JournalVolume 1, №2Approximation of large scale unstable systems by the implicit restart techniques (pages 175-189) Nisar Ahmed Abstract. This paper is concerned with the model reduction problem of large-scale unstable systems. Such systems are developed when employing approximation methods for non-linear, partial differential equations that represent engineering systems. The application of finite element and finite difference techniques produces ordinary differential equations representative of the system but with a number of states that is prohibitive from control and simulation point of view. Krylov subspace techniques are used with the Arnoldi methods to alleviate such over-modeling by performing model reduction on the large-scale problems. The deficiencies of the Krylov techniques are that they come with no guarantee on the final model stability and usually contain redundant information that is unacceptable from control design point of view. One way of overcoming this problem is to implement a restart scheme in an iterative procedure. In this work, algorithms are developed based on oblique projections followed by implicit restart in order to obtain an approximation of original unstable systems. We have proposed two procedures, based on the implicit restart scheme, in order to obtain reduced order model for unstable systems. The first uses stable-antistable decompositions while the latter one uses co-prime factorisation technique. Finally, the developed algorithms are implemented on a largescale unstable system and it has been shown that how a computationally efficient and inexpensive reduced order model is produced as compared to existing techniques.
|
|