check or pats Special Issue Incerational Journal of ‘Acciveturl Computing Computational design, engineering = * © The Author(s) 2020 and manufacturing of a material- ‘Arid revi pele: sagspub.comourralermisions efficient 3D printed lattice structure pou w7serpers Jounakcsagepub corto SAGE Roberto Naboni®, Anja Kunic and Luca Breseghello® Abstract Building with addicive manufacturing is an increasingly relevant research topic in the field of Construction 4.0, where designers are seeking higher levels of automation, complexity and precision compared to conventional construction methods, As an answer to the increasing problem of scarcity of resources, the presented research exploits the potential of Fused Deposition Modelling in the production of a lightweight load-responsive cellular lattice structure at the architectural scale. The article offers an extensive insight into the computational processes involved in the design, ‘engineering, analysis, optimization and fabrication of a material-efficient, fully 3D printed, lattice structure. Material, structure and manufacturing features are integrated within the design development in a comprehensive computational workflow. The article presents methods and results while discussing the project as a material-efficient approach to complex structures. Keywords ‘Automated design, cellular lattice, digital fabrication, additive manufacturing, computational workflow Introduction In the last five years, the emerging advancements in the world of 3D printing have induced a considerable interest among architects, engineers and builders, encouraging the conception of novel architectural solu- tions.'? While the adoption of additive manufacturing (AM) is virtually leading to enhanced design freedom, its use at the building scale is still challenging, and several technologies are currently being explored towards applications in constructions. Current approaches to AM-based construction are focusing on continuous printing of monolithic building, prefabrication of building components, and spatial extrusion of lightweight structures.” The last one, in particular, arose in combination with data-driven design, which has allowed the study of weight-efficient configurations with increased mechanical properties, such as strength, stiffness and energy absorption.‘ The main advantage lies in the high strength-to-mass ratio, as well as in the possibility to optimally distribute and adapt the material organization throughout a volume.’ These features have stimu lated researchers to look into cellular structures as a potential replacement of conventional construction, with University of Southern Denmark (SDU) ~ Section for Civil and Architectural Engineering, Odense, Denmark Corresponding auth: Roberto Naboni, University of Southern Denmark, (SDU), Section of Civil and Architectural Engineering, Campusve| $5, Odense 5230, Denmark, Emall: [email protected]2 International Journal of Architectural Computing 00(0) Figure 1. (2) Space truss prototype, Xtreet; (b) Studies in recursive lattices, OEZCT; (c) Mesh Mould, Gramazio Kohler Research Group, ETH; (4) Cellular fabrication, Branch Technology; (¢) Daedalus Pavilion, Ai Build; (9) SHoP Pavilion, Branch Technology: (g) Rapid Liquid Printing of lattice structures, Self-Assembly Lab, MIT: (h) Digital cellular materials, MIT research. a perspective of reducing material usage, However, scaling up these structures to the realm of architectural construction poses several challenges in terms of material, geometry, fabrication and structural continuity at large scale. In this article, we discuss the implementation and application of a computational workflow for the design and additive manufacturing of performance-driven lattice structures at the construction scale, through the development of a fully automated process for (1) lattice generation in response to defined load- ing conditions; (2) material and topology optimization on several hierarchical levels, that is, layers, cells, components, clusters; (3) adaptive toolpath generation in function of mechanical performance and printing quality; (4) fabrication management and control. The work integrates all the design-to-construction phases into a unique, computationally controlled process with continuous information flow. This workflow is dem- onstrated with an architectural scale construction, namely Trabeculae Pavilion, building upon fundamental concepts of the emerging paradigm of Construction 4.0, such as full product customization, rapid-design generation, design-feedback and design-to-construction automation. Additive manufacturing of lattice-based constructions ‘The literature review on lattice-based construction highlights a number of studies in the last years, which are here classified on the base of their different functional, material and manufacturing features (Figure 1) A first approach employs 3D printing for the manufacturing of lattice moulds subsequently filled with struc- tural material, Philippe Morel and Thibault Schwartz* used sand printing for the demonstrative construction of three-dimensional spatial lattices made of concrete; a few years later, the French company XtreeE explored the direct 3D printing of semifiuid materials such as clay and concrete to create a one-piece formwork for a concrete lattice in the project Space Truss Prototype. In this case, due to the material characteristics which limited the production of overhanging geometties, the lattice formwork required supporting material, which in turn have resulted in a simplified design solution.’ The second approach is based on the spatial extrusion of thermoplastics to form lattice structures. Gramazio Kobler Research Group from ETI Zurich*® andNaboni et al 3 Figure 2. Trabeculze Pavilion is a 3D printed lattice structure at construction scale Source: Gabriele Seghiz Branch Technology" used robotically extruded lattice meshes both as a reinforcement and a formwork to contain infill material for the construction of walls; the same technique was also employed to realize larger scale self-supporting gridshell structures." This approach guarantees quick production, but a limited con- trol on construction details such as the joinery between elements. Furthermore, new manufacturing tech- niques were investigated at MIT to challenge the limitations of current printing methods. An example is from Hajash et al."® who employed Rapid Liquid Printing (RLP), a system for the robotic injection of a two-part chemically cured material, temporarily supported by a granular gel contained in a tank to achieve unlimited geometric freedom. Differently, Cheung!” introduced an approach for the additive mamufacturing of fibre- reinforced composites, where modular lattice struts are put together by robots to form low-weight assem- blies. In our research, we focus on the development of hollow lattice structures fabricated through fused deposition modelling (FDM), in which material extrusion and lattice topology parameters are defined as a function of specific and varying loading conditions. Through this approach, we aim to generate lightweight architecture with efficient material distribution and seamlessly integrated joinery solutions Research aim The presented work outlines an exploration on the use of AM at construction scale through the design and fabrication of a component-based, structurally optimized lattice structure, using an FDM 3D printing farm, Specifically, the research focuses on multi-hierarchical material optimization, the design engineering of a component-based structure, the integration of a seamless joinery system, the implementation of high-resolu- tion 3D printing. These key features are described through the material research, design, engineering and fabrication development of a fully 3D printed functionally graded lattice structure, the Trabeculae Pavilion (Figure 2). The experimental case study is based on a computational workflow that entangles processes of material characterization, geometric and design optimization and fabrication,4 International Journal of Architectural Computing 00(0) Figure 3. Fabrication process of a cellular component with fused deposition modelling. Source: Roberto Nabon Fabrication and material system Digital fabrication: features and constraints ‘The geometric, mechanical and aesthetic features of large FDM printed objects are directly influenced by the control over the material deposition and the printing settings that are utilized. An iterative process is needed to fine-tune the extrusion process to obtain physical objects that match precisely the actual model in the digital space. A deep understanding of the machinic characteristics and material behaviour is an indispensable el ment for developing a fabrication-aware design workflow. Within the context of existing FDM systems, we have opted for commercially available mid-size delta 3D printers. These were preferred over the more com- mon Cartesian machines because of their light architecture which offers increased acceleration and travel speed, where time plays an important role. As well, in the production of discontinuous geometries such as lattice structures, the high travel speed during printing can significantly improve the overall fabrication time and productivity of the fabrication system (Figure 3). Moreover, the printers were equipped with a high-per- formance extruder developed by an industrial partner during the process, which can depose material at a high rate, During this work, an adaptive layer thickness that varies from 1.2 to 2mm was utilized in order to rein- force locally the highly stressed areas of the structure without the need for multiple shells, A printing farm consisting of delta FDM printers was implemented to produce all of the construction, components. The printers shared the same architecture and extruder features, ensuring constant printing quality and mechanical characteristics, obtaining an effective reduction of the fabrication time, mistakes and tolerances. Within this framework, two printing parameters play an essential role: the control over the print- ing angles and the adhesion between layers."* For this reason, the initial phase of the research focused on the production of a large number of printing tests performed to define the optimal fabrication settings. These include layer extrusion, geometric constraints and path planning, as well as the settings related to the specific ‘material behaviour, that is, the temperature of the printing bed and the nozzle, the printing and travellingNaboni et al 5 Figure 4. Tensile cesting of the custom lattice strut sample, Source: Roberto Naboni speed, and the material flow rates. While in principle these parameters are common to all the FDM machines, the settings must be tailored for the specific combination of hardware, material, scale and application, Analyzing the overhanging angles independently from the material employed, we concluded that printing ‘geometries with overhangs between 0° and 45° from the vertical axis are achievable without compromising the fabrication quality, as they display minor alterations up to 65° while considerably compromising layer cohesion up to 80°, This information is translated into the computational workflow to define the fabrication