[sa64r]

Hi Everyone,

I'm working on a masters project which is to understand the aerodynamic effects on a projectile that is launched from a railgun on a space-bound trajectory.

I'm having difficulties simulating flows around projectiles that have velocities of around 7+ km/s.

For context using an inviscid model, the density based solver and air to be an ideal gas my residuals never converge, in fact they keep diverging until the simulation cuts off due too high of a divergence.

Any help would be greatly appreciated

[aerosayan]

Can't say for sure what's causing the divergence without seeing the simulation setup, but most likely your mesh is either too coarse, or your timestep is too large while you may be trying to solve for transient cases, or your physics setup is wrong (least likely).

Try to gain experience by solving for transonic, supersonic and low hypersonic flows first.
Try to gain experience by solving only for steady state cases first, then for transient cases.
Try to use explicit solvers with a low timestep if you're solving for transient cases, as a high CFL and high timestep value used for implicit solver maybe failing to capture the flow physics. Although do be careful because explicit solvers can be slow if your timestep is too low.

You can use the case study of supersonic/hypersonic flow over a compression corner or wedge. You can validate your results from a gas table book for oblique shocks over compression corners.

[sa64r]

Thanks for your reply Sayan.

I have solely tried steady state as I am quite new to CFD and don't have an extensive background in fluids simulation.

I think you're right about simulating lower mach numbers first, hopefully this can determine which mach number my simulation fails at.

[aerosayan]

Cool! Basically the idea behind simulating low mach no. flows is that, they're less computationally intense, and there's a lot of literature you can review to validate the results. But the most important benefit in your case, will be to identify the problems that may occur while simulating supersonic/hypersonic flows, and how you can solve them by changing the simulation setup.


Also, find a paper on your research topic and see how they have setup, executed, and validated their simulations.

[sa64r]

Ok yes this makes sense.

Thanks for the advice, I have been stuck on this for a while now :| but you've given me some avenues to explore so thanks for that.

[sa64r]

I've performed a few more simulations this morning and I think one of the issues I keep encountering is when I use the density based model the residuals diverge compared to when I use the pressure based solved.

Any tips on how to mitigate that?

[aero_head]

Hello Sagar,

Which numerical model are you using for your simulations (i.e. SST)?

Also, how many iterations have you gone through, and what is your time step? What is your Courant number?

EDIT: Courant Number/CFL

[sa64r]

Hey Kira,

I'm using Spalart - Allamaras (1eqn).

Well at the moment the number of iterations is irrelevant as my simulation errors out with floating point error when I use the density solver.

Excuse me if I am not correct but since I am working on a steady state sim, I don't have courant number or time-step.
Please correct me if I am wrong haha.

UPDATE:
So I discovered there was a Pseudo Transient setting selected in the solution methods tab, I've unckeched that and it seems to allow the simulation to run without crashing after 9 iterations. However still my residuals diverge greatly

[agd]

The physics of a 7km/s flowfield are extremely complex. What solver are you using? What chemistry model are you using? Just starting a flow simulation at 7km/s can be almost impossible with many solvers. Consider the basic physics of the flow at the base of a projectile during startup. If you initialize your flow field to a 7km/s freestream and then turn on the simulation, you get a very large instantaneous pressure drop in the base region as the fluid moves away from the base. This large pressure drop (which is an artifact of the starting process) can break a solver, particularly if you assume ideal gas. A real flow undergoing that expansion would experience physical/chemistry effects that would limit the pressure and density drop to some extent. You might try ramping the Mach number up over a number of iterations. But if you don't have a good understanding of the physics of hypersonic flow and how the flow solver works in that regime you may never find a good solution.

[sadra2003]

Hello everybody, I would like to simulate a three phase reactor and study its hydrodynamics. I would like to know which type of software is better? I have a good command of ANSYS Fluent and I am really convenient with it but i have no experience of working with OPEN FOAM or CFX ...
Do you recommend to use another one? please let me know your opinion.
Thanks

[sa64r]

Originally Posted by sadra2003

Hello everybody, I would like to simulate a three phase reactor and study its hydrodynamics. I would like to know which type of software is better? I have a good command of ANSYS Fluent and I am really convenient with it but i have no experience of working with OPEN FOAM or CFX ...
Do you recommend to use another one? please let me know your opinion.
Thanks

Hi, I think you may want to post this as an isolate thread instead of inside this one if you want help

[sa64r]

Originally Posted by agd

The physics of a 7km/s flowfield are extremely complex. What solver are you using? What chemistry model are you using? Just starting a flow simulation at 7km/s can be almost impossible with many solvers. Consider the basic physics of the flow at the base of a projectile during startup. If you initialize your flow field to a 7km/s freestream and then turn on the simulation, you get a very large instantaneous pressure drop in the base region as the fluid moves away from the base. This large pressure drop (which is an artifact of the starting process) can break a solver, particularly if you assume ideal gas. A real flow undergoing that expansion would experience physical/chemistry effects that would limit the pressure and density drop to some extent. You might try ramping the Mach number up over a number of iterations. But if you don't have a good understanding of the physics of hypersonic flow and how the flow solver works in that regime you may never find a good solution.

