Averaging element values (saved as Numpy array) at common nodes

[arun-maniam]

Hello,

I have element values saved as numpy array. To plot the element values graphically, I have done something like this (thanks to @germa89 for input from a previous discussion)

from ansys.mapdl.core import launch_mapdl

mapdl = launch_mapdl()

# Generate model
# 

# Genertate values of each elements
#

grid = mapdl.mesh.grid
scalars = values_for_each_element
grid.plot(scalars=scalars)

Now I would like to average the values at the common nodes. Previously this is possible in APDL when using PLETAB, where one can choose to average at common nodes. Is there any pymapdl that can peform this? If not, any ideas on how i could do this?
Thanks in advance

Arun

[germa89]

I'm not sure about his one. Pinging @mikerife for the MAPDL side and @AlejandroFernandezLuces for the pyvista side.

To summarise you want to average the elements (cells) values in the nodes (points) and plot its results.

However, I believe you could accomplish something similar by just using nodal values in the scalar field in add_mesh, instead of using the elemental values. I do not think the nodal values are the average in the nodes of the elemental values (what you want), but, it should be close to, specially if the nodal values are elemental derived quantities (stresses, strains, etc)

...
import pyvista as pv

grid = mapdl.mesh._grid

nodal_values = mapdl.post_processing.nodal_displacement()

pl = pv.Plotter()
pl.add_mesh(grid, scalars = nodal_values, camp="bwr")
pl.show()

[arun-maniam]

Yes you are right but the problem is i have done some calculations in the post processing with elemental values only. So i dont have the nodal solutions for that :(

[germa89]

The "hardway" would be to find a mapping nodes-> element (you have the opposite one in: mapdl.mesh.elem[ielem][10:]).

elem_ids = mapdl.mesh.enum 
elem_to_nodes = []
for i in elem_ids:
    elem_to_nodes.append(mapdl.mesh.elem[i-1][10:])

Then you can calculate for each node the average value, since you have the element values ( nodal_value = np.mean(element_values[element_for_node_id]).

# because you have same number of nodes per elements, you can do this:
elem_to_nodes = np.array(elem_to_nodes)

nodes_id = mapdl.mesh.nnum
values_nodes = []
for node in nodes_id:
    node_values_ = []
    for elem_i, elem_map in zip(elem_ids,elem_to_nodes):
        if node in elem_map:
            node_values_.append(elem_values[elem_i-1])
    
    values_nodes.append(np.mean(node_values_))  # averaging

values_nodes = np.array(values_nodes)  # For convenience

I hope it helps!

[arun-maniam]

Thanks for showing the 'hardway'. What do you actually mean by 'hardway'? Do you mean that it is tedious or it is time intensive?
Regarding the line in the code

for i in elem_ids:
    elem_to_nodes.append(mapdl.mesh.elem[i-1][10:])

Could you explain what is actually done here? I dont quite get it why the array is being sliced from the 10th element.

[germa89]

I meant it is more manual.. because you need to code everything... there are not much helpers in PyMAPDL for it (as far as I know, my memory is pretty bad). @mikerife is giving a good suggestion too.

mapdl.mesh.elem is giving an output similar to ELIST where the nodes of each element are the latest 10 values of the array. You can check the docs in: https://mapdl.docs.pyansys.com/version/0.67/api/_autosummary/ansys.mapdl.core.mesh_grpc.MeshGrpc.elem.html#ansys.mapdl.core.mesh_grpc.MeshGrpc.elem
So basically we are just storing in elem_to_nodes the nodes of each element.

[mikerife]

Hi @arun-maniam & @germa89
I was testing out the PyVista "cell_data_to_point_data" and I think this does what you are looking for. Watch out for derived data like equivalent stress; I think that MAPDL might average the components first, then compute the equivalent stress. Whereas PyVista is doing a straight averaging.

I was testing a simple model and the results looked good (except for derived).

import matplotlib as mpl
cmap1 = (mpl.colors.ListedColormap(['blue','royalblue', 'cyan', '#00FA9A','#7CFC00', '#ADFF2F','yellow', 'orange','red'])
        .with_extremes())

mapdl.post1()
mapdl.set('last')
mapdl.etable('sy','s','y')
mapdl.view(1,1,1,1)
cmd = '''
/show,png
pletab,sy
pletab,sy,avg
/show
/replot
'''
_ = mapdl.run_multiline(cmd)

cell_data = mapdl.pretab('sy').to_list()
cell_sy = []
for item in cell_data:
    cell_sy.append(item[1])
mesh = mapdl.mesh.grid
mesh["StressY"] = cell_sy
mesh.plot(scalars = "StressY", cpos = 'iso', cmap = cmap1)

mesh2 = mesh.cell_data_to_point_data()
mesh2.plot(scalars = 'StressY', cpos = 'iso', cmap = cmap1)

The two averaged plots are then:

They are a little different but I'm not sure if the difference is much of an issue.

Mike