In this technical publication (link) by Marshall Space Flight Center of NASA, topology optimization for compliant mechanisms is studied. Genetic Algorithm based optimization is used, also featuring subdivision to smooth the geometry, also geometry relaxation, so that the mesh is not constant but changes shape. There are many things not clear to me, I would be happy if a knowledgeable reader can enlighten us in the comments or by e-mail. I am not questioning the usefulness of this research, it is for sure one step for the humanity, even if it may not have immediate practical applications. Why not use higher resolution but subdivision and geometry relaxation? How is the result better from the one that TopOpt mechanism design module which is a free web based application, can produce? And it is stated that the results for a 47×24 design space were obtained by 5 months of continuous calculations, where as TopOpt gives results in minutes or seconds. Is it because as the article states:
“Although many researchers are using density relaxation, it is well known that the relaxed density parameter CM design problems do not lead to useful designs”
I will research this last statement to understand the differences.