Abaqus Simulation Run Time

[mhdiab]

Hey all,

I am running a wave propagation simulation on Abaqus that requires a large number of element and a very small time step(increment). I am using the Dynamic Implicit analysis type. When I run the simulation ,it take a very long time. I even tried using a faster computer(32 GB of RAM) but that did not reduce the time by much! Is there anything I can do to reduce the time it take that simulation to complete?
Also, the .odb file is very big(about 1 GB) although I only have requested one Field Output(displacement) and one History Output(also displacement). Any idea if that is normal ?
Thank you for your help!

Regards.

 

[burningDiesel]

1. If you have a large number of elements and small time step, it's normal to have a long simulation.
Things to take into account:
If you don't have enough memory, you will start using swap memory and things will slow down bigtime. But 32GB of RAM should be more than enough.
32 GB of RAM is measure of quantity of memory, not speed. So, unless you are using more than the available amount, CPU and RAM speed will determine actual speed (and I/O if you are using parallel computing).
2. Your mesh should be fine enough and not too fine to fulfill CFL and Blake's criteria
3. With fixed time incrementation in implicit dynamics you can try setting NOHAF to save computational time (but check your solution!)
4. Are you sure you are also not asking the default outputs? Else you truly have a huge mesh. Also, determine at what frequency you want output, and set that frequency (instead of at all increments).

 

[mhdiab]

2.I have chosen the appropriate mesh and time increment size based on CFL and Blake's criteria. I am trying to simulate with a high excitation frequency(100Khz) and I would need about 16000 time increments in addition to a high number of elements. Last time I ran the simulation I had 3000 increments and that took more than 8 hours. So I thought I might be doing something wrong!
3. How do I change that?
4.Yes I have only selected only U3 from the History Output Request.
I don't understand why the .odb file is so big even though I only requested U3 for one node only !

 

[burningDiesel]

add ",NOHAF" (without quotes) after the *DYNAMIC keyword
if nonlinearities are not important you can also set NLGEOM=NO in the *step option

16000 increments is a lot for an implicit solver, maybe you can just use *dynamic, explicit.

For the output, do you need 16000 datapoints? you can also use filtering of output (and less datapoints).