Space Frame Optimization 1

[noahfox]

Hey Guys,

I am designing the chassis for my school's Formula SAE car. We use a welded steel tube space frame and in the past, the design has been done intuitively then manually optimized using FEA. I was wondering, since a space frame chassis is relatively simple to analyze and there are many rule constraints, if it would be reasonable to try and set up an optimization program of some sort to arrive at a closer-to-optimal design which could then be fine tuned using FEA. How complex would it be to set this up? Are there any useful resources to consult on the topic? Has anyone implemented something like this in the past?

Thanks for all your help.

 

[3DDave]

The first difficulty is determining what 'optimal' is.

Generally there is a single basis for optimizing tradeoff, such that increasing or decreasing the item leads to poorer performance. Weight is one of these in a stress related study = more material increases the inertia loads and stresses, less material is insufficient to manage the stresses. Fuel load causes increased range, but also adds weight and requires a larger, heavier fuel tank which decrease fuel economy, which tend to decrease range.

Some studies are just local minimum or maximum, such as minimizing cost. They are valuable, but aren't optimizing because they have only one side. Spending too little money causes failure, but more than the minimum is rarely detrimental.

One approach is genetic alternative generation. Where variations are randomly generated and the variants evaluated.

Check "Automated Design Using Darwinian Evolution and Genetic Programming" on YouTube. Judging the book by its cover, it looks to have a suitable content.

 

[GregLocock]

I have done it in the past using GA to select from a palette of tube sizes. I wrote it all in Matlab, based on OpenFEM. However I am told that genesis uses sizing optimization which may do the job

https://www.vrand.com/Genesis.html

Think carefully about your target function, as it will drive your design. For FSAE for example a high overall torsional stiffness for the frame (say 10000 lb ft/deg) is not necessary and will merely weigh more than a sensible design. If your old car was ok then you could use that as a target and see how much weight you can pull out.





Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?

 

[noahfox]

Thank you for your responses. From my research, a GA does seem like the way to go. To clarify my original question, I have a stiffness goal and need to optimize the chassis to meet that goal with minimum mass. I am considering trying to set up optimization using MATLAB to create designs and Ansys structural to analyze stiffness and mass. Have any of you all set up something like this before? Thank you for the advice.

 

[GregLocock]

Once upon a time I wrote a program in quickbasic that assembled the model data file from slabs of test with some text editing, and then called nastran to run it, and then qbasic picked up the results and looked at them. I strongly recommend not doing it that way.

The same job in matlab is easier.

However, why not hit altair for a copy of genesis? It'll take less time to learn a new interface than write a matlab textfile analyser.

Cheers

Greg Locock


New here? Try reading these, they might help FAQ731-376

http://eng-tips.com/market.cfm?