eufrat
Materials
- Feb 1, 2018
- 3
Hey guys,
I am working on a simulation of the impregnation of carbon fibers with resin. Therefore i have built a geometric model which contains the cross section of 4 carbon fibers and an area which represents the air and another area which represents the resin (I added an image in the attachements). Its my very desire to simulate the diving in of the carbon fibers into the resin for a time span of 5 seconds.
The geometric size of the model is very small, as I modelled the carbon fibers accordingly to their real dimensions, which are about 8 micrometers in diameter. I built the modell in mm, so my units are N, tonne, mJ, MPa, etc.
Unfortunately, with the correct material parameters, my stable time increment is with 2.7e-11 much too small to be anywhere near an acceptable computation time. I ran the simulation for a time span of about three weeks and at that point, the programm only managed to get to a total time of ~ 5e-4 seconds of the desired 5 seconds job time. Its just brutally slow. The computer which runs the simulation is a quite potent one as it is one of the institutes computers where I am working - So the problem cannot be a lack of hardware power. Anyway, the job runs with an Abaqus Teaching Licence which only allows the parallel use of 4 cpu cores...which is not that bad either.
The model itself is correct and there are no methodic mistakes in it as far as my knowledge goes.
At the very beginning of my work on this subject, the material parameters were wrong, as they followed the unit system meter with N,kg,J,Pa etc.
With these wrong parameters, the simulation is much faster with an stable time increment of ~ 1e-7. With this computational speed, the job is completed within 3-4 days, which is an acceptable amount of time. I understand the reason for this, as Abaqus calculates its time increment size in Explicit with the speed of a dilatational wave trough the model which the density and other material parameters have a direct impact on. So roughly spoken, the bigger my numbers for the material parameters, the higher the stable time increment. Sadly the model is simply wrong with these parameters.
Mass scaling doesnt work in this simulation because of two reasons: 1. It is not supported for Eulerian elements and 2. It is not supported for materials associated with EOS properties either.
Regarding the amount of elements used: My mesh is fairly fine, but not THAT fine. Its appropiate for the simulations goals. Anyway, of course I tried to lower the amount of elements used and watched the impact on the stable time increment. If I increase my global element size by the factor 10, which leads to an extremely coarse mesh, the stable time increment increases by ~ factor 3 and I get an increment size of about 7e-11. Still muuuch too slow, so I came to the conclusion that the extremely low time increment size is somewhat immanent to the model and its material parameters and is not caused by any mesh difficulties.
A few other information which (maybe) could be helpful:
1. The model has two loads: Gravity which applies to the whole model and body force on the fibers which is supposed to push the fibers downwards into the resin
2. The Boundary Conditions I used are of the type Displacement/Rotation
3. There are three materials in the model: Carbon (Density, Youngs Modulus, Poissons Ratio), resin (Density, EOS (Us-Up), Viscosity (Newton)) and air (density, EOS (ideal gas), specific heat, viscosity (Newton))
4. I built the model as thin as possible to reduce the computation time. In the mesh, its just one element in thickness.
So that was quite a bit of explaining and I thank anyone in advance who takes the time to read through this.
My question to you guys is: Is there any way to cope with such small increment sizes in Abaqus Explicit besides Mass Scaling (which doesnt work for the reasons stated above)? Maybe some of you have some other idea how I get to increase the time increment size significantly? Or is CEL just not suited for this type of application because of the tiny model size?
I am thankfull for any insights!
Greetings
I am working on a simulation of the impregnation of carbon fibers with resin. Therefore i have built a geometric model which contains the cross section of 4 carbon fibers and an area which represents the air and another area which represents the resin (I added an image in the attachements). Its my very desire to simulate the diving in of the carbon fibers into the resin for a time span of 5 seconds.
The geometric size of the model is very small, as I modelled the carbon fibers accordingly to their real dimensions, which are about 8 micrometers in diameter. I built the modell in mm, so my units are N, tonne, mJ, MPa, etc.
Unfortunately, with the correct material parameters, my stable time increment is with 2.7e-11 much too small to be anywhere near an acceptable computation time. I ran the simulation for a time span of about three weeks and at that point, the programm only managed to get to a total time of ~ 5e-4 seconds of the desired 5 seconds job time. Its just brutally slow. The computer which runs the simulation is a quite potent one as it is one of the institutes computers where I am working - So the problem cannot be a lack of hardware power. Anyway, the job runs with an Abaqus Teaching Licence which only allows the parallel use of 4 cpu cores...which is not that bad either.
The model itself is correct and there are no methodic mistakes in it as far as my knowledge goes.
At the very beginning of my work on this subject, the material parameters were wrong, as they followed the unit system meter with N,kg,J,Pa etc.
With these wrong parameters, the simulation is much faster with an stable time increment of ~ 1e-7. With this computational speed, the job is completed within 3-4 days, which is an acceptable amount of time. I understand the reason for this, as Abaqus calculates its time increment size in Explicit with the speed of a dilatational wave trough the model which the density and other material parameters have a direct impact on. So roughly spoken, the bigger my numbers for the material parameters, the higher the stable time increment. Sadly the model is simply wrong with these parameters.
Mass scaling doesnt work in this simulation because of two reasons: 1. It is not supported for Eulerian elements and 2. It is not supported for materials associated with EOS properties either.
Regarding the amount of elements used: My mesh is fairly fine, but not THAT fine. Its appropiate for the simulations goals. Anyway, of course I tried to lower the amount of elements used and watched the impact on the stable time increment. If I increase my global element size by the factor 10, which leads to an extremely coarse mesh, the stable time increment increases by ~ factor 3 and I get an increment size of about 7e-11. Still muuuch too slow, so I came to the conclusion that the extremely low time increment size is somewhat immanent to the model and its material parameters and is not caused by any mesh difficulties.
A few other information which (maybe) could be helpful:
1. The model has two loads: Gravity which applies to the whole model and body force on the fibers which is supposed to push the fibers downwards into the resin
2. The Boundary Conditions I used are of the type Displacement/Rotation
3. There are three materials in the model: Carbon (Density, Youngs Modulus, Poissons Ratio), resin (Density, EOS (Us-Up), Viscosity (Newton)) and air (density, EOS (ideal gas), specific heat, viscosity (Newton))
4. I built the model as thin as possible to reduce the computation time. In the mesh, its just one element in thickness.
So that was quite a bit of explaining and I thank anyone in advance who takes the time to read through this.
My question to you guys is: Is there any way to cope with such small increment sizes in Abaqus Explicit besides Mass Scaling (which doesnt work for the reasons stated above)? Maybe some of you have some other idea how I get to increase the time increment size significantly? Or is CEL just not suited for this type of application because of the tiny model size?
I am thankfull for any insights!
Greetings