Fortran Project - Finite Element Solver

FEA_Stress · October 14, 2022, 1:31am

It has been difficult to find people that can program in Fortran, so I thought I would ask here. If anyone is interested in helping out with a Finite Element solver, have a look at https://www.mystran.com. FEA (Finite Element Analysis) has been instrumental in improving the state of engineering, science, and medicine. It was popularized by NASA in the 70’s and 80’s. The MYSTRAN project was developed by former NASA engineers (as a variation of NASTRAN), then became a commercial code, and then later became an open source project. We have a forum and a Discord chat (which is now more active) as well. For now, we have a couple bugs that could use smashing and some performance issues. But there are also bigger goals if we can get some help.

Arjen · October 14, 2022, 6:29am

Welcome to the forum! FEA is not my area of expertise and there is plenty of other stuff I am currently involved in, but I am sure there are people here who can help out. It does sound like an interesting project.

(I noticed a small typo on your website: documenation )

FEA_Stress · October 14, 2022, 7:04am

Thanks, got the typo fixed up. Hopefully there is some interest. Its a worthwhile project IMO. Its just becoming harder and harder to find people who know Fortran.

Aurelius_Nero · October 14, 2022, 10:08am

I plan to do my masters in Computational mechanics which involves a fair bit of FEM, and I intend to use Fortran heavily. I could consider collaborating with you once I have some more expertise in the subject.
However, for now let me introduce @Ashok to you who knows some Fortran and has expertise on FEM as well.

Ashok · October 14, 2022, 10:43am

Thanks @Aurelius_Nero for the referral.
Hi @FEA_Stress, I am a Ph.D in structural engineering with expertise in FEM and Dynamics. I am also looking for some opportunities to code a FEA solver in Fortran - for now I am writing some toy solvers for myself. I am happy to collaborate on this - let me know your requirements.
Thanks,
Ashok

FEA_Stress · October 14, 2022, 12:24pm

That seems great…we would appreciate any help.

Well…there are many options. One of the more challenging ones would be to add Bathe’s MITC shell elements and shell buckling to MYSTRAN. There are some comments here:
https://www.mystran.com/forums/showthread.php?tid=85

Another thing that would help is to look at BUGS/Item 2 in this file:

github.com

MYSTRANsolver/MYSTRAN_Documentation/blob/main/Developer_Docs/Actions_and_Bugs/Actions.md

DOCUMENTATION
1.	(9/2022) Demonstrate (or develop a new procedure) for how to create the user manual with bookmarks. The prior versions had bookmarks (Bill used his procedure), but the updated user docs do no have bookmarks. This needs to be addressed. See “MYSTRAN-Users-Manual-how to create a PDF.doc” for Bill’s procedure (which may be out of date for a newer version of Acrobat).
2.	(9/2022) The Install and Run manual may need a TOC update.
3.	(9/2022) Investigate the Craig-Bampton Demo (See the Preface for further discussion). It is not clear what other problems may be used. This example is relatively specific so it is put in the developer side for now.
4.	(9/2022) The “Source_Code_Structure” folder is a data dump for now. It seems there are 2 versions, but neither one may be up to date (though one may be closer than the other).
5.  (9/2022) Update the change log here: https://github.com/MYSTRANsolver/MYSTRAN_Documentation/blob/main/Developer_Docs/ChangeLog.md
6.  (9/2022) Need OP2 general documentation and documentation that needs to go in the user manual. There is some preliminary information in the following link, but we also need something more formal for the user manual.
https://github.com/MYSTRANsolver/MYSTRAN_Documentation/tree/main/User_Docs/OP2_Preliminary
7.  (10/2022) Create a user document with some tips that can help users (things that may not be common). For example, SKIPMGG may be considered for problems where a mass matrix is not needed.
8.  (10/2022) Create a user document with sample BDF/DAT files that shows the setup for a MYSTRAN/NASTRAN deck in a simple way, including comments.
Also, within the sample, show how some common flags that may be not be obvious are used (for example, PARAM, SOLLIB, BANDED and PARAM, SOLLIB, SPARSE).

BENCHMARKS
1.	(9/2022) Add CSHEAR models to the benchmarks
2.  (9/2022) Add in the BUSH, BAR, etc. benchmarks that Toto was working on with Bill

GENERAL 
1.  (9/2022) Long-term modularization effort for Sparse Solvers
2.  (9/2022) Look into adding a CSV output. For starters, this could be a default output that is hard coded (no BDF input required).
The user would make changes at the Fortran level and then rebuild to create a "custom" CSV. After that, it could be adjusted to accept customization via BDF.

This file has been truncated. show original

We believe this bug was introduced between 13.2 and 13.3 (but discovered recently) and the windows EXE and source for the older versions. There is a MYSTRAN_Releases on GitHub that has these versions, but since I am a new user, I am limited to just 2 links per post.

There are also some performance issues, but the developers on the Discord channel are more familiar with the details of that. Bill was the primary developer, but he passed away. Our other developers are better at python and other languages, but struggle in Fortran.

Also, there was another request regarding how to deal with some output files, but that discussion was ongoing in the Discord chat.

Let me know what I can do to assist you further.

