hi All,
I am currently generate a RVE model with Periodic boundary condition for composite material and calculate in the ABAQUS/Explicit. the model dimension is 5um*5um*1um with an interface of 0.139um, the element type for the fiber and matrix is C3D6 and for the interface is COH3D8. I used the engineering constant to define the elasticity of fiber and define a local coordinate for the fiber direction, and use isotropic material for matrix then define the plastic behaviour with Drucker Prager, and define the interface with cohesive element. then i created periodic boundary condition with 18702 equations. i used the mass scaling by setting the target time increment 6e-6 (if below) and used the default increment setting (which is global) then I ran this model with double precision and full nodal output precision with 16 cores for 30hrs on HPC. however only 0.102 was computed then I checked the dat file and found that the stable increment stays at e-7 or e-8. I tried larger time increment 6e-5, resulting in unstable simulation and tried fixed time increment but simulation failed. some friends suggested using the set to set in the equation to reduce the number of equations, but the problem is still there.
could anyone give any suggestion or advice regarding this computing problem? thanks very much
https://www.sciencedirect.com/search?qs=ABAQUS%20model%20of%20RVE%20for%20composite%20material&show=25&sortBy=relevance
hi Md Washim Akram , thanks for your reply. my periodic boundary condition is exactly applied according to wang2017, and your provided papers are very useful. do you have any idea about the application of cohesive element with a certain thickness in the RVE without very expensive computation cost? thanks
Dear Lei Wan
I think you don't need to setup mass scaling for this analysis, let's keep it as default. You also can increase your increments to short your computing time.
Can you share your file here. I can try to check it.
hi Quang,
thanks so much for your suggestion. if we keep the mass scalling as default the stable increment timestep will be 4e-10, which is huge computation. that is why i set the target stable timestep as 5e-6. the file is attached, thanks for your help.
i got another problem when calculating the model, when i submitted the job to HPC with 8 cores, there is an error called "illegal memory reference (signal 11)" then i can only run with 1 core; when i submitted the job to workstation with more than 1 cpu, there is another error called " The executable explicit.exe aborted with system error code 3". do you have any idea about these error? thanks for your help.
hi Ahmed Boukar
I used a 5e-6 timestep with 0.1s loading time. you need to be cautious with the ratio of internal energy and kinetic energy to ensure your simulation is stable. also, the element number was chosen as around 50000 to compromise between computing time and result accuracy.
hope it helps
- Badges
- Science topic