Computational Design of Microvascular Biomimetic Materials

185 p.Thesis (Ph.D.)--University of Illinois at Urbana-Champaign, 2010.In this work, we propose to design these materials using Genetic Algorithms (GAs), the most common methodology within a broader category of Evolutionary Algorithms (EAs). GAs can be combined with a Pareto-selection mechanism to create Multi-Objective Genetic Algorithms (MOGAs), which enable the optimization of an arbitrary number of objective functions. As a result, a Pareto-optimal front is obtained, where all candidates are optimal solutions to the optimization problem. Adding a procedure to deal with constraints results in a powerful tool for multi-objective constrained optimization. The method allows the use of discrete variable problems and it does not require any a priori knowledge of the optimal solution. Furthermore, GAs search the entire decision space so the optimal solutions found are likely to be global. TheMOGAoptimization framework is also combined with a physical solver based on advanced finite element methods to study the thermal behavior of these materials. Because the MOGA requires a vast number of individual evaluations, emphasis is placed on the computational efficiency of the solver. Thus, a simplified formulation is used to take into account the cooling effect of the fluid, instead of solving the conjugate heat transfer problem for obtaining the temperature field in both solid and fluid domains. The Generalized Finite Element Method (GFEM) is adopted because accurate finite element approximations of the temperature field can be obtained on finite element meshes that are independent of the geometry of the embedded network. Numerical experiments of multi-physics optimization involving flow efficiency, void volume fraction and thermal control are presented. Results show that the tradeoffs between conflicting objectives is well captured so that the optimal design is readily available to the analyst.U of I OnlyRestricted to the U of I community idenfinitely during batch ingest of legacy ETD