l=0 U l(x)t l (ξ) p(x, ξ) = l=2 l=0 p l(x)t l (ξ) 1 1 ξ 2 reminder T 0 (ξ) = 1 T 1 (ξ) = 2ξ T 2 (ξ) = 2(2ξ 2 1) Continuity equation div (U(x, ξ)) = div ( l=2 ) l=0 U l(x)t l (ξ) div (U(x, ξ)) = l=2 l=0 div(u l(x))t l (ξ)

59 INTRUSIVE POLYNOMIAL CHAOS Example 1 (2/3) Steady incompressible flow. Continuity equation. One uncertain parameter (angle of attack or...) following D cf (ξ) = 1 Π 1 1 ξ 2 Continuity equation div (U(x, ξ)) = l=2 l=0 div(u l(x))t l (ξ) = 0 Galerkin projection for T 0 (ξ),t 1 (ξ) and T 2 (ξ) (For T 1 for example) div (U(x, ξ)) T 1 (ξ)d cf (ξ) = P l=2 l=0 div(u l(x))t l (ξ)t 1 (ξ)d cf (ξ) = 0 A = R 1 1 div (U(x, ξ)))t 1(ξ)D cf (ξ) = 0

60 INTRUSIVE POLYNOMIAL CHAOS Example 1 (3/3) Steady incompressible flow. Continuity equation. One uncertain parameter (angle of attack or...) following D cf (ξ) = 1 Π 1 1 ξ 2 Continuity equation div (U(x, ξ)) = l=2 l=0 div(u l(x))t l (ξ) = 0 Galerkin projection for T 0 (ξ),t 1 (ξ) and T 2 (ξ) (For T 1 for example) A = R 1 P l=2 1 l=0 div(u l(x))t l (ξ)t 1 (ξ)d cf (ξ)dξ = 0 A = P l=2 l=0 div(u l(x)) R 1 1 T l(ξ)t 1 (ξ)d cf (ξ)dξ = div(u 1 (x)) = 0 Using orthonormality property : div(u 1 (x)) = 0 (of course div(u 0 (x)) and div(u 2 (x)) are also proved to be equal to zero using corresponding Galerkin projection)

61 INTRUSIVE POLYNOMIAL CHAOS Example 2 (1/3) Unsteady incompressible flow. Euler equations. Momentum equation. One uncertain parameter following D cf (ξ) = 1 Π 1 1 ξ 2 Order 2 expansion of primitive variables ( final number of equations multiplied by three) U(x, t, ξ) = l=2 l=0 U l(x, t)t l (ξ) p(x, t, ξ) = l=2 l=0 p l(x, t)t l (ξ) Momentum equation U t + (U. )U + p = 0 t (P l=2 l=0 U l(x, t)t l (ξ))+ P l=2 l=0 P n=2 n=0 (U l(x, t). )U n (x, t)t l (ξ)t n (ξ)+ P l=2 l=0 p l(x, t)t l (ξ) = 0

62 INTRUSIVE POLYNOMIAL CHAOS Example 2 (2/3) Steady incompressible flow. Euler equations. Momentum equation. One uncertain parameter following D cf (ξ) = 1 Π Momentum equation t (P l=2 + P l=2 l=0 1 1 ξ 2 l=0 U l(x, t)t l (ξ)) P n=2 n=0 (U l(x, t). )U n (x, t)t l (ξ)t n (ξ) + P l=2 l=0 p l(x, t)t l (ξ) = 0 Multiplying by T k k [0, 2] and taking the expected value for D cf t (P l=2 + P l=2 l=0 l=0 U l(x, t) < T l T k > P n=2 n=0 (U l(x, t). )U n (x, t) < T l T n T k > + P l=2 l=0 p l(x, t) < T l T k >= 0

63 INTRUSIVE POLYNOMIAL CHAOS Example 2 (3/3) Steady incompressible flow. Euler equations. Momentum equation. One uncertain parameter following D cf (ξ) = 1 Π Momentum equation t (P l=2 + P l=2 l=0 1 1 ξ 2 l=0 U l(x, t) < T l T k >) P n=2 n=0 (U l(x, t). )U n (x) < T l T n T k > + P l=2 l=0 p l(x, t) < T l T k >= 0 Using orthogonality property t U k(x, t) + P l,n=2 l,n=0 (U l(x, t). )U n (x, t) < T l T n T k > + p k (x, t) = 0 Set of p.d.e for (p 0, p 1, p 2 ) (U 0, U 1, U 2 ). Coupling through momentum equation

64 INTRUSIVE POLYNOMIAL CHAOS Stochastic post-processing Expand primitive variables on PC basis U(x, t, ξ) = l=m 1 l=0 U l (x, t)p l (ξ) Straightforward post-processing at continuous level E(U(x, t, ξ)) = U 0 (x, t) E((U(x, t, ξ) U 0 (x, t)) 2 ) = l=m 1 l=1 U l (x, t) 2 Actual stochastic post-processing at discrete level Use corresponding formulas at discrete level Derive mean and variance of (linearized) outputs of interest

65 INTRUSIVE POLYNOMIAL CHAOS Compressible Euler equations Including several technical details Pseudo spectral approach for cubic terms Truncation of higher order terms Three applications Feasibility for large codes?

66 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

67 EXTENSION TO HIGH DIMENSIONS Basics of multidimensional probability (1) Several R values stochastic variables a sample space Ω R d events i.e. element of Ω or set of elements of Ω elementary event ξ R d valued vector event space A, set of subsets of Ω, stable by union, intersection, including null set and Ω (also called σ-algebra) For the sake of simplicity Ω R 2 Define, cumulative distribution of (ξ 1, ξ 2 ), probabilty density function, marginal distributions, independance, covariance

68 EXTENSION TO HIGH DIMENSIONS Basics of multidimensional probability (2) Cumulative density function Φ(u 1, u 2 ) = P ({ξ 1 u 1 } {ξ 2 u 2 }) Probabilistic density function D ξ (ξ 1, ξ 2 ) = 2 Φ ξ x y (ξ 1, ξ 2 ) probability of ξ [ξ 1 dx/2, ξ 1 + dx/2][ξ 2 dy/2, ξ 2 + dy/2] is D ξ (ξ 1, ξ 2 )dxdy

69 EXTENSION TO HIGH DIMENSIONS Basics of multidimensional probability (3) Marginal distribution w.r.t. the two variables Φ 1 (ξ 1 ) = Φ 2 (ξ 2 ) = Z ξ1 Z + Z + Z ξ2 «D ξ (x, y)dy dx «D ξ (x, y)dx dy Marginal probabilistic density function D 1 (ξ 1 ) = Z + D ξ (ξ 1, y)dy D 2 (ξ 2 ) = Z + D ξ (x, ξ 2 )dx

70 EXTENSION TO HIGH DIMENSIONS Basics of multidimensional probability (4) Independant ξ 1 and ξ 2 = one variable provides no information about the other : φ ξ (ξ 1, ξ 2 ) = Φ 1 (ξ 1 )Φ 2 (ξ 2 ) D ξ (ξ 1, ξ 2 ) = D 1 (ξ 1 )D 2 (ξ 2 ) Null covariance is ensured for independant stochastic variable. Null covariance is (by far) a less stronger property than independance of stochatic variables Example : independant variables ξ 1 ξ 2 on [ 1, +1] 2 : D ξ (ξ 1, ξ 2 ) = D ξ (ξ 1, ξ 2 ) = (1. ξ2 1) (1. ξ2 2) 2

