Commutation Error in Reduced Order Modeling of Fluid Flows

TL;DR

This paper examines whether differentiation and spatial filtering commute in reduced order models of fluid flows, revealing that the commutation error exists and impacts ROM accuracy mainly at low Reynolds numbers.

Contribution

It provides a theoretical and computational analysis of the commutation error in ROMs, specifically for the Laplacian and two ROM filters, and assesses its significance across different flow regimes.

Findings

The commutation error exists in ROMs of fluid flows.

The commutation error significantly affects ROMs at low Reynolds numbers.

At higher Reynolds numbers, the impact of the commutation error diminishes.

Abstract

For reduced order models (ROMs) of fluid flows, we investigate theoretically and computationally whether differentiation and ROM spatial filtering commute, i.e., whether the commutation error (CE) is nonzero. We study the CE for the Laplacian and two ROM filters: the ROM projection and the ROM differential filter. Furthermore, when the CE is nonzero, we investigate whether it has any significant effect on ROMs that are constructed by using spatial filtering. As numerical tests, we use the Burgers equation with viscosities $\nu=10^{-1}$ and $\nu=10^{-3}$ and a 2D flow past a circular cylinder at Reynolds numbers $Re=1$ and $Re=100$. Our investigation shows that: (i) the CE exists, and (ii) the CE has a significant effect on ROM development for low Reynolds numbers, but not so much for higher Reynolds numbers.