MOGA-DAKOTA inverse design.

[Mr. Tariq Khamlaj 0]

Hi All,

 

I have done an optimization study on ducted wind turbines using the MOGA in DAKOTA for two objective functions. Everything seems to work as expected. Now, I am working on another task where I have to pick a random design from the previous simulation and perform an inverse design for a verification purpose. After roughly two weeks, DAKOTA came up with nine designs (Pareto Front) and when I compare them with the one that I picked, they looked a little bit different. My expectation is that at least one or two should match the one that I picked. Is this an expected behavior in the MOGA if we have five design variables and two objective functions?

 

Any suggestions would be appreciated.

 

Thank you very much.

[Jörg 29]

Hi Tariq,

 

Actually, I don't really get it. You picked a random design from a previous simulation, from another task? And you compare it with a MOGA result?

[Mr. Tariq Khamlaj 0]

Hi Joerg,

 

Thank you very much for the reply. Sorry for the confusion! I picked a random design from the simulation of the ducted turbine that I did so that I can do an inverse design and verify that the MOGA comes up with an exact design. 

 

I hope that it is clear now!

 

Thanks again.

[Mr. Carsten Fuetterer 9]

Hello Tariq,

 

I think it is still not clear to us. Here is what I understand:

 

• you have a baseline geometry and perform a Moga study and you come up with a pareto front
• now you pick a design from this run (not on the pareto front) and perfom another MOGA with the same variable setup and come up with another pareto front
• now you are wondering why the designs of the two pareto fronts look different?

best regards

 

Carsten

[Mr. Tariq Khamlaj 0]

Hello Carsten, 

 

Thank you very much for the efforts that you made in order to explain what I was supposed to explain. Well done! I was expecting the MOGA to come up with the design but it didn't. My main purpose is to verify the MOGA by doing an inverse design. Is there any other way to verify it? I tried the algorithm on the FON function and it works fine but it didn't perform well when I did inverse design.

 

Thanks again for your time and help.

 

Kind regards,