71 EXTENSION TO HIGH DIMENSIONS Basics of multidimensional probability (5) Example : independant variables ξ 1 ξ 2 on [ 1, +1] 2 : D ξ (ξ 1, ξ 2 ) = D ξ (ξ 1, ξ 2 ) = (1. ξ2 1) (1. ξ2 2) 2 Example = dependant variables ξ 1 ξ 2 on [ 1, +1] 2 : D a ξ (ξ 1, ξ 2 ) = D ξ (ξ 1, ξ 2 ) = 3/4 (1 1/4 (ξ 1 ξ 2 ) 2 ) D b ξ(ξ 1, ξ 2 ) = D ξ (ξ 1, ξ 2 ) = 3/4 (1 1/4 (ξ 1 + ξ 2 ) 2 ) Exercise = calculate all previous quantities : Dξ a 1 (ξ 1 ) = 3/4((5/6 1/2ξ1) 2 Dξ a 2 (ξ 2 ) = 3/4(5/6 1/2ξ2) 2 Dξ b 1 (ξ 1 ) = 3/4((5/6 1/2ξ1) 2 Dξ b 2 (ξ 2 ) = 3/4(5/6 1/2ξ2) 2 E a (ξ 1 ) = 0 E a (ξ 2 ) = 0 E b (ξ 1 ) = 0 E b (ξ 2 ) = 0 cov a (ξ 1, ξ 2 ) = 1/6 cov b (ξ 1, ξ 2 ) = 1/6

72 EXTENSION TO HIGH DIMENSIONS Changes from 1D to dimension d Several uncertainties considered simultaneously ξ is a vector ξ = (ξ 1, ξ 2.., ξ d ) Caution : subscript = components <> exponent = number in a sampling What does not change: Monte-Carlo method What does not change very much: Monte-Carlo plus metamodel Polynomial chaos method for independant uncertain ξ k and tensorial basis of polynomials. Size of polynomial basis (collocation grid) G d What does change: Polynomial chaos method for dependant uncertain ξ k or degree-m multivariate polynomials

73 EXTENSION TO HIGH DIMENSIONS Monte-Carlo method R d Basic Monte-Carlo No difference between R and R d Monte-Carlo plus Kriging, RBF, SVM/SVR... Number of inner parameters of the meta-model prop. to d Cost of the definition of an accurate meta-model increasing with d Monte-Carlo plus adjoint-based first or second order Taylor expansion First order : solve one adjoint equation to compute first-order derivative Second order : solve d direct equations plus one adjoint equation to compute first- and second-order derivatives

74 EXTENSION TO HIGH DIMENSIONS Independant uncertain variables. Tensorial-product polynomial basis (1/2) D(ξ) = D 1 (ξ 1 )D 2 (ξ 2 ) For 2D problem (i generic grid index) W (i, ξ) P CW (i, ξ) = l 1 =M,l 2 =M l 1 =0,l 2 =0 C l1,l 2 (i)p l1 (ξ 1 )Q l2 (ξ 2 ) P orthonormal polynomials associated to D 1 (ξ 1 ) Q orthonormal polynomials associated to D 2 (ξ 2 ) Calculate C l1,l 2 (i) by collocation Calculate C l1,l 2 (i) by quadrature using : Z C l1,l 2 (i) =< P CW, P l1 Q l2 >= P CW (i, ξ)p l1 (ξ 1 )Q l2 (ξ 2 )D 1 (ξ 1 )D 2 (ξ 2 )dξ 1 dξ 2 C l1,l 2 (i) < W, P l1 Q l2 >= Z W (i, ξ)p l1 (ξ 1 )Q l2 (ξ 2 )D 1 (ξ 1 )D 2 (ξ 2 )dξ 1 dξ 2 using a G G grid based on Gaussian quadrature points

75 EXTENSION TO HIGH DIMENSIONS Independant uncertain variables. Tensorial-product polynomial basis (2/2) Accuracy issue. How close is W (i, ξ)p l1 (ξ 1 )Q l2 (ξ 2 ) from a (2G 1) in ξ 1 times a (2G 1) polynomial in ξ 2? Stochastic post-processing done for the expansion P CW instead of W straightforward evaluation of mean value E(P CW (i, ξ)) =< P CW (i, ξ), 1 >= C 0,0 (i) straightforward evaluation of variance E((P CW (i, ξ) E(P CW (i))) 2 ) = l 1 =M,l 2 =M l 1 =0,l 2 =0 (l 1,l 2 )<>(0,0) C l 1,l 2 (i) 2

76 EXTENSION TO HIGH DIMENSIONS Independant uncertain variables. Tensorial-grid probabilistic collocation (1/2) D(ξ) = D 1 (ξ 1 )D 2 (ξ 2 ) Lagrangian polynomial expansion For 2D problem W (i, ξ) SCW (i, ξ) = l 1 =N,l 2 =N H l (ξ j ) = Π m=n m=1,m<>l l 1 =1,l 2 =1 W l 1,l 2 (i)h l1 (ξ 1 )H l2 (ξ 2 ) (ξ j ξ j,m ) (ξ j,l ξ j,m ) (j = 1 or 2) Note that actually SCW (i, ξ l1, ξ l2 ) = W l1,l 2 (i). no coefficient computation, define W l1,l 2 (i) = W (i, ξ l1, ξ l2 ) Definition of (ξ j,1, ξ j,2,..., ξ j,n ) (j = 1 or 2) (most often, not absolutely necessary) N points of the N-point quadrature associated to D j (ξ j )

77 EXTENSION TO HIGH DIMENSIONS Independant uncertain variables. Tensorial-grid probabilistic collocation (2/2) Stochastic post-processing (mean and variance) done for the expansion SCW instead of W straightforward evaluation of mean value E(SCW (i, ξ)) =< SCW (i, ξ), 1 >= l 1 =N,l 2 =N l 1 =1,l 2 =1 ω 1,l 1 ω 2,l2 W l1,l 2 (i) straightforward evaluation of variance E((SCW (i, ξ) E(SCW (i))) 2 ) = l 1 =N,l 2 =N l 1 =1,l 2 =1 ω 1,l 1 ω 2,l2 W l1,l 2 (i) 2 ( l 1 =N,l 2 =N l 1 =1,l 2 =1 ω 1,l 1 ω 2,l2 W l1,l 2 (i)) 2

78 EXTENSION TO HIGH DIMENSIONS Dependant uncertain variables. True degree M polynomials in R d Number of terms for a degree M polynomial in dimension d (M+d)! M!d! Number of continuous equations for intrusive approach Multiplied by (M+d)! M!d! instead of (M + 1) Computation of coefficients for non-intrusive approach Collocation : from at least (M+d)! M!d! flow computations Exact quadrature for degree-m polynomial [Smolyak 1963]... Define clever enrichment... Huge interest the community. Huge literature......attend (UQ) course level two

79 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

80 Some outputs of NODESIM-CFD M. Lazareff NODESIM-CFD = EU project Some important outputs of NODESIM-CFD Split (UQ) issues and CFD analysis issues. Test (UQ) propagation method on well defined CFD test cases AND mathematical models Compare standard deviation due to uncertainty to numerical errors and modelization errors Need for accurate definition of the input p.d.f. Actually, most often non intrusive methods for complex test cases

