This thesis consists of an introduction and sevenindependent, but closely related, papers which all deal withproblems in structural optimization. In particular, we considermodels and methods for global optimization of problems intopology design of discrete and continuum structures. In the first four papers of the thesis the nonconvex problemof minimizing the weight of a truss structure subject to stressconstraints is considered. First itis shown that a certainsubclass of these problems can equivalently be cast as linearprograms and thus efficiently solved to global optimality.Thereafter, the behavior of a certain well-known perturbationtechnique is studied. It is concluded that, in practice, thistechnique can not guarantee that a global minimizer is found.Finally, a convergent continuous branch-and-bound method forglobal optimization of minimum weight problems with stress,displacement, and local buckling constraints is developed.Using this method, several problems taken from the literatureare solved with a proof of global optimality for the firsttime. The last three papers of the thesis deal with topologyoptimization of discretized continuum structures. Theseproblems are usually modeled as mixed or pure nonlinear 0-1programs. First, the behavior of certain often usedpenalization methods for minimum compliance problems isstudied. It is concluded that these methods may fail to producea zero-one solution to the considered problem. To remedy this,a material interpolation scheme based on a rational functionsuch that compli- ance becomes a concave function is proposed.Finally, it is shown that a broad range of nonlinear 0-1topology optimization problems, including stress- anddisplacement-constrained minimum weight problems, canequivalently be modeled as linear mixed 0-1 programs. Thisresult implies that any of the standard methods available forgeneral linear integer programming can now be used on topologyoptimization problems. <b>Keywords:</b>topology optimization, global optimization,stress constraints, linear programming, mixed integerprogramming, branch-and-bound.

<p>This thesis consists of an introduction and sevenindependent, but closely related, papers which all deal withproblems in structural optimization. In particular, we considermodels and methods for global optimization of problems intopology design of discrete and continuum structures.</p><p>In the first four papers of the thesis the nonconvex problemof minimizing the weight of a truss structure subject to stressconstraints is considered. First itis shown that a certainsubclass of these problems can equivalently be cast as linearprograms and thus efficiently solved to global optimality.Thereafter, the behavior of a certain well-known perturbationtechnique is studied. It is concluded that, in practice, thistechnique can not guarantee that a global minimizer is found.Finally, a convergent continuous branch-and-bound method forglobal optimization of minimum weight problems with stress,displacement, and local buckling constraints is developed.Using this method, several problems taken from the literatureare solved with a proof of global optimality for the firsttime.</p><p>The last three papers of the thesis deal with topologyoptimization of discretized continuum structures. Theseproblems are usually modeled as mixed or pure nonlinear 0-1programs. First, the behavior of certain often usedpenalization methods for minimum compliance problems isstudied. It is concluded that these methods may fail to producea zero-one solution to the considered problem. To remedy this,a material interpolation scheme based on a rational functionsuch that compli- ance becomes a concave function is proposed.Finally, it is shown that a broad range of nonlinear 0-1topology optimization problems, including stress- anddisplacement-constrained minimum weight problems, canequivalently be modeled as linear mixed 0-1 programs. Thisresult implies that any of the standard methods available forgeneral linear integer programming can now be used on topologyoptimization problems.</p><p><b>Keywords:</b>topology optimization, global optimization,stress constraints, linear programming, mixed integerprogramming, branch-and-bound.</p>

