“When our simulations get to be a certain size, it is absolutely essential that we run on high-performance computing (HPC) clusters on the cloud to support a more digestible workflow. Ansys HPC applications on Microsoft Azure have been the best experience running an HPC service. We have access to a lot of Ansys tools, but we could not have, in any reasonable amount of time, gotten anywhere using them without this service.”
— Nathan Moshman, Founder, Moshman Research
Moshman Research partner Avigation is developing a heavy-lift semi-autonomous unmanned aerial vehicle (UAV) for rapid wildfire response, the Hellbender Wildland Fire Response System. The system features a mountaintop-launched aerial platform design with a sheet spray misting system to enable rapid response and optimal suppression. Using simulation to optimize thrust efficiency of the fan blades, Moshman (in support of Avigation) hoped to maximize system operation time for better fire containment that leads to less destruction of forests, property damage, and loss of life.
For Moshman, the challenge was to perform a proof-of-concept shape optimization of a ducted fan blade primarily using computational fluid dynamics (CFD) to demonstrate improved performance at a relevant operating point. It was a task involving a large model that included tens of millions of elements. Even with a simplification of the full geometry, modeling was still done in three dimensions, adding another layer of complexity. This resulted in hours-long runtimes on local machines punctuated with stops and starts during analysis, which was detrimental to maintaining workflow. Additionally, the team needed to consider adjoint solve time, deformation settings, and degrading mesh quality over design iterations during shape morphing to fully optimize the efficiency and performance related to fan blade tip shape and curvature.
Researchers benchmarked Ansys CFD on HBv3 virtual machines powered by AMD EPYC™ 7V73X (code name: Milan-X) processors, utilizing 96 cores on Microsoft Azure. Taking advantage of Ansys HPC applications on Microsoft Azure for production runs enabled the team to efficiently minimize necessary computational load on local, low-core count machines to support more complex CFD analysis.
With a quasi-fixed fan blade tip, the shape morph process yielded a 32.6% increase in thrust produced by the blade and led to increased efficiency overall. Freeing up the blade tip led to a shape producing a greater net thrust but no change to efficiency.