$H_2$ model reduction for diffusively coupled second-order networks by convex-optimization

TL;DR

This paper introduces an $H_2$ optimal convex relaxation method for reducing diffusively coupled second-order network systems, preserving key structures and stability while enabling interpretability as a smaller network.

Contribution

It presents a novel $H_2$ optimal reduction technique that maintains second-order structure and stability, along with a graph reconstruction approach for interpretability.

Findings

Effective reduction of large-scale mass-spring-damper networks

Preservation of stability and diffusive coupling in reduced models

Validation through numerical experiments on networked systems

Abstract

This paper provides an $H_2$ optimal scheme for reducing diffusively coupled second-order systems evolving over undirected networks. The aim is to find a reduced-order model that not only approximates the input-output mapping of the original system but also preserves crucial structures, such as the second-order form, asymptotically stability, and diffusive couplings. To this end, an $H_2$ optimal approach based on a convex relaxation is implemented to reduce the dimension, yielding a lower order asymptotically stable approximation of the original second-order network system. Then, a novel graph reconstruction approach is employed to convert the obtained model to a reduced system that is interpretable as an undirected diffusively coupled network. Finally, the effectiveness of the proposed method is illustrated via a large-scale networked mass-spring-damper system.