Proper orthogonal decomposition method for multiscale elliptic PDEs with random coefficients (Q2297081)

The authors develop an efficient multiscale reduced basis method to solve the multiscale elliptic boundary value problem \[ -\nabla\cdot(a^\varepsilon(x,\omega)\nabla u^\varepsilon(x,\omega))=f(x),\ x\in D,\,\omega\in \varOmega, \quad u^\varepsilon(x,\omega)=0, \] where \(D\subset \mathbb{R}^d\) is bounded, \((\varOmega,\mathcal{F},\mathbb{P})\) is a probability space. The coefficient \(a^\varepsilon(x,\omega)=(a^\varepsilon(x,\xi_1(\omega),\dots,\xi_r(\omega))\) is parametrized by \(r\) independent random variables. Convergence analysis of the proposed method and numerical results are presented showing the accuracy and efficiency of the method for several multiscale problems with or without scale separation.