Evolutionary topology redesign for performance by large admissible perturbations.

An evolutionary topology design methodology for performance is developed and implemented in this research. The performance constraints are treated by the method of LargE Admissible Perturbations (LEAP) so that finite element calculations are reduced dramatically. An evolutionary algorithm is developed to solve the uni-material topology redesign problem, where material properties are limited to voids or a specified Young's modulus Es. Most structures in reality are not homogeneous and often composed of different materials. In this dissertation, a second algorithm is developed for the bi-material topology redesign problem, where material properties are limited to voids, an original Young's modulus E0, or E u. The evolved topology with such material specifications may not be so difficult to manufacture. The desired topology/material is achieved in 4--12 iterations for changes in performance by a factor of 2. Benchmark applications show the capability of the methodology for handling multiple performance constraints including static displacements, natural frequencies, static stresses and forced response amplitudes. Further more, 3D problems are solved with static, modal dynamic, and simultaneous static and modal dynamic constraints, which shows that the developed methodology can be easily applied to 3D problems to achieve large topology changes within a small number of iterations.Ph.D.Applied SciencesOcean engineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/124269/2/3137901.pd