Simulating the behavior of flapping-wing micro-aerial vehicles (MAVs) is computationally expensive. Identifying bottlenecks in the computational process and tracking the accuracy and efficiency of various computational fluid dynamics (CFD) packages is a necessary factor in making these codes readily available for use by designers and engineers. Plans for this project, which began in the last half of 2009, involve formalizing two-dimensional plunging airfoil studies, establishing correlations with experiments and other simulation efforts; developing guidelines for grid resolution, spatial accuracy, and time accuracy; and studying the performance of the code for the moving grid case with various numbers of message passing interface (MPI) groups and open multiprocessing (OpenMP) threads, paying particular attention to the performance of the portions of the code that deal with grid motion.

Investigations of NASA’s OVERFLOW code has shown that the geometric operations required by moving grid calculations do not pose a computational bottleneck under the conditions studied. An analysis of Stanford’s AERO-F code showed four functions that consume most of the computational time and are possible targets for optimization work.

Preliminary two-dimensional airfoil analyses have been performed, using OVERFLOW2.1 on the NASA Pleiades system (40,000 cores) to find solutions for a rigid NACA 0012 airfoil undergoing generalized pitching and plunging motion. A new grid movement technique has been implemented and validated to model correctly a two-dimensional combined pitching and plunging motion. Grid refinement studies are being done to establish the optimal domain and grid sizing, and these studies are guiding selection of the optimal temporal solution parameters, a key requirement to limit the computational cost, especially for optimization, which will require large numbers of flow solutions.