Time for Linear Static Analysis 1

[JeremySh]

Hi everyone, I've just started using eng-tip and I like to say it is one of the best engineering forums about!

As I'm sitting here waiting for my analysis to complete I was thinking has anyone correlated average times for analysis to complete vs. number of nodes, elements and DOF's. I'm fairly new to complex assembly analysis. 6hrs in and no indication of how much longer to go. I have quite a few models to run through so I was thinking I might set up a dedicated FEA PC I have lying around as I'm using my laptop at the moment and it's not to good for portability.

I'm not running an awesome system (Laptop) and the desktop is fairly similar:
- Core 2 Duo @ 2GHz
- 4GB RAM
- Windows 7 Build 7077 (is nice and stable at the moment)
- Nvidia 8400GM
- Solidworks 09 with Cosmos Advanced Pro.

Anyone got any thoughts, ideas, suggestions???

 

[TheTick]

There is a button on the control window that comes up that will show a convergence plot. Keep an eye on it to see if your analysis is in fact converging.

I was taught years ago to start with a coarse mesh to be sure the setup is working, then refine the mesh as needed.

 

[GregLocock]

In theory solve time will increase roughly proportional to the number of nodes, squared, for a well connected structure.

6 hours seems like a lot, check and see if the model is all in RAM, if not fit more RAM.

My downmarket laptop runs 100000 node problems in about 3 hours, as I have enough RAM to run them in memory.

If you do not hav enough RAM your computer will spend the entire time talking to the disk and this can increase execution time by a factor of 8.







Cheers

Greg Locock

SIGlease see FAQ731-376 for tips on how to make the best use of Eng-Tips.

 

[JeremySh]

Thanks for the replies so far, I've added another Solid state harddrive for page file memory so hopefully that speeds up the analysis a bit. I have a feeling the connections I'm using aren't completely constraining the parts.

This leads to another question will the bond connection between 2 surfaces of 2 parts, (1 part be fully fixed) completely constrain the parts?

 

[JeremySh]

Just incase anyone was following this. The reason I was running out of memory and analysis taking that long was the simultaneous equations weren't being solved i.e. Not converging. The analysis is happening in around 3.5hrs now and I'm getting results that are reasonable.

Thanks for all the comments.

 

[TheTick]

Wen I used MSC NASTRAN, I had a second hard rive installed to handle pagefile and also NASTRAN's process files. It helped a lot.

I don't think COSMOS lets you direct where temp files go, so the effect probably isn't as great.