Rocket Nozzle Startup CFD

[0FlyingSwede0]

what cfd is this?

[doverdx]

have you done any runs where you've seen asymmetric shedding across the nozzle? just wondering since that startup transient can be a sizable side load. thanks!

[asinkan]

Hi, have you written some article about this,or can you recomend me some? Thx

[phph4v31aye]

Hello,

I'm new at this and as far as I've seen the shocks should be outside the nozzle... We'll maybe I'm getting it all wrong. Could you please comment on this. Thank you in advance!

[byustanfordboy]

@phph4v31aye

Internal shock waves are common in nozzles as are external ones. This simulation focuses on the former.

[phph4v31aye]

@byustanfordboy Thank you!!

[Buddhabr0t]

is it 2D or axisymmetric?

[byustanfordboy]

@Buddhabr0t

This is 2D calculation, reflected about the center line.

[ayushsaurabh]

Hey which software you used for the post-processing visualization..

[MrPubicfaceFTW]

SO you're using Finite difference compact schemes with artifical diffusion?

What kind of shock sensor are you using?

Have you considered weno finite difference/central schemes.? THeyre usually faster and well suited for LES. with strong shocks.

[gasturbinefreak]

very impressive..... It must have taken you a long time to write that code........

[mihir1310]

which code is this btw '?

[byustanfordboy]

Its my own hand rolled research code....

-10th order compact spatial derivatives

-5 stage RK4 temporal integration

-high-order artificial viscosity for shocks

-general curvilinear structured grid

[nonnonee]

how big is the stencil of the 10th order spatial?

[bramiozo]

@byustanfordboy

Do you use FCT or do you calculate the flux directly over the solution vector ?

Why 10th order space and 4th in time and how do you maintain the order near the boundaries ? Did you think about lowering the spatial order and using Riemann solvers or was it your intention to avoid Riemann solvers as a research topic ?

TVD RK of course ?

[byustanfordboy]

@bramiozo

No FCT... just standard FD fluxes. 10th order has spectral like resolution meaning very low numerical dissipation in regions of vortex dominated flow (eg. shear layer, turbulent BL. etc) The formal accuracy clearly drops near the boundaries but local resolution is maintained. Shock capturing with these types of schemes is a research topic.

[bramiozo]

@byustanfordboy

How do you ensure conservation ? Do you monitor any mass/energy loss ?

[bramiozo]

@bramiozo

Perhaps you can run a ridiculously fine mesh with low order accuracy to compare

with regard to oscillatory behavior (+ computation times of course).

Do you use mimetic discretisation as part of the research ?

Results look very nice btw :) .

[byustanfordboy]

@bramiozo

Formal conservation is maintained by forming lower-order conservative stencils near the mesh boundaries which ensure fluxes in/out of comp. domain only occur at the comp. boundaries. eg, a no slip wall will have zero flux (mass,mom.,energy... assume adiabatic) and the original equations are formally conservative. Have tested this but for other cases. Also, haven't looked at mimetic discretisation.

[bramiozo]

@byustanfordboy

Thanks for the response.

How about global formal conservation? Is it nearing machine accuracy ?

[byustanfordboy]

@bramiozo

For a periodic problem or fully wall bounded problem yes... with unsteady outflows this quantity fluctuates as expected.

[mihir1310]

is this using openfoam ??? how did you treat the reflections at the outlet ?

[byustanfordboy]

Not open foam... outlet is supersonic fixed pressure.

[mihir1310]

ohh ,, supersonic, then its ok .. i expected a subsonic outlet & hence -ve eigenvalues which require non-reflective boundary conditions

[velaroful]

Please give a few conditions, e.g. p_0, T_0, R (test gas) , p_E! Thanks.

[byustanfordboy]

The flow conditions are dimensionless but the ratio p_0/p_E is set to give and ideally expanded nozzle. At/Ae = 2.0 and test gas is air.