Using DRESP2 and deequation for multiobjective optimization in OptiStruct

Jiaxin ZHOU

Hi, everyone,

I am using optistruct to do multiobjective topology optimization.

The responses are assiged in dresp2 and their relationship is defined using /DEQATN.

However, the optimized results always have uniform density in design area.

(FIY: the response is the shear compliance calculated from bulk cube with PBC)

Is there any solution for this problem?

I am looking forward to your suggestion.

Best regards,

Jiaxin ZHOU

 

 

Adriano_Koga

hi @Jiaxin ZHOU 

unfortunately it is hard to tell only from your description. Maybe adding a few pictures and also attaching the .out file might help others to comment better.

additionally, if this is a multiobjective function, I'd rather start optimizing for its individual objectives, and make sure they make sense, and later on, combine it into a multiobjective. If you're using some weighting factors for your equation, maybe starting with 0.0 weighting for one of the responses, and test it. then, increase the weighting factor a little bit and try it again.

Also in multiobjective optimization you may need to make sure that the responses have similar numerical magnitude, so that one objective does not overcomes the other. If this is not the case, you could normalize their values, or just use the objective reference entity in order to get some normalized values.