Matrix Krylov subspace methods for large scale model reduction problems (Q856114): Difference between revisions

The paper presents matrix Krylov subspace methods for solving large coupled Lyapunov equations with low rank right-hand sides of the form \[ AP +PA^T + BB^T = 0, \quad A^TQ +QA + C^TC = 0, \] where \(A\) is a real, large, and sparse matrix with \(A,P,Q \in \mathbb R^{n \times n}\), \(B \in \mathbb R^{n \times s}\), \(C^T \in \mathbb R^{n \times r}\) \(\text{rank}(B)=s\), \(\text{rank}(C)=r,\) and \(s\ll n\), \(r\ll n\). The coupled Lyapunov equations play an important role in model reduction problems. The posed iterative processes are based on the global Arnoldi and Lanczos methods. An upper bound for the residual norm that can be used to stop the iterations in the Lyapunov global Arnoldi and Lanczos processes without having to compute extra products involving the matrix \(A\) is also given. Moreover, the paper presents the approach for obtaining a reduced order model for a linear time invariant system by approximating the corresponding transfer functions. In order to approximate the transfer function, the global full orthogonal method is employed. Numerical experiments illustrate the effectiveness of the methods.