Load Cases vs Configurations

philh

Hi Victor. 

As a user of the Load Case feature in v6, which I liked a lot, I have just tried out v7 Configurations.  A v6 model that previously contained 7 separate load cases opened with 7 configuration tabs automatically in v7, so backward compatibility looked good.  However, now it appears each configuration, each containing a load case, has to be separately run.  Each run seems to re-solve the stiffness matrix and boundary conditions, so what took x time to run now appears to take 7x.  The model file (177000 elements, 51800 nodes), including results, is 50% larger and the time taken to simply open my model has gone from 1 minute to almost 10 minutes in v7 .  My computer became so sluggish after running the model that I may have been encountering some kind of memory leakage (if that's the right term!)  

It seems a shame that such as productive feature as load cases is be substituted by configurations which take so much longer to implement.  I had seen load cases as a great aspect of Mecway; one that was the envy of my ABAQUS colleagues.  Is it not possible to retain the capability of having a single solve of the stiffness and boundaries if there is no change in geometry?  Or retain the option to run load cases in a single run?

philh

Sorry, should have kept under the Version Beta string

Sergio

Is there a way to keep the results outiside the liml file maybe? I have the same problems back in v6 with long runs (no lineal analysis with too much increments and contacts), and yes, open big files can be a nightmare, even worst if is to modify the model and run again, where 10 min wasted.

Victor

Thanks for bringing up these problems. They include the issues that I'm most concerned about with the new configurations. Here are my thoughts:

File open time: I've seen a smaller increase in load time (50%) with a similar sized model. The 10x increase you have seems too high. Perhaps it now uses too much RAM and Windows is swapping things to disk. Do you mind checking Task Manager -> Performance to see if there's a lot of page file usage compared to v6? This would typically continue after closing Mecway and make the whole computer slow down for a while until other applications' data is loaded back into RAM.

Solving one by one: I'll try to add a button in v7 to solve all configurations with a single click. It will still be done sequentially and you won't be able to use the modeler while it's running but at least you can do something else without coming back every few minutes to press solve again.

File size and opening time: For version 8, I hope to make field variables optional so you can exclude ones you don't need and to allow more new ones. Enabling this is one of the reasons for removing load cases, since they had complexity that was making it more difficult.

Solving time: Have the solver intelligently decide if several configurations can be solved simultaneously (only differ in their loads) and solve them the same way load cases were. This is more remote and depends on how many people complain.

Sergio

And could be possible to keep load cases inside a configuration? I meant, in NX there is the possibility of with the same mesh/materials, create different load cases (that are solved in one time as the only difference for each load case are the loads). Otherwise if you want to change the boundary conditions, or the kind of analysis (modal, no lineal...), one can create a new Analysis that will be solved separately. I'm not sure if CCX allow solving load cases without recomputing the stiffness matrix

Another interesting option playing with lineal load cases, is the possibility of create new results combining several cases, without using the solver, just working with the results.

Regards

Sergio

If you add the possibility of selecting variables to reduce the results size, can you add also some options to save results every x number of iterations for no lineal analysis? That's a good way not only to reduce the size, but also to make it easier to work with the timeline.

Victor

I won't be putting load cases back in together with configurations. That's too much conceptual complexity. As for combining linear results in postprocessing, that's part of what this change is leading towards - being able to define a new field variable which is a mathematical function of other ones.

Here's what I hope is the final version 7. It has a Tools -> Solve all configurations menu item to make that a little easier.

http://mecway.com/download/mecway70rc.msi

philh

Sorry for delay responding to query re paging to file.  I was away. 

Paging to file isn't significant; just the occasional flicker.  But I think running Mecway on this computer in a client office (Windows 7 Enterprise 64bit with only 4MB RAM) is probably the problem.  Typically the paging is at about 450MB with V7 but this doesn't appear to be different to V6.  However, almost all RAM is consumed and remains consumed by Mecway after running is complete.  Background RAM is at about 32% with Mecway unloaded, 85%-90% with a Mecway model loaded but doing nothing.  Is that right that the model is held in RAM when it is not doing anything?  I have noticed that after running Mecway (and the run is finished), my memory usage keeps climbing.  I have no other applications open other than Mecway and Acrobat (to keep transferring my licence key between v6 and v7).  Opening or running any other application with Mecway open is a non-starter; just too slow.

Not sure if of any help.  I think the slow opening and slow running I'm experiencing is probably a function of this PC not being up to the task rather than an issue with Mecway per se.  See attached graphic of my little time and motion study with task manager.

Notwithstanding, the need for each config requiring it's own run time (albeit with a single click)  instead of all load cases run together makes Mecway analysis a slower process for me now.  I'm drinking a lot more coffee!

Regards

Victor

4GB is a bit low but it should work perfectly well with small models. Your graph shows that almost all the RAM is used simply by opening Mecway, without any model at all. I see approximately 0 memory use when doing that, so something's abnormal there. Does the whole computer slowing down happen even when you don't open a model? If not, then the graph is misleading and the apparent memory use is harmless. If so, then there seems to be a bug.

You can activate both versions at the same time on the same computer.

philh

Hi Victor. 

The computer only slows once Mecway has a model loaded. I'll just close Mecway down to use other applications. 

Thanks