Model order reduction by using improved approximation techniques

Abstract EN

A simplified approach for model order reduction (MOR) is presented in this article using the balanced singular perturbation approximation (BSPA) approach applicable to large-scale linear dynamical (LSLD) systems.

The reduced system was so designed to preserve complete parameters of the original system with reasonable accuracy, employing MOR.

The approach is based on the retention of the dominant states of the system and comparatively less important ones.

The reduced system comes from the preservation of the dominant states (say 'desirable states') of the original system and, thus, from stability to preservation.

The key disadvantage of the Balanced Truncation approach is that the ROM steady-state values do not correspond with the higher-order systems.

This drawback has been eliminated in the proposed approach, which leads to hybridisation of balanced truncation and singular perturbation approximation into a novel reduction method without the loss of retaining its dynamic behaviour.

The proposed approach has been tested on LSLD systems and the results obtained show the efficacy of the approach.

The methodology presented has been tested on two typical numerical examples taken from the literature review to examine the performance, precision and comparison with other available standard order reduction methods.