[Cafard]

Hello everyone,

I have simulate a supercritical CO2 in centrifugal compressor,and the simulation work well,the boundary conditions were:

inlet T/P : 500K/8MPa
mass : 1.35 kg/s
speed : 80000RPM
mesh : 100e4

material data :
min/max T : 200/700K
min/max P : 7/11MPa

when I want to simulate 350K/8MPa(2.35kg/s) ,the solver would occur error : fatal overflow in linear solver,and the first steps occur notice:
****** Notice ****** |
| While evaluating |
| Density Derivative wrt Pressure at Constant Temperature |
| on domain "R1", |
| the variable |
| Absolute Pressure |
| went outside of its upper limit. Its maximum value was |
| 2.3393E+07. The bounds error was handled by clipping. |
| If this situation persists, consider increasing the table range. |

I have thought out a method to solve this :
"enlarge the memory size(now is 16GB) ,then enlarge the material data base range"
but does it slove the fatal overflow problems to?
I have reed the FAQ,and I thik the boundary conditios that I set were right.

[ghorrocks]

No, you need to enlarge the material properties table size, not the memory allocation.

With tricky material properties like this you have to expect convergence difficulties. Make sure you read the FAQ carefully as the tips it has will be important for this model.

[Cafard]

I tried to enlarge the material data base,
but still occur error : fatal overflow.
(by the way,timestep was 7.5e-5)

Maybe the conditions was too close critical point [304K/7.38MPa] ,
so as the critical condition occur in simulation,
the solver would be error??

[ghorrocks]

OK, so it looks like the table size is not the cause of the instability. In that case my previous comment is the thing to look at:

With tricky material properties like this you have to expect convergence difficulties. Make sure you read the FAQ carefully as the tips it has will be important for this model.

[Cafard]

I have read the FAQ about overflow again,and check my mesh again.
I used BladeGen&TurboGrid to created my compressor mesh,
and I found the mesh quality near blade (boundary layer) is bad,
the aspect ratio is large.
After I read the tutorial, I found the large aspect ratio near boundary layer is normal phenomenon.

And strangely,why 500K/8MPa worked well, 350K/8Mpa occured error??

[ghorrocks]

It sounds like as you move towards the critical point the numerics becomes more unstable. Or you may have regions which fall below the critical point and then you have phase change stuff happening. Either way, it means a simulation which is stable at 550K is not necessarily stable at 350K. You will need to be extra careful with your 350K model to make sure the mesh is as good as you can get, the time step selection is correct, double precision numerics and all the other tricks are used to get convergence.

[Cafard]

After check my 500K/8MPa simulation,
I found some region where the state is lower than critical point,
but it condition can be simulated.

I don't know why 350K/8MPa can't be simulated!

[ghorrocks]

I cannot answer that - you have given no information about what you are modelling or what the results look like - and we cannot diagnose that sort of stuff over the forum anyway.

Maybe in the lower temperature case the fluid goes further into the multiphase regime. Maybe a phase change occurs near a critical point in the flow (like a shock wave or separation). Could be lots of reasons.

[Cafard]

1.jpg
this is my centrifugal model (1/6)
use Opening and Outlet boundary
2.JPG
and this is the mesh quality
3.jpg INLET
4.JPG OUTLET
5.JPG Marerial table

[ghorrocks]

As I said in my previous post there is no way you can diagnose something as complex as your model on the forum. You are just going to have to use the general principles described in the FAQs I linked to work the problem out yourself.

[Opaque]

I am very surprised by your inlet boundary conditions. Any particular reason you are not running Total Pressure+ Flow Direction, Specified Turbulence Levels, and Total Temperature ?

The ANSYS CFX guidelines for boundary conditions indicate those are the most stable combinations for compressible flows.

[Cafard]

I will try it,thanks.