Recursive one-way Navier Stokes equations with PSE-like cost

TL;DR

This paper introduces a new variant of the one-way Navier-Stokes equations that significantly reduces computational cost while accurately capturing all downstream-traveling flow modes, bridging the gap between PSE and global methods.

Contribution

A novel OWNS variant that approximates PSE-like cost while retaining all downstream modes, using an efficiently computable projection operator.

Findings

Method reduces CPU time and memory requirements.

Accurately captures all downstream modes in flow simulations.

Demonstrated effectiveness on shear and boundary layer flows.

Abstract

Spatial marching methods, in which the flow state is spatially evolved in the downstream direction, can be used to produce low-cost models of flows containing a slowly varying direction, such as mixing layers, jets, and boundary layers. The parabolized stability equations (PSE) are popular due to their extremely low cost but can only capture a single instability mode; all other modes are damped or distorted by regularization methods required to stabilize the spatial march, precluding PSE from properly capturing non-modal behavior, acoustics, and interactions between multiple instability mechanisms. The one-way Navier-Stokes (OWNS) equations properly retain all downstream-traveling modes at a cost that is a fraction of that of global methods but still one to two orders of magnitude higher than PSE. In this paper, we introduce a new variant of OWNS whose cost, both in terms of CPU time and memory requirements, approaches that of PSE while still properly capturing the contributions of all downstream-traveling modes. The method is formulated in terms of a projection operator that eliminates upstream-traveling modes. Unlike previous OWNS variants, the action of this operator on a vector can be efficiently approximated using a series of equations that can be solved recursively, i.e., successively one after the next, rather than as a coupled set. In addition to highlighting the improved cost scaling of our method, we derive explicit error expressions and elucidate the relationship with previous OWNS variants. The properties, efficiency, and accuracy of our method are demonstrated for both free-shear and wall-bounded flows.