81 RAE2822 profile Configuration characteristics RAE2822 airfoil Re = , M = 0.734, α = 2.79 o Classical test case of high-re transonic CFD Complex flow with upper and lower shocks Strong dependance on turbulence model and mesh density

82 RAE2822 profile, k ω model, Re = around M = 0.734, α = 2.79 o TPS C L response surface with SST correction medium CFD grid, low-density collocation (blue spheres)

83 RAE2822 profile, k ω model, Re = around M = 0.734, α = 2.79 o TPS C L response surface without (transparent) and with (opaque, lower) SST correction medium CFD grid, low-density collocation (white/blue spheres)

84 NASA Rotor37 compressor Configuration characteristics Isolated rotor blade. Transonic flow. Subsonic outlet. Uncertainty propagation exercises β distribution (bounded support) uncertainty on outlet static-pressure and tip-clearance uncertainty quantification for several points of characteristic curve (several outlet static pressure <> several mass flows) polynomial chaos (gaussian quadrature) applied to functional outputs good quadrature convergence is observed computed order-4 moments almost identical to order-3 ones use order-3 Jacobi polynomials (associated to β distributions) only 4 CFD points needed for near-complete characterisation

85 NASA Rotor37 compressor Barplots for uncertainties on outlet pressure (left) and tip clearance (right) variations

86 NASA Rotor37 compressor Tip clearance : order-1 (mean) moments for global quantities

87 NASA Rotor37 compressor Tip clearance : order-2 (stdev) moments for global quantities

88 NASA Rotor37 compressor Application of intrusive polynomial chaos method Assessment of intrusive and non-intrusive non-deterministic CFD methodologies based on polynomial chaos expansions. [Dinsecu C, Smirnov S, Hirsch C, Lacor C. IJESMS 2 (2010)] Comparison between intrusive PC and probabilistic collocation NASA Rotor 37 (Compressor map) Uncertain parameters : outlet static pressure, height of tip clearance Output parameters : local static pressure (middle gridline, mid-span), pressure ration, isentropic efficiency Very consistent results for mean and standard deviation

89 OUTLINE Introduction Probability basics and Monte-Carlo Orthogonal/orthonormal polynomials Non-intrusive polynomial chaos methods in 1D Intrusive polynomial chaos method in 1D Extension to high dimensions Examples of application Conclusion

90 Conclusion Uncertainty quantification more and more interest and projects (EU, RTO... ) get definition of industry relevant problems robust design optimization process Methods 1D definition multi-d, sparsity, clever enrichment... Challenges Get relevant p.d.f. for uncertain parameters Deal efficiently with large number of uncertain parameters Deal efficiently with geometrical uncertainties

Dinesh Kumar, Mehrdad Raisee and Chris Lacor

Dinesh Kumar, Mehrdad Raisee and Chris Lacor Dinesh Kumar, Mehrdad Raisee and Chris Lacor Fluid Mechanics and Thermodynamics Research Group Vrije Universiteit Brussel, BELGIUM dkumar@vub.ac.be; m_raisee@yahoo.com; chris.lacor@vub.ac.be October, 2014

More information

Uncertainty Management and Quantification in Industrial Analysis and Design

Uncertainty Management and Quantification in Industrial Analysis and Design Uncertainty Management and Quantification in Industrial Analysis and Design www.numeca.com Charles Hirsch Professor, em. Vrije Universiteit Brussel President, NUMECA International The Role of Uncertainties

More information

Polynomial chaos expansions for sensitivity analysis

Polynomial chaos expansions for sensitivity analysis c DEPARTMENT OF CIVIL, ENVIRONMENTAL AND GEOMATIC ENGINEERING CHAIR OF RISK, SAFETY & UNCERTAINTY QUANTIFICATION Polynomial chaos expansions for sensitivity analysis B. Sudret Chair of Risk, Safety & Uncertainty

More information

Fast Numerical Methods for Stochastic Computations

Fast Numerical Methods for Stochastic Computations Fast AreviewbyDongbinXiu May 16 th,2013 Outline Motivation 1 Motivation 2 3 4 5 Example: Burgers Equation Let us consider the Burger s equation: u t + uu x = νu xx, x [ 1, 1] u( 1) =1 u(1) = 1 Example:

More information

A Non-Intrusive Polynomial Chaos Method For Uncertainty Propagation in CFD Simulations

A Non-Intrusive Polynomial Chaos Method For Uncertainty Propagation in CFD Simulations An Extended Abstract submitted for the 44th AIAA Aerospace Sciences Meeting and Exhibit, Reno, Nevada January 26 Preferred Session Topic: Uncertainty quantification and stochastic methods for CFD A Non-Intrusive

More information

Efficient Sampling for Non-Intrusive Polynomial Chaos Applications with Multiple Uncertain Input Variables

Efficient Sampling for Non-Intrusive Polynomial Chaos Applications with Multiple Uncertain Input Variables Missouri University of Science and Technology Scholars' Mine Mechanical and Aerospace Engineering Faculty Research & Creative Works Mechanical and Aerospace Engineering 4-1-2007 Efficient Sampling for

More information

Introduction to Uncertainty Quantification in Computational Science Handout #3

Introduction to Uncertainty Quantification in Computational Science Handout #3 Introduction to Uncertainty Quantification in Computational Science Handout #3 Gianluca Iaccarino Department of Mechanical Engineering Stanford University June 29 - July 1, 2009 Scuola di Dottorato di

More information

A Polynomial Chaos Approach to Robust Multiobjective Optimization

A Polynomial Chaos Approach to Robust Multiobjective Optimization A Polynomial Chaos Approach to Robust Multiobjective Optimization Silvia Poles 1, Alberto Lovison 2 1 EnginSoft S.p.A., Optimization Consulting Via Giambellino, 7 35129 Padova, Italy s.poles@enginsoft.it

More information

Adjoint based multi-objective shape optimization of a transonic airfoil under uncertainties

Adjoint based multi-objective shape optimization of a transonic airfoil under uncertainties EngOpt 2016-5 th International Conference on Engineering Optimization Iguassu Falls, Brazil, 19-23 June 2016. Adjoint based multi-objective shape optimization of a transonic airfoil under uncertainties

More information

Uncertainty Quantification in Computational Models

Uncertainty Quantification in Computational Models Uncertainty Quantification in Computational Models Habib N. Najm Sandia National Laboratories, Livermore, CA, USA Workshop on Understanding Climate Change from Data (UCC11) University of Minnesota, Minneapolis,

More information

A Non-Intrusive Polynomial Chaos Method for Uncertainty Propagation in CFD Simulations

A Non-Intrusive Polynomial Chaos Method for Uncertainty Propagation in CFD Simulations Missouri University of Science and Technology Scholars' Mine Mechanical and Aerospace Engineering Faculty Research & Creative Works Mechanical and Aerospace Engineering --6 A Non-Intrusive Polynomial Chaos

More information

Polynomial chaos expansions for structural reliability analysis

Polynomial chaos expansions for structural reliability analysis DEPARTMENT OF CIVIL, ENVIRONMENTAL AND GEOMATIC ENGINEERING CHAIR OF RISK, SAFETY & UNCERTAINTY QUANTIFICATION Polynomial chaos expansions for structural reliability analysis B. Sudret & S. Marelli Incl.

