The doctoral thesis presents a generalized essential boundary condition sensitivity analysis based implementation of FE2 and mesh-in-element (MIEL) multi-scale methods and the application of these methods in multi-scale optimization algorithms. The implementation is derived as an alternative to standard implementations of multi-scale analysis, where the calculation of the Schur complement of the microscopic tangent matrix is needed to bridge different scales. The thesis presents a unified approach to the development of an arbitrary MIEL or FE2 computational scheme for an arbitrary path-dependent material model. Implementation is based on efficient first and second order analytical sensitivity analyses, for which an automatic-differentiation-based formulation (ADB) of essential boundary condition sensitivity analysis is derived. A fully consistently linearized two-level path-following algorithm is introduced as a solution algorithm for multi-scale modeling. Sensitivity analysis allows each macro step to be followed by an arbitrary number of intermediate micro steps while retaining quadratic convergence of the overall solution algorithm.
The implementation of multi-scale optimization algorithms is described, where a gradient-based optimization algorithm was wrapped around a multi-scale solution procedure. The versatility of sensitivity analysis for connecting scales and optimization purposes was proven. Examples using a developed optimization algorithm are presented. Through optimal distribution and opening size across the domain, fascinating mechanical properties can be achieved taking into account different optimization criteria. This process was used to design metamaterials for optimal energy dissipation.
|