Ashok · October 15, 2022, 11:26am

Oh God! What a coincidence. I was exactly looking for implementing shell buckling in my toy solver. Fine, I will be happy to work on this. I will contact you further in the mystran forum.

mecej4 · October 15, 2022, 12:02pm

In contrast, most of us here may struggle to understand how the various pieces of a highly complex package such as Nastran fit together. Therefore, it may be helpful if you (and your developers) can describe specific problems that you are seeing with the Fortran modules/subroutines in your codebase. Doing so would reduce the risk that one of us may offer a fix that breaks the functioning of the remaining body of the package.

FEA_Stress · October 15, 2022, 12:38pm

Thanks for the responses. Adding the MITC shells and shell buckling would be a major improvement (its really the biggest missing ingredient now). We have shells now, but they are not the best and they are only first order (3-node tri and 4-node quad).

MYSTRAN is a complex package, but still much easier to work with than the NASA open package NASTRAN-95, which has been abandoned. We have a big challenge though. The main developer passed away in 2021. There are some files here (though I am not sure what is the most current that may be of assistance): MYSTRAN_Documentation/Developer_Docs/Source_Code_Structure at main · MYSTRANsolver/MYSTRAN_Documentation · GitHub

Starting on page 11 in the following document, there is also some information:MYSTRAN_Documentation/MYSTRAN_Installation_and_Run_Manual_1.0.pdf at main · MYSTRANsolver/MYSTRAN_Documentation · GitHub

But to get into the details, it may be better to go to our Discord chat or post on our forum since it could easily get complex if we try to do everything in the thread here. And while some developers are good at python, none of us are very familiar with the MYSTRAN code base itself (we have relied on the primary author in the past). For now, we are scrambling to figure things out more. Perhaps we should make better documentation for that.

FEA_Stress · October 17, 2022, 2:06pm

Sounds good…either the forum or Discord chat. We can talk in more detail about it then.

Arjen · October 18, 2022, 9:23am

FWIW, out of curiosity, I tried building the package on Windows. My experiences are less than satisfying:

As suggested I tried building via MSYS2 (or rather MinGW-w64). That failed because despite a message that CMake 3.23 would be installed, I am stuck with CMake 3.17.
Instead, I switched to Cygwin, which does have an up-to-date version on my machine.
The configuration step started, but then I got a complaint that submodules had not been downloaded. No indication of what submodules are to be downloaded. There is no indication in the BUILD.md file either.
A quick check of the MakeLists.txt file showed a few packages are needed, but there is no information that I can see about where to get the various packages.
I was also surprised to see f2c - is that really necessary?
The information in BUILD.md suggests though the script will do it. Is that outdated information?

Sorry to have all these complaints. Just thought you might like to know the experiences of a layman via-à-vis this FEA package.

FEA_Stress · October 18, 2022, 12:32pm

This definitely needs to be looked at and we will work on it. There is also a process to use CodeBlocks. Would that be something you find useful? If so, I will try to update that document and post a link.

Arjen · October 18, 2022, 12:39pm

Well, I am not a CodeBlocks user , but I am proficient enough with CMake to get through the various steps, even if I need to read the CMakeLists.txt files. It is just that there is more to it than “download, unpack, run CMake, run make.”

FEA_Stress · October 18, 2022, 12:46pm

Understood. We want to make it as user friendly as possible so we will look into it. Thanks for the feedback.

Arjen · October 18, 2022, 1:01pm

I will try to continue with this, but it may take a while.

Side note: I find a fairly recent version of CMake in the installation but it does not do much (no output, stops immediately - missing DLLs?). In the MinGW-w64 console a different version is invoked that actually works.

Arjen · October 19, 2022, 5:38am

I downloaded the source code for superlu and after being puzzled about the same error, I realised that the build script assumes you are building in the source directory. That is kind of unexpected - I usually build in a separate directory of my own choosing to keep the source directory clean. But now it is building, with the object files and the like appearing in separate directories.

FEA_Stress · October 19, 2022, 5:53am

Great We had some initial feedback from developers who said this.
“The initial response is to use a modern version of msys2 and install the correct version of cmake. This will be added to the trouble shoot section of the BUILD.MD file.”

Also, another comment: “I got same or similar message when compiling v14.01 on macOS. I had to copy superlu on the mystran directory since the zip from github doesn’t include it. After that everything went as expected.”

Arjen · October 19, 2022, 6:20am

I do have a relatively recent version of MinGW-w64/MSYS2 (installed in 2020 and updated this week) but I find the package manager/management a trifle confusing. That is my problem Anyway, I got it built via Cygwin.

One remark: the compiler issued a bunch of warnings on the parse_command_line function. They may be specific to the platform (Windows in particular):

/cygdrive/d/fortran/mystran/superlu/TESTING/cdrive.c: In function ‘parse_command_line’:
/cygdrive/d/fortran/mystran/superlu/TESTING/cdrive.c:533:18: warning: ‘optarg’ redeclared without dllimport attribute: previous dllimport ignored [-Wattributes]
533 | extern char *optarg;
|
I have no idea if that has any serious impact. As it is part of superlu, you may not have any control over this.

FEA_Stress · October 26, 2022, 3:08am

Hi @Ashok, have you been able to take a look at MYSTRAN? We could certainly use the help

Topic		Replies	Views
Altair Radioss - FEM solver goes open source	12	1475	September 13, 2022
New to Fortran!	12	696	September 18, 2021
Contributions to Open Source Engineering Projects	2	257	February 2, 2024
Hi! looking for new Fortran resources	15	1046	October 7, 2022
Replacing legacy code with Julia	15	884	July 8, 2022

Fortran Project - Finite Element Solver

Related Topics