A stopping criterion for Krylov-subspace based model order reduction techniques

Modeling of large scale systems is a numerically costly and time consuming procedure. Model Order Reduction techniques seek to replace such large scale systems by considerably smaller ones while preserving the input-output dynamics. Krylov-subspace based reduction techniques are numerically efficient techniques for reducing very large systems with orders ranging from tens of thousands to millions. However, one of the disadvantages with these techniques is that there is no stopping criterion to automatically select a suitable order for the reduced system. This paper presents an efficient stopping criterion which can be embedded in the iterative algorithms used conventionally for generating the ortho-normal basis for the Krylov subspace.