CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FloEFD, FloWorks & FloTHERM (https://www.cfd-online.com/Forums/floefd-floworks-flotherm/)
-   -   CFD-driven modelling? (https://www.cfd-online.com/Forums/floefd-floworks-flotherm/27160-cfd-driven-modelling.html)

Bill Porter April 21, 2008 00:00

CFD-driven modelling?
 
Hello,

I saw a demonstration of Vanderplaats' Genesis software where the specific needs, materials, and constraints of a part were inputted, and the software then generated a model. The software optimized the shape to the specified criteria.

Is this possible with any CFD-related software? If so, which?

I'm designing a simple elbow for airflow to pass through two known flanges fixed in space and dimension. I am looking to optimize this elbow by having software tell me the geometry between the two flanges rather than have me take SWAG's over and over as I test them via CFD and crudely hone in on a design.

Thanks in advance for any help!

BP


Johannes April 27, 2008 16:31

Re: CFD-driven modelling?
 
Hi Bill, since years I am carrying the idea to do work about shape optimization by CFD with me. Unfortunately, I got no time to pursue it seriously, sofar. In fact I am considering to reopen this box this years as I got EFD on my desk.

Mechanical structure optimization seems to work a bit different from CFD structural optimization and is better established by standard procedures.

If it was just one (input) parameter which controls the shape of the body to be optimized (output parameter) and is responsible for the result (e.g. pressure drop), you could run a series of cloned projects and look for the optimum by plotting data in a x-y graph.

For multiparameter problems, I am fond of the work of Prof. Ingo Rechenberg in Berlin (http://en.wikipedia.org/wiki/Evolution_strategy) who e.g. optimized airfoil wings experimentally (analog, not digital) simply by throwing dice and tuned hardware control screws to modify the shape of a wing. He invented a series of evolutionary algorithms (ES) for mutation and selection. The experiments could and should be replaced by EFD calculations. For the first trial runs the input parameters for any ES algorithm should be modified and controlled manually before we think to let be done by an algorithm. I am very interested to pursue such an approach.


All times are GMT -4. The time now is 21:22.