More information

Keywords: Sonic boom analysis, Atmospheric uncertainties, Uncertainty quantification, Monte Carlo method, Polynomial chaos method.

Keywords: Sonic boom analysis, Atmospheric uncertainties, Uncertainty quantification, Monte Carlo method, Polynomial chaos method. Blucher Mechanical Engineering Proceedings May 2014, vol. 1, num. 1 www.proceedings.blucher.com.br/evento/10wccm SONIC BOOM ANALYSIS UNDER ATMOSPHERIC UNCERTAINTIES BY A NON-INTRUSIVE POLYNOMIAL CHAOS

More information

Overview. Bayesian assimilation of experimental data into simulation (for Goland wing flutter) Why not uncertainty quantification?

Overview. Bayesian assimilation of experimental data into simulation (for Goland wing flutter) Why not uncertainty quantification? Delft University of Technology Overview Bayesian assimilation of experimental data into simulation (for Goland wing flutter), Simao Marques 1. Why not uncertainty quantification? 2. Why uncertainty quantification?

More information

Stochastic Spectral Approaches to Bayesian Inference

Stochastic Spectral Approaches to Bayesian Inference Stochastic Spectral Approaches to Bayesian Inference Prof. Nathan L. Gibson Department of Mathematics Applied Mathematics and Computation Seminar March 4, 2011 Prof. Gibson (OSU) Spectral Approaches to

More information

Stochastic analysis of a radial-inflow turbine in the presence of parametric uncertainties

Stochastic analysis of a radial-inflow turbine in the presence of parametric uncertainties ICCM2015, 14-17 th July, Auckland, NZ Stochastic analysis of a radial-inflow turbine in the presence of parametric uncertainties *A. Zou¹, E. Sauret 1, J.-C. Chassaing 2, S. C. Saha 1, and YT Gu 1 1 School

More information

Estimating functional uncertainty using polynomial chaos and adjoint equations

Estimating functional uncertainty using polynomial chaos and adjoint equations 0. Estimating functional uncertainty using polynomial chaos and adjoint equations February 24, 2011 1 Florida State University, Tallahassee, Florida, Usa 2 Moscow Institute of Physics and Technology, Moscow,

More information

SENSITIVITY ANALYSIS IN NUMERICAL SIMULATION OF MULTIPHASE FLOW FOR CO 2 STORAGE IN SALINE AQUIFERS USING THE PROBABILISTIC COLLOCATION APPROACH

SENSITIVITY ANALYSIS IN NUMERICAL SIMULATION OF MULTIPHASE FLOW FOR CO 2 STORAGE IN SALINE AQUIFERS USING THE PROBABILISTIC COLLOCATION APPROACH XIX International Conference on Water Resources CMWR 2012 University of Illinois at Urbana-Champaign June 17-22,2012 SENSITIVITY ANALYSIS IN NUMERICAL SIMULATION OF MULTIPHASE FLOW FOR CO 2 STORAGE IN

More information

Performance Evaluation of Generalized Polynomial Chaos

Performance Evaluation of Generalized Polynomial Chaos Performance Evaluation of Generalized Polynomial Chaos Dongbin Xiu, Didier Lucor, C.-H. Su, and George Em Karniadakis 1 Division of Applied Mathematics, Brown University, Providence, RI 02912, USA, gk@dam.brown.edu

More information

Application of Dual Time Stepping to Fully Implicit Runge Kutta Schemes for Unsteady Flow Calculations

Application of Dual Time Stepping to Fully Implicit Runge Kutta Schemes for Unsteady Flow Calculations Application of Dual Time Stepping to Fully Implicit Runge Kutta Schemes for Unsteady Flow Calculations Antony Jameson Department of Aeronautics and Astronautics, Stanford University, Stanford, CA, 94305

More information

Algorithms for Uncertainty Quantification

Algorithms for Uncertainty Quantification Algorithms for Uncertainty Quantification Lecture 9: Sensitivity Analysis ST 2018 Tobias Neckel Scientific Computing in Computer Science TUM Repetition of Previous Lecture Sparse grids in Uncertainty Quantification

More information

Random Matrix Eigenvalue Problems in Probabilistic Structural Mechanics

Random Matrix Eigenvalue Problems in Probabilistic Structural Mechanics Random Matrix Eigenvalue Problems in Probabilistic Structural Mechanics S Adhikari Department of Aerospace Engineering, University of Bristol, Bristol, U.K. URL: http://www.aer.bris.ac.uk/contact/academic/adhikari/home.html

More information

Solving the steady state diffusion equation with uncertainty Final Presentation

Solving the steady state diffusion equation with uncertainty Final Presentation Solving the steady state diffusion equation with uncertainty Final Presentation Virginia Forstall vhfors@gmail.com Advisor: Howard Elman elman@cs.umd.edu Department of Computer Science May 6, 2012 Problem

More information

Uncertainty Evolution In Stochastic Dynamic Models Using Polynomial Chaos

Uncertainty Evolution In Stochastic Dynamic Models Using Polynomial Chaos Noname manuscript No. (will be inserted by the editor) Uncertainty Evolution In Stochastic Dynamic Models Using Polynomial Chaos Umamaheswara Konda Puneet Singla Tarunraj Singh Peter Scott Received: date

More information

Sampling and low-rank tensor approximation of the response surface

Sampling and low-rank tensor approximation of the response surface Sampling and low-rank tensor approximation of the response surface tifica Alexander Litvinenko 1,2 (joint work with Hermann G. Matthies 3 ) 1 Group of Raul Tempone, SRI UQ, and 2 Group of David Keyes,

More information

Effect of Geometric Uncertainties on the Aerodynamic Characteristic of Offshore Wind Turbine Blades

Effect of Geometric Uncertainties on the Aerodynamic Characteristic of Offshore Wind Turbine Blades the Aerodynamic Offshore Wind, Henning Schmitt, Jörg R. Seume The Science of Making Torque from Wind 2012 October 9-11, 2012 Oldenburg, Germany 1. Motivation and 2. 3. 4. 5. Conclusions and Slide 2 / 12

More information

Robust optimization for windmill airfoil design under variable wind conditions

Robust optimization for windmill airfoil design under variable wind conditions Center for Turbulence Research Proceedings of the Summer Program 2012 241 Robust optimization for windmill airfoil design under variable wind conditions By H. Tachikawa, D. Schiavazzi, T. Arima AND G.