parameters, Material testing and characterization Material selection is another crucial factor in pursuing the realization of large scale lightweight lattice struc~ tures. Thermoplastics are a relatively unusual class of materials for construction, and moreover, the use of FDM requires an investigation to identify the most suitable polymer and define its mechanical properties Given the specificity of the application, an extensive series of mechanical testing was performed. Initially, different materials were involved to identify the performance with respect to the mechanical resistance and thermal behaviour as well as to an adequate printability. The material characterization was performed through tensile testing of custom specimens that resemble the hollow lattice struts with the minimum diam- ter utilized in the structure ~ 10mm and 1.4mm wall section, Initially, polymers were selected on the base of their nominal density and stiffness. All the specimens were tested with a load applied at a speed of I mm/s, acting upon an area of 37.82mm?, while the analyzed strut length was 60mm (Figure 4), The testing series has been conducted in three stages: Material comparison (i): The ultimate tensile strength (UTS) and Elastic modulus of different polymers and composite filaments have been measured in a specific configuration of hollow struts of the lattice structure, The tested polymers included: Polylactic acid (PLA), Polycarbonate + Acrylonitrile butadiene styrene (PC ABS), Long-chain nylon (LCN), Carbon nylon (CN), Acrylonitrile styrene acrylate (ASA),6 International Journal of Architectural Computing 00(0) Polypropylene (PP), and Polyethylene terephthalate (PETG). The second set of specimens was printed and coated with epoxy resin, PLA showed the highest values of ultimate strength and elastic modulus, The addition of epoxy resin coating improved the strength of the material in all the tested cases, with an average increase of 20%. LCN resulted in the highest percentage of elongation at break Material and printing refinement (ii): Considering the results of the material comparison, Bio-based Polylactic Acid was identified as the most appropriate material, thanks to its optimal printing quality and intralayer adhesion that reflects in a high mechanical resistance both in compression and in tension. Subsequently, variations in its chemical composition were tested and compared to further improve its characteristics. On the one hand, tensile tests were performed on specimens printed with different angles to simulate real-case conditions; on the other hand, the same tests were conducted while exposing the specimens to different environmental temperatures in a heating chamber varying from 20°C to 80°C. In this process, different types of additives were added to offset the heat deflection temperature (HDT) of basic PLA. After different trials, a specific chemical composition which offered increased mechanical characteristics up to a threshold temperature of 55°C was selected, Anisotropy testing (iii): Controlling the mechanical behaviour of printed components under different inclination angles is a challenging but crucial aspect in FDM 3D printing, as the overall quality of an object and the layer adhesion can be compromised. Therefore, itis critical to introduce a method for pre- dicting the influence of angled printing on mechanical performance, especially for its use in non-standard cellular structures with very differentiated mechanical conditions. In response to this need, we introduced an intralayer optimization: the material is deposited with variable thickness in order to compensate for the effect of angled printing, preventing layers discontinuity and guaranteeing a sufficient overlap among layers. This material optimization was analyzed through the tensile testing of custom specimens produced under the angles of 0°, 30°, 45° and 60° (Figure 5) from the vertical axis. All the samples broke for layer failure, as expected. However, they showed neglectable differences in terms of UTS, which allowed us to conveniently consider the material as isotropic in further structural analysis. Computational methodology for the design, analysis and manufacturing of a 3D printed Pavilion Biomimetic studies on the bone structure ‘Nature is providing several examples of efficient construction based on cellular structures, such as the hon- eycomb structure of the wasp nest, the closed-cell construction found in wood and cork, the open-cell con- figuration of sponges and corals, or the structural lattice found in the trabecular bone." To better comprehend how cellular structures behave and what ate the potentials of employing them into innovative building models, many architects and structural engineers have looked into Nature as an inspiration, which mastered the processes of an optimized material organization over the course of more than 3.8billion years. In the realm of biology, form, structure and performance are intertwined parts of one coherent development. Thanks to the quasi-complete geometric freedom and control of material allocation in space, AM offers the possibil- ity to apply similar morphogenetic logics in man-made structures. In this framework, material waste can be programmatically reduced through informed design development. Among numerous examples of cellular structures found in Nature, this research looks at the hierarchical structure of bones as a role model to chal- Jenge current design paradigms of lightweight and structurally efficient architecture. More specifically, we focus on its intemal structure, namely trabecular bone, where individual bone cells are varying in