[Draft] Hyperelasticity support (based on work by Joseph Beckett) #184

mohamed82008 · 2024-10-11T23:15:12Z

Summary

This draft PR is made from the hyperelastic branch of the fork https://github.com/estradalab/TopOpt.jl made by @jbecktt. I just wanted to see how many changes were made in a PR. The merge of master into this branch was clean so I hope no breakages were introduced and that you (@jbecktt) can continue working off of this branch jb/hyperelastic in the main TopOpt.jl repo. I invited you to be a collaborator in this repo.

Adding functions relevant to the optimzaiton of mechanical test data richness and experimental certainty.

Converted the name from JosephLoadCantilever to TensionBar, added a schematic for the documentation, and deleted a requirements related to enforcing an even number of elements along the y/z directions that was unnecessary.

Fixes bug where all of the quadrature points were not iterated over all of the basis functions. Also adds missing quadrature weight and determinant of the Jacobian to the formulation

Ignore .ipynb, .jl, and .vtu files frequently developed during testing.

Added a tension bar problem type where x displacement are the only enforced condition on the bar. Two additional points have contraints to prevent translation of the sample and to prevent rigid body rotation about the x-axis.

All FEA currently done in TopOpt assumes plane strain. This small fix just clears up the ambiguous 3D terms of defgrad tensors of 2D problem types (i.e., F33=0 and F31=F32=F13=23=0)

Add detailed comments and make code more concise. The primary focus of the edits were to the sensitivity related functions (i.e., not the entropy related functions).

Making a simple indexing fix to fix a mutation error from Zygote when calling autodiff.

Adding basic fundamentals of a hyperelastic solver. Still needs plenty of cleanings and comments. The main issue currently is autodiff still needs to be upgraded to use ImplicitDiff.jl to handle the iterative nature of the solver. The code still needs additional validation, but the bones are there.

Only ignore .jl files in top level of directory. Also ignore .png and .zip files.

Additional validation is pending regarding the time stepping code, but confidence is fairly high that it is implemented correctly. Stability seems to be excellent so long as sufficiently small time steps are used to achieve the solution. This also includes some of the foundation for implenting a nearly-incompressible hyperelastic solver that is not implemented yet.

Nearly incompressible HE is still only partially implemented. The smart time stepping is not adjustable yet via kwargs.

mohamed82008 · 2024-10-11T23:21:30Z

src/Functions/goodness.jl

+           if make_plot==true
+       #Below line creates figure slightly differently, using a method that is not the same as what was used to get h
+               # hist = histogram2d(K2_Vector, K3_Vector, nbins=n_bins, color=:viridis, xlims=K2_edges, ylims=K3_edges, background="black")
+               p=(heatmap(h))


It would be better to separate the plotting to be done outside this function, perhaps in the test folder instead of src. I don't think adding the heavy Plots package as a dependency for a single function is a good idea when a user can just load Plots externally. Alternatively, if you would like to provide more convenience plotting functionality, we have a Makie extension which we can extend further to support additional plots.

mohamed82008 · 2024-10-11T23:23:51Z

src/Functions/goodness.jl

+Outputs: stressTerms[i,j] (i.e. ∂²W(K₁,K₂,K₃)/∂Kᵢ∂ξⱼ) and kinTerms[i,j,k] (this is equivalent to d^3W/dK_i/dXi_j/dK_kk) =#
+
+function sensitivityFieldFncs(matlModel::Symbol)
+    @variables α K[1:3] # K[1:3] are the orthogonal strain invariants


Let's discuss if use of Symbolics is really required here. I wonder if ForwardDiff could be used to get the derivatives numerically instead of symbolically.

mohamed82008 · 2024-10-11T23:26:50Z

src/TopOptProblems/problem_types.jl

@@ -496,7 +496,7 @@ function LBeam(
    add!(ch, dbc)
    close!(ch)

-    t = T(0)
+    t = T(1)


is this intentional?

mohamed82008 · 2024-10-11T23:27:14Z

src/TopOptProblems/problem_types.jl

@@ -656,7 +656,7 @@ function TieBeam(
    add!(ch, dbc2)
    close!(ch)

