search

dynamicslab / hydrogym

An RL-Gym for Challenge Problems in Data-Driven Modeling and Control of Fluid Dynamics.
https://hydrogym.readthedocs.io
MIT License
45 stars 10 forks source link

Create cylinder meshes #1

Closed jcallaham closed 2 years ago

jcallaham commented 2 years ago

Create meshes comparable to both Noack et al (2003) and Sipp & Lebedev (2007). The latter has a much larger domain, but may be relatively slow and have higher resolution than is necessary for the purposes of a modeling and control testbed.

Once the meshes are generated, implement a steady-state solver and compare drag coefficients at Re=40 and Re=100 (also compare to Nek5000 and IBPM implementations). The final decision between the two can wait until stability analysis is implemented... possibly it would be better to leave both options in place.

jcallaham commented 2 years ago

Generated Sipp-Lebedev mesh and did preliminary simulation (see examples/cylinder)

jcallaham commented 2 years ago

Also see #3 for aerodynamic coefficients

jcallaham commented 2 years ago

https://fenicsproject.discourse.group/t/transitioning-from-mesh-xml-to-mesh-xdmf-from-dolfin-convert-to-meshio/412/65

jcallaham commented 2 years ago

Generated mesh similar to Noack et al, also added force calculation and steady-state solve (can do default direct solve for the for the Jacobian inverse since it's in 2D: even the larger mesh only requires ~10Gb RAM). The Sipp & Lebedev mesh is a good match for reported results, but has 175k elements compared to only 17k for Noack et al. Also the larger domain would probably be better for things like adjoint modes.

Probably the best thing to do will be to leave both meshes in place and have the ability to switch back and forth. The smaller domain would be better for quick runs (reinforcement learning, for instance), but the larger domain might be better for modal analysis and model-based control.

Benchmarks from Taira & Colonius (2007)

image image

FEniCS (Re=40)

Mesh CD
Sipp & Lebedev 1.53
Noack et al 1.81

Next steps

Closing this out, but it would be good to add CD/CL for Re=200 once the time-stepping is settled (#2).