## General Mechanical

#### Constraints for topology optimization

• spakerare
Subscriber

Hi everyone,

I am currently working with version 2019 R2 and I have following problem in a topology optimization for a plattform like shown. The middle hole is fixed while the ones in the corners are external displacement locked in z-direction.

I have several static structural systems connected and my objectives are to minimize compliance for each static structural system as well as a mass reduction. As a response constraint I used global von mises stress with 2,75e^7 Pa.

But when I add a validation system the maximum stress is something like 4e^6 Pa, which is compared to the constraint very low. Thats why I would like to increase this value, resulting in further mass reduction as well. Unfortunately I have no idea, which parameters to change for obtaining a better solution.

Mesh convergence, solution convergence are already checked.

Any help is much appriciated

best regards!

• Sandeep Medikonda
Ansys Employee

Isn't the constraint not to exceed the specified limit? If so, the optimization algorithm is achieving that...

In the topology optimization, the mean value of the element is used. So, take a look at the elemental mean value in the static structural analysis.

In your static structural analysis, how many load steps are you working with? If it is >1 then I would recommend looking at the stresses at all the intermediary load steps as well.

• spakerare
Subscriber

yes, it is achieving this but the maximum stress is much lower in the optimized topology. If I use the new topology and run another optimization loop I am achieving to reduce the mass of 30 more volume precent. Except the geomerty I changed nothing.

I work with 4 load steps to simulate a rotating system. so every loadstep has the same force, but with an offset of 90°.

I am not sure about this, but it seems like the solver calculates a mean stress of all load steps, which would be wrong since there is just one force per time step. any idea how to solve this?