Fatigue, or failure of material due to repeated cyclic loading, is one of the most common causes of mechanical failures. The risk of fatigue in a load carrying component is often lowered by adding material, thereby reducing stresses. This increases the component weight, reducing the performance of the component and increasing its manufacturing cost. There is thus a need to design components to be as light as possible, while keeping the risk of fatigue at a low enough level, i.e. there is a need for optimization of the component subject to fatigue constraints.  This thesis deals with design against fatigue using topology optimization, which is a form of structural optimization where an optimal design is sought by using mathematical programming to decide which parts of a design domain should be filled with material, and which should not.  To predict fatigue, accurate representation of the geometry and accurate stress computation are of utmost importance. In this thesis, methods for imposing constraints such as minimum inner radii and minimum member sizes in the form of four new density filters are proposed. The filters are able to generate a very sharp representation of the structural boundary. A method for improving the accuracy of stress results at the structural boundary is also proposed, based on extrapolation of results from the interior of the structure. The method gives more accurate stresses, which affects the resulting structures when solving optimization problems.  A formulation for fatigue constraints in topology optimization is proposed, based on the weakest link integral. The formulation avoids the problem of choosing between accurate but costly local constraints, and efficient but approximate aggregated constraints, and gives a theoretical motivation for using expressions similar to the p-norm of stresses.  For verifying calculations of the fatigue probability of an optimized structure, critical plane criteria are commonly used. A new method for evaluating such criteria using optimization methods is proposed, and is proved to give results within a user given error tolerance. It is shown that compared to existing brute force methods, the proposed method evaluates significantly fewer planes in the search of the critical one. / <p>QC 20150504</p>

Topology optimization is commonly used to distribute a given amount of material to obtain the stiffest structure, with predefined fixed loads. The present work investigates the result of applying stress constraints to topology optimization, for problems with design-depending loading, such as self-weight and pressure. In order to apply pressure loading, a material boundary identification scheme is proposed, iteratively connecting points of equal density. In previous research, design-dependent loading problems have been limited to compliance minimization. The present study employs a more practical approach by minimizing mass subject to failure constraints, and uses a stress relaxation technique to avoid stress constraint singularities. The results show that these design dependent loading problems may converge to a local minimum when stress constraints are enforced. Comparisons between compliance minimization solutions and stress-constrained solutions are also given. The resulting topologies of these two solutions are usually vastly different, demonstrating the need for stress-constrained topology optimization.

Topology optimization is commonly used to distribute a given amount of material to obtain the stiffest structure, with predefined fixed loads. The present work investigates the result of applying stress constraints to topology optimization, for problems with design-depending loading, such as self-weight and pressure. In order to apply pressure loading, a material boundary identification scheme is proposed, iteratively connecting points of equal density. In previous research, design-dependent loading problems have been limited to compliance minimization. The present study employs a more practical approach by minimizing mass subject to failure constraints, and uses a stress relaxation technique to avoid stress constraint singularities. The results show that these design dependent loading problems may converge to a local minimum when stress constraints are enforced. Comparisons between compliance minimization solutions and stress-constrained solutions are also given. The resulting topologies of these two solutions are usually vastly different, demonstrating the need for stress-constrained topology optimization.

