Application of the VORTFIND algorithm for the identification of vortical flow features around complex three‐dimensional geometries

Abstract

SUMMARYAccurate prediction of the hydrodynamic forces and moments acting on a manoeuvring marine vehicle using Reynolds Averaged Navier Stokes (RANS) simulations requires sufficient mesh resolution to capture offbody vortical structures. Since the path of these structures is not known a priori, a vortex identification and capture strategy is required alongside an iterative mesh adaption process. An improved version of the VORTFIND algorithm which can identify multiple vortices of variable strength and rotational direction using a K-means algorithm is described. The algorithm is applied to velocity fields generated from RANS simulations to increase the mesh resolution in the vortex core region, ensuring sufficient mesh density to capture the downstream propagation of the vortex for a submarine hull at drift and ship propeller-rudder interaction.

Uncertainty assessment of coupled Digital Image Correlation and Particle Image Velocimetry for fluid-structure interaction wind tunnel experiments

“…Therefore, if you rank all the vectors by their proximity to different sectors you can determine how close they are to a vortex centre. This approach allows the position of a single vortex structure to be located and is simple to implement, but can be extended to cope with multiple vortex structures if required (Phillips and Turnock, 2013).…”

Feasibility study into a computational approach for marine propeller noise and cavitation modelling

“…As shown in Figure 2, a relatively good agreement may be observed between the predicted and measured location of the cavitation regions. One of the drawbacks, however, is the lack of the tip vortex extending downstream of the propeller This is caused by lack of appropriate refinement of the mesh away from the propeller blade and by the fact that RANS methods in general tend to introduce too much dissipation and thus cause the vortices to disappear much sooner than they would in reality unless appropriate local mesh refinement is applied (Turnock et al, 2006, Phillips & Turnock., 2013.…”