MASS SCALING

[kengwit]

I am trying to use ABAQUS Explicit to do a static analysis followed by a dynamic analysis.

I am trying to model the effects of impact load at the ground surface on a buried pipe. The Young's modulus of the pipe is very high relative (1000 times higher) to the soil elements.

Here's the problem:

In the analysis there are 2 steps:

In Step 1, I want to do a quasi-static analysis to obtain a reasonable static. To speed up run time I tried using mass scaling *FIXED MASS SCALING and the static solution was reasonable based on a comparison with the static implicit case. The mass scaling adjusts the masses so that a single reasonably large time increment is used.

Now, in Step 2, the load is immediately stepped an additional 30% at the start of the step and then kept constant at that level. In this step, the scaled masses are reset by reissuing the *FIXED MASS SCALING. But what I am observing is that there are very high accelerations emanating from the pipe at the start (time=0)of the step. This doesn't make sense because my disturbance hasn't propagated and reached the pipe.

I think it is because when I reset (reduced) my masses back to the original values, it is doing something with the accelerations: a=M^(-1)*F

So when I reduced the mass the accelerations jumped significantly. As I mentioned before, the pipe stiffness is very high relative to the soil, so during the first step, the mass has to be reduced drastically to obtain a reasonable time step, and when the mass change during the reset in Step 2 is also huge.

Has anybody encountered such a situation? Is my reasoning correct? Right now, I a running the quasi-static without mass scaling but ramping the load up slowly and smoothly. Any suggestions on speeding things up will be appreciated.

Thanks.

 

[mizzjoey]

Hi kengwit.

Hmm... when we run Explicit with mass scaling in my company, we always run in a single step. The energy and acceleration get too high at the end of the 1st step which will spoil the 2nd step. So what we do is run separate analyses, one for each step. You can try running the 1st step in one analysis, import the result into a second analysis (which is really the 2nd step), and so forth. A lot of work, copying the model and modifying for each step, but better than the energy spike from the previous step making the results from subsequent steps unreliable.

Don't know any other way to go about this issue. I'm interested to find out other methods too...

hope this helps,
mizzjoey

 

[rmg1978]

Hi, kengwit.

I encountered something similar in attempting to model high impact forming with subsequent steps.

I found it useful to add in an extra step with nominal time. Applying full motion constraints to all nodes I was able to remove the k.e.

I was then able to continue with my subsequent time step, without having to copy to new models

Hope this helps.

Regards

RMG