frame analysis run is taking all night

[whatttup_G]

and still not done.. any suggestions on easing the workload? is this why i should have learned stand alone mode? p-mesh this first?


its a framework type assembly, and reducing the geometry, changing the shapes or suppressing features would really eliminate the worth of the results.. in other words, i dont think it can get much more simplified and still test what i need to see


suggestions from anyone with tons of time on mechanica to simplify this? its running me out of memory every time and like i said, leaving it overnight doesnt even get me to the p-loop passes.. i think i'll buy the rest of the RAM today and put myself upto 4 gigs, but any help in the mean time would be appreciated and honestly i dont know if memory alone will fix this


maybe theres enough here to choke a cray for a week, i dont know, but i'd like to not spend a month per run if possible


thanks in advance

 

[acook]

Hey whatttup_G,

Alot of things can cause this "problem" (I have seen) or it may be just the way things are. Firstly, what Pro/Engineer Wildfire version are you using? When I had my old workstation (1 Gb Ram, 1 GHz PIII Processor), Pro/Mechanica (WF 2.0 M050) on fairly complex geometries and loading schemes would go through all the passes until an edge order of about 6 or 7 and simply stop. I left it on there for a day and monitoring the disk and processor activity, realized that it was not workign at all. I should have looked for a bug report on this issue but I got a new workstation now (3 GB, 3 Ghz P4) and the problem does not exist.

I used to use standalone all of the time (mostly in the pre-Wildfire releases). Standalone mode will not help you all that much it seems from the description in your original post. Standalone mode (to most) simply forces the user to create a somewhat simplified model due to its more primative modeling nature than Pro/E. However, it does contain some FEA specific features that Integrated Pro/Mechanica mode does not. Here are some examples (this list is certainly non-inclusive) of features that, when present, draw computation times out (you may already know most of these).

1. Remove small fillets and rounds.

2. Small holes that lie close together can be hard to mesh between hole spacings.

3. Excessive curvatures.

4. Thin, hollow shells. (This may be where you are getting stuck, but it seems inevitable in your case, since eliminiating it would destroy your analysis)

5. Small holes (generally under 1/2 in. in diameter) have trouble meshing around the "rim" of the hole at the model surface.

The points above have been found by me meshing the model with AutoGen and using a custom mesh before running a multi-pass computation. Using AutoGen to produce a custom mesh,as opposed to a computer generated one, can prove very useful in finding meshing errors especially if you have any of the features listing above and cannot remove them. This will reduce the time until the results converge. Meshing takes alot of time (trial and error on some really complex models) but you usually arrive at a very favorable compromise. I have found out the worst thing you can do is enter a model analysis with a mesh that is computer generated and that contains errors. These results may take up to edge order 9 to complete (if at all).

If you want, shoot me a Private Message with your e-mail address and I will look for some PDFs in our company library concerning AutoGEN and Mechanica. They may be from Pro/Engineer Wildfire 1.0 but little has changed between the Pro/Mechanica versions (except for maybe the elimination of the Menu Manager after Wildfire 1.0).

For the record, I would recommend at least 2 GB of RAM for every workstation doing any type of Mechanica work as it is a memory hog no matter how simplified you get your model and it may stop working when the memory runs out (this may have been MY problem described above).

Hope this helps.




Edited by: acook

 

[whatttup_G]

thx for the reply


spoke with a guy that has run a ton of mechanica today, he immediately led me to a microsoft tech bulletin.. or maybe it was a ptc post.. regarding a 3GB switch that one adds to the boot.ini file when running XP, this basically unleashes some memory limits inherant without that switch employed, i'll spare the rehash here cuz its not infront of me... search for that subject on either site for tons of info


i'm on M150 and trying to run on full autopilot (as if that existed ) in integrated mode.. i've been trying to pound the model thru as is since i work all day and am moonlighting side work at night... the times on my posts are a big hint, its typically midnight before i get anywhere


anyways, found a few other things i should have known but didn't catch, the largest screwup on my part was although i left the rounds on the exterior of the frame members, they are assembled in a way that has the round facing a flat.. if you think of an asymptote like graph, i'm told this is how mechanica treats joints like that and it drives it nuts because of the infinate nature of the round approching the flats


i also thought that the line to line mates made at the junctions of the bars were a problem, but i'm told that is not bad, so long as there are no rounds doing what i described above


further though, my buddy says WF2 has a great fastener function that you can do for bolts or bolted type joints, and its killer... he was busy with some 64 bit testing with WF3 so i didnt get a chance to pursue that... i'll try to look at it tonite if i can stay awake


also there are some smallish holes that seem suspect based on your comments on mesh and autogem... i might suppress those just to see if it helps since you offered that advice up


thanks either way, i appreciate the help... if you find anything worth sending my way, post a short note here and i'll PM you my email... i've just rebooted and am itching to restart my run with all this new ammo


ttyl