Reduced Krylov basis methods for parametric partial differential equations (Q6974682)

In this long work, the authors introduce a novel family of reduced basis methods (RBMs), referred to as reduced Krylov basis methods (RKBMs), which are constructed upon preconditioned Krylov subspace methods. They develop several practical and computationally efficient RKBMs variants tailored for the solution of parametric partial differential equations (PDEs). The reduced basis is generated by using a limited number of preconditioned Krylov subspace iterations, thereby significantly reducing computational cost while maintaining accuracy. A convergence analysis is provided for the case of linear PDEs with two parametric coefficients, assuming exact inverse preconditioning at a fixed parameter value. The proposed methods are supported through comprehensive numerical examples.