Many studies involving damage models which are not present in Abaqus are generally implemented with UMAT (Abaqus Standard) and VUMAT (Abaqus Explicit) Fortran subroutines. However such subroutines are relatively complex to implement and debug. Could USDFLD subroutine be employed for such a purpose?
Dear Francesco,
My answer to your question would be "probably not". But not with full surety.
In USDFLD you can play with the behaviour by assigning a field variable dependent property. But how will you assign field dependent property to a whole plastic dataset. One yield point you can do, but then that would be homogeneous plastic. Elastic properties can be easily played around by USDFLD alone.
The model you are talking about perhaps contains yield criteria like GTN. In order to do that and calculate void fractions from stresses/strains you will need to write UMAT or VUMAT. Then further complexity can be imposed if you use UMAT/VUMAT with USDFLD.
And UMAT/VUMAT is not that complex when you try it. Once its done I am sure you will not even think about using USDFLD alone for the said purpose.
Find the attached file for more details.
All the best.
Regards,
Abhishek
Dear Abhishek
first thank you for you prompt reply and the file you have attached. I've already red such a pdf and it appears evident that UMAT and VUMAT subroutines would be much more complicated as compared to USDFLD. In addition, I've red a lot of comments on UMAT which confirm the complexity in the model debug which discouraged me to deal with it. Have you some experience with such subroutines?
On the other hand, I've found some papers which introduce the damage parameter by USDLFD: "In the crack propagation analysis, a field variable was defined as damage parameter (β) using the USDFLD subroutine within ABAQUS. " from "Assessment of ductile fracture resistance of structural steel using a simple numerical approach " and "The first one is the USDFLD subroutine to re-define field variables. By defining as state variables, information on the mean normal (hydrostatic) stresses and equiv- alent stresses/strains at gauss points is passed into the UHARD subroutine. In the UHARD subroutine, accumulated damage is calculated according to Eq. (4) using information passed from the USDFLD subroutine. " from "A finite element ductile failure simulation method using stress-modified fracture strain model".
what do you think?
Dear Sir,
Frankly speaking I dont have much knowledge about RnT model at the moment. What I could gather is that its void volume fraction dependent on stress/strains. I am not sure whether any yield criteria like GTN is applied there.
If you dont have to change Yield criteria you can apply USDFLD alone.
I have written VUMAT for GTN and crack growth, so from that experience I can tell you that VUMAT/UMAT gives you more freedom than USDFLD.
Further, people have done crack growth using UMAT in conjugation with XFEM. But I could not do that due to some ABAQUS error. Even in XFEM there is not much freedom with the fracture criteria and crack opening direction.
These were the issues I had which is the reason for my affinity for VUMAT/UMAT.
Moreover, I also think that if you deal with USDFLD a little more complicated would be VUMAT/UMAT. One more thing is that in VUMAT you dont have to make Jacobian; so unless you try to implement non-locality of damage, VUMAT would be better than UMAT, because element deletion option is available for VUMAT.
These are of course my views.
With regards,
Abhishek
VUMAT is probably the simplest way; use explicit time integration to deal with any rate equation. Much easier than UMAT, no iterations required, no convergence issues...
It depends on how complex the material behavior is, as well as on limitations associated with built-in constitutive laws e.g. the hyperelastic laws do not allow for damage but the viscoplastic do. Therefore VUSDFLD through field variables will not help when using a built-in hyper(visco)elastic law and in this case VUMAT/UMAT is necessary. However debugging the VUMAT routines through simple models i.e. single elements does not guarantee reasonable behavior in the real application model especially when large strains & element deletion take place. You have to account for sufficient energy dissipation to avoid such problems. The following paper gives a practical-easy solution for modelling large strain fracture for non-linear materials in ABAQUS:
SKAMNIOTIS, C. G.; ELLIOTT, M.; CHARALAMBIDES, M. N. On modeling the large strain fracture behaviour of soft viscous foods. Physics of Fluids, 2017, 29.12: 121610.
- Badges
- Science topic