In der fleischverarbeitenden Industrie gibt es eine Vielzahl von Schneidwerkzeugen. Kuttermesser stehen hierbei in der Prozesskette weit hinten und haben einen direkten Einfluss auf die Qualität des Endprodukts. Der Prozess des Kutterns ist bislang nicht komplett analytisch geklärt. Während des Vorgangs durchläuft das Schneidgut (i.A. das Fleisch bzw. das Brät) wechselnde Aggregatzustände von fester (leicht gefrorener) Form hin zum zähviskosen Zustand. Weiterhin ist es permanentem korrosiven Kontakt ausgesetzt. Die Komplexität macht eine analytische Herangehensweise äußerst aufwendig, sodass sich mittels der FEM durch numerisches Vorgehen und Lastannahmen aus Untersuchungen diesem Problem gewidmet wird. Dabei sind bislang nicht bekannte Potentiale zu erkennen. Hierbei wurden verschiedene Vernetzungsstrategien (p- und h-Methode) der FEM angewandt und verglichen. Es sind dabei Materialreduktionen bis knapp 30% ersichtlich.

This bachelor thesis aims to describe the work performed for the design of the undertray for the Kungliga Tekniska Högskolan Formula Student(KTHFS) race car. The goal was to achieve an aerodynamically optimized undertray design that follows the regulations of the competition and the targets set by KTHFS concerning the weight, the size, the materials needed for its manufacture and costs. After some research on previous work, the concept, on which we decided that the undertray would rely on, is Venturi tunnels inspired by the Aston Martin Valkyrie, chosen for its ability to provide a large amount of downforce with a negligible amount of drag using ”ground effect”. Numerous CAD design models were created in Solid Edge and a finalized design was then ported over to Siemens NX to be analyzed using Star-CCM+ and its Design Manager feature. The CFD analyses and optimization was performed in Star-CCM+ with regards to pressure gradient, streamline velocity and downforce. These were done with variable parameters in areas such as expansion height, inlet area and ride height. Contained within this report is a more detailed description of how the CFD analysis was performed as well as suggestions for manufacturing said undertray. Given the time constraints and the societal impacts of COVID-19, manufacturing had to be removed from the scope of the project, however, a step-by-step manufacture guide is provided within. Analysis of uur final design showed 428 N of downforce, a weight of 2.55 kg and a production cost of approximately 2320 SEK. It therefore passes the requirements for weight, cost and ride-height rule regulations set by Formula Student and internal KTHFS targets. / Detta kandidatexamesarbete syftar till att beskriva arbetet som utförts för konstruktionsdesignen av Kungliga Tekniska Högskolan Formula Student (KTHFS) racerbils underrede. Målet var att uppnå en aerodynamisk optimerad underredes design som följer de regler och krav fastställda av KTHFS gällande vikt, storlek, material som behövs till tillverkningen och kostnader. Efter en litteraturstudie på tidigare arbete, blev Venturi tunnlar, inspirerade av Aston Martin Valkyrie, konceptet som vi beslutade att uderreden skulle bygga på och valda på grund av deras förmåga att förbättra bilens prestanda genom sitt nedkraftsbildande och försumbar mängd drag med hjälp av ”ground effect”. Många CAD-designmodeller skapades i Solid Edge och en slutgiltig design överfördes sedan till Siemens NX för att analyseras med Star CCM+ och dess Design Managerfunktion. CFD-analyserna och optimeringen utfördes i Star CCM+ med avseende på tryckgradient, strömlinjehastighet och nedkrafter. Dessa gjordes med variabla parametrar i områden som utvidgningshöjd, inloppsarea och frigångshöjd. I denna rapport finns en mer detaljerad beskrivning av hur CFD-analysen utfördes samt förslag för tillverkning. Med tanke på tidsbegränsningarna och samhällseffekterna av COVID-19 fick vi ta bort tillverknink från projektets omfattning, men en steg-för-steg tillverkningsguide tillhandahålls i rapporten. Analyser av vår slutgiltiga design visade på 428N downforce, en vikt på 2,55 kg och en produktionskostnad på cirka 2320 SEK. Den överenstämmer därför kraven för vikt, kostnad och frigångshöjd som fastställdes av Formula Student.

