Time-and-spatially adapting simulations for efficient dynamic stall predictions

Abstract

The ability to accurately and efficiently predict the occurrence and severity of dynamic stall remains a major roadblock in the design and analysis of conventional rotors as well as new concepts for future vertical lift. Several approaches to reduce the cost of these dynamic stall simulations for airfoils and finite wings are investigated. Temporal error controllers, variable time step sizes, and feature-based near-body mesh adaptation are evaluated for their ability to more cost-effectively predict dynamic stall on three different configurations. A fourth-order temporal controller has been observed to provide a balanced cost-accuracy ratio, as a maximum of three to four orders of magnitude convergence of the Newton subiterations is obtained during much of the dynamic stall cycle. Larger times steps can be applied, in particular during the attached upstroke portion of the dynamic stall cycle with fourth-order temporal convergence. Mesh reductions via a feature-based two-level adaptation provided a 50% reduction in computational costs with comparable accuracy to a fixed, refined mesh size. Additional refinements may be warranted just after the dynamic stall onset to capture the complex flow field.