[Filly]

Hi everyone,


I'm in the middle of a uni project where I have to simulate the cooling channel of the Space Shuttle Main Engine and I'm having difficulties getting similar results to those from various literatures.


Since I am quite the beginner when it comes to CFD, I thought I'd ask for input here from more experienced people.



I have the geometry, and the values for the inlet temperature, pressure of the cooling fluid (Hydrogen) as well as the mass-flow thru the channel.
I also have the heat-flux in form of a profile that I use as a boundary condidtion on the hot-gas wall.


How would you guys start with the simulation?

• Pressure-based or density based?
  • The fluid is supercritical throughout the whole channel with varying density. So I tried using the density based solver since it's a compressible fluid. But the simulation isn't really converging (residuals around 1e-2).
  • When I try to use the pressure based solver, the convergence is better, but the results still deviate too much from literature data.

• Inlet/outlet types?
  • With the pressure-based solver: I tried pressure inlet (gauge pressure = literature value) with a mass-flow outlet. The results are 'similar' to those of the literature, or at least the trend is similar. But the T_max is still about 100K too high and the pressure drop at the outlet is 4MPa too high.
  • When I try the same setup with the Density based solver, it's not converging.
  • When I use the velocity-inlet and mass-flow outlet with the density based solver, the pressure drop along the channel is basically non-existent. Can I not use this inlet/outlet combination together?
  • My uni supervisor told me not to use the pressure-outlet condition since the pressure drop thru the channel is part of the desired results and defining the outlet pressure would defeat the purpose of the simulation.

• Hydrogen properties?
  • I can't use constant values, so I tried using different real-gas options.
  • I tried using a UDRGM, but I keep getting a warning that the c_p values are negative. I think that's because the c_p polynomial coefficients are only valid for temperatures above 300K. (the inlet temperature is 53K). However, I can't find any polynomial values for the temperatures below 300K.
  • Instead I used the Real-Gas-Aungier-Redlich-Kwong method for the density, and kinetic-theory for c_p and thermal conductivity. Power-law for the viscosity. But I'm actually not sure about some of the values used with these methods.
  • I keep getting the ("temperature limited to 1.000 in 757 cells")-message when I use the density solver with the above mentioned settings.

I've pretty much been thrown into the cold water with this project, so I'm open for any ideas/tips/input, even if they seem trivial to you.


Also, I welcome any books or other resources that you can recommend to get a better understanding.


Cheers.

[CFDKareem]

To start, have you run a mesh sensitivity study? Convergence issues are often due to a poor mesh. For heat transfer simulations where you are trying to get very accurate data, it is vital to capture the thermal boundary layer at the interface between the solid and fluid. This would involve adding inflation at the walls to make sure the thermal boundary layer is correctly resolved. The error "temperature being limited..." is often indicative of a poor quality mesh at some location.

Start with a mesh sensitivity study before playing with solver settings. If you're still having trouble then we can look at the solver settings.

[Filly]

Thank you for the quick reply.


I haven't really done a complete mesh sensitivity study yet. I'm using the student version on my computer which means I am limited in the maximum mesh elements.

I do have a bit low quality of orthogonality in some areas of the boundary layer.
I used inflation layers for my mesh to get a y+ near 1.

If my calculations are correct I need a first layer height of 6e-5mm.
But with these settings I need to keep the mesh of the rest of the geometry comparatively coarse (0.3mm) to stay under the max. element limit. Am I correct in assuming the big difference is the source for the poor orthogonality?


My plan was to use the university's server and full ansys license to simulate the finer meshes once I get the solver settings right. But it sounds like I have to do all the calculations with the full license.


This is probably a stupid question, but if I need to make the mesh sensitivity study before I can dial-in the solver settings, what solver settings do I use for the sensitivity study? The same settings that gave me vaguely similar results?