29th Jul, 2020
University of Texas at Arlington
Question
Asked 28th Jul, 2020
Can I assign different constitutive laws to different parts in FEniCS?
I am working on multiscale modeling of composites and I want to explore the capability of FEniCS in modeling composites. Specifically, Can I assign different constitutive laws to different parts in mesh using FEniCS? Can I access the integration points to change the material behavior in a nonlinear analysis?
Thank you.
Most recent answer
Hi Dr. Helfer and Dr. Bleyer,
Thank you for answering this question. I think this project is very interesting and meaningful. I will keep following the latest updates from you.
All Answers (6)
This us feasible but with some limitations. For example, you may need to allocate a field with thé maximum number of components to hold the internal statue variables.
1 Recommendation
Dr. Bleyer. Thank you for your reply. I was thinking about general cases. For example, in a fiber-reinforced composite I may want to use a elasto-plastic model for matrix and linear elastic model for fibers. Or in a textile composite I may want to use a elasto-plastic model for matrix and a continuum damage model for yarns. The key point is to connect an user-defined constitutive model to the integration points in FEniCS. Something similar to Abaqus UMAT.
Thank you!
Jeremy Bleyer we could do this with MGIS (MFrontGenericInterfaceSupport) but not mgis.fenics. However, we shall do some plumbering and I can't guarantee that post-processing of internal state variables would ne easily feasible.
For the first case, you can check the following discussion https://fenicsproject.discourse.group/t/two-weak-forms-in-single-domain/3833/2
Right now the second case cannot be handled smoothly but still remain feasible with some work as suggested by Thomas Helfer . We are currently working on the future dolfin-x version to extend the mgis.fenics module and possibly show how to interface seamlessly FEniCS with any user-defined constitutive model. But this is not ready yet.
Similar questions and discussions
Multiple abaqus jobs submitted through python script on a cluster is running one-by-one instead of parallel at the same time?
Hello,
I'm using for loops in a python script to run a parameter study in Abaqus, that is I submit a job using:
mdb.jobs[jobName].submit(consistencyChecking=OFF)
I then change some parameters and submit a new job with a different jobname, etc. When I'm doing this on my own pc Abaqus is able to utilize all the resources I have available, for example if I have a 6 core 12 thread cpu and submit a bunch of 2 cpu jobs, I can have 6 jobs running simultaneously.
If I try the same method on my institution's cluster, I request 12 cpu cores for a slurm job and submit my python script for analysis, there's only 1 job running at a time instead of 6, meaning 1 parameter job starts and finishes before the next one starts (I'm not using the .waitforcompletion() method).
I'm aware it's a rather open question, but have anyone experienced this unintentional non-parallel behaviour on a cluster before, and how might I be able to solve it?
Best regards, Jens
Related Publications
This document describes the software developed for use in calculating K, the 4th order parameter tensor used in ALE3D's anisotropic plasticity model. The multi-scale modeling method developed for this calculation begins with orientation imaging microscopy (OIM) data. The program OIMA3D characterizes the sizes and crystal orientation of the grains f...
The hierarchical multi-scale procedure is analysed in this paper. A local multi-scale model has been studied with respect to the macro-level mesh size and meso-level cell size dependency. The material behaviour has been analysed in case of linear-elasticity, hardening and softening. Though the results show no dependency in cases of linear-elasticit...
Presents a range of numerical results obtained from the geometrically nonlinear analysis of a cantilevered cylindrical shell. Shows that, while the fine-mesh solution involves no limit points, as the mesh is coarsened, an increasing series of “false limit points” is encountered.