Implementing euler equation custom boundaries

[Nadexterbrown]

Hello All,
I am new here and trying to get the hang of using Pyro to solve the Euler equations. I have been able to follow and run the Sod test. Currently, I am a bit confused about how to set up custom boundary conditions. I am attempting to solve the Euler equation for a channel with an imposed velocity at the left-hand side (similar to a piston accelerating). However, from my understanding, to set a Dirichlet boundary condition value, one must go through the multigrid solvers, requiring one to restate the problem again using the equation types provided. Am I understanding this correctly? Any assistance on this would be greatly appreciated. Thank you and have a nice day.
Best,
Nolan Dexter

[zingale]

I think that the closest we have to what you are interested in is with the double Mach reflection problem, ramp. The boundary conditions for that are here:

pyro2/pyro/compressible/BC.py

Line 183 in a3da1f1

elif bc_name == "ramp":

I would start with something like that, but if it doesn't work well, we might want to try to do a proper inflow using a one-sided stencil and specifying the interface state outside of the domain directly.

[Nadexterbrown]

Hello again, I have been attempting to implement the piston boundary condition using both of the approaches mentioned in the prior message. Currently, the boundary ghost cells appear to be behaving as expected, imposing the velocity onto the gas within the system. However, after a few time steps there appears a spike in the energy that then drives a spike in the velocity above the piston velocity, which is not physical. This behavior is observed occuring as the pushed gas begins to form a shock wave ahead of the piston within the domain. I have tried various approaches, setting the piston state a few cells into the domain at simulation start, setting a relaxed ramp up time, and decreasing the cfl number. Nothing so far has resolved this problem. Any assistance or recommendations would be extremely appreciated. I have uploaded simulation results showing this behavior (https://github.com/Nadexterbrown/1D-Eulerian-Piston-Euler-Solver/tree/main/Constant-Velocity-Piston-Figures/Comparison), with the aformentioned phenomena occuring just after the 1000th time step.

[zingale]

boundary conditions are always tricky. One issue might be that you are setting the velocity in the ghost cells, but what you actually want is for the ghost cell values to give you the correct velocity on the interface itself when you do the reconstruction of the data to the interfaces. This requires understanding the stencil used by the hydro itself.