Adaptive ANOVA decomposition of stochastic incompressible and compressible flows

Realistic representation of stochastic inputs associated with various sources of uncertainty in the simulation of fluid flows leads to high dimensional representations that are computationally prohibitive. We investigate the use of adaptive ANOVA decomposition as an effective dimension-reduction technique in modeling steady incompressible and compressible flows with nominal dimension of random space up to 100. We present three different adaptivity criteria and compare the adaptive ANOVA method against sparse grid, Monte Carlo and quasi-Monte Carlo methods to evaluate its relative efficiency and accuracy. For the incompressible flow problem, the effect of random temperature boundary conditions (modeled as high-dimensional stochastic processes) on the Nusselt number is investigated for different values of correlation length. For the compressible flow, the effects of random geometric perturbations (simulating random roughness) on the scattering of a strong shock wave is investigated both analytically and numerically. A probabilistic collocation method is combined with adaptive ANOVA to obtain both incompressible and compressible flow solutions. We demonstrate that for both cases even draconian truncations of the ANOVA expansion lead to accurate solutions with a speed-up factor of three orders of magnitude compared to Monte Carlo and at least one order of magnitude compared to sparse grids for comparable accuracy.

[1]  W. J. Whiten,et al.  Computational investigations of low-discrepancy sequences , 1997, TOMS.

[2]  W. Hoeffding.  A Class of Statistics with Asymptotically Normal Distribution , 1948 .

[3]  Daniel M. Tartakovsky,et al.  Numerical Methods for Differential Equations in Random Domains , 2006, SIAM J. Sci. Comput..

[4]  Daniele Venturi,et al.  Stochastic low-dimensional modelling of a random laminar wake past a circular cylinder , 2008, Journal of Fluid Mechanics.

[5]  Henryk Wozniakowski,et al.  When Are Quasi-Monte Carlo Algorithms Efficient for High Dimensional Integrals? , 1998, J. Complex..

[6]  Xiang Ma,et al.  An adaptive hierarchical sparse grid collocation algorithm for the solution of stochastic differential equations , 2009, J. Comput. Phys..

[7]  B. Chu.  On Weak Interaction of Strong Shock and Mach Waves Generated Downstream of the Shock , 1952 .

[8]  Alexander Yu. Gelfgat,et al.  Different Modes of Rayleigh–Bénard Instability in Two- and Three-Dimensional Rectangular Enclosures , 1999 .

[9]  Daniele Venturi,et al.  Stochastic bifurcation analysis of Rayleigh–Bénard convection , 2010, Journal of Fluid Mechanics.

[10]  Xiang Ma,et al.  An adaptive high-dimensional stochastic model representation technique for the solution of stochastic partial differential equations , 2010, J. Comput. Phys..

[11]  Baskar Ganapathysubramanian,et al.  Sparse grid collocation schemes for stochastic natural convection problems , 2007, J. Comput. Phys..

[12]  George Em Karniadakis,et al.  Random roughness enhances lift in supersonic flow. , 2007, Physical review letters.

[13]  M. Kendall.  Statistical Methods for Research Workers , 1937, Nature.

[14]  D. Hunter.  Valuation of mortgage-backed securities using Brownian bridges to reduce effective dimension , 2000 .

[15]  Alexander J. Smola,et al.  Kernels and Regularization on Graphs , 2003, COLT.

[16]  P. Hall,et al.  The smooth transition to a convective régime in a two-dimensional box , 1978, Proceedings of the Royal Society of London. A. Mathematical and Physical Sciences.

[17]  M. Lighthill.  XVIII. The flow behind a stationary shock , 1949 .

[18]  Habib N. Najm,et al.  Natural Convection in a Closed Cavity under Stochastic Non-Boussinesq Conditions , 2005, SIAM J. Sci. Comput..

[19]  Daniel M. Tartakovsky,et al.  Multivariate sensitivity analysis of saturated flow through simulated highly heterogeneous groundwater aquifers , 2005, J. Comput. Phys..

[20]  C. Schwab,et al.  Sparse high order FEM for elliptic sPDEs , 2009 .

[21]  George Em Karniadakis,et al.  Stochastic modeling of random roughness in shock scattering problems: Theory and simulations , 2008 .

[22]  Li Jun Jiang,et al.  An Adaptive Hierarchical Sparse Grid Collocation Method , 2014 .

[23]  I. Sobol.  Global sensitivity indices for nonlinear mathematical models and their Monte Carlo estimates , 2001 .

[24]  Minseok Choi,et al.  Error Estimates for the ANOVA Method with Polynomial Chaos Interpolation: Tensor Product Functions , 2012, SIAM J. Sci. Comput..

[25]  M. Griebel.  Sparse Grids and Related Approximation Schemes for Higher Dimensional Problems , 2006 .

[26]  Ian H. Sloan,et al.  Why Are High-Dimensional Finance Problems Often of Low Effective Dimension? , 2005, SIAM J. Sci. Comput..

[27]  George Em Karniadakis,et al.  Anchor Points Matter in ANOVA Decomposition , 2011 .

[28]  Lee A. Segel,et al.  Distant side-walls cause slow amplitude modulation of cellular convection , 1969, Journal of Fluid Mechanics.

[29]  George E. Karniadakis,et al.  Multi-element probabilistic collocation method in high dimensions , 2010, J. Comput. Phys..

[30]  Chi-Wang Shu,et al.  Efficient Implementation of Weighted ENO Schemes , 1995 .

[31]  Jasmine Yen-teng Foo,et al.  Multi-element probabilistic collocation in high dimensions: Applications to systems biology and physical systems , 2008 .

[32]  Renato Machado Cotta,et al.  Integral transform solutions of transient natural convection in enclosures with variable fluid properties , 2000 .

[33]  Bernhard Schölkopf,et al.  Cluster Kernels for Semi-Supervised Learning , 2002, NIPS.

[34]  Renato Machado Cotta,et al.  Integral Transforms in Computational Heat and Fluid Flow , 1993 .

[35]  S. Rahman,et al.  A generalized dimension‐reduction method for multidimensional integration in stochastic mechanics , 2004 .