An efficient hybrid method for uncertainty quantification (Q2132433)

The paper focuses on coupling intrusive and non-intrusive uncertainty quantification methods for hyperbolic systems of equations, which are applied in different parts of the computational domain separated by interfaces. The intrusive method is based on the stochastic Galerkin projection combined with a polynomial chaos basis, which is efficient for slowly changing stochastic problems. The non-intrusive method employs a quadrature rule and the probability density functions of the uncertainties, which is well-suited for rapidly varying problems. The proposed coupling, based on a suitable projection onto an intermediate glue grid and penalization, is shown to be stable and accurate. Numerical experiments are presented for one- and two-dimensional domains to illustrate the efficiency and applicability of the method.