In this case study, we examine the effect of airfoil shape/camber on the formation and existence of stall cells. A series of experiments using a NACA0012 wing with a trailing-edge flap has been carried out over a range of angles of attack ($11.5^\circ$–$21^\circ$), flap angles ($0^\circ$, $5^\circ$, $10^\circ$) and chord-length-based Reynolds numbers (100 000–500 000). The influence of these parameters on stall cell formation has been explored. Tufts have been used to identify the flow behaviour near the surface, while forces have been measured to relate the surface flow behaviour and wing performance. The results from the tuft analysis on the wing indicated that two different Reynolds-number regimes exist with respect to stall cell formation criteria. A preliminary estimate of the airfoil shape influence on stall cell formation is presented. A data-driven approach is used to relate the aerodynamic parameters (lift coefficient and lift-curve slope) to the formation criteria of stall cells for the wing (and flap angles). The lift coefficient can be used to implicitly take into account the change in airfoil shape in addition to the angle of attack and the Reynolds number. It is hoped that the results presented in this case study could be extended to various other airfoil shapes and that the stall cell formation angle of attack can be deduced from just the mean lift behaviour.

This application paper describes a novel, cluster-based, semi-automatic, stream surface placement strategy for structured and unstructured computational fluid dynamics (CFD) data, tailored towards a specific application: The BLOODHOUND jet and rocket propelled land speed record vehicle. An existing automatic stream surface placement algorithm(8), is extensively modified to cater for large unstructured CFD simulation data. The existing algorithm uses hierarchical clustering of velocity and distance vectors to find potential stream surface seeding locations. This work replaces the hierarchical clustering algorithm, designed to work with small regular grids, with a K-means clustering approach suitable for large unstructured grids. Modifications are made to the seeding curve construction algorithm, improving the smoothness and distribution of the discretised curve in complex cases. A new distance function is described which allows the user to target particular characteristics of simulation data. The proposed algorithm reduces the required memory footprint and computational requirement compared to previous work(8). The performance and effectiveness of the proposed algorithm is demonstrated, and CFD domain expert evaluation is provided describing the value of this approach.