Altair RISE

A program to recognize and reward our most engaged community members
Nominate Yourself Now!

huge temporary files

User: "Luca_00"
Altair Community Member

Hello Altair Communuity,

I built a large finite element (FE) model and want to evaluate the stresses in the components. The problem is that HyperWorks creates a large number of temporary files during the calculation, which causes the calculation to break down when only a few elements are output (less than 100,000).
If the calculation runs successfully, the result files are not large either. (e.g., 8 GB).
The load case is a modal transient shock load (250 time steps, one shock direction) on a metal and carbon structure.
Does anyone know why this large number of temporary files is created? This does not occur with other models that have a similar number of elements and are made of metallic materials only.

Thanks in advance

Find more posts tagged with

Sort by:
1 - 2 of 21

    how many modes or frequency range are you requesting? Maybe there are too many modes in the frequency range of your modal base. Also if you request too many results and result timesteps, it will require a lot of storage/scratch.

    Are you using AMSES (EIGRA) for eigen modes extraction? it should be much faster.

    User: "Luca_00"
    Altair Community Member
    OP

    Yes, I am using EIGRA for eigenmode extraction. I use a frequency range of 0 to 1000 Hz and a number of roots (ND) of 500. I have already varied this number down to ND = 50 without significant changes.
    The Global Output Request Card contains cstress (only principal stresses), stress, and displacements for specific sets of elements that I chose (combined output less than 200k elements). I also usually request significantly more time steps than 250, like in this model, with much shorter calculation times.

    In addition to the large temporary files, which often exceed my 1.5 TB disk space, the calculation takes much longer than usual. It takes up to 12 hours, whereas normal shocks without the use of carbon components take a maximum of ~2.5 hours.

    it is hard to tell. Maybe from the .stat file you could compare and narrow down what it the bottleneck for this model.

    .stat file show the overall time for each routine to be performed….

    But the only thing that came to y mind was having more modes calculated than the usual runs…