juliankarlbauer / mechkit Goto Github PK

Add overview tables following suggestion by reviewer.

Problems:
In rendered docs,
the svg-images show up

• a bit small
• without arrow tips.

It might be good to

• enlarge the images a bit
• fix arrow tips.

It might be better to switch to png-files with low resolution (keep data small).

When you run

python setup.py develop

Pytest is not installed and you can't run the test suite.

Extention of the doc string of "mechkit.notation.ExplicitConveter"
by representations of stiffnesses and compliances in

• Voigt
• UMAT
• VUMAT

notation would be a huge improvement.

Currently, some placeholder "todos" show where additional views may be placed.

Inspiration for the representations may be given by these views (actually the stiffness-representation for Voigt notation is included in the below screenshot)

• Abaqus UMAT notation
  • stress
  • strain
  • stiffness
  • compliance
• Abaqus VUMAT notation (stress, strain, )
  • stress
  • strain
  • stiffness
  • compliance

Exchange of completely symmetric tensors, such as fiber orientation tensors of fourth order in a
minimal, i.e., nearly redundancy-free notation is a common task for engineers working with short fiber reinforced composites.

Support of

tensors = <bunch of  fourth order fiber orientation tensors>

con = mechkit.notation.ExplicitConverter()
con.convert(
inp=tensors, source="tensor", target="exchange", quantity="fiber_orientation"
)

with "exchange", e.g., being

[
N_1111, N_2222, N_3333,
N_2233, N_1133, N_1122,
N_1123, N_2223, N_3323, 
N_1113, N_2213, N_3313, 
N_1112, N_2212, N_3312, 
]

See also equations (38) to (40) in
Variety of fiber orientation tensors, Bauer and Böhlke 2021

It is clear, that, e.g., N_3333 could be omitted, as only 14 components are independent, but
not omitting it, will enable usage for, e.g., diffusion-weighted magnet resonance tomography tensors of fourth order,
which have no normalized trace.

Current implementation focuses programmatically on extensibility and clarity
(e.g. dicts containing converter functions and scaling factors).
However, this extensibility is not truly existent.
Converter-commands like .to_mandel6, .to_tensor, .to_mandel9
identify the source format by analysis of the dimensions of the input quantity.

Possible structural changes:

• Switch from Converter to tensor-class
  • Keep underlying extensible structure of converter internally and
    supply information of the source notation explicitly.

Addressing this comment, the contributing instructions should be improved and switch from "Github flow" to "Fork and Pull request".

Addressing this review comment, a comparison with multi-purpose tensor packages should be contained within the mechki-JOSS-paper.