code_saturne User's Forum

Hello everyone,
I have some interrogations about convergence, i am trying to reproduce some results from an article (which gives HTC and temperature from a cube heated in an air flow) but the results I obtain are not really equivalent. Moreover my residuals are not in the usual range (under 10^-6).
Here is a picture of my norm residuals : I also put some probes that seems to be stables : To show you the theoretical results from Defraeye et al ("CFD analysis of convective heat transfer at the surfaces of a cube immersed in a turbulent boundary layer") in comparison to the results I produced (in blue) : If you have any suggestion of how to obtain a better result, and how to have a good convergence .

Regards

Quentin

Here are the infos from my simulation :
CS : 4.0

I attached the files requested.

Regards,

Quentin

Hello,

The problem that you want to simulate is a 3D problem and you are solving a "2D" (only 1 cell in the z-axis), this will change the flow around the cube. Additionally, the paper mention that the copper cube is cover with a epoxy layer of 1.5 mm and at its interior surface, a constant temperature of 75 C is imposed, which is not the same that impose the 75 C at the faces exposed to the flow. Also, they use a low-reynolds mesh.

Regards,

Luciano

Thank you Luciano,
I can understand that I have not exactly the same behaviour cause the physic is not exactly the same .
But speaking about convergence, do you have any idea why my norm. residuals are so high compared to the 10^-6 criteria ? Do I have to change something in my simulation parameters or so ?

Thank you,
Regards,

Quentin

Or maybe am I looking for the wrong convergence criteria ?
Can we assume convergence only with probes in the domain ?
Regards,

Quentin

Hello,

By default in CS 5.0 the precisions are the following (you can view in setup.log)
epsrgr 1e-05 Gradient reconstruction precision
epsrsm 1e-07 Rhs reconstruction precision
epsilo 1e-08 Resolution precision (Iterative solver)

You can try to change these values using the subroutine usipsu inside of the cs_user_parameters.f90. Also, you only set one monitoring point upstream of the square block, I think that you have to set more monitoring point downstream of the block where the turbulence plays a fundamental role. Also, try to change the turbulence model (K-epsilon, K-omega).

Regards,

Luciano