Dual approach using a variant perimeter constraint and efficient sub-iteration scheme for topology optimization

To prevent numerical instabilities associated with the mesh-dependence, checkerboards and grey regions in topology optimization, a variant perimeter-constrained version of the SIMP algorithm is proposed using a smooth and quadratic function. In order to have an efficient implementation and to make sure the strict satisfaction of such an upper-bound perimeter constraint, a diagonal quadratic approximation of the perimeter constraint is used in the construction of each explicit optimization subproblem. The latter is then solved by a dual sub-iteration scheme. Numerical results show that the incorporation of such a sub-iteration scheme leads to a convergent solution without needs of move-limits or artificial control parameters. In addition to this, it is found that successive relaxations of the perimeter constraint by increasing the upper-bound tend to regularize the topology solution and result in a checkerboard free and satisfactory design solution without grey regions.