No presente trabalho, foi implementada uma formulação de otimização topológica com o objetivo de encontrar o mínimo volume de estruturas contínuas bidimensionais, em estado plano de tensão, sujeitas à restrição de tensão de von Mises. Foi utilizado o Método dos Elementos Finitos para discretizar o domínio, com o elemento não conforme de Taylor. A tensão foi suavizada, calculando-se um valor de tensão para cada nó do elemento. O fenômeno da singularidade foi contornado através do método de relaxação da tensão, penalizando-se o tensor constitutivo. Foi usada uma única medida de tensão global, a normap, resultando na redução do custo computacional do cálculo das sensibilidades. As sensibilidades da função objetivo e da restrição de tensão foram calculadas analiticamente. O problema de otimização topológica foi resolvido por um algoritmo de Programação Linear Sequencial. Os fenômenos da instabilidade de tabuleiro e da dependência da malha foram contornados pela utilização de um filtro de densidade linear. A formulação desenvolvida foi testada em 3 casos clássicos. No primeiro deles, foi testada uma viga curta em balanço, submetida a 3 diferentes tipos de penalização da função objetivo, obtendo-se uma estrutura com 27% do volume inicial, com reduzido número de elementos com densidades intermediárias. No segundo caso, foi testada a mesma estrutura submetida à flexão, chegandose a uma topologia bem definida no formato de duas barras, com 16,25% do volume inicial. No terceiro caso, em que foi utilizado um componente estrutural em formato de “L”, justamente por favorecer o surgimento de concentração de tensão em sua quina interna, o otimizador gerou uma estrutura bem definida, permanecendo, contudo, uma pequena região de concentração de tensão na topologia final. / A topology optimization formulation to search for the minimum volume of twodimensional linear elastic continuous structures in plane stress, subject to a von Mises stress constraint, was implemented in this study. The extended domain was discretized using Taylor nonconforming finite element. Nodal values of the stress tensor field were computed by global smoothing. A penalized constitutive tensor stress relaxation method bypassed the stress singularity problem. A single p-norm global stress measure was used to speed up the sensitivity analysis. The sensitivities of the objective function and stress constraints were derived analytically. The topology optimization problem was solved by a Sequential Linear Programming algorithm. A linear density filter avoided the checkerboard and the mesh dependence phenomena. The formulation was tested with three benchmark cases. In the first case, a tip loaded short cantilever beam was optimized using a sequence of three different objective function penalizations. The converged design had approximately 27% of the initial volume, with a small proportion of intermediate densities areas. In the second case, the same domain was subjected to shear, resulting a well defined two-bar design, with 16.25% of the initial volume. In the third case, an L-shape structure was studied, because it has a stress concentration at the reentrant corner. In this last case, the final topology was well-defined, but the stress concentration was not completely removed.

No presente trabalho, foi implementada uma formulação de otimização topológica com o objetivo de encontrar o mínimo volume de estruturas contínuas bidimensionais, em estado plano de tensão, sujeitas à restrição de tensão de von Mises. Foi utilizado o Método dos Elementos Finitos para discretizar o domínio, com o elemento não conforme de Taylor. A tensão foi suavizada, calculando-se um valor de tensão para cada nó do elemento. O fenômeno da singularidade foi contornado através do método de relaxação da tensão, penalizando-se o tensor constitutivo. Foi usada uma única medida de tensão global, a normap, resultando na redução do custo computacional do cálculo das sensibilidades. As sensibilidades da função objetivo e da restrição de tensão foram calculadas analiticamente. O problema de otimização topológica foi resolvido por um algoritmo de Programação Linear Sequencial. Os fenômenos da instabilidade de tabuleiro e da dependência da malha foram contornados pela utilização de um filtro de densidade linear. A formulação desenvolvida foi testada em 3 casos clássicos. No primeiro deles, foi testada uma viga curta em balanço, submetida a 3 diferentes tipos de penalização da função objetivo, obtendo-se uma estrutura com 27% do volume inicial, com reduzido número de elementos com densidades intermediárias. No segundo caso, foi testada a mesma estrutura submetida à flexão, chegandose a uma topologia bem definida no formato de duas barras, com 16,25% do volume inicial. No terceiro caso, em que foi utilizado um componente estrutural em formato de “L”, justamente por favorecer o surgimento de concentração de tensão em sua quina interna, o otimizador gerou uma estrutura bem definida, permanecendo, contudo, uma pequena região de concentração de tensão na topologia final. / A topology optimization formulation to search for the minimum volume of twodimensional linear elastic continuous structures in plane stress, subject to a von Mises stress constraint, was implemented in this study. The extended domain was discretized using Taylor nonconforming finite element. Nodal values of the stress tensor field were computed by global smoothing. A penalized constitutive tensor stress relaxation method bypassed the stress singularity problem. A single p-norm global stress measure was used to speed up the sensitivity analysis. The sensitivities of the objective function and stress constraints were derived analytically. The topology optimization problem was solved by a Sequential Linear Programming algorithm. A linear density filter avoided the checkerboard and the mesh dependence phenomena. The formulation was tested with three benchmark cases. In the first case, a tip loaded short cantilever beam was optimized using a sequence of three different objective function penalizations. The converged design had approximately 27% of the initial volume, with a small proportion of intermediate densities areas. In the second case, the same domain was subjected to shear, resulting a well defined two-bar design, with 16.25% of the initial volume. In the third case, an L-shape structure was studied, because it has a stress concentration at the reentrant corner. In this last case, the final topology was well-defined, but the stress concentration was not completely removed.

