Abstract :
Dual optimization algorithms for the topology optimization of continuum structures in discrete variables are
gaining popularity in recent times since, in topology design problems, the number of constraints is small in
comparison to the number of design variables. Good topologies can be obtained for the minimum compliance
design problem when the perimeter constraint is imposed in addition to the volume constraint. However,
when the perimeter constraint is relaxed, the dual algorithm tends to give bad results, even with the use of
higher-order nite element models as we demonstrate in this work. Since, a priori, one does not know what
a good value of the perimeter to be speci ed is, it is essential to have an algorithm which generates good
topologies even in the absence of the perimeter constraint. We show how the dual algorithm can be made
more robust so that it yields good designs consistently in the absence of the perimeter constraint. In particular,
we show that the problem of checkerboarding which is frequently observed with the use of lower-order nite
elements is eliminated