More information

Reducing uncertainties in a wind-tunnel experiment using Bayesian updating

Reducing uncertainties in a wind-tunnel experiment using Bayesian updating Reducing uncertainties in a wind-tunnel experiment using Bayesian updating D.J. Boon, R.P. Dwight, J.J.H.M. Sterenborg, and H. Bijl Aerodynamics Group, Delft University of Technology, The Netherlands We

More information

Finite volume method on unstructured grids

Finite volume method on unstructured grids Finite volume method on unstructured grids Praveen. C praveen@math.tifrbng.res.in Tata Institute of Fundamental Research Center for Applicable Mathematics Bangalore 560065 http://math.tifrbng.res.in/~praveen

More information

A Crash-Course on the Adjoint Method for Aerodynamic Shape Optimization

A Crash-Course on the Adjoint Method for Aerodynamic Shape Optimization A Crash-Course on the Adjoint Method for Aerodynamic Shape Optimization Juan J. Alonso Department of Aeronautics & Astronautics Stanford University jjalonso@stanford.edu Lecture 19 AA200b Applied Aerodynamics

More information

UNCERTAINTY QUANTIFICATION AND ROBUST OPTIMIZATION FOR THROUGHFLOW AXIAL COMPRESSOR DESIGN

UNCERTAINTY QUANTIFICATION AND ROBUST OPTIMIZATION FOR THROUGHFLOW AXIAL COMPRESSOR DESIGN Proceedings of 11 th European Conference on Turbomachinery Fluid dynamics & Thermodynamics ETC11, March 23-27, 2015, Madrid, Spain UNCERTAINTY QUANTIFICATION AND ROBUST OPTIMIZATION FOR THROUGHFLOW AXIAL

More information

Uncertainty Analysis of Computational Fluid Dynamics Via Polynomial Chaos

Uncertainty Analysis of Computational Fluid Dynamics Via Polynomial Chaos Uncertainty Analysis of Computational Fluid Dynamics Via Polynomial Chaos Rafael A. Perez Dissertation submitted to the faculty of the Virginia Polytechnic Institute and State University in partial fulfillment

More information

Quadrature for Uncertainty Analysis Stochastic Collocation. What does quadrature have to do with uncertainty?

Quadrature for Uncertainty Analysis Stochastic Collocation. What does quadrature have to do with uncertainty? Quadrature for Uncertainty Analysis Stochastic Collocation What does quadrature have to do with uncertainty? Quadrature for Uncertainty Analysis Stochastic Collocation What does quadrature have to do with

More information

EFFICIENT SHAPE OPTIMIZATION USING POLYNOMIAL CHAOS EXPANSION AND LOCAL SENSITIVITIES

EFFICIENT SHAPE OPTIMIZATION USING POLYNOMIAL CHAOS EXPANSION AND LOCAL SENSITIVITIES 9 th ASCE Specialty Conference on Probabilistic Mechanics and Structural Reliability EFFICIENT SHAPE OPTIMIZATION USING POLYNOMIAL CHAOS EXPANSION AND LOCAL SENSITIVITIES Nam H. Kim and Haoyu Wang University

More information

Uncertainty Quantification of an ORC turbine blade under a low quantile constrain

Uncertainty Quantification of an ORC turbine blade under a low quantile constrain Available online at www.sciencedirect.com ScienceDirect Energy Procedia 129 (2017) 1149 1155 www.elsevier.com/locate/procedia IV International Seminar on ORC Power Systems, ORC2017 13-15 September 2017,

More information

arxiv: v1 [math.na] 3 Apr 2019

arxiv: v1 [math.na] 3 Apr 2019 arxiv:1904.02017v1 [math.na] 3 Apr 2019 Poly-Sinc Solution of Stochastic Elliptic Differential Equations Maha Youssef and Roland Pulch Institute of Mathematics and Computer Science, University of Greifswald,

More information

Random Eigenvalue Problems Revisited

Random Eigenvalue Problems Revisited Random Eigenvalue Problems Revisited S Adhikari Department of Aerospace Engineering, University of Bristol, Bristol, U.K. Email: S.Adhikari@bristol.ac.uk URL: http://www.aer.bris.ac.uk/contact/academic/adhikari/home.html

More information

Efficient Modelling of a Nonlinear Gust Loads Process for Uncertainty Quantification of Highly Flexible Aircraft

Efficient Modelling of a Nonlinear Gust Loads Process for Uncertainty Quantification of Highly Flexible Aircraft Efficient Modelling of a Nonlinear Gust Loads Process for Uncertainty Quantification of Highly Flexible Aircraft SciTech, 8 th -12 th January 2018 Kissimmee, Florida Robert Cook, Chris Wales, Ann Gaitonde,

More information

Uncertainty Quantification in Computational Science

Uncertainty Quantification in Computational Science DTU 2010 - Lecture I Uncertainty Quantification in Computational Science Jan S Hesthaven Brown University Jan.Hesthaven@Brown.edu Objective of lectures The main objective of these lectures are To offer

More information

Stochastic Collocation Methods for Polynomial Chaos: Analysis and Applications

Stochastic Collocation Methods for Polynomial Chaos: Analysis and Applications Stochastic Collocation Methods for Polynomial Chaos: Analysis and Applications Dongbin Xiu Department of Mathematics, Purdue University Support: AFOSR FA955-8-1-353 (Computational Math) SF CAREER DMS-64535

More information

A reduced-order stochastic finite element analysis for structures with uncertainties

A reduced-order stochastic finite element analysis for structures with uncertainties A reduced-order stochastic finite element analysis for structures with uncertainties Ji Yang 1, Béatrice Faverjon 1,2, Herwig Peters 1, icole Kessissoglou 1 1 School of Mechanical and Manufacturing Engineering,

More information

Robust Optimal Control using Polynomial Chaos and Adjoints for Systems with Uncertain Inputs

Robust Optimal Control using Polynomial Chaos and Adjoints for Systems with Uncertain Inputs 2th AIAA Computational Fluid Dynamics Conference 27-3 June 211, Honolulu, Hawaii AIAA 211-369 Robust Optimal Control using Polynomial Chaos and Adjoints for Systems with Uncertain Inputs Sriram General

More information

Sparse polynomial chaos expansions in engineering applications

Sparse polynomial chaos expansions in engineering applications DEPARTMENT OF CIVIL, ENVIRONMENTAL AND GEOMATIC ENGINEERING CHAIR OF RISK, SAFETY & UNCERTAINTY QUANTIFICATION Sparse polynomial chaos expansions in engineering applications B. Sudret G. Blatman (EDF R&D,

More information

Suggestions for Making Useful the Uncertainty Quantification Results from CFD Applications

Suggestions for Making Useful the Uncertainty Quantification Results from CFD Applications Suggestions for Making Useful the Uncertainty Quantification Results from CFD Applications Thomas A. Zang tzandmands@wildblue.net Aug. 8, 2012 CFD Futures Conference: Zang 1 Context The focus of this presentation

More information

Steps in Uncertainty Quantification

Steps in Uncertainty Quantification Steps in Uncertainty Quantification Challenge: How do we do uncertainty quantification for computationally expensive models? Example: We have a computational budget of 5 model evaluations. Bayesian inference

More information

Accepted Manuscript. SAMBA: Sparse approximation of moment-based arbitrary polynomial chaos. R. Ahlfeld, B. Belkouchi, F.