Structural Optimization techniques are widely used in product development process in ‘modern industry’ to generate optimal designs with only sufficient material to serve the purpose of the component. In conventional design problems, the design process usually generates overdesigned components with excess material and weight. This will in turn increase the life time cost of machines, both in terms material wastage and expense of usage. The thesis “Design of a weight optimized casted ADI component using topology and shape optimization” deals with redesigning a component from a welded steel plate structure into a castable design for reduced manufacturing cost and weight reduction. The component “Drill Steel Support” mounted in front of the drilling boom of a Face Drilling Machine is redesigned during this work. The main objective of the thesis is to provide an alternative design with lower weight that can be mounted on the existing machine layout without any changes in the mounting interfaces. This thesis report covers in detail procedure followed for attaining the weight reduction of the “Drill Steel Support” and presents the results and methodology which is based on both topology and shape optimization. / Strukturoptimering används ofta i produktutvecklingsprocessen i modern industri för att ta fram optimala konstruktioner med minsta möjliga materialåtgång för komponenten. Konventionella konstruktionsmetoder genererar vanligtvis överdimensionerade komponenter med överflödigt material och vikt. Detta ökar i sin tur livstidskostnaderna för maskiner både i termer av materialavfall och användning. Avhandlingen "Konstruktion av viktoptimerad gjuten ADI-komponent" behandlar omkonstruktionen av en komponent från en svetsad stålplåtstruktur till en gjutbar konstruktion med minskad tillverkningskostnad och vikt. Komponenten “Borrstöd” monterad i framkant av bommen på en ortdrivningsmaskin är omkonstruerad under detta arbete. Huvudsyftet med avhandlingen är ta fram en alternativ konstruktion med lägre vikt och som kan monteras på befintlig maskinlayout utan någon ändring i monteringsgränssnittet. Denna avhandling innehåller en detaljerad beskrivning av förfarandet för att uppnå viktminskningen av "borrstödet" och presenterar resultaten samt metodiken som baseras på både topologi- och parameter- optimering.

The present-day automotive industry is striving to design lightweight components by optimizing the design for minimization of weight. The topology optimization is used widely for the design of lightweight components. The casting process is time and cost effective for mass production and widely adopted within the automotive industry. Generally, castability is not considered in the weight optimization process. These weight optimized components are optimized once again in the later stage for cost-effectiveness in the casting process. The modified design usually weighs more than the weight optimized design. The design can be optimized for weight and castability simultaneously in the early stage of design and this thesis report presents an optimization process for the same. The optimization process presents effective usage of the topology optimization to design lightweight components without compromising castability. It is a three-step process where thetopology optimization is integrated with solidification simulation along with DFX -castability evaluation. The reliability of the conceptual design is predicted based on the mapping of solidification and stress hotspots. The process is implemented to design three components of Scaniatruck and weight is reduced by 15% to 25%. / Dagens bilindustri strävar efter att utforma lätta komponenter genom att optimera designen för att minimera vikt. Topologioptimering används i stor utsträckning för design av lätta komponenter. Gjutningsprocessen är tids-och kostnadseffektiv för massproduktion och allmänt vedertagen inom bilindustrin. Generellt ingår inte gjutbarhet i viktoptimeringsprocessen. Dessa viktoptimerade komponenter optimeras igen i ett senare skede för kostnadseffektivitet vid massproduktion. De flesta gånger är viktoptimerade koncept modifierade för att erhålla kostnadseffektivitet vid gjutning genom att lägga till extra vikt. I den här rapporten presenteras enoptimeringsdriven designprocess för att få pålitlig lättviktsgjutbar design. Optimeringsprocessen presenterar effektiv användning av topologioptimering för att utformalätta komponenter utan att kompromissa med gjutbarheten. Det är en trestegsprocess där topologioptimering integreras med förstärkningssimulering tillsammans med utvärdering avDFX-sårbarhet. Tillförlitligheten hos den konceptuella designen förutses baserat på kartläggningav stelningen och spänninggskoncentrationer. Processen är implementerad för att optimera utformningen av tre komponenter i Scania-lastbilar och vikten minskas med 15% till 25%.

