Adaptive sampling for nonlinear dimensionality reduction based on manifold learning

We make use of the non-intrusive dimensionality reduction method Isomap in order to emulate nonlinear parametric flow problems that are governed by the Reynolds-averaged Navier-Stokes equations. Isomap is a manifold learning approach that provides a low-dimensional embedding space that is approximately isometric to the manifold that is assumed to be formed by the high-fidelity Navier-Stokes flow solutions under smooth variations of the inflow conditions. The focus of the work at hand is the adaptive construction and refinement of the Isomap emulator: We exploit the non-Euclidean Isomap metric to detect and fill up gaps in the sampling in the embedding space. The performance of the proposed manifold filling method will be illustrated by numerical experiments, where we consider nonlinear parameter-dependent steady-state Navier-Stokes flows in the transonic regime.

[1]  Andy J. Keane,et al.  Engineering Design via Surrogate Modelling - A Practical Guide , 2008 .

[2]  J. Halton.  On the efficiency of certain quasi-random sequences of points in evaluating multi-dimensional integrals , 1960 .

[3]  Alexander P. Kuleshov,et al.  Tangent Bundle Manifold Learning via Grassmann&Stiefel Eigenmaps , 2012, ArXiv.

[4]  Thomas Franz,et al.  Reduced-order modeling for steady transonic flows via manifold learning , 2015 .

[5]  Norbert Kroll,et al.  The DLR Flow Solver TAU - Status and Recent Algorithmic Developments , 2014 .

[6]  J. Tenenbaum,et al.  A global geometric framework for nonlinear dimensionality reduction. , 2000, Science.

[7]  A. Patera,et al.  Reduced basis approximation and a posteriori error estimation for affinely parametrized elliptic coercive partial differential equations , 2007 .

[8]  R. Pinnau.  Model Reduction via Proper Orthogonal Decomposition , 2008 .

[9]  V. T. Rajan.  Optimality of the Delaunay triangulation in ℝd , 1994, Discret. Comput. Geom..

[10]  Forrester T. Johnson,et al.  Modi cations and Clari cations for the Implementation of the Spalart-Allmaras Turbulence Model , 2011 .

[11]  A. Patera,et al.  Reduced basis approximation and a posteriori error estimation for affinely parametrized elliptic coercive partial differential equations , 2007 .

[12]  V. T. Rajan,et al.  Optimality of the Delaunay triangulation in Rd , 1991, SCG '91.

[13]  Eric O. Postma,et al.  Dimensionality Reduction: A Comparative Review , 2008 .

[14]  Yoshua Bengio,et al.  Exploring Strategies for Training Deep Neural Networks , 2009, J. Mach. Learn. Res..

[15]  Lawrence Cayton,et al.  Algorithms for manifold learning , 2005 .

[16]  K. Willcox,et al.  Aerodynamic Data Reconstruction and Inverse Design Using Proper Orthogonal Decomposition , 2004 .

[17]  M. Damodaran,et al.  Proper Orthogonal Decomposition Extensions For Parametric Applications in Transonic Aerodynamics , 2003 .

[18]  Jiri Blazek,et al.  Computational Fluid Dynamics: Principles and Applications , 2001 .

[19]  M. J. D. Powell,et al.  Radial basis function methods for interpolation to functions of many variables , 2001, HERCMA.

[20]  R. Zimmermann,et al.  Interpolation-based reduced-order modelling for steady transonic flows via manifold learning , 2014 .

[21]  M. Spivak.  A comprehensive introduction to differential geometry , 1979 .

[22]  Stefan Görtz,et al.  Improved extrapolation of steady turbulent aerodynamics using a non-linear POD-based reduced order model , 2012, The Aeronautical Journal (1968).