Optimization with stress constraint?

For topology optimization the stress constraint is considered to be a global constraint that only takes into account elements with relative density equal to one. When elements have densities lower than one, the constraint is turned off for those elements. This is because you cannot calculate the stress for elements that have had their local stiffness matrices penalized (SIMP). If a a single stress constraint is imposed on your model, the end result may have only a few elements that have a relative density equal to one. Because of the nature of gradient descent methods used in hypermesh/optistruct, this will lead you to poor results.

If you want to do a stress constrained optimization with mass minimization you will need to consider a different optimization algorithm. Stress constraint in combination with other constraints such as displacements, volume fraction or member size control is a suggested practice (in hypermesh).

Hope this helps,

-Eric