[pt] Neste trabalho, são apresentadas algumas aplicações da otimização topológica para problemas de autovalor onde o principal objetivo é maximizar um determinado autovalor, como por exemplo uma frequência natural de vibração ou uma carga crítica linearizada, usando elementos finitos poligonais em domínios bidimensionais arbitrários. A otimização topológica tem sido comumente utilizada para minimizar a flexibilidade de estruturas sujeitas a restrições de volume. A ideia desta técnica é distribuir uma certa quantidade de material em uma estrutura, sujeita a carregamentos e condições de contorno, visando maximizar a sua rigidez. Neste trabalho, o objetivo é obter uma distribuição ótima de material de maneira a maximizar uma determinada frequência natural (para mantê-la afastada da frequência de excitação externa, por exemplo) ou maximizar a menor carga crítica linearizada (para garantir um nível mais elevado de estabilidade da estrutura). Malhas poligonais construídas usando diagramas de Voronoi são empregadas na solução do problema de otimização topológica. As variáveis de projeto, i.e. as densidades do material, utilizadas no processo de otimização, são associadas a cada elemento poligonal da malha. Vários exemplos de otimização topológica, tanto para problemas de frequências naturais de vibração quanto para cargas críticas linearizadas, são apresentados para demonstrar a funcionalidade e a aplicabilidade da metodologia proposta. / [en] In this work, we present some applications of topology optimization for eigenvalue problems where the main goal is to maximize a specified eigenvalue, such as a natural frequency or a linearized buckling load using polygonal finite elements in arbitrary two-dimensional domains. Topology optimization has commonly been used to minimize the compliance of structures subjected to volume constraints. The idea is to distribute a certain amount of material in a given design domain subjected to a set of loads and boundary conditions such that to maximize its stiffness. In this work, the objective is to obtain the optimal material distribution in order to maximize the fundamental natural frequency (e.g. to keep it away from an external excitation frequency) or to maximize the lowest critical buckling load (e.g. to ensure a higher level of stability of the structures). We employ unstructured polygonal meshes constructed using Voronoi tessellations for the solution of the structural topology optimization problems. The design variables, i.e. material densities, used in the optimization scheme, are associated with each polygonal element in the mesh. We present several topology optimization examples for both eigenfrequency and buckling problems in order to demonstrate the functionality and applicability of the proposed methodology.

<p>Advancements in future technologies such as artificial intelligence, electric vehicles, and renewable energy create a consistent need for more powerful and smaller electronic devices and systems. As a result, thermal management components such as heat sinks need to remove higher heat loads from more compact spaces to keep electronics within their operational temperature limits. Constraints imposed by conventional manufacturing processes restrict the design of heat sinks to simple geometries with limited cooling performance. Recent widespread commercialization of metal additive manufacturing (AM) tools offers new potential for leveraging the design freedom of these manufacturing technologies to design and fabricate heat sinks with improved performance. </p> <p>In AM, three dimensional parts are created through layer-by-layer depositing of materials, which allows fabrication of complex geometries that would be impossible or too costly using conventional subtractive methods. Many novel heat sink geometries have been proposed in literature which incorporate features such as manifolds, flow mixers, and curved channels using engineering intuition to reduce pressure drop or enhance heat transfer. Although such designs have been shown to offer improved performance, mathematical design algorithms such as topology optimization (TO) have been shown to outperform engineering intuition. Topology optimization optimizes the material distribution within a given design space, guided by physics-based simulations, to achieve a user-defined objective such as minimization of thermal resistance. Previous TO approaches have used penalization methods to ensure the final designs are composed of macroscopic and non-porous features due to the past precedent of fabrication capabilities. This traditional penalization approach is well-suited to the constraints of conventional manufacturing methods; however, microstructures and porous features are easily fabricable with additive manufacturing. There is a need to develop TO approaches that are better suited for leveraging AM for the design of heat sinks. In this thesis, a homogenization approach to topology optimization is proposed wherein the material distribution is represented as parametrized microstructures. This formulation allows design of thermal management components that have sub-grid features and leverages AM for fabrication. The focus of this thesis is the development of the homogenization approach for TO of heat sinks, as well as the exploration of the design problems it can address, the performance benefits made available, and the two-phase flow physics that it uniquely allows to be incorporated into the topology optimization process.