Accepted Manuscript. SAMBA: Sparse approximation of moment-based arbitrary polynomial chaos. R. Ahlfeld, B. Belkouchi, F. Accepted Manuscript SAMBA: Sparse approximation of moment-based arbitrary polynomial chaos R. Ahlfeld, B. Belkouchi, F. Montomoli PII: S0021-9991(16)30151-6 DOI: http://dx.doi.org/10.1016/j.jcp.2016.05.014

More information

Uncertainty Quantification of Two-Phase Flow in Heterogeneous Porous Media

Uncertainty Quantification of Two-Phase Flow in Heterogeneous Porous Media Uncertainty Quantification of Two-Phase Flow in Heterogeneous Porous Media M.Köppel, C.Rohde Institute for Applied Analysis and Numerical Simulation Inria, Nov 15th, 2016 Porous Media Examples: sponge,

More information

Uncertainty Quantification in MEMS

Uncertainty Quantification in MEMS Uncertainty Quantification in MEMS N. Agarwal and N. R. Aluru Department of Mechanical Science and Engineering for Advanced Science and Technology Introduction Capacitive RF MEMS switch Comb drive Various

More information

Simultaneous Robust Design and Tolerancing of Compressor Blades

Simultaneous Robust Design and Tolerancing of Compressor Blades Simultaneous Robust Design and Tolerancing of Compressor Blades Eric Dow Aerospace Computational Design Laboratory Department of Aeronautics and Astronautics Massachusetts Institute of Technology GTL Seminar

More information

This ODE arises in many physical systems that we shall investigate. + ( + 1)u = 0. (λ + s)x λ + s + ( + 1) a λ. (s + 1)(s + 2) a 0

This ODE arises in many physical systems that we shall investigate. + ( + 1)u = 0. (λ + s)x λ + s + ( + 1) a λ. (s + 1)(s + 2) a 0 Legendre equation This ODE arises in many physical systems that we shall investigate We choose We then have Substitution gives ( x 2 ) d 2 u du 2x 2 dx dx + ( + )u u x s a λ x λ a du dx λ a λ (λ + s)x

More information

CERTAIN THOUGHTS ON UNCERTAINTY ANALYSIS FOR DYNAMICAL SYSTEMS

CERTAIN THOUGHTS ON UNCERTAINTY ANALYSIS FOR DYNAMICAL SYSTEMS CERTAIN THOUGHTS ON UNCERTAINTY ANALYSIS FOR DYNAMICAL SYSTEMS Puneet Singla Assistant Professor Department of Mechanical & Aerospace Engineering University at Buffalo, Buffalo, NY-1426 Probabilistic Analysis

More information

TECHNISCHE UNIVERSITÄT MÜNCHEN. Uncertainty Quantification in Fluid Flows via Polynomial Chaos Methodologies

TECHNISCHE UNIVERSITÄT MÜNCHEN. Uncertainty Quantification in Fluid Flows via Polynomial Chaos Methodologies TECHNISCHE UNIVERSITÄT MÜNCHEN Bachelor s Thesis in Engineering Science Uncertainty Quantification in Fluid Flows via Polynomial Chaos Methodologies Jan Sültemeyer DEPARTMENT OF INFORMATICS MUNICH SCHOOL

More information

An Empirical Chaos Expansion Method for Uncertainty Quantification

An Empirical Chaos Expansion Method for Uncertainty Quantification An Empirical Chaos Expansion Method for Uncertainty Quantification Melvin Leok and Gautam Wilkins Abstract. Uncertainty quantification seeks to provide a quantitative means to understand complex systems

More information

Uncertainty Quantification and Validation Using RAVEN. A. Alfonsi, C. Rabiti. Risk-Informed Safety Margin Characterization. https://lwrs.inl.

Uncertainty Quantification and Validation Using RAVEN. A. Alfonsi, C. Rabiti. Risk-Informed Safety Margin Characterization. https://lwrs.inl. Risk-Informed Safety Margin Characterization Uncertainty Quantification and Validation Using RAVEN https://lwrs.inl.gov A. Alfonsi, C. Rabiti North Carolina State University, Raleigh 06/28/2017 Assumptions

More information

Aeroelastic Gust Response

Aeroelastic Gust Response Aeroelastic Gust Response Civil Transport Aircraft - xxx Presented By: Fausto Gill Di Vincenzo 04-06-2012 What is Aeroelasticity? Aeroelasticity studies the effect of aerodynamic loads on flexible structures,

More information

Analysis and Simulation of Blood Flow in the Portal Vein with Uncertainty Quantification

Analysis and Simulation of Blood Flow in the Portal Vein with Uncertainty Quantification Analysis and Simulation of Blood Flow in the Portal Vein with Uncertainty Quantification João Pedro Carvalho Rêgo de Serra e Moura Instituto Superior Técnico Abstract Blood flow simulations in CFD are

More information

Stochastic Solvers for the Euler Equations

Stochastic Solvers for the Euler Equations 43rd AIAA Aerospace Sciences Meeting and Exhibit 1-13 January 5, Reno, Nevada 5-873 Stochastic Solvers for the Euler Equations G. Lin, C.-H. Su and G.E. Karniadakis Division of Applied Mathematics Brown

More information

A Multi-Dimensional Limiter for Hybrid Grid

A Multi-Dimensional Limiter for Hybrid Grid APCOM & ISCM 11-14 th December, 2013, Singapore A Multi-Dimensional Limiter for Hybrid Grid * H. W. Zheng ¹ 1 State Key Laboratory of High Temperature Gas Dynamics, Institute of Mechanics, Chinese Academy

More information

Discrete and continuous adjoint method for compressible CFD J. Peter ONERA

Discrete and continuous adjoint method for compressible CFD J. Peter ONERA Discrete and continuous adjoint method for compressible CFD J. Peter ONERA J. Peter 1 1 ONERA DMFN October 2, 2014 J. Peter (ONERA DMFN) Adjoint method for compressible CFD October 2, 2014 1 / 60 Outline

More information

Compressible Potential Flow: The Full Potential Equation. Copyright 2009 Narayanan Komerath

Compressible Potential Flow: The Full Potential Equation. Copyright 2009 Narayanan Komerath Compressible Potential Flow: The Full Potential Equation 1 Introduction Recall that for incompressible flow conditions, velocity is not large enough to cause density changes, so density is known. Thus

More information

Dimension-adaptive sparse grid for industrial applications using Sobol variances

Dimension-adaptive sparse grid for industrial applications using Sobol variances Master of Science Thesis Dimension-adaptive sparse grid for industrial applications using Sobol variances Heavy gas flow over a barrier March 11, 2015 Ad Dimension-adaptive sparse grid for industrial

More information

Transonic Aerodynamics Wind Tunnel Testing Considerations. W.H. Mason Configuration Aerodynamics Class

Transonic Aerodynamics Wind Tunnel Testing Considerations. W.H. Mason Configuration Aerodynamics Class Transonic Aerodynamics Wind Tunnel Testing Considerations W.H. Mason Configuration Aerodynamics Class Transonic Aerodynamics History Pre WWII propeller tip speeds limited airplane speed Props did encounter

More information

International Journal of Modern Trends in Engineering and Research e-issn No.: , Date: 2-4 July, 2015

International Journal of Modern Trends in Engineering and Research  e-issn No.: , Date: 2-4 July, 2015 International Journal of Modern Trends in Engineering and Research www.ijmter.com e-issn No.:2349-9745, Date: 2-4 July, 2015 CFD Analysis of Airfoil NACA0012 Pritesh S. Gugliya 1, Yogesh R. Jaiswal 2,

More information

Monte Carlo Methods for Uncertainty Quantification

Monte Carlo Methods for Uncertainty Quantification Monte Carlo Methods for Uncertainty Quantification Mike Giles Mathematical Institute, University of Oxford Contemporary Numerical Techniques Mike Giles (Oxford) Monte Carlo methods 1 / 23 Lecture outline

More information

Masters in Mechanical Engineering. Problems of incompressible viscous flow. 2µ dx y(y h)+ U h y 0 < y < h,