density as a function of the strain and in orientation aligning with the main stress trajectories, to withstand tensile and compressive forces and therefore reducing their weight” (Figure 6). This configuration is emerging from theNaboni et al Figure 5. Material testing and characterization: The image shows specimens after testing and various typologies of their fracture depending on the printing angle; from left: 0°, 30°, 45° and 60°. Source: Roberto Naboni, Low coy High oxcsty neuTRAL as COMPRESSION Figure 6. (2) The section shows trabecular organization within a human femur with varying degrees of porosity and orientations based on the external loading conditions; (b) Principal stress trajectories along which bone’s microstructure gets organized. Source: Roberto Naboni8 International Journal of Architectural Computing 00(0) so-called bone remodelling process,” a highly-coordinated homeostatic process for the formation and con- stant adaptation of the bones’ tissue. The simultaneous action of two different cells called osteoblasts and osteoclasts in the bone remodelling process has the function of measuring and moderating the local stress levels in the bone by removing or adding material, in a dynamic process of structural optimization. This process was demonstrated by J.S. Tumer” in a comprehensive algorithm, which was taken as an inspiration in this work and translated into a computational workflow for the design and generation of a load-responsive lattice structure. Computational workflow Following the morphogenetic logics of the trabecular bone, a functionally graded cellular structure was generated where topology, porosity and geometric orientation of the lattice structure respond to the external loading conditions. This has been integrated into a comprehensive workflow of computational design, multi- criteria optimization and fabrication of an architectural shelter, fully exploiting the advantages of AM. In the case study described in this article, an anticlastic shell working both in compression and tension was obtained as a minimal surface connecting a few boundary curves. The custom-developed workflow takes this as an input shell geometry and automatically generates a lattice structure informed by the loading scenario as well as by physical constraints of the material and the fabrication system used. The goal is to consolidate a versa~ tile design and construction method for a material-efficient architecture that can be applied to a wide range of loading conditions and morphologies. To determine the most efficient material distribution for the specific anticlastic shell design, an iterative ‘mechanical analysis based on Finite Element Methods (FEM) is implemented in an analogy with the bone remodelling process. The multi-level design and optimization workflow is developed within the digital envi- ronment of Grasshopper for Rhinoceros. The lattice structure is a translation of the material allocation pro- cess, based on the calculation of the stiffiness factor values across the shell, which fine-tunes the structural response to the specific shape and loading case varying topology, thickness and cell orientation alike the trabecular bone. The critical input parameters influencing the lattice morphogenetic process relate not only to the geometric characteristics and loading scenario but also to the fabrication setup, the material properties and the inherent manufacturing constraints. Resulting from the variations in any of these parameters is a dif= ferent configuration of the generated load-responsive lattice structure (Figure 7). ‘The crucial benefit of the established computational workflow is its potential generalization for any mate- rial and additive fabrication setup, as it is developed to automate design generation and structural lattice optimization based on the input material properties and fabrication constraints. Moreover, the approach is especially favourable for mechanically inferior materials, as it relies on the lattice topology and its relative density rather than on the material strength to achieve high mechanical performance in construction, Mechanical analysis at cell scale In biological systems, redundancy is used as a strategy to achieve high-levels of structural performance.”* Cellular structures as well behave in this way and use large numbers of cells to distribute loads, being struc- turally efficient yet sustaining relatively small stresses. In such configurations, damage or removal of a sin- gle or a small group of elements do not influence the overall mechanical behaviour and stability of the structure, due to a large number of interconnections between the cells. In bones, the cellular structure is in the order of 100 to 400m,” and in mechanics applications, lattices are mostly within the centimetre. Different parameters were taken into account to define the appropriate cell scale for this application, includ~ ing manufacturing precision, time of production of such scale elements, as well as the overall weight of the structure. After the mechanical tests performed on struts, cells and nodes, the range between 30mm andNaboni et al 9 Figure 7. Computational workflow diagram, showing an interdependent multiscalar approach between two parallel processes: form definition and structural analysis on one side and fabrication and assembly process on the other one. Source: Roberto Naboni 120mm resulted as the most appropriate for the scale of application. A series of cell configurations were designed and compared to assess their printability, relative density, visual permeability and mechanical resistance. After thorough investigations, the