Solution adaptive refinement of cut‐cell Cartesian meshes can improve FDA nozzle computational fluid dynamics efficiency

Abstract

AbstractTo be an effective tool in studying cardiovascular disease and designing new treatments, computational fluid dynamics (CFD) needs to move from the realm of “high performance” to “high throughput” computing by improving efficiency while retaining accuracy. Solution adaptive mesh refinement (AMR) has the potential to decrease simulation time and benefits would be greater if the mesh could dynamically adapt throughout a heartbeat. This study used cut‐cell Cartesian grids with AMR at each time step. The refinement criteria was based on subgrid scale (SGS). The potential efficiency improvements from AMR were investigated with the FDA nozzle benchmark case at Re 500, 3500 and 6500. Using AMR with SGS = 1% mean throat velocity simulations could be performed in under 24 h on 16 CPUs with high accuracy compared to both FDA particle image velocimetry experiments and refined uniform grid CFD. This was an efficiency improvement of over an order of magnitude compared to uniform grids and other AMR SGS values. Using AMR with SGS = 5 or 0.2% mean throat velocity did not result in large efficiency improvements. Lastly, simulations of pulsatile flow suggest that performance improvements for typical cardiovascular flows may be even greater than were found simulating the statistically stationary FDA nozzle experiments.