-    t = T(0)
+    t = T(1)


same comment

mohamed82008 · 2024-10-11T23:27:29Z

src/TopOptProblems/problem_types.jl

@@ -754,7 +754,7 @@ function RayProblem(
        add!(ch, dbc)
    end
    close!(ch)
-    t = T(0)
+    t = T(1)


looks intentional :) but why

mohamed82008 · 2024-10-11T23:28:01Z

Project.toml

@@ -9,6 +9,7 @@ ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 ColorSchemes = "35d6a980-a343-548e-a6ea-1d62b119f2f4"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 Einsum = "b7d42ee7-0b51-5a75-98ca-779d3107e4c0"
+FFMPEG = "c87230d0-a227-11e9-1b43-d7ebe4e7570a"


This is not used I think.

mohamed82008 · 2024-10-11T23:28:37Z

Project.toml

@@ -36,7 +39,9 @@ Setfield = "efcf1570-3423-57d1-acb7-fd33fddbac46"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
+StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"


where is this used?

mohamed82008 · 2024-10-11T23:28:59Z

Project.toml

+PlotlyJS = "f0f68f2c-4968-5e81-91da-67840de0976a"
+Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"


Let's consider using Makie if possible since we already have an extension for Makie.

tolulade2023 and others added 15 commits May 13, 2024 11:41

Goodness intialization

d16be9f

Adding functions relevant to the optimzaiton of mechanical test data richness and experimental certainty.

Add DBC Problem Type and 2D DefGrad/Goodness support

da3cf0d

Clean up TensionBar problem type

baa5295

Converted the name from JosephLoadCantilever to TensionBar, added a schematic for the documentation, and deleted a requirements related to enforcing an even number of elements along the y/z directions that was unnecessary.

Fixed quadrature formulation in defgrad

32ba84e

Fixes bug where all of the quadrature points were not iterated over all of the basis functions. Also adds missing quadrature weight and determinant of the Jacobian to the formulation

Update .gitignore for dev work

2cdeae5

Ignore .ipynb, .jl, and .vtu files frequently developed during testing.

Add volume correction in defgrad

ee1ea1a

Add TensionRoller problem type

5a93b46

Added a tension bar problem type where x displacement are the only enforced condition on the bar. Two additional points have contraints to prevent translation of the sample and to prevent rigid body rotation about the x-axis.

Add plane strain components 2D defgrad tensors

4e408a4

All FEA currently done in TopOpt assumes plane strain. This small fix just clears up the ambiguous 3D terms of defgrad tensors of 2D problem types (i.e., F33=0 and F31=F32=F13=23=0)

Clean and restructure goodness.jl

e5e1ed9

Add detailed comments and make code more concise. The primary focus of the edits were to the sensitivity related functions (i.e., not the entropy related functions).

Patch for autodiff compatibility of FToK123()

a4511d4

Making a simple indexing fix to fix a mutation error from Zygote when calling autodiff.

Update .gitignore

02a72fd

Only ignore .jl files in top level of directory. Also ignore .png and .zip files.

Add def grad export and smart time stepping

edef8ae

Nearly incompressible HE is still only partially implemented. The smart time stepping is not adjustable yet via kwargs.

Merge branch 'master' into hyperelastic

f73911a

mohamed82008 commented Oct 11, 2024

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

[Draft] Hyperelasticity support (based on work by Joseph Beckett) #184

[Draft] Hyperelasticity support (based on work by Joseph Beckett) #184

mohamed82008 commented Oct 11, 2024

mohamed82008 Oct 11, 2024

mohamed82008 Oct 11, 2024

mohamed82008 Oct 11, 2024

mohamed82008 Oct 11, 2024

mohamed82008 Oct 11, 2024

mohamed82008 Oct 11, 2024

mohamed82008 Oct 11, 2024

mohamed82008 Oct 11, 2024

		PlotlyJS = "f0f68f2c-4968-5e81-91da-67840de0976a"
		Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"

[Draft] Hyperelasticity support (based on work by Joseph Beckett) #184

Are you sure you want to change the base?

[Draft] Hyperelasticity support (based on work by Joseph Beckett) #184

Conversation

mohamed82008 commented Oct 11, 2024

Summary

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment