geomechanics / mpm Goto Github PK

The extended MPM (XMPM) strategy is introduced to describe shear band generation in soil.

Describe the bug
geomechanics wiki doesn't have the details for using "isotropic" flag with particles stresses. Right now it only shows:

        "particles_stresses": {
          "type": "file",
          "file": "particles-stresses.txt"
        },

I'm having trouble getting the "isotropic" flag to work on my own.

If correct syntax is provided here, I'll update the wiki.

This issue aims to implement the consistent tangent matrix of the small strain Mohr-Coulomb model for the implicit solver.

Included in on-going PR
This bug will be fixed in #62. Opening issue to make everyone aware of known bug.

Describe the bug
Garbage results for 3D simultions with absorbing boundary.

To Reproduce
Steps to reproduce the behavior:

1. Compile with standard WIKI instructions
2. Run on local machine with -p 1
3. Use 3D input files from mpm-examples repo

Expected behavior
See "Results" in absorbing boundary README.md

Additional context
Need to remove Tdim in following code and replace with 2.

Describe the bug
I have used the MPMExplicit2D solver for a 2D slope failure problem and MPMExplicit3D solver for a water dam break problem. Unfortunately, I encountered material points crossing the fixed boundaries (velocity set to 0 in specified directions) in both cases. I double-checked the setting of the boundary conditions and did not find a problem.

I modified the source code of the program to output the coordinates of these out-of-bounds particles. And I found that the coordinates of these particles are like (-1.75539e-18, 0.999947, 3.17375e-07) when I have a fixed bound in x=0 and the particles should have been located within 0<x<0.2. A screenshot of my console is shown.

ED2Q4 and ED3Q8 elements are used for 2D and 3D problems, respectively. Both usf and musl update schemes are tested and both take a time step dt=1.0E-4. Only velocity constraints are set.

To Reproduce
Steps to reproduce the behavior:

1. The code is compiled on Ubuntu 20.04 on WSL2 without mpi. The compilation results passed all tests.
2. The code runs on Ubuntu 20.04 on WSL2.
3. Take the water dam break problem as an example. As shown in the following figure, I meshed a box of 0.2x1.0x1.0 and generated a set of particles forming a box of 0.2x0.2x0.4. All six faces of the meshed box are set zero-velocity constraints in the normal direction.

Expected behavior
Figure out what causes the bug and fix it.

Runtime environment (please complete the following information):

• OS/Docker image: Ubuntu 20.04 on WSL2
• Branch: master

Additional context
None

Following PR #14 being merged, I am opening the issue to start looking at implementing the LME shape function to our code. Current development is going to be pushed under the branch element/lme.

Reference:
Arroyo, M., & Ortiz, M. (2006). Local maximum‐entropy approximation schemes: a seamless bridge between finite elements and meshfree methods. International journal for numerical methods in engineering, 65(13), 2167-2202.

A BSpline element is implemented for 2D and 3D MPM problems. I have checked that this works currently for explicit and implicit single-phase and explicit and semi-implicit two-phase codes. In the upcoming days, I will check the parallel performance and test the accuracy.

In this issue, we implement Newton-Raphson iteration for the implicit nonlinear scheme.

Describe the bug
After the merge of PR #37, the nightly build fails as these assert checks do not pass.

I will make a copy of such tests in our new benchmark repo and move the test to our repo instead of the main branch in CB-Geo.

There is a bug in the output timestamp in all of the implemented schemes. For example:

• output at step 0 is actually at 1*dt.
• output at step 100 is actually 101*dt

Add nodal accerlation support as boundary condition.

Useful for

• Directly applying accerlation time histories

Context

In our last meeting Kenichi seemed pretty adamant that the code should be called "CB-Geo-Berkeley". His thoughts were it draws a clear distinction from classic CB-Geo while also making it clear where most of the base structure is from.

Places

I don't think we should change the repo name (geomechanics is nice), but we can add CB-Geo-Berkeley in a few places.

Here are the places I've thought of:

1. Top of REAME.md
2. Wiki welcome page
3. Top of AUTHORS.md

Questions

1. What do others feel about this name?
2. Are the above places appropriate to put the name to make Kenichi happy?
3. Am I not considering another certain "good" location?

Describe the bug
There are two detected bugs in our two-phase semi-implicit code running on MPI.

1. After merging of #6, there is a new error associated with the kinematic fields of the shared nodes. The magnitude will be half compared to what it should be. While the pressure field is ok, this is something to be fixed.
   
   I am reverting PR #6 back as it should be the right one considering the reduction of masses at the shared nodes.

2. The pressure changes once a particle is transferred from one rank to another. This is caused by the non-transferred projection parameter related to the scheme.
   

To Reproduce
Steps to reproduce the behavior:

1. Compile geomechanics/mpm/master
2. Run a two-phase semi-implicit code on a linux ubuntu machine.
3. Activate MPI.
4. See error.

Screenshots
Provided above.

Runtime environment (please complete the following information):

• OS: Linux Ubuntu 20.04
• Branch: master

Additional context
A new branch bugfix/mpi_issues should fix this issue. A PR will be made soon.

Describe the bug
There is a new bug after PR #37 is merged. I am so sorry for not checking this before.

Here is the problem: the new update of displacement uses the old particle velocity with the factor of a half. This is not good since the boundary condition might not be imposed appropriately nearby the boundary.

I am currently finding a way to mitigate this but at the same time correct the error. In the meantime, I would propose that we revert PR #37 since it is breaking.

Simulation results before and after PR #37.

Improvement Idea

Use either:

1. input markdown file; or
2. the github wiki feature

to improve the documentation of features added to geomechanics repo.

Untracked Changes

Features already merged into geomechanics (without documentation) mentioned in Dec. 2 dev meeting:

• Implicit MPM
• 2 phase

@RyotaHashimoto , let's start discussing the finite strain implementation here. I am thinking that other than providing an option for users to choose small/finite strain, we need to do the following:

1. Material model with finite strain
2. Volume update with finite strain
3. Geometric stiffness matrix (should be added to the default tangent matrix)
4. Add treatment of volumetric locking (HKU team's work on smoothing the Deformation Gradient)
5. Code validation

Any further thoughts?

There is a bug in the current explicit MPM scheme which produces an error of 1/2 a*t^2 in the displacement field.

Add direct linear solver of Eigen for implicit and semi-implicit

This issue is for the implementation of B-bar element for infinitesimal strain formulation.
The implementation contains.

1. Adding a Boolean for the B-bar element usage.
2. Modification of dndx (explicit) or B matrix (implicit) when the Boolean is active.

Describe the bug
Updated versions of Ubuntu (21.10 and superiors) installed libc6 (glibc) 2.34. SIGSTKSZ is now no longer constant, and can’t be used directly to define array sizes. According to documentation:

"The constant SIGSTKSZ is defined to be large enough to cover the usual size requirements for an alternate signal stack".

See similar PR's in other people's codes:

• doctest git
• json git

And this is a proposed fix to the bug:

• json git

Affected file in the code:

• mpm/external/catch.hpp

Any thoughts @bodhinandach ?

To Reproduce
These are my machine specifications: "Ubuntu 22.04 LTS" and "gcc 11.2.0"

Screenshots