octahedral cell unit was found as the most appropriate topology for a lightweight architecture with FDM technology. In fact, the absence of horizontal elements makes it highly suitable for manufacturing and its low relative density makes it lightweight, yet maintaining high load-transferring capacity. To estimate the loading capacity of the cellular structure, we performed tests on ‘one component with simple octahedron cells and compared it to one with a reinforced octahedron cellular structure, The first specimen, weighing 185 g, presented a more elastic behaviour, and withstood a compres- sive load of about 3200N; the second specimen, weighing 220g, reacted as a more rigid material and per- formed three times better, reaching a compressive load of 10,000N, equivalent to 4500 times its own weight (Figure 8) Topology and material optimization An anticlastic shell obtained through a minimal shape form-finding process was used as a case study to develop a demonstrator testing the approach and the system developed. The shell is analysed through FEM, considering a distributed load of S00N/m? in addition to its own weight, considering an elastic modulus of 3.5GPa, a Poisson ratio of 0.36 and a density of 1251.5Kg/m’. From the analysis, we derive information about the stress distribution and stress orientation. In particular, the main compressive trajectories are discre- tized into a spatial configuration of the lattice over the shell. Under the same conditions, an iterativelo International Journal of Architectural Computing 00(0) et 8 @ Figure 8. Compression testing of the cellular components with two different topologies and comparison of their deformation at the beginning of the test, at Yield point and at Fracture point — (up) simple octahedron cell, (down) ‘octahedron cell reinforced with vertical struts Source: Roberto Nabon algorithm for the optimization of material distribution is then run on the mesh of the shell. This is based on the solid isotropic microstructure with penalization (SIMP) interpolation scheme, an approach to topology optimization (TO) introduced by Bendsoe and Kikuchi?” and Rozvany et al.2* SIMP aims at predicting mate- rial distribution given design space, a load case, boundary conditions and performance requirements. Unlike the traditional approach in topology optimization, where a domain is discretized into a grid of finite elements which are either filled with material or emptied according to the load case using binary values 0 and 1, the introduction of a continuous relative density distribution function allows the assignment of a gradient of densities ranging from empty to full. Graphically, the results are represented with a greyscale gradient rang- ing from black where the material is not required to white where the full density of the material is most needed (Figure 9), The results of the topology optimization are exploited in two ways: first, a series of additional vertical struts are added accordingly, increasing the stiffness and strength gradually in whiter areas and maintaining a more porous (elastic) topology in the black sectors (Figure 10 left); second, the corresponding 0-to-1 value is assigned to cach of the $5,363 lattice struts, and their thickness is parameterized accordingly in a range between 10mm and 35mm. These values are then used as a scalar field to generate a polygonal mesh from the lines lattice configuration through @ Marching Cubes algorithm”? (Figure 10 right). ‘The final lattice configuration was evaluated and verified through two distinct FEM software packages, Sofistik and ANSYS. The numerical analysis considered the dead load with the bottom restrained nodes (Figure 11), as the structure cannot simply rest on the ground but needs to be hinged to prevent lateralNaboni et al u wares iu we Figure 9. The image shows a sequence of computing iterations of the topology optimization process Source: Roberto Naboni Figure 10. Lattice variation based on the results from stress analysis: left — varying topology (number of lattice struts); right — varying porosity (lattice thickness). Source: Roberto Nabon, movements and to resist minor tension. In both cases, the lattice structure was modelled as beam elements, undergoing a geometric nonlinear analysis, The output values proved that the specific lattice arrangement with additional vertical elements has significantly enhanced the global structural behaviour, reducing the flexural contribution, The resulting stresses are well below allowable limits, and therefore the structure is confirmed as safe.2 International Journal of Architectural Computing 00(0) Li SE me nrc e162 _ Figure 1. Structural analysis and verification of the overall lattice structure: up (SOFISTIK — Von Mises stresses for a dead load; down (Ansys) — (a) Structure’s boundary conditions; (b) Max stresses in frames (axial=flexural) (KPa); {6) Flexural over total stress ratio, Source: Valentin Koslow, Francesco Martell Multi-objective geometry optimization ‘A method for discretizing the designed anticlastic shell into construction components was developed to comply with manufacturing and operational requirements, The need for subdividing the shell into compo- nents derives both to fit in the available machines and for more agile handling of the structure off and on-site. The number and dimensions of the available 3D printers were used to set the maximum dimension of the components as well as to keep a proportional amount of components for each used printer size. In the pre- sented research, the setup composed of one large and three medium-sized 3D printers constrained three- quarters of the