You have a lot of good points, and yes I don't have a strong understanding of Hypersonics or the physics at such high speed flows.

I perhaps should explore these first before even attempting working with CFD.

[aerosayan]

Originally Posted by sa64r

Hey Kira,

I'm using Spalart - Allamaras (1eqn).

Well at the moment the number of iterations is irrelevant as my simulation errors out with floating point error when I use the density solver.

Excuse me if I am not correct but since I am working on a steady state sim, I don't have courant number or time-step.
Please correct me if I am wrong haha.

UPDATE:
So I discovered there was a Pseudo Transient setting selected in the solution methods tab, I've unckeched that and it seems to allow the simulation to run without crashing after 9 iterations. However still my residuals diverge greatly

Your solver most likely uses local tiimestepping to accelerate the convergence of the steady state simulation. That itself requires the CFL value to be given, Please check if your CFL value is over 1. That can cause an error.


And as Kira mentioned, the flow physics at such a high speed is extremely complicated, and you'll have such high temperatures, that the gas itself will dissociate into ions, and form a plasma around the projectile.


Run the simulation on Mach 1, 2, 3, 4, 5, ..., N for a few thousand iterations and see at which Mach No., the solutions starts to diverge.


Based on that, you can survey the literature to figure out what kind of initial conditions you need to setup, and what kind of physics setup you need, As Kira mentioned, the solver might break due to the sudden change in pressure and energy.

[aerosayan]

Originally Posted by sa64r

You have a lot of good points, and yes I don't have a strong understanding of Hypersonics or the physics at such high speed flows.

I perhaps should explore these first before even attempting working with CFD.

Keep trying to run the CFD solution along with studying more about your research topic.
You'll learn more by trying and failing, and get more experience with the software so that will be helpful.


I'm also interested in flows with shockwaves, so I'm making my own solver, but the progress is slow. Really slow. So, don't get discouraged along the way, and take things one step at a time.


Instead of a 3D simulation, try with an axisymmetric 2D simulation, and see if that works. Expect your simulation to run for a long time, so try to be patient. Try to see if your CFD solver works on structured grids, and runs faster in structured grids. If so, you can surely use that to accelerate your CFD solution.



Since you're doing Master's you probably only have 2 years of time. In my experience, CFD projects can take some time, so it might be helpful to you if you start as soon as possible, and try to finish your project six months before your deadline.


If you drag the deadline until the last minute, you might not finish in time.

[sa64r]

Hey Sayan,

Thanks so much for the advice you have given.

I am indeed simulating the flow as a 2D Axisymmetric flow, but my masters is 1 year long and I've only around 4 months to complete my project :|

I have read around what challenges are common when simulating high speed flow and like many people here have said, the intermolecular physics is non-negligible for this type of flow and so simulating it is extremely complex.

I think I underestimated this project given my lack of CFD experience, I thought it was going to be relatively straight forward haha, how naive.

I really chose to simulate railgun projectiles out of interest of the technology rather than my ability to carry out this project lol.

[sa64r]

Hey Sayan & Kira, I just found my CFL number and it is 5. So yes, definitely over 1.

[JBeilke]

Here are two OpenFOAM solvers with tutorials, which are suitable for you:

https://hisa.gitlab.io/archive.html
https://github.com/unicfdlab/hybridCentralSolvers

Mesh generation for this type of flow is also very important. Check the GridPro blog.

https://blog.gridpro.com
https://blog.gridpro.com/know-your-m...ntry-capsules/

[Mik7]

Originally Posted by aerosayan

Your solver most likely uses local tiimestepping to accelerate the convergence of the steady state simulation. That itself requires the CFL value to be given, Please check if your CFL value is over 1. That can cause an error.


And as Kira mentioned, the flow physics at such a high speed is extremely complicated, and you'll have such high temperatures, that the gas itself will dissociate into ions, and form a plasma around the projectile.


Run the simulation on Mach 1, 2, 3, 4, 5, ..., N for a few thousand iterations and see at which Mach No., the solutions starts to diverge.


Based on that, you can survey the literature to figure out what kind of initial conditions you need to setup, and what kind of physics setup you need, As Kira mentioned, the solver might break due to the sudden change in pressure and energy.

The suggestion here makes no sense. Flow physics is extremely important to tweaking the solution, and M=1, 5, and 20 all have completely different flow physics.