Energy equation convergence acusolve (thermal analysis - natural convection)

Sebastian Yang

Hi all,

 

I am having a tough time trying to reach convergence regarding temperature residuals in a 3D natural convection problem. It consists of a box cointaining an hole in the middle which releases heat flux per unit area into the fluid.

The BCs are:

Fixed temp on the outer wall.

Heat flux on the walls of the square hole.

 

The solver settings are:

Boussinesq density model

laminar flow

relaxation factor 0.2

 

 

My questions are:

1. Are there separate relaxation factors for each one of the residuals (velocity, pressure, temperature) ? I only see 1 global relaxation factor.

2. How can I check the regions with bad convergence? In this post,

https://community.altair.com/community?id=community_question&sys_id=88b600ba1b2bd0908017dc61ec4bcbf1

ydigit explains that there is an option of outputting advanced varaiables. ¿How can I do that and how do I visualize the "bad" regions?

3. Any recommendation on large models? I can get convergence in a small model of size 0.1 m x 0.1 m, where residuals of temperature are <1e-3. However, I have a model of 1.5 m x 1 m size, I can not afford putting the same element size I used in the small model and the temperature residuals I am getting are around 0.1 after testing various mesh densities. On the other hand, Is it possible or is it a usual practice to do similarity analysis?

Thank you

acupro

Sebastian Yang said:

Thank you, this is really helpful!

The absolute value of the residual is probably better - as the residuals can be + or -.

Can also use the HWCFD Convert tool to create the H3D file - indicate to use extended output and select the desired variables to convert including the residuals - and open that H3D file in HWCFD Post.

ydigit

Could you post some pictures of your mesh? 

Hopefully you have activated gravity for natural convection. If possible upload your .inp and .Log file to start with. 

 

 

ydigit

HyperWorks CFD: Enable Nodal residual variables in „Fields Outputs“

Visualization of these residuals currently only with HyperView. Enable Extended Nodaly Outputs for
AcuSOlve Results Reader (either via .Log file or H3D if exported accordingly)

Possible to create iso surfaces to visualize regions of bad residuals. (ignore values in the screenshot
legend)

 

 

Sebastian Yang

ydigit_21275 said:

HyperWorks CFD: Enable Nodal residual variables in „Fields Outputs“

Visualization of these residuals currently only with HyperView. Enable Extended Nodaly Outputs for
AcuSOlve Results Reader (either via .Log file or H3D if exported accordingly)

Possible to create iso surfaces to visualize regions of bad residuals. (ignore values in the screenshot
legend)

 

 

Thank you, this is really helpful!

Sebastian Yang

ydigit_21275 said:

Could you post some pictures of your mesh? 

Hopefully you have activated gravity for natural convection. If possible upload your .inp and .Log file to start with. 

 

 

Hi ydigit, yes, I did remember to activate gravity for natural convection.

The final geometry is shown in the next picture, this is a simplification of the problem I am solving.

The outer box´s dimensions are 1.4x0.8x0.8 m

The inner box´s dimensions are 0.2x0.2x0.2 m

 

Here are some pictures of my mesh, this is a rather coarse mesh with 5 millon elements but I think it serves the purpose of illustrating the general mesh configuration I choose.

Picture 1: general view

 

Picture 2: detailed view of the boundary layer in the center.

 

Picture 3: mesh around the corner of the outer box

 

The mesh parameters are the followings:

Inner surface mesh size: 0.003 m

Outer surface mesh size: 0.01 m

Boundary layer:

 

Attached you will find the .log, .inp and .hm files.

- The hm has no volume mesh, it is intented for you to check the geometry if necessarily.

- The HM version is 2019, I chose that for convenience, because running on 2020 will take way more time due to computational resources availability problem.

 

I know the mesh can be refined, I tried to add more elements by limiting the maximum size of the tetras or reducing the size of 2d elements, but this results in a number of elements of tens of millons and it did not work in my real problem.

 

Another thing that I notice is that there are strange hotspots when I process the results, please check the next figure:

Thanks!

acupro

Sebastian Yang said:

Thank you, this is really helpful!

The absolute value of the residual is probably better - as the residuals can be + or -.

Can also use the HWCFD Convert tool to create the H3D file - indicate to use extended output and select the desired variables to convert including the residuals - and open that H3D file in HWCFD Post.

Sebastian Yang

acupro_21778 said:

The absolute value of the residual is probably better - as the residuals can be + or -.

Can also use the HWCFD Convert tool to create the H3D file - indicate to use extended output and select the desired variables to convert including the residuals - and open that H3D file in HWCFD Post.

Thanks for the tips, I now realise that I can use the derived results functionality to compute the absolute value of the residuals. 

I need to learn HWCFD though, too used to AcuConsole and HM