Optimization of Structure Other Than Panels

[dlj2_msfc]

I know that HS is great for panel structures, but I am trying to use it to optimize truss configurations.  The basic difference, that I've seen, within HS is the use of statistical average loads.  While I understand the necessity of using average loads for panels, the same is not true for the bar elements of a truss.  Is there a way to use the peak loads imported in from the FEM to size for beam buckling instead of the average loads?

[Phil]

First of all, you can tell HyperSizer to perform strength and local buckling analyses for "peak" loads by choosing "Element Based" from the FEA Statistical Loading Method selection on the Design-To Loads tab of the sizing form.  This will tell HyperSizer to do an analysis for every element and return the lowest margin of safety for any element and therefore capturing the "peak" load.

For Beam Buckling, that is, buckling of the entire beam as with Euler column buckling, HyperSizer will use the average compressive load, regardless of the selection of statistical loading method.  This still seems to be appropriate as this type of failure mode, column buckling, would not be susceptible to load peaks but rather is an average effect for load in the overall column.

For more localized buckling effects, such as buckling of the web of a beam, or for circular tube beams, cylindrical buckling of the wall, these are considered  local buckling effects and would actually use the same loads as those used for strength.   So if you turned on element based, then the local buckling analyses will be done element by element.

Are these the kind of buckling effects that you are considering when you talk about beam buckling? 

As of now, there is no way to tell HyperSizer to use peak element loads for the overall column buckling.

[dlj2_msfc]

Thanks for the reply.  The issue I am having seems to be within the Euler Buckling of the beam.  I have attached a couple of pictures to illustrate what I am running into.

What you see in slide 1 is the axial force of one of the properties of beams I have in my model.  Each of the eight beams has eight elements per beam.  One thing to note is that any one beam in this property must able to withstand the worst case load because of the loading scenario.  So all eight must be identical, hence the logic to make them all the same property.  From what I can tell, HS takes the average of all of the compressive forces across the property, not just the compressive forces in one beam. 

Slide 2 is just the HS Design-To Loads tab which shows the correspondence to the FEM when I ran it with the Element Based option.

The solution I see is to break out the individual beams into individual properties in the FEM, instead of one property for all eight beams.  Then HS would read each beam as a component and size it for element based loads.  Care would have to be taken to make sure that all eight members retain the sizing of the worst case component.  With close to 270 beams in my truss structure, that makes for a lot of properties in the FEM, but this seems to be the only logical solution.  Any other advice would be helpful.

[Phil]

I believe you hit the solution right on the head.  In your earlier post where you talked about averaging, I was thinking you were talking about averaging the loads from one of the beams (say 8 elements) and I think it is correct to do that averaging.

However, in your case, it would be much more appropriate to split those 8 beams into 8 different components.  Then create a group and put all 8 components into the group.  Now, select the button on the Variables tab that says "Link All Variables".  This will cause HyperSizer to size all eight beams to the same exact design (all variables and materials) and will ensure that all 8 components have positive margins.  Some of these components will have relatively high margins and some components should be close to zero margins.

If you are going this route, probably no reason to turn on the "Element Based" analysis on the Design-To Loads Tab.  Because if each component is only one of your beams (composed of eight elements), it now definitely seems like average loading is the way to go.