How can I make the analysis finish faster

[alkhal4]

Hello


Is there a way of performing structural analysis faster on ProE,


I have an assembly model of a heat exchanger that contains 168 tubes within a cylindrical pressure vessel, I have been trying to do some structural analyais to study the stresses.


I am new to ProE, when I run my analyais I use Single pass method, the problem here is that whenever I run the analysis the computer seem to work for hours and hours, I have the analysis running now for 10 hours, is this Normal?


Is there anyway of making this go faster?


If the computer crash does ProE pickup where it left off if I re-run the analysis?

Could anybody give me an advise!


I am using a computer with windows platform, 1 GB Ram. I need to really speed this up a little, it's really frustrating

especially when I get a fatal error message after 4 or 5 hours of run.


Thanks for reading

 

[walkerone]

Hi,

Can you start to run the analysis, then look at the summary and give some more information from there. It sounds like the analysis is too large for Mechanica to handle, so if we can at least check that the analysis isn't trying to figure out thousands of elements, etc

Post up summary and I'll try to give more advice.

 

[alkhal4]

Hello


I had to stop the analysis this morning after giving up,


I had a look at the summary file and it seems that Mechanica is trying to figure out many elements, even when I try to mesh the model it takes too long and doesn't complete, I have attached the analysis file that I stopped, I would appreciate any advice,





Thanks


2007-07-23_053000_model.zip

 

[walkerone]

I'd expect this analysis to run for an age, if it ever got through. You need to re-think the analysis that you are trying to run here - no FEA package that I've ever used would get through this in a short time period.


Am I right in assuming that the outer jacket is a shell, and the central heat exchange core has been modelled as a solid model? You need to break the problem down into smaller elements and then use Pro/Mechanicas built in features (symmetry, etc) to reduce that further into an analysis that runs successfully and still gives meaningful results.


If you are primarily interested in the heat aspects, I'd be looking at using a plane stress model through the cross-section of the exchanger initially. Then repeat for the end of the heat-exchanger at 90

 

[alkhal4]

Thanks for the advise,


You are correct, I tried to model the exchanger as a solid model,


But I also have an assembly model that does take time to process, I will try to do more reading about this area within the shot time I have,


I have a question,


If I scale my model down to a much smaller one and re-draw it would this make a difference?

 

[walkerone]

I hadn't really thought of that as an approach. I believe that the number of elements won't change, they'll just get proportionally smaller as the model is "shrunk".


Pro/Mechanica sizes the elements to suit the geometry i.e. a large plate with large features will have less elements than alarge plate with small features - you need more elements around the small features to get a good mesh.

 

[Kaz Z06]

alkhal4,


It appears your model can take advantage of 1/4 symmetry. This would certainly cut the mesh & solve time down significantly. If your loads and constraints are distributed evenly front-to-back and left-to-right (as they appear to be based on your image file), you can cut it in half from front to back and again from left to right. No matter how large or small your model is, symmetry is the first thing to look for before anything else.


Scaling your model wil not help as the elements are created based on the aspect ratio available in the volume/surface that you're meshing.


Good luck,


Kaz Z06

 

[alkhal4]

Thanks for having a look Kaz Z06,


I know how to create a cross sectionof the model using view managerbut I have never realy used this with any analysis before, the only thing that I am thinking about here is if I create a cross section then move to Mechanica mode, Apply the forces and run the analysis, is it going to run based on the cross section that I have or will it still take in consideration the reminder of the model?


I am trying to understand how to create the 1/4 symmetry to simplify the model, I will appreciate any quick tips.


Thanks

 

[pjw]

One way of decreasing run time in thermal simulation situations is to take advantage of known heat transfer equations and use them in conjunction with modifying geometry so that a coarser mesh might be applied.


For example, in the Fourier's law of heat conduction, we know that Q = -kA dt/dx.


If we thicken a given section by 10X (so that a coarser mesh can be used) and multiply the conduction coefficient by the same amount, then we know that the overall heat transfer (Q, Joules) will be the same in both cases. In more complex models, this might be a bit trickier to apply, but this is just an example of how one could go about improving run times.


As is always the case, ignore the factors that will have little impact on simulation results, i.e. convection heat transfer may be negligible in comparison to conduction heat transfer. Use shell elements where possible and one dimensional heat flow (where possible).


Hope this helps.


Phil