components to fit in the latter group of machines. The tessellation of the shell is based on shingled rows, where at each intersection three components meet. Each component has a planar base forNaboni et al B { Figure 12. Scaggered tiling of the components, The geometric construction of the five-edges element guarantees a planar base, optimal for manufacturing with FDM 3D printing. Source: Roberto Nabon\ optimal positioning on the printing bed without the need for support material, resulting in each construction component being inscribed in a closed polygon with five planar surfaces (Figure 12). The cellular lattice structure and the tessellation are then adjusted and optimized in sizing and orientation through a multi-objective genetic algorithm, where the variables include the lattice configuration (cell size and inclination), density, length and amount of material to be employed. First, the overall structure is subdi- vided into components with size fitting into the printers’ volumes; second, the number of struts that exceed the critical printing inclination angle of 65° from the vertical axis is reduced; third, the material usage is minimized. The final generated solution results in a lattice system with a relative density equal to 0.034 (p = 0.034), an average cell size ratio of 0.8, and inclination angle ranging between 15° and 65° (Figure 13). Automated detailing and joinery design ‘The system of connections in a component-based lattice structure is crucial as it has to provide a seamless continuity in the force-flow with a small amount of contact surface between the adjacent components. In fact, the contact areas are mostly punctual, located in the nodes, that is, the intersections between the lattice struts, Moreover, the load-responsive mesh of the lattice makes every node unique in their topology and size, as they are defined by the optimization process, The nodes are automatically generated within the Grasshopper design workflow, with the only addition in the assembly phase of an embedded system of bolt-nut connec- tion to tighten them together. Discrete components are joined together in two directions — vertically and laterally (Figure 14), with a specific connection logic for the different node typologies, which reflects the complex topological articulation of the system. Four different typologies of intersection have to be imple- mented depending on the position and number of struts meeting in the node. The first case is that of the (1) corner where three different components meet — they are joined together through a male-female configura tion to guarantec a good amount of friction and increase the strength of the node; the second node typology applies to (2) the vertical connections in the middle of the construction components — here the overlap4 International Journal of Architectural Computing 00(0) Figure 13. Multi-criteria Lattice structure optimization through genetic algorithms. The gradient defines the percentage of struts exceeding inclination over a critical angle of 65°. Source: Roberto Naben Vertical Connection Lateral Connection Figure 14. Connection deta left — joinery type between components connecting vertically right — joinery type between components connecting laterally. Source: Roberto Nabon connection is implemented; the third typology is (3) the connection among the vertical struts which are added as structural reinforcement - similarly to the second typology, the overlap joint is used; the last case is in the (4) lateral joints — these are connections between adjacent side faces of components, and they are simply aligned and tightened with a bolt transversal to the load direction. After a series of experiments and mechanical tests (Figure 15), a variable printing flow is embedded in the code to strengthen the nodes but keep the lightweight nature of the structure. Higher extrusion flow is used in the area of connection, increas- ing the quantity of material extruded, thus stiffening those walls, whereas a normal flow is gradually reached outside the area of the node, Moreover, the nodes are further stiffened with internal surfaces placed parallelNaboni et al Is Figure 15. Tensile cesting of the connection. Source: Roberto Naboni to the direction of the compression force, enhancing the resistance of the connections and in turn of the overall structure. ‘Automated design-to-fabrication workflow AM is increasingly applied in many industrial sectors for its flexibility and ability to produce complex designs with no extra costs. To comply with the relative slowness of the process, design and manufacturing workflow have to be optimized to reduce the time of fabrication, amount of material used, and the risk of errors in the whole production chain, Taking from the dimensional analysis previously described, a custom Cif algorithm run in Grasshopper is creating a production schedule, assigning components to one of the print- cers of the available printing farm, according to their dimensions and balancing printing time to have an even distribution of production. In this process, the construction components are algorithmically categorized into six production batches, ranging from the smallest components fitting into the smaller printer to the biggest part that needs extra rotation to fit their volume into the biggest available printer (Figure 16). Each compo- nent is labelled and a .xm file is synchronized and updated at any change throughout the process. In parallel, the construction components are oriented such that their base is aligned to the printing bed and placed in the correct position to fit in the printing volume by matching their