cancel
Showing results for 
Search instead for 
Did you mean: 
cancel
Showing results for 
Search instead for 
Did you mean: 

Optimization accuracy / sensitivity

Highlighted
Newbie

Optimization accuracy / sensitivity

Hi all,

I've basically got an assembly with a piston, crank and conrod and am trying to run an optimisation to maximise / minimise the volume using my component tolerances as my variables.

I have used copy geometries to put together a solid component which I am using as my volume to optimize.

My issue comes that when one of my variables has a limit of less than +/-0.1 I get a message saying that the model is already optimized. Also - if I try and optimize from the nominal value I get the same message, but if I start at the upper or lower limit it will work?

All my accuracies are set to absolute 0.001. Any ideas?

Thanks

Alex

3 REPLIES 3
Highlighted

Re: Optimization accuracy / sensitivity

First of all, and you probably already did this, you need to apply your Creo Simulate dimension variations to the nominal dimension, not the tolerance parameter.

The optimization accuracy will need to be set to a smaller value; this is changed in the tolerance design study definition form under "Options", see picture.

Recall that your model's mesh/convergence needs to be in at least as good as your optimization tolerance. So if you require 0.1% optimization tolerance, then your model should calculate stresses etc. with at least that accuracy. This is a very tough requirement. Try doing a global sensitivity study to verify that your small design changes produce a change in your goal function that is greater than the analysis accuracy. If so, then I guess it is meaningful to move on with an optimization.

Capture.PNG

Highlighted

Re: Optimization accuracy / sensitivity

Thanks for your reply,

Apologies I don't think I was clear on my original post. So I am using the optimize/feasibility function within creo, outside of simulate to optimize for volume rather than calculate any stresses etc.

I have my convergence set to 0.001% - as you suggested I measured the % change in volume between my nominal and max / nominal and min and it is within the tolerance of my convergence accuracy.

I made a dummy part and it seems that once the model size / volumes get to a certain size the solver does not iterate?

Does this make sense?

Thanks

Alex

Highlighted

Re: Optimization accuracy / sensitivity

Alex, in this case you do not have any contradicting requirements, so the optimizer will (I guess) simply reduce your d3 dimension to the minimum (or max if it's a feature that cuts away material) in one step.

In this simple case I would make a sensitivity study (I think its called "feasibiily study" in BMX) and plot your volume vs d3.

For the optimizer to make sense, you need some contradicting requirements, for example minimizing mass(or volume) while at the same time keeping some other variable within a specified limit, a "design constraint". Otherwise the optimization becomes trivial, "as thin as possible" i.e. the minimum of your specified interval, 135.575.

Announcements