</p> <p>A topology optimization algorithm using the homogenization approach is developed by representing the material distribution as arrays of pin fins with varying gap sizes. To this end, the pin fins are modeled as a porous medium with volume-averaged effective properties. Height-averaged two-dimensional flow and non-equilibrium thermal models for porous media are developed for transport in the pin fin array. Through multi-objective optimization, TO designs are generated for an example case involving a hotspot over a uniform background heat input. The resulting topologies have porous-membrane-like designs where the liquid is transported through a fractal network of open, low-hydraulic-resistance manifold pathways and then forced across tightly spaced arrays of pin fins for effective heat transfer. The TO designs are revealed to offer significant performance improvements relative to the benchmark straight microchannel (SMC) heat sink with features optimized under the same multi-objective cost function. A series of microchannel heat sinks are fabricated using direct metal laser sintering to investigate the printing capabilities and to experimentally demonstrate the performance of topology optimized designs. Advantages of the homogenization approach over the penalization approach can be summarized as follows: (1) reduced computational costs due to its ability to create sub-resolution features, (2) intrinsically fabricable parts using available metal AM tools, and (3) easier to use due to significantly reduced number of hyperparameters (e.g., penalization factors) that are controlled by the user. </p> <p>Topology optimization has been applied to thermal management methods involving single-phase flows such as natural convection, forced air cooling, and pumped liquid cooling. Compared to these conventional heat sink technologies, flow boiling offers very high heat transfer coefficients and effective heat capacities, making it a promising candidate for future cooling electronics applications. The final goal of this thesis is to enable topology optimization of flow boiling heat sinks. However, TO of flow boiling heat sinks has been avoided due to difficulties in modeling the boiling phenomena; of note, there are no examples of TO being applied to the design of heat sink under flow boiling throughout the literature. Multi-dimensional two-phase flow models require prior knowledge of friction factor and heat transfer coefficients. Correlations are available in literature but are not universal and depend significantly on channel/fin geometries, surface roughness, and operating conditions. Given that traditional penalization-based TO approach results in fin and channel geometries with unknown shapes, dimensions, and alignment before the optimization is completed, this prohibits their use for optimization of flow boiling heat sinks. However, the homogenization approach to topology optimization developed in this thesis enables the optimization of flow boiling heat sinks. As it relies on user-defined microstructures with known shapes, alignments, and ranges of geometric dimensions, a universal correlation for flow boiling in microchannels is not needed. Instead, correlations for the user-defined microstructures are sufficient to simulate flow boiling in TO designs generated using the homogenization approach. To this end, a predefined microstructure geometry is chosen for which two-phase flow correlations exist and therefore topology optimization can be performed. Topology optimized heat sink designs under flow-boiling are generated and investigated at various heat inputs, topology optimization grid sizes, and maximum vapor quality constraints. Topology optimized heat sinks designed for single-phase versus two-phase flow are compared.  There are significant differences in hydraulic and thermal responses of the single-phase and two-phase designs due to high effective heat capacity rates and high heat transfer coefficients of flow boiling. The algorithm demonstrated in this work extends the capabilities of topology optimization to two-phase flow physics, and thereby enables the design of various two-phase flow components such as evaporators, condensers, heat sinks, and cold plates.</p> <p>The flow and heat transfer of the TO algorithm for microchannel heat sinks under flow boiling use a two-phase mixture model featuring an effective porous medium formulation. However, closure of the governing equations requires empirical correlations for pressure drop and heat transfer that are specific to the operating conditions, microstructure geometry, and surface finish. Therefore, it must be demonstrated these available correlations can be successfully calibrated over a range of microstructural variations present within the homogenization framework, so as to attain the required prediction generality and accuracy needed to ensure the resulting designs achieve Pareto-optimality. To this end, a set of uniform pin fin calibration samples are additively manufactured and experimentally tested under flow boiling at various flow rates and heat inputs for model calibration. All of the unknown/free coefficients in the adopted correlations are determined by minimizing the error between the model predictions and the experimental measurements using gradient-based optimization. The calibrated topology optimization algorithm is then used to generate a Pareto-optimal set of heat sinks optimized for minimum pressure drop and thermal resistance during flow boiling. Experimental characterization of these additively manufactured heat sinks, unseen during the model coefficient calibration process, reveals that the measured Pareto optimality curve matches that predicted by the topology optimization algorithm. Lastly, a heat sink design is generated for a design space involving multiple hot spots and background heating to showcase the capability of the experimentally calibrated two-phase topology optimization algorithm at handling complex boundary conditions. The optimized heat sink intelligently distributes an adequate amount of coolant flow to each of the heated regions to avoid local dry-out. This work demonstrates a complete framework for two-phase topology optimization of heat sinks through experimental calibration of flow boiling correlations to the porous medium used by the homogenization approach. </p> <p>The major contribution of this thesis is the development of a homogenization approach for TO of additively manufactured microchannel heat sinks under single- and two-phase flows. Not only does the homogenization approach provide several advantages over the traditional penalization approaches such as reduced computational costs, intrinsic fabricability using AM, and ease of use, but it also enables TO of heat sinks under flow boiling and potentially TO of other two-phase thermal management components. The work discussed in this thesis serves a comprehensive end-to-end guide on TO of microchannel heat sinks using the homogenization approach with experimental demonstrations for validation.</p>

Without lubrication, machines are not imaginable to perform over a long period of time and complete their designated operations. With its omnipresent availability, the air is capable of functioning as a lubricant in long operations very efficiently. Moreover, thrust bearings support axial loads and transmit power at the same time under heavy loads. Therefore, to provide separation under heavy loads in lubricated rotating devices such as thrust pad bearings keeping the power losses at a minimum, film thickness and pressure distribution are very important to investigate at the bearing interfaces. Thrust pad gas (air) bearings are being used in very high-speed rotating machines. Usages of these air bearings are increasing nowadays in industries. In this thesis project, simulations of lubricated contacts of a thrust pad air bearing are performed utilizing multiphysics phenomena and surface textures as mathematical functions. Structural mechanics and fluid mechanics physics are used to model multiphysics functionality. Ideal surface texture models defined by mathematical functions are utilized. More efficient techniques such as homogenization techniques to model the influences of surface roughness are introduced for multiscale study. The current work also presents the Reynolds equation for incompressible and iso-viscous Newtonian fluid flow and formulation for a stationary study. The air bearing with three pads is presented and a virtual twin of this model is built for simulation in COMSOL Multiphysics software. Simulation results are obtained using a single pad from the air bearing considering periodicity of the mathematical formulation. Numerical solutions for pressure build-up and film thickness distributions are achieved from a stationary study performed in COMSOL Multiphysics. MATLAB is used for rigid body solutions. Numerical verification is carried out between the rigid body solutions from MATLAB and fluid physics solutions from COMSOL Multiphysics only for the simulations with tilting pad configuration. Obtained rigid body solutions are also compared to the trends of thrust pad bearing design diagrams to verify the modelling approach and the results. A tilting pad lubricating configuration is used for the thrust pad bearing first. Then pocket geometries for optimization of the bearing pads are explored. For that purpose, separate digital models of the bearing pad are built in COMSOL and analysed for the best performances. Material properties of steel AISI 4340 and Polylactic Acid (PLA) material are used to model virtual bearing pads. To understand the performance of the bearing better, its performance parameters such as load carrying capacity (LCC), relative power loss, and coefficient of friction torque (COT) solutions from the simulations of lubricated contacts of the thrust pad air bearing are analysed. To characterize the performance of the bearing, dimensionless LCC, relative power loss, and COT are explicitly formulated and computed from the pressure and film thickness solutions obtained in the simulations. Relative power loss and COT are resulted from the development of shear stresses in the lubricating fluid due to motion. Parametric analysis is also performed for these parameters in COMSOL Multiphysics. Additionally, performances of several pocket geometry design configurations are also analysed for the best values reached such as the maximum LCC. Pockets with shallower depths are found to have provided higher LCC in general than deeper pocket geometries and plane pads with tilting pad lubricating configuration. Finally, a physical model of an air thrust pad bearing with 3D-printed bearing segments made of PLA material is tested. The physical bearing performed very well in achieving full film separation in the test.

In the past decades of years, huge amounts of people chose to move to big cities for better education and medical service, which also makes many cities are very crowded and noisy. Moreover, the house rent in city center is some kind too expensive for many people, especially for the youth. In this sense, more people are willing to live in suburb instead of city center. Due to the larger distance between home and office, people’s requirement for a faster public transportation method is enormous. Elon Musk first publicly mentioned the concept of Hyperloop in 2012[1], which is a sealed tube or system of tubes with nearly vacuum condition through which a pod can transport people or objects at super high velocity. With the linear induction motor and magnetic levitation technology, the drag force on the pod can be reduced tremendously, thus increasing the peak velocity to 1200 km/h. To gather more ideas for this concept, SpaceX holds the Hyperloop Pod Competition where worldwide teams will design their own Hyperloop pod to demonstrate their technical feasibility of new ideas [2]. A Hyperloop system is currently in development by the Integrated Transport Research Lab (ITRL) at KTH Royal Institute of Technology to participate in the upcoming Hyperloop Pod Competition. KTH Hyperloop group has some primary design of chassis and shell. However, they have no idea how good of their current design is. Furthermore, since the velocity is the only criteria for this competition, they also want to reduce the mass as much as possible. In this sense, some finite element analysis and optimization analysis are necessary. The objective of this master’s thesis is to analyze the current shell and chassis design to assess the quality of the attachments and integrity of the design and to reduce the total mass while keeping the stiffness within the safety range. The used tools are HyperMesh, Optistruct and HyperView which are parts of the software HyperWorks from Altair. / Ett Hyperloop-system utvecklas för närvarande av Integrated Transport Research Lab (ITRL) vid KTH Royal Institute of Technology för att delta i den kommande Hyperloop Pod-tävlingen. Hyperloop-gruppen vid KTH har utvecklat en primärkonstruktion av chassi och skal. De har dock ingen aning om hur bra deras nuvarande design är. Eftersom hastigheten är de enda kriterierna för denna tävling, vill de också minska massan så mycket som möjligt. I detta avseende är det nödvändigt med finita element- och optimeringsanalyser. Syftet med denna masteruppsats är att analysera den aktuella skal- och chassikonstruktionen för att utvärdera kvaliteten på dess fästen och integriteten hos designen, samt att minska den totala massan samtidigt som styvheten uppfyller specificerat krav. De använda verktygen är HyperMesh, Optistruct och HyperView som är delar av programvaran HyperWorks från Altair.

Additive Manufacturing (AM) using the Powder Bed Fusion (PBF) is a relatively new manufacturing method that is capable of creating shapes that was previously practically impossible to manufacture. Many think it will revolutionize how manufacturing will be done in the future. This thesis is about some aspects of when and how to Design for Additive Manufacturing (DfAM) when using the PBF method in metal materials. Designing complex shapes is neither easy nor always needed, so when to design for AM is a question with different answers depending on industry or product. The cost versus performance is an important metric in making that selection. How to design for AM can be divided into how to improve performance and how to improve additive manufacturability where how to improve performance once depends on product, company and customer needs. Using advanced part shaping techniques like using Lattices or Topology Optimization (TO) to lower part mass may increase customer value in addition to lowering part cost due to faster part builds and less powder and energy use. Improving PBF manufacturability is then warranted for parts that reach series production, where determining an optimal build direction is key as it affects many properties of PBF parts. Complex shapes which are designed for optimal performance are usually more sensitive to defects which might reduce the expected performance of the part. Non Destructive Evaluation (NDE) might be needed to certify a part for dimensional accuracy and internal defects prior use. The licentiate thesis covers some aspects of both when to DfAM and how to DfAM of products destined for series production. It uses design by Lattices and Topology Optimization to reduce mass and looks at the effect on part cost and mass. It also shows effects on geometry translation accuracies from design to AM caused by differences in geometric definitions. Finally it shows the effect on how different NDE methods are capable of detecting defects in additively manufactured parts.

<p dir="ltr">Multiscale topology optimization is a numerical method that enables the synthesis of hierarchical structures, offering greater design flexibility than single-scale topology optimization. However, this increased flexibility also incurs higher computational costs. Recent advancements have integrated machine learning models into MSTO methods to address this issue. Unfortunately, existing machine learning-based multiscale topology optimization (ML-MSTO) approaches underutilize the potential of machine learning models to surrogate the inner optimization, analysis, and numerical homogenization of arbitrary non-periodic microstructures. This dissertation presents an ML-MSTO method featuring displacement-driven topology-optimized microstructures (TOMs). The proposed method solves an outer optimization problem to design a homogenized macroscale structure and multiple inner optimization problems to obtain spatially distributed, non-periodic TOMs. The inner problem formulation employs the macroscale element densities and nodal displacements to define constraints and boundary conditions for microscale density-based topology optimization problems. Each problem yields a free-form TOM. To reduce computational costs, artificial neural networks (ANNs) are trained to predict their homogenized constitutive tensor. The ANNs also enable sensitivity coefficients to be approximated through a variety of standard derivative methods. The effect of the neural network-based derivative methods on topology optimization results is evaluated in a comparative study. An explicit dehomogenization approach is proposed, leveraging the TOMs of the ML-MSTO method. The explicit approach also features two post-processing schemes to improve the connectivity and clean the final multiscale structure. A 2D and a 3D case study are designed with the ML-MSTO method and dehomogenized with the explicit approach. The resulting multiscale structures are non-periodic with free-form microstructures. In addition, a second implicit dehomogenization approach is developed in this dissertation that allows the projection of homogenized mechanical property fields onto a discrete lattice structure of arbitrary shape. The implicit approach is capable of dehomogenizing any homogenized design. This is done by incorporating an optimization algorithm to find the lattice thickness distribution that minimizes the difference between a local target homogenized property and a corresponding lattice homogenized stiffness tensor. The result is a well-connected, functionally graded lattice structure, that enables control over the length scale, orientation, and complexity of the final microstructured design.</p>