Development of New Frequency Weighted Model Reduction Technique

Abstract

Derivation of mathematical model from any dynamic system has a vital role in design, analysis and simulation. Higher order complex models are obtained from physical systems. Partial differential equations (PDEs) and Ordinary differential equations (ODEs) are used to represent these complex models. As these models should be simplify for the ease in solution, reduced order models (ROMs) are needed which approximates with the actual system as much close as possible. In the area of model order reduction (MOR), considerable amount of work has been carried out. Existing techniques show large approximation error, error bound and also have stability issues. In this thesis, a new frequency weighted MOR technique is proposed for continuous time systems. The proposed technique guarantees in reduced order system (ROS) and yield low approximation error in the presence of double sided weightings. The significance of proposed technique is intimated by numerical problems and comparison with the existing frequency MOR techniques. Practical applications of MOR are: • Real time applications in industry • In missile system analysis • Image compression