projected barycentre on the horizontal plane with the centre of the printing bed. Subsequently, easy-to-remove supports are generated for the few hori- zontal surfaces that exist in the area of the joints. Additive manufacturing and construction ‘The overall design to fabrication process is developed and managed through an algorithmic procedure in the same digital environment, For every component, the mesh is sliced in horizontal curves, that is, layers, to16 International Journal of Architectural Computing 00(0) Figure 16. The construction components are subdivided into different production batches and assigned to printers according to their volumetric dimension. Source: Roberto Nabon, generate a machining code (.geode) directly in Grasshopper. A seamless workflow within a single software is preferred to the existing commercial solutions as it offers a direct transition from digital geometry and fabrication and allows to fine-tune the movement strategy of the machine to optimize the printing times and quality. This also allowed to extrude a quantity of material proportional to the printing angle, which makes the 3D printed parts isotropic as previously described. The tool paths are transferred to a printing farm pro- ducing approximately seven components per day, at an average extrusion speed of 60mms (Figure 17) ‘The parallel production process was run for 4352h to produce 352 construction components. These were first joined into larger clusters of approximately 2.5m? area, to be easily transportable and handled by opera~ tors without the need for heavy lifting machinery. The lightweight nature of the structure gave the possibility for a fairly easy manual handling and assembly process. This was carried out in three steps, where the bot- tommost clusters were first anchored through a series of bolts to a plywood base and subsequently, the sec- ond row of clusters is added and fixed, forming two self-standing side walls; as a final step, the top saddle is placed from above and fixed to the side walls.” Results This article discusses several advancements in FDM large-scale direct 3D Printing in construction. From a material point of view, this work introduces a design and fabrication strategy to make material properties in 3D printed structures homogeneous, independently from their printing orientation. A methodology for material testing of hollow structures in polymer printing is utilized for the first time to evaluate hollow samples of different materials. We have assessed that printing quality is more relevant than intrinsic mate- rial properties in determining the final mechanical performance of an object. In this sense, a bio-based compound developed with an industrial partner resulted in higher mechanical performance compared to polymers with superior nominal strength. The research establishes a set of computational procedures for the integrated design and engineering of large scale lattice structures, with an optimized strength to weight ratio. An innovative load-responsive process inspired by the bone internal structure is applied as a work- flow that allows for a multi-level material and structural optimization, connecting structural design and fabrication into a unique process. The workflow developed in this research is applied to the design andNaboni et al 7 Figure 17. The printing farm consists of five delta printers capable of producing an average of seven components a aay. Source: Roberto Naboni, making of a 36m? semi-permanent lattice structure installed in the University Campus of Politecnico di Milano (Figure 18). The established computational workflow allowed for a seamless data exchange and process automation in the construction of an extremely resistant and lightweight structure with a variable topology resulting in a weight-to-area ratio from 6 to 10kg/m?, shaped by a 112km-long extrusion (Figure 19), This innovative approach introduces a radically different strategy to the realization of lightweight architecture by adopting complex cellular organizations emerging as a result of a specific material, fabrica- tion and architectural setup, Conclusion and discussion This research outlines an approach based on principles of Construction 4.0 that involve design and fabrica~ tion automation, AM-based design process, and design tailoring. The developed construction takes full advantage of the use of FDM in the realization of a large scale structure with an unpaired control on material organization at multiple hierarchical levels. The methodology of this work can be implemented in design and fabrication at different scales, with prospective applications into material-efficient construction. ‘The approach also demonstrates that FDM can be successfully utilized in lightweight construction, by introduc- ing advanced design and redundant structural systems. Further development in this direction will involve advancements in material composition and the study of long-fibre reinforced FDM; the implementation of such system into an extended fabrication volume which would allow to print larger elements and take full advantage of the low weight of the system; the integration of real-time adaptive optimization of the machin- ing toolpath to account for effective material control and the establishment of a digital-twin toward a fully Industry 4.0 workflow:International Journal of Architectural Computing 00(0) Figure 18. The image showcases lattice variation derived from topology optimization pattern. source: Gabriele Seghiz, AAA Wekewemanics WR CEEOCORL ESE: Be eabuuinmuounasacedeececee Be sebbubiamenononiedadeces VB hha 4, tia eaeany eae Figure 19. The image