Short fiber example: fiber orientation tensor

[FedericoNegri]

No description provided.

[FedericoNegri]

@janvonrickenbach @greschd I downloaded the html doc artifact and was surprised to see that the Axx and Ayy plot have inverted titles!

I've already noticed in the past the order of saved images to be different than the actual commands, and I'm fine with that. But this is an issue. Ever seen it before?

These are the expected plots

[janvonrickenbach]

@janvonrickenbach @greschd I downloaded the html doc artifact and was surprised to see that the Axx and Ayy plot have inverted titles! I've already noticed in the past the order of saved images to be different than the actual commands, and I'm fine with that. But this is an issue. Ever seen it before?

These are the expected plots

@PProfizi: We get strange results when plotting results on the same mesh. To me this looks related to ansys/pydpf-core#632.

[PProfizi]

@janvonrickenbach @greschd I downloaded the html doc artifact and was surprised to see that the Axx and Ayy plot have inverted titles! I've already noticed in the past the order of saved images to be different than the actual commands, and I'm fine with that. But this is an issue. Ever seen it before?
These are the expected plots

@PProfizi: We get strange results when plotting results on the same mesh. To me this looks related to pyansys/pydpf-core#632.

Hi @janvonrickenbach, looking at the scalar bars between your "expected" and the "current", not only are the titles inverted but the maximum and minimum are not the same. I will take a look but I need to know how you got your "expected" reference. Is that with DPF-Core 0.7.4 or is that when run locally vs generated by Sphinx?

[FedericoNegri]

@PProfizi Reference results were obtained by running the notebook locally step-wise in a venv with DPF-Core 0.7.4.

[greschd]

Another way I managed to produce the wrong results is by creating three different meshes to plot on (while keeping the rest the same). Still running the script locally:

I'm not sure if this is supposed to work, but it might be an easier way to get to the root cause since it takes Sphinx out of the equation.

@@ -50,6 +50,12 @@ data_sources.set_result_file_path(composite_files_on_server.rst)
 model = dpf.Model(data_sources=data_sources)
 mesh = model.metadata.meshed_region
 
+model2 = dpf.Model(data_sources=data_sources)
+mesh2 = model2.metadata.meshed_region
+
+model3 = dpf.Model(data_sources=data_sources)
+mesh3 = model3.metadata.meshed_region
+
 # %%
 # Plot input data
 # ~~~~~~~~~~~~~~~
@@ -112,6 +118,16 @@ aXX = fiber_orientation_tensor_fc.select_component(0)[0]
 aYY = fiber_orientation_tensor_fc.select_component(1)[0]
 aXZ = fiber_orientation_tensor_fc.select_component(5)[0]
 
+# %%
+# Plot Axx
 mesh.plot(aXX, title="Axx", text="Axx plot")
-mesh.plot(aYY, title="Ayy", text="Ayy plot")
-mesh.plot(aXZ, title="Axz", text="Axz plot")
+
+
+# %%
+# Plot Ayy
+mesh2.plot(aYY, title="Ayy", text="Ayy plot")
+
+
+# %%
+# Plot Axz
+mesh3.plot(aXZ, title="Axz", text="Axz plot")

[PProfizi]

Hi @greschd @FedericoNegri, I did not manage to reproduce the issue when running 9_short_fiber_orientation_tensor.py, even with the three different meshes. I am using:

• Windows
• Python 3.10
• This branch of pydpf-composites
• I tried with Ansys 231 (DPF 5.0) and the latest 232 (DPF 6.1)
• I have this list of packages installed:

ansys-dpf-composites     0.2.dev0  D:\ANSYSDev\Sandbox\pydpf-composites
ansys-dpf-core           0.7.4                                         
ansys-dpf-gate           0.3.1                                         
ansys-dpf-gatebin        0.3.1                                         
ansys-grpc-dpf           0.7.1                                         
appdirs                  1.4.4                                         
argcomplete              3.0.5
cachetools               5.3.0
certifi                  2022.12.7
charset-normalizer       3.1.0
click                    8.1.3
colorama                 0.4.6
contourpy                1.0.7
cycler                   0.11.0
fonttools                4.39.3
google-api-core          2.11.0
google-api-python-client 2.83.0
google-auth              2.17.1
google-auth-httplib2     0.1.0
googleapis-common-protos 1.59.0
grpcio                   1.53.0
httplib2                 0.22.0
idna                     3.4
imageio                  2.27.0
importlib-metadata       6.1.0
kiwisolver               1.4.4
matplotlib               3.7.1
numpy                    1.24.2
packaging                23.0
Pillow                   9.5.0
pip                      21.3.1
pipx                     1.2.0
protobuf                 4.22.1
psutil                   5.9.4
pyasn1                   0.4.8
pyasn1-modules           0.2.8
pyparsing                3.0.9
python-dateutil          2.8.2
pyvista                  0.36.1
requests                 2.28.2
rsa                      4.9
scipy                    1.10.1
scooby                   0.7.1
setuptools               60.2.0
six                      1.16.0
tqdm                     4.65.0
uritemplate              4.1.1
urllib3                  1.26.15
userpath                 1.8.0
vtk                      9.2.6
wheel                    0.37.1
zipp                     3.15.0

Is there something I am missing?

PS: I have added a while ago this step right before all of our PyDPF testing or documentation jobs. It has proved to make debugging a lot easier.

[greschd]

@PProfizi I realized the difference is not due to the three different meshes, but due to running via the pydpf-composites Docker container or directly from the unified install.

The unified install gives the correct output, whereas the Docker container does not. I'm not sure if this means it's a Linux-only issue, or if something is wrong only in the Docker container. I'll test a unified install on Linux to compare.

Debug instructions:

• run docker run -d -p 21002:50052 -e ANSYSLMD_LICENSE_FILE=<license_server> -e ANSYS_DPF_ACCEPT_LA=Y ghcr.io/pyansys/pydpf-composites:latest
• modify the example with

  diff --git a/examples/9_short_fiber_orientation_tensor.py b/examples/9_short_fiber_orientation_tensor.py
  index d28c8a1d..36f790ce 100644
  --- a/examples/9_short_fiber_orientation_tensor.py
  +++ b/examples/9_short_fiber_orientation_tensor.py
  @@ -32,7 +32,7 @@ from ansys.dpf.composites.server_helpers import connect_to_or_start_server
   #
   # * Mechanical APDL result (RST) file containing the simulation results
   # * Mechanical APDL input file (DS.DAT) containing the fiber orientation tensor data
  -server = connect_to_or_start_server()
  +server = connect_to_or_start_server(port=21002)
   composite_files_on_server = get_short_fiber_example_files(server, "short_fiber")
  
   # %%

[greschd]

The difference seems to be independent of plotting. I've modified the example to just print a result, and run with both a local and container server here.

Output on Windows:

<local server output>
Result local:  0.8476316975728202
Result container:  0.08833158599895692

[greschd]

@PProfizi FYI, we figured out the root cause for these issues. In the example, we're accessing some data by index instead of by label, and the order in which the arrays are added is not deterministic.

No further action needed from your side, thanks for your help.

[greschd]

@FedericoNegri please review the changes made to the example. The main change is using the field variable name to correctly identify them (since the index is non-deterministic).