Infiltration models #190
Infiltration models #190
Conversation
|
Documentation preview: https://applied-material-modeling.github.io/neml2/pr-preview/pr-190 |
| void diagnose(std::vector<Diagnosis> &) const override; | ||
|
|
||
| protected: | ||
| virtual void update_forces() override; |
There was a problem hiding this comment.
Typicallyvirtual + override is considered redundant, because override by itself means the function is virtual. So:
| virtual void update_forces() override; | |
| void update_forces() override; |
| /** | ||
| * The value of the driving force, depending on `_control` this is either the prescribed strain or | ||
| * the prescribed stress. | ||
| */ | ||
| Scalar _driving_force; | ||
| VariableName _driving_force_name; |
There was a problem hiding this comment.
We used the name "driving force" because for solid mechanics, the force could either be strain or stress. But I suppose there is a more specific name for this, maybe something like "_inlet_mass_flow_rate"?
There was a problem hiding this comment.
We will also need doxygen documentation for most, if not all, members. They are the comments right above each member. There is a specific format for the comment, see eg https://www.doxygen.nl/manual/docblocks.html
There was a problem hiding this comment.
Also the variable name could be const, i.e.
Scalar _driving_force;
const VariableName _driving_force_name;| LiquidInfiltrationDriver::expected_options() | ||
| { | ||
| OptionSet options = TransientDriver::expected_options(); | ||
| options.doc() = "Driver for Liquid Infiltration Process"; |
There was a problem hiding this comment.
| options.doc() = "Driver for Liquid Infiltration Process"; | |
| options.doc() = "Driver for the liquid infiltration process"; |
| OptionSet options = TransientDriver::expected_options(); | ||
| options.doc() = "Driver for Liquid Infiltration Process"; | ||
|
|
||
| options.set<CrossRef<torch::Tensor>>("prescribed_liquid_inlet_rate"); |
There was a problem hiding this comment.
| options.set<CrossRef<torch::Tensor>>("prescribed_liquid_inlet_rate"); | |
| options.set<CrossRef<Scalar>>("prescribed_liquid_inlet_rate"); |
| options.set("prescribed_liquid_inlet_rate").doc() = | ||
| "Liquid saturate rate (mol/V) introducted to the RVE at the inlet"; |
There was a problem hiding this comment.
Maybe say this rate has units of molar mass per volume. "mol/V" might be a misleading. Also I don't think we should mention "RVE" in particular, because the driver could work for any infiltration model, not just ones derived from an RVE.
There was a problem hiding this comment.
Updated. Fixed for all other files as well.
| } | ||
|
|
||
| neml_assert(found, "No buffer named '", tensor_name, "' in file ", fname); | ||
| t = t.to(torch::kFloat64); |
There was a problem hiding this comment.
We should try not to hard-code the dtype to be float64.
There was a problem hiding this comment.
So we can issue a comments/ errors to the user to make sure they get the correct dtype?
| { | ||
| OptionSet options = FromTorchScript::expected_options(); | ||
|
|
||
| options.set<Size>("batch_dim") = -1; |
There was a problem hiding this comment.
Why default to -1?
There was a problem hiding this comment.
I feel like when people provided the tensors, I don't think they have a specific batch dimension in mind. I could change it to {}.
There was a problem hiding this comment.
We don't need to check in this file, do we?
There was a problem hiding this comment.
I was thinking using this file as an example to show how TorchScript can be used to generate files as user input tensors. Might be helpful or we can add this code as an example in the documentation.
There was a problem hiding this comment.
If we reuse most of the code from regression_solid_mechanics, we should think about how to reduce code duplication.
hugary1995
changed the title
…t functionality and updated NewtonWithLineSearch file
Co-authored-by: Gary (Tianchen) Hu <[email protected]>
…e, add in *.h Doxygen
…gfucnction file, add in *.h Doxygen
…gfucnction file, move files around, add in *.h Doxygen,
@hdt5kt feel free to edit the description of this PR.