No presente trabalho, foi implementada uma formulação de otimização topológica com o objetivo de encontrar o mínimo volume de estruturas contínuas bidimensionais, em estado plano de tensão, sujeitas à restrição de tensão de von Mises. Foi utilizado o Método dos Elementos Finitos para discretizar o domínio, com o elemento não conforme de Taylor. A tensão foi suavizada, calculando-se um valor de tensão para cada nó do elemento. O fenômeno da singularidade foi contornado através do método de relaxação da tensão, penalizando-se o tensor constitutivo. Foi usada uma única medida de tensão global, a normap, resultando na redução do custo computacional do cálculo das sensibilidades. As sensibilidades da função objetivo e da restrição de tensão foram calculadas analiticamente. O problema de otimização topológica foi resolvido por um algoritmo de Programação Linear Sequencial. Os fenômenos da instabilidade de tabuleiro e da dependência da malha foram contornados pela utilização de um filtro de densidade linear. A formulação desenvolvida foi testada em 3 casos clássicos. No primeiro deles, foi testada uma viga curta em balanço, submetida a 3 diferentes tipos de penalização da função objetivo, obtendo-se uma estrutura com 27% do volume inicial, com reduzido número de elementos com densidades intermediárias. No segundo caso, foi testada a mesma estrutura submetida à flexão, chegandose a uma topologia bem definida no formato de duas barras, com 16,25% do volume inicial. No terceiro caso, em que foi utilizado um componente estrutural em formato de “L”, justamente por favorecer o surgimento de concentração de tensão em sua quina interna, o otimizador gerou uma estrutura bem definida, permanecendo, contudo, uma pequena região de concentração de tensão na topologia final. / A topology optimization formulation to search for the minimum volume of twodimensional linear elastic continuous structures in plane stress, subject to a von Mises stress constraint, was implemented in this study. The extended domain was discretized using Taylor nonconforming finite element. Nodal values of the stress tensor field were computed by global smoothing. A penalized constitutive tensor stress relaxation method bypassed the stress singularity problem. A single p-norm global stress measure was used to speed up the sensitivity analysis. The sensitivities of the objective function and stress constraints were derived analytically. The topology optimization problem was solved by a Sequential Linear Programming algorithm. A linear density filter avoided the checkerboard and the mesh dependence phenomena. The formulation was tested with three benchmark cases. In the first case, a tip loaded short cantilever beam was optimized using a sequence of three different objective function penalizations. The converged design had approximately 27% of the initial volume, with a small proportion of intermediate densities areas. In the second case, the same domain was subjected to shear, resulting a well defined two-bar design, with 16.25% of the initial volume. In the third case, an L-shape structure was studied, because it has a stress concentration at the reentrant corner. In this last case, the final topology was well-defined, but the stress concentration was not completely removed.