shows a close-up view of the functionally graded cellular structure. Source: Gabriele Seghiz,Naboni et al 19 Acknowledgements ‘The authors wish to thank the entire team that has worked on the development of Trabeculae Pavilion, a research project, promoted by ACTLAB: Prof. Roberto Naboni (CREATE ~ University of Southern Denmark) and Prof. Ingrid Paoletti Politecnico di Milano). Research, Design and Development Team: Roberto Naboni (Main Investigator and Project Leader), Anja Kunié (Computational Design), Luca Breseghello (Computational Design), Fabrication and Construction Development: Mithun Kumar Thiyagarajan, Gabriele Seghizzi. In collaboration with Francesco Martelli (Structural Analysis) and ITKE - University of Stuttgart: Valentin Koslowski, Jan Knippers (Structural Analysis + Material Testing). Collaborators: Verley Henry Coco Ir., Rabul Sehgal, Elena Kriklenko, Hamed Abbasi, Maia Zheliazkova, Francesco Pasi Declaration of conflicting interests ‘The author(s) declared no potential conflicts of interest with respect tothe research, authorship, and/or publication of this article Funding ‘The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: The research has been developed with industrial partners: WASP (3D Printers), and FILOALFA® (Polymeric Materials). The project has been supported by contributions of Saperlab - Laboratorio Unico Dipartimento ABC (Politecnico di Milano) and MADE Expo. ORCID iDs Roberto Naboni © https:/orcid.org/0000-0001-9647-0426 Luca Breseghello (© https://orcid.org/0000-0001-S063-4595 References 1. Labonnote N, Reanguist A, Manum B, et al, Additive construction: state of the ar, challenges and opportunites Autom Constr 2016; 72(0t. 3): 347-366, 2. Delgado Camacho D, Clayton P, O'Brien W, et al. Applications of additive manufacturing in the construction industry ~a forward-looking review. Autom Constr 2018; 89: 110-119. 3. NaboniR, Breseghello L and Kunic A. Multiscale design and fabrication of the Trabeculae Pavilion. Addit Manuf 2019; 27: 305-317. 4. Gibson LI and Ashby MF. Cellular solids: structure and properties. 2nd ed. Cambridge: Cambridge University Press, 1997, p. 8 5. Liang H, David R, Chunze Y, Ahmed H and Philippe Y. Design and additive manufacturing of cellular lattice srue- tures. In: The international conference on advanced research in virtual and rapid prototyping (VRAP) (eds Bartolo PS, etal), 28 September October 2011, Leiria, Portugal 6. Morel P and Schwartz T. Automated casting systems for spatial concrete latices. In: Thomsen MR, Tamke M, Gengnagel C, eta. (eds) Modelling behaviour ~ design modelling symposium. Switzerland: Springer International Publishing, 2015, pp. 213-225. 7. Gaudiliére N, Duballet R, Bouyssou C, et al. Large-scale additive manufacturing of ultta-high-performance eon- crete of integrated formwork for Truss-Shaped Pillars. In: Willman J, Block P, Hutter M, etal. (eds) ROBARCH 2018, Robotic fabrication in architecture, art and design 2018, Switzerland: Springer Nature Switzerland AG, 2019, pp. 459-472 8. Hack N and Lauer WV. Mesh mould: roboticaly fabricated spatial meshes as reinforced concrete formwark. Archit Design 2014, 84: 44-53 9. Hack N, Lauer WV and Gramazio F. Mesh mould: differentiation for enhanced performance. In: Rethinking com- prehensive design: speculative counterculture. Proceedings ofthe 19th international conference on computer aided architectural design research in Asia CAADRIA 2014 (eds Gu N, Watanabe S, Fshan H, et a.), Kyoto, Japan, 14-16 May 2014, pp. 139-148. Kyoto: Kyoto Institute of Technology.20 International Journal of Architectural Computing 00(0) 10 u 12, 19, 20. au 22. 2B 24, 26. 20 28. 29. Branch Technology. Cellular fabrication, https://www branch technology/blogs/2017/8/3 /how-to-start-building- like-nature (2015, accessed 7 May 2020). Branch Technology and MIT, Digital Structures Lab. Optimized bench, https://www branch.technology/projects- 1/2017/6/9/03ql2cdrm68r8613whatjwSucl9dlu (2017, accessed 7 May 2020). Branch Technology SHoP Architect, SHoP Pavilion. Design Miami Pavilions. Miami, FL, https:/iwww.branch. technology/projects-1/2017/6/9/shop (2016, accessed 7 May 2020. Branch Technology and Range Projects, Site-Civil Engineering. One city pavilion. Nashville, T? branch.technology/projects-1/2018/11/8/one-city-pavilion (2018, accessed 7 May 2020), Ai Build | Arup. Daedalus Pavilion, Amsterdam, The Netherlands, https:/ai-build.com/daedalus.html (2016, accessed 7 May 2020), Ai Build. Concrete formwork. London, UK, https:/ai-build.com/concreteformwork html (2017, accessed 9 May 2020). Hajash K, Sparrman B, Guberan C, et al, Large-scale rapid liquid printing. 3D Print Addit Manuf 2017, 4(3) 122-131 Cheung KC, Digital cellular solids: reconfigurable composite materials, Doctoral thesis, Massachusetts Institute of Technology, Cambridge, MA, 2012. Hambali RH, Colik HK, Smith PC, ct al. Effect of build orientation on FDM parts: a case study for validation of deformation behaviour by FEA. In: iDECON 2010 - International conference on design and concurrent engineer- ing (eds Hashim, Melaka and Pembuatan ), 20-21 September 2010, Melaka, Malaysia, pp. 224-228. Malaysia: Penerbit Universiti, Universiti Teknikal Malaysia, Gibson LJ, Ashby MF and Harley BA. Cellular materials in nature and medicine. Cambridge: Cambridge University Press, 2010, p. 166, Benyus JM. Biomimicry. Innovation inspired by nature. Reissued in Perennial 2002, New York: HarperCollins, 2002, Raggatt LJ and Partridge NC. Cellular and molecular mechanisms of bone remodeling. J Bio Chem 2010; 285(33): 25103-25108, Turner JS, Evolutionary architecture? Some perspectives from biological design, Archