Improving the Performance and Flexibility of Grid-Based Vorticity-Velocity Solvers for General Rotorcraft Flow Analysis

Accurate flow prediction is essential to rotorcraft development, and in recent years hybrid CFD methods, where a body-fitted CFD solver is coupled to a vortex method, have been developed to address the cost and vorticity preservation issues associated with full-unsteady rotorcraft simulation. Recent work developing the serial version of the VorTran-M2 grid-based vorticity-velocity method for application to hybrid CFD analysis of rotorcraft addressed critical formulational issues such as maintaining a solenoidal vorticity field, general multi-resolution adaptive gridding, consistent and efficient implementation of the viscous terms and computation of the unsteady pressure contribution. This paper describes the development of a simplified CFD/VorTran-M2 coupling interface and an efficient parallel implementation of this method, with particular emphasis on dynamic domain decomposition. VorTran-M2, an adaptive grid solver, only has cells in regions of significant vorticity. This compact grid representation permits very low cell-counts, but adds significant complexity when developing scalable parallel implementations. Adaptive grid methods are becoming more common in the CFD community, though VorTran-M2 is a particularly extreme example since predictions start with zero cells, rather than from a coarse grid. This lack of initial grid to form the basis of initial domain decomposition combined with increasing problem size and parallelizing the Biot-Savart solution of the Poisson equation, makes efficient scaling challenging and something that is only now being demonstrated.