Additive manufacturing (AM) is a particularly useful manufacturing method for components designed using topology optimization (TO) since it allows for a greater part complexity than any traditional manufacturing method. However, the AM process potentially leads to anisotropic material properties due to the layer-by-layer buildup of parts and the fast and directional cooling. For Ti6Al4V tensile specimens built using electron beam melting (EBM), it has been observed that flat built specimens show superior strength and elastic moduli compared to top built specimens. Designs with the loading direction parallel to the build layers are therefore expected to show greater reliability. In this thesis a procedure is developed to optimize the AM build orientation considering anisotropic elastic material properties. A transversely isotropic material model is used to represent the in-plane and out-of-plane characteristics of AM produced parts. Two additional design variables are added to the TO formulation in order to control the orientation of the material using a coordinate transformation. Sensitivity analysis for the material direction variables is conducted for compliance as well as maximum von-Mises stress using a -norm stress aggregation function. The procedures for the AM build orientation optimization and stress constraints are implemented in the finite element software TRINITAS and evaluated using a number of examples in 2D and 3D. It is found that the procedure works well for compliance as well as stress but that a combination of these may lead to convergence issues due to contradicting optimal material orientations. An evaluation of the -norm stress aggregation function showed that a single global stress measure in combination with a stress correction procedure works well for most problems given that the mesh is refined enough to resolve the stresses accurately.

Thesis (MEng)--Stellenbosch University, 2014. / ENGLISH ABSTRACT: We investigate the simultaneous analysis and design (SAND) formulation of the topology optimisation problem. The characteristics of the formulation are presented considering the simple compliance/weight constrained problem and the more complex local stress constrained case. The problems are solved in an efficient sparse sequential approximate optimisation (SAO) framework with the SAND formulation showing an significant reduction in computational requirements compared to the traditional and inherently expensive nested analysis and design (NAND) approach. In SAND the state equations are included in the optimisation problem as a set of equality constraints and not solved exactly in each iteration, as would be the case in NAND. Decision and state variables are thus independent, resulting in an immensely sparse optimisation problem. The availability of simple exact analytic expressions for all the constraint functions (via the finite element method) allows for the construction of accurate approximate subproblems with little computational effort. Furthermore, material can be removed completely from the design domain with few complications, resulting in a decrease in subproblem size as the algorithm progresses, further reducing computation time. The inclusion of void material in the design domain leads to the formulation of stress constraints as so-called ‘vanishing’ constraints. Furthermore, the SAND formulation provides a new perspective on the infamous singularity problem. Amongst other results, we present some test cases that seem to scale linearly in computational requirements for a specific range of problem sizes. / AFRIKAANSE OPSOMMING: Die formulering van die topologie optimerings probleem as ’n gelyktydige analise en ontwerp (simultaneous analysis and design (SAND)) formulering word ondersoek. Die eienskappe van die formulering word bespreek in die konteks van die eenvoudig begrensde styfheid/gewig geval en die meer komplekse plaaslike spanning begrensde geval. Die probleme word opgelos in ’n sekwenti¨ele benaderde optimering (SBO; sequential approximate optimisation (SAO)) raamwerk met die SAND formulering, wat lei tot ’n wesenlike vermindering in berekenings vereistes benodig in vergelyking met die tradisionele en inherente duur geneste analise en ontwerp (nested analysis and design (NAND)) geval. In SAND word die vergelykings wat die respons van die struktuur beskryf met gelykheidsbegrensings in die optimerings probleem verteenwoordig. Die respons van die struktuur word dus nie presies opgelos in elke iterasie nie, soos in die geval van NAND wel gebeur. Alle optimerings veranderlikes is dus onafhanklik en lei tot ’n baie yl optimerings probleem. Deur middel van die eindige element metode is die analitiese vorm van alle begrensings beskikbaar en kan dit gebruik word om akkurate benaderde subprobleme op te stel sonder ekstra berekenings koste. Verder kan materiaal heeltemal verwyder uit van die ontwerpsgebied met weinig komplikasies. Dit lei tot ’n verkleining van subprobleme soos die algoritme vordering maak wat berekenings tyd nog meer verminder. Die feit dat materiaal heeltemal verwyder kan word van die ontwerp gebied lei tot die formulering van spannings begrensings as sogenaamde ‘verdwynende’ begrensings. Verder gee die SAND formulering ’n nuwe uitsig op die bekende singulariteitsprobleem. Met verskeie ander resultate word daar ook gewys dat dit voorkom of ’n spesifieke stel toetsprobleme lineˆer skaal in berekenings tyd.