Masters in Mechanical Engineering. Problems of incompressible viscous flow. 2µ dx y(y h)+ U h y 0 < y < h, Masters in Mechanical Engineering Problems of incompressible viscous flow 1. Consider the laminar Couette flow between two infinite flat plates (lower plate (y = 0) with no velocity and top plate (y =

More information

Hyperbolic Polynomial Chaos Expansion (HPCE) and its Application to Statistical Analysis of Nonlinear Circuits

Hyperbolic Polynomial Chaos Expansion (HPCE) and its Application to Statistical Analysis of Nonlinear Circuits Hyperbolic Polynomial Chaos Expansion HPCE and its Application to Statistical Analysis of Nonlinear Circuits Majid Ahadi, Aditi Krishna Prasad, Sourajeet Roy High Speed System Simulations Laboratory Department

More information

Solving the stochastic steady-state diffusion problem using multigrid

Solving the stochastic steady-state diffusion problem using multigrid IMA Journal of Numerical Analysis (2007) 27, 675 688 doi:10.1093/imanum/drm006 Advance Access publication on April 9, 2007 Solving the stochastic steady-state diffusion problem using multigrid HOWARD ELMAN

More information

Evaluation of Non-Intrusive Approaches for Wiener-Askey Generalized Polynomial Chaos

Evaluation of Non-Intrusive Approaches for Wiener-Askey Generalized Polynomial Chaos 49th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference 6t 7 - April 28, Schaumburg, IL AIAA 28-892 Evaluation of Non-Intrusive Approaches for Wiener-Askey Generalized

More information

Solving the Stochastic Steady-State Diffusion Problem Using Multigrid

Solving the Stochastic Steady-State Diffusion Problem Using Multigrid Solving the Stochastic Steady-State Diffusion Problem Using Multigrid Tengfei Su Applied Mathematics and Scientific Computing Advisor: Howard Elman Department of Computer Science Sept. 29, 2015 Tengfei

More information

arxiv: v2 [physics.comp-ph] 15 Sep 2015

arxiv: v2 [physics.comp-ph] 15 Sep 2015 On Uncertainty Quantification of Lithium-ion Batteries: Application to an LiC 6 /LiCoO 2 cell arxiv:1505.07776v2 [physics.comp-ph] 15 Sep 2015 Mohammad Hadigol a, Kurt Maute a, Alireza Doostan a, a Aerospace

More information

Is My CFD Mesh Adequate? A Quantitative Answer

Is My CFD Mesh Adequate? A Quantitative Answer Is My CFD Mesh Adequate? A Quantitative Answer Krzysztof J. Fidkowski Gas Dynamics Research Colloqium Aerospace Engineering Department University of Michigan January 26, 2011 K.J. Fidkowski (UM) GDRC 2011

More information

AProofoftheStabilityoftheSpectral Difference Method For All Orders of Accuracy

AProofoftheStabilityoftheSpectral Difference Method For All Orders of Accuracy AProofoftheStabilityoftheSpectral Difference Method For All Orders of Accuracy Antony Jameson 1 1 Thomas V. Jones Professor of Engineering Department of Aeronautics and Astronautics Stanford University

More information

Beyond Wiener Askey Expansions: Handling Arbitrary PDFs

Beyond Wiener Askey Expansions: Handling Arbitrary PDFs Journal of Scientific Computing, Vol. 27, Nos. 1 3, June 2006 ( 2005) DOI: 10.1007/s10915-005-9038-8 Beyond Wiener Askey Expansions: Handling Arbitrary PDFs Xiaoliang Wan 1 and George Em Karniadakis 1

More information

DG Methods for Aerodynamic Flows: Higher Order, Error Estimation and Adaptive Mesh Refinement

DG Methods for Aerodynamic Flows: Higher Order, Error Estimation and Adaptive Mesh Refinement HONOM 2011 in Trento DG Methods for Aerodynamic Flows: Higher Order, Error Estimation and Adaptive Mesh Refinement Institute of Aerodynamics and Flow Technology DLR Braunschweig 11. April 2011 1 / 35 Research

More information

Large-eddy simulations for wind turbine blade: rotational augmentation and dynamic stall

Large-eddy simulations for wind turbine blade: rotational augmentation and dynamic stall Large-eddy simulations for wind turbine blade: rotational augmentation and dynamic stall Y. Kim, I.P. Castro, and Z.T. Xie Introduction Wind turbines operate in the atmospheric boundary layer and their

More information

Differentiation and Integration

Differentiation and Integration Differentiation and Integration (Lectures on Numerical Analysis for Economists II) Jesús Fernández-Villaverde 1 and Pablo Guerrón 2 February 12, 2018 1 University of Pennsylvania 2 Boston College Motivation

More information

Final abstract for ONERA Taylor-Green DG participation

Final abstract for ONERA Taylor-Green DG participation 1st International Workshop On High-Order CFD Methods January 7-8, 2012 at the 50th AIAA Aerospace Sciences Meeting, Nashville, Tennessee Final abstract for ONERA Taylor-Green DG participation JB Chapelier,

More information

Transition Modeling Activities at AS-C²A²S²E (DLR)

Transition Modeling Activities at AS-C²A²S²E (DLR) Transition Modeling Activities at AS-C²A²S²E (DLR) Andreas Krumbein German Aerospace Center (DLR) Institute of Aerodynamics and Flow Technology (AS) C²A²S²E - Center for Computer Applications in AeroSpace

More information

Uncertainty Quantification for multiscale kinetic equations with high dimensional random inputs with sparse grids

Uncertainty Quantification for multiscale kinetic equations with high dimensional random inputs with sparse grids Uncertainty Quantification for multiscale kinetic equations with high dimensional random inputs with sparse grids Shi Jin University of Wisconsin-Madison, USA Kinetic equations Different Q Boltmann Landau

More information

Accuracy, Precision and Efficiency in Sparse Grids

Accuracy, Precision and Efficiency in Sparse Grids John, Information Technology Department, Virginia Tech.... http://people.sc.fsu.edu/ jburkardt/presentations/ sandia 2009.pdf... Computer Science Research Institute, Sandia National Laboratory, 23 July

More information

Spectral Representation of Random Processes

Spectral Representation of Random Processes Spectral Representation of Random Processes Example: Represent u(t,x,q) by! u K (t, x, Q) = u k (t, x) k(q) where k(q) are orthogonal polynomials. Single Random Variable:! Let k (Q) be orthogonal with

More information

New issues in LES of turbulent flows: multiphysics and uncertainty modelling

New issues in LES of turbulent flows: multiphysics and uncertainty modelling New issues in LES of turbulent flows: multiphysics and uncertainty modelling Pierre Sagaut Institut Jean Le Rond d Alembert Université Pierre et Marie Curie- Paris 6, France http://www.lmm.jussieu.fr/~sagaut

More information

Mestrado Integrado em Engenharia Mecânica Aerodynamics 1 st Semester 2012/13

Mestrado Integrado em Engenharia Mecânica Aerodynamics 1 st Semester 2012/13 Mestrado Integrado em Engenharia Mecânica Aerodynamics 1 st Semester 212/13 Exam 2ª época, 2 February 213 Name : Time : 8: Number: Duration : 3 hours 1 st Part : No textbooks/notes allowed 2 nd Part :

More information

Enabling Advanced Automation Tools to manage Trajectory Prediction Uncertainty

Enabling Advanced Automation Tools to manage Trajectory Prediction Uncertainty Engineering, Test & Technology Boeing Research & Technology Enabling Advanced Automation Tools to manage Trajectory Prediction Uncertainty ART 12 - Automation Enrique Casado (BR&T-E) enrique.casado@boeing.com

More information

Fast Numerical Methods for Stochastic Computations: A Review

Fast Numerical Methods for Stochastic Computations: A Review COMMUNICATIONS IN COMPUTATIONAL PHYSICS Vol. 5, No. 2-4, pp. 242-272 Commun. Comput. Phys. February 2009 REVIEW ARTICLE Fast Numerical Methods for Stochastic Computations: A Review Dongbin Xiu Department

More information

UNCERTAINTY ASSESSMENT USING STOCHASTIC REDUCED BASIS METHOD FOR FLOW IN POROUS MEDIA

UNCERTAINTY ASSESSMENT USING STOCHASTIC REDUCED BASIS METHOD FOR FLOW IN POROUS MEDIA UNCERTAINTY ASSESSMENT USING STOCHASTIC REDUCED BASIS METHOD FOR FLOW IN POROUS MEDIA A REPORT SUBMITTED TO THE DEPARTMENT OF ENERGY RESOURCES ENGINEERING OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT

More information

Uncertainty Propagation and Global Sensitivity Analysis in Hybrid Simulation using Polynomial Chaos Expansion

Uncertainty Propagation and Global Sensitivity Analysis in Hybrid Simulation using Polynomial Chaos Expansion Uncertainty Propagation and Global Sensitivity Analysis in Hybrid Simulation using Polynomial Chaos Expansion EU-US-Asia workshop on hybrid testing Ispra, 5-6 October 2015 G. Abbiati, S. Marelli, O.S.

More information

Uncertainty Quantification and hypocoercivity based sensitivity analysis for multiscale kinetic equations with random inputs.

Uncertainty Quantification and hypocoercivity based sensitivity analysis for multiscale kinetic equations with random inputs. Uncertainty Quantification and hypocoercivity based sensitivity analysis for multiscale kinetic equations with random inputs Shi Jin University of Wisconsin-Madison, USA Shanghai Jiao Tong University,

More information

Efficient Hessian Calculation using Automatic Differentiation

Efficient Hessian Calculation using Automatic Differentiation 25th AIAA Applied Aerodynamics Conference, 25-28 June 2007, Miami Efficient Hessian Calculation using Automatic Differentiation Devendra P. Ghate and Michael B. Giles Oxford University Computing Laboratory,

More information

Final Report: DE-FG02-95ER25239 Spectral Representations of Uncertainty: Algorithms and Applications

Final Report: DE-FG02-95ER25239 Spectral Representations of Uncertainty: Algorithms and Applications Final Report: DE-FG02-95ER25239 Spectral Representations of Uncertainty: Algorithms and Applications PI: George Em Karniadakis Division of Applied Mathematics, Brown University April 25, 2005 1 Objectives

More information

Implicit Solution of Viscous Aerodynamic Flows using the Discontinuous Galerkin Method

Implicit Solution of Viscous Aerodynamic Flows using the Discontinuous Galerkin Method Implicit Solution of Viscous Aerodynamic Flows using the Discontinuous Galerkin Method Per-Olof Persson and Jaime Peraire Massachusetts Institute of Technology 7th World Congress on Computational Mechanics

More information

PART IV Spectral Methods

PART IV Spectral Methods PART IV Spectral Methods Additional References: R. Peyret, Spectral methods for incompressible viscous flow, Springer (2002), B. Mercier, An introduction to the numerical analysis of spectral methods,

More information

Introduction to Computational Stochastic Differential Equations

Introduction to Computational Stochastic Differential Equations Introduction to Computational Stochastic Differential Equations Gabriel J. Lord Catherine E. Powell Tony Shardlow Preface Techniques for solving many of the differential equations traditionally used by

More information

Collocation based high dimensional model representation for stochastic partial differential equations

Collocation based high dimensional model representation for stochastic partial differential equations Collocation based high dimensional model representation for stochastic partial differential equations S Adhikari 1 1 Swansea University, UK ECCM 2010: IV European Conference on Computational Mechanics,

More information

BACKGROUND NOTES FYS 4550/FYS EXPERIMENTAL HIGH ENERGY PHYSICS AUTUMN 2016 PROBABILITY A. STRANDLIE NTNU AT GJØVIK AND UNIVERSITY OF OSLO

BACKGROUND NOTES FYS 4550/FYS EXPERIMENTAL HIGH ENERGY PHYSICS AUTUMN 2016 PROBABILITY A. STRANDLIE NTNU AT GJØVIK AND UNIVERSITY OF OSLO ACKGROUND NOTES FYS 4550/FYS9550 - EXERIMENTAL HIGH ENERGY HYSICS AUTUMN 2016 ROAILITY A. STRANDLIE NTNU AT GJØVIK AND UNIVERSITY OF OSLO efore embarking on the concept of probability, we will first define

More information

Benjamin L. Pence 1, Hosam K. Fathy 2, and Jeffrey L. Stein 3

Benjamin L. Pence 1, Hosam K. Fathy 2, and Jeffrey L. Stein 3 2010 American Control Conference Marriott Waterfront, Baltimore, MD, USA June 30-July 02, 2010 WeC17.1 Benjamin L. Pence 1, Hosam K. Fathy 2, and Jeffrey L. Stein 3 (1) Graduate Student, (2) Assistant

More information

Divergence Formulation of Source Term

Divergence Formulation of Source Term Preprint accepted for publication in Journal of Computational Physics, 2012 http://dx.doi.org/10.1016/j.jcp.2012.05.032 Divergence Formulation of Source Term Hiroaki Nishikawa National Institute of Aerospace,

More information

Preliminary Study of Single Particle Lidar for Wing Wake Survey. M.Valla, B. Augère, D. Bailly, A. Dolfi-Bouteyre, E. Garnier, M.

Preliminary Study of Single Particle Lidar for Wing Wake Survey. M.Valla, B. Augère, D. Bailly, A. Dolfi-Bouteyre, E. Garnier, M. Preliminary Study of Single Particle Lidar for Wing Wake Survey M.Valla, B. Augère, D. Bailly, A. Dolfi-Bouteyre, E. Garnier, M. Méheut Context of research Clean Sky Joint Technology Initiative : Aims

More information

Prediction of airfoil performance at high Reynolds numbers.

Prediction of airfoil performance at high Reynolds numbers. Downloaded from orbit.dtu.dk on: Nov 04, 2018 Prediction of airfoil performance at high Reynolds numbers. Sørensen, Niels N.; Zahle, Frederik; Michelsen, Jess Publication date: 2014 Document Version Publisher's

More information

Uncertainty Quantification for multiscale kinetic equations with random inputs. Shi Jin. University of Wisconsin-Madison, USA

Uncertainty Quantification for multiscale kinetic equations with random inputs. Shi Jin. University of Wisconsin-Madison, USA Uncertainty Quantification for multiscale kinetic equations with random inputs Shi Jin University of Wisconsin-Madison, USA Where do kinetic equations sit in physics Kinetic equations with applications

More information