I’m a FEA analyst and a long-time Abaqus user, mainly focusing on failure analyses. By following the development of Calculix, I recently discovered what great work has been done with PrePoMax.
The reason I’m still using Abaqus is the lack of fracture mechanics features in Calculix. However, there is a great open source fracture mechanics focused FE solver called Warp3D (http://www.warp3d.net), but with no pre or post-processor. It has a similar input based syntax as Abaqus and the results are output as binary or text-based formats. It has a different material model library and more specialized features related to failure modelling of metals than Calculix (such as J-integrals).
Currently I have built my own adhoc Python scripts to convert Abaqus-style meshes to match the Warp3D syntax and added the loads etc. manually to the input. Then I have a second script that converts the Warp3D output stress and strain files to the frd-format readable by PrePoMax.
My feature request is to make PrePoMax support Warp3D. The capability to export meshes and sets or whole models in the Warp3D format and the capability to read in the Warp3D output files (text or binary) would help a lot! Having the solver integrated in the PrePoMax GUI might be asking too much.
As an example of the file formats, attached is a test model of an uniaxial bar with results both in Warp3D format and .frd format.
Have you heard about the newly introduced fracture mechanics capabilities in CalculiX ? The 2.19 version brought HCF crack propagation procedure (*HCF) and, from what I remeber (unfortunately older release notes are replaced with new ones), the 2.18 version brought support for LCF analyses (*CRACK PROPAGATION). I haven’t tested these functionalities yet but they seem interesting.
The power of Abaqus in terms of fracture mechanics analyses lies in its XFEM implementation which is pretty much unrivalled. There are some open-source codes that utilize this method such as OpenXfem++ and GetFEM++ but they can’t be easily integrated with advanced preprocessor like PrePoMax.
What you said about the use of Warp3D in conjunction with PrePoMax is very interesting. I’ve seen some examples in which open-source topology optimization solvers where used with models created in PrePoMax but I diidn’t know that it’s also possible with crack propagation analyses.
I have to disagree a bit that fracture mechanics simulations would be only crack growth modelling, with XFEM or other techniques.
I don’t usually consider crack growth with FE simulations, because it is often tackled with analytical methods with sufficient accuracy. The functionality I’m after is the possibility to evaluate contour integrals (J-integral - Wikipedia) for static cracks, which is not available in CalculiX or most other open source FE codes. The crack propagation functionality in CalculiX, as the name suggests, considers only fatigue-induced growth without actually calculating the contour integrals.
Also, apart from user material subroutines, CalculiX does not have any damage models, such as cohesive elements, strain-based softening or the GTN model. Warp3D is focused in this field.Having PrePoMax export the meshes and sets in a Warp3D-compatible format would save a lot of manual conversion work and being able to read the Warp3D binary results in would be even better.
Warp3D has a conversion routine for the Exodus II format (.exo) that can be opened with ParaView, but PrePoMax is way superior to ParaView in processing finite element modelling results. Maybe supporting this format would be easier.
In my previous post, I focused on crack propagation modeling but you are right that contour integral and SIF evaluation for stationary cracks is also important and missing in most open-source codes. CalculiX’s fracture mechanics capabilities are very new and will likely be further extended in future releases so there’s a chance that contour integrals will also be included.
I’ve also seen other examples where J-integral was calculated in postprocessing in the software not having any fracture mechanics functionalities implemented. It was based on strain energy output but more automated since the software supported user-defined output based on existing variables and various operators (including integration). Maybe that could be utilized in PrePoMax as well (there were requests for equations operating on output variables to create custom results already).
I was considering how to support other solvers in the past. Integrating them into the source code is problematic since a detailed knowledge of the solver is needed to correctly prepare the data and I do not have/will not have such knowledge/time.
So the idea is, to have support for some kind of add-ins in the shape of exe files. The add-in would have to import the inp created by PrePoMax and then export the required file format. Then use the same principle for parsing the results. PrePoMax could be used to run the solver and the necessary plugins before and after the solver.
Thanks for the suggestions! This discussion goes a bit off-topic, but the mentioned examples use the general definition of the energy release rate, by calculating it via the change in energy of two crack lengths. It has practical use only in simple 2D cases and requires the construction of two distinct models and results in an average of the energy release rate for the crack. It cannot separate different loading modes and contributions to the ERR.
Warp3D and Abaqus use the virtual crack extension approach that allows the SIF or J to be calculated from a single model and at all nodes along the crack front.
During my studies, I tried to reproduce the contour integral procedure in Abaqus with only limited success (it worked in simple cases but taking into account thermal and residual stress fields required access to detailed element-wise results not available to the user). I followed the procedure proposed in (PDF) Numerical Evaluation of Domain and Contour Integral for Nonlinear Fracture Mechanics: Formulation and Implementation Aspects .The developers of Warp3D have devoted a lot of effort in their contour integral routines, which actually outperform those in Abaqus. So the contour integral calculation is a difficult task and I prefer to use existing and qualified routines.
may in the future PrePoMax will be expanded to support many FE solver develop by each implementer (e.g PrePoMax & CalculiX, Code_Aster, OOFEM, OpenSees, etc).
separation may required since a fully integration in one packages could make doubt and confused for general user. each FE solver has specific domains and behavior, may not compatible between of them.
fortunately, the limitation of material library in CalculiX could be possible to fulfilled by MFront integration.