Exploring a Novel Adaptive Mesh Refinement Strategy for High-Speed CFD

Adaptive Mesh Refinement (AMR) techniques to efficiently and robustly achieve grid independent solutions on multi-element unstructured grids is a topic of practical interest within the CFD community. The current effort focuses on the efficiency of a novel Adaptive Mesh Refinement (AMR) strategy that is developed and evaluated for transonic high-speed flows using Ansys Fluent. The algorithm for marking cells for adaptation is designed to systematically reduce local truncation errors based on the curvature of the primitive vector field. The algorithm for marking cells for adaptation is described in sufficient detail to be portable to other flow solvers that offer AMR. The relative importance of each primitive vector variable within the scheme is evaluated using both equal-weighting and optimized-weighting approaches. Variations of the proposed algorithm that use flow gradients or limit adaptation regionally are also investigated. The negative consequences of adaptation without enforcing the original smooth surface shape are demonstrated. An equal-weighted, primitive vector curvature-based strategy is shown to typically produce near-grid-independent results with an order of magnitude less grid required than classic grid refinement.