Harbir Antil - Research

Reduced Order Modeling of Nonlinear and Parametric Systems


Projection based methods lead to reduced order models (ROMs) with dramatically reduced numbers of equations and unknowns. However, for nonlinear or parametrically varying problems the cost of evaluating these ROMs still depends on the size of the full order model and therefore is still expensive. The Discrete Empirical Interpolation Method (DEIM) further approximates the nonlinearity in the projection based ROM. The resulting DEIM ROM nonlinearity depends only on a few components of the original nonlinearity. If each component of the original nonlinearity depends only on a few components of the argument, the resulting DEIM ROM can be evaluated efficiently at a cost that is independent of the size of the original problem. For systems obtained from finite difference approximations, the $i$th component of the original nonlinearity often depends only on the $i$th component of the argument. This is different for systems obtained using finite element methods, where the dependence is determined by the mesh and by the polynomial degree of the finite element subspaces. This paper describes two approaches of applying DEIM in the finite element context, one applied to the assembled and the other to the unassembled form of the nonlinearity. We carefully examine how the DEIM is applied in each case, and the substantial efficiency gains obtained by the DEIM. In addition, we demonstrate how to apply DEIM to obtain ROMs for a class of parameterized system that arises, e.g., in shape optimization. The evaluations of the DEIM ROMs are substantially faster than those of the standard projection based ROMs. Additional gains are obtained with the DEIM ROMs when one has to compute derivatives of the model with respect to the parameter.



Related Publications

  • H. Antil, M. Heinkenschloss, and D. C. Sorensen.
    Application of Discrete Empirical Interpolation Method to Reduced Order Modeling of Nonlinear and Parametric Systems.
    A. Quarteroni and G. Rozzas (eds.), Reduced Order Methods for Modeling and Computational Reduction, Model. Simul.& Appl. Vol. 9, 2014, pp. 101-136, Springer Italia, Milan.