Problem solving a thermal simulation with API pyMAPDL !!

[MASV1610]

Hi, I am new to pyMAPDL and i need urgent help with this code. If you have any questions or need more context, please do not hesitate to contact me.

Context

The main goal of this project is to simulate a multi-layered thermal system using ANSYS MAPDL through Python. The simulation includes three distinct volumes: a wooden base table, an EPS (Expanded Polystyrene) block acting as an insulating layer, and a top RC material sample that exchanges heat with the environment via solar absorption, convection, and long-wave radiative cooling.

The geometry is defined using block(...) commands, creating structured regions for each material with precise spatial coordinates. The materials have different thermal conductivities, which allows for the study of how heat propagates through the layers and how the final temperature of the upper material (RC) depends on environmental conditions such as ambient temperature (Taire), wind speed (v_vent), dew point (Tdp), relative humidity (HR), time of day (t), and solar irradiation (Irrad). These variables are read from an Excel file (design_points.xlsx) and processed row by row.

To analyze the thermal behavior of the system, a custom function called Tmostra(...) iteratively calculates the surface temperature of the RC material until convergence is reached. This function takes into account:

• Natural and forced convection coefficients

• Heat fluxes from absorbed solar radiation

• Long-Wave Radiative losses (qLWR(...), imported from another module)

• Conductive coupling between layers

At the end of each iteration, nodal temperatures are extracted using get_array("NODE", item1="TEMP"), and the median or mean is calculated to represent the steady-state temperature of the sample (Tmostra_i).

Problem

I've been doing some research, but after a long time I still don't know what I'm doing wrong. I am new to pyMAPDL and this is my first code. Any help is welcome, thank you very much.

Code
codi_def_2realvariables.txt

[mikerife]

Hi @MASV1610
Well I see two major issues just reading through the script. The first is that the volumes/mesh are not connected to each other. Neither by shared CAD (i.e. two volumes sharing 1 or more faces) nor by contact. The second is that MAPDL allows one load/bc per element face. I can't quite keep track in my head of all the node selection going on for the loads, but it looks like at least some are defining multiple loads to the same set of faces. If you want to apply multiple loads you need to use surface effect elements to apply the 2nd, 3rd etc load to the face.

I'll trim down the input file to run just one instance of the model to confirm. But that will take a while.

Mike

[MASV1610]

Hi Mike, thanks for your reply.

I don't understand at all why you say that the volumes are not connected through the mesh, because when I do the mapdl.eplot I see that everything is unified. Also, the mapdl.vsel(‘ALL’) command selects all the volumes created and then mapdl.vmesh(‘ALL’) meshes it all as one, isn't need to link and make a master - slave relationship between nodes, correct me if I'm wrong.

I don't understand this: If you want to apply multiple loads you need to use surface effect elements to apply the 2nd, 3rd etc load to the face.

However, the code is already working, as I've been tinkering with it these days. It seems that the real reason for the high temperature error was because I was applying too much thermal load in general, both on the EPS block faces and on the wooden table. The main problem is that the Solar irradiation does not apply everywhere at the same time, because there are shadows and it will depend on the tilt/position of the sun and the time of day.

Now, if I only halve the thermal load and apply it only to two side faces (or 1) and to the half of the table (or a quarter), the code works without errors, the only thing is that it gives me an error relative to the temperatures of the original experiment, about 10%, but I guess it is refining this last thing I said about the position of the sun during the day.

What do you think, do you think that could have been the problem? I'll share the actual code for you to take a look. :)

Thank you very much!!

codi_def_2realvariables.txt

[mikerife]

Hi @MASV1610
The mesh sizing just happens to be defined so that the nodes are coincident at the interface between the volumes. But there are duplicate nodes. If you have node numbering turned on when plotting the elements you would see the coincident nodes easier (in MAPDL interactive). Selecting the volumes and meshing does not automatically create a connected mesh between volumes. The volumes need to share area/lines/keypoints for the mesh to share nodes.

Perhaps you have an APDL macro you wrote named CMAP? For example I found this in the script:

mapdl.cmap("mesa_nodos_sup", "eps_nodos_inf", "NODE", "TEMP")  # Acoplamos temperaturas

However CMAP in PyMAPDL is its version of the APDL command /CMAP (deals with color mapping table). If I am right, I don't have access to your macro library so it makes sense that the FEM I see is different than yours.

Mike

p.s. the same goes for qLWR_calculator python library. A google search returned nothing so I guess it is a library you wrote....??

[MASV1610]

Hi again Mike,
So should I add "tolerances" to the Z coordinate when creating the bulk box? So that it shares part of the volume with the table (botto

m surface) and part with the sample (top surface).

About his: mapdl.cmap("mesa_nodos_sup", "eps_nodos_inf", "NODE", "TEMP") # Acoplamos temperaturas; an AI ​​provided it to me, is used to join nodes between the different volumes, but I have commented on it, because supposedly, according to what chatGPT told me, it is not necessary because the nodes already communicate automatically.

The first code is the one that executes the other two codes: the first one I post it before and also, indirectly, the one that is named qLWR, which is the last one I passed. I also attach the excel file from which the data is taken.
design_points.xlsx
corre codi def 2real.txt
qLWR_calculator.txt

Thank you!

[mikerife]

Hi @MASV1610 "So should I add "tolerances" to the Z coordinate when creating the bulk box? So that it shares part of the volume with the table (bottom surface) and part with the sample (top surface)." No. MAPDL does not automatically join/glue/add/etc CAD entities as they are created. Unlike say DesignModeler which does, unless the body is in a 'frozen' state. Or if the newly created bidy is created as 'frozen'.

MAPDL has several options to deal with two or more CAD entities that are disconnected but need to be connected. There are Boolean opterators; merging of nodes or keypoints; etc.

ChatGPT did you wrong! Not sure how it got that so wrong; see here.

Thanks for the qLWR library.

[MASV1610]

Hello @mikerife

All right, I will check this out. If you can give me some idea or example of how to connect the different geometries/volumes, you do me a favour.

Thanks.

[germa89]

If nodes are coincident (or within tolerance), you can use nummrg. This command merges nodes, elements, areas, etc..

If you need interface behaviour, like sliding or cohesion, you need to use CONTACT elements. You should check the contact guide from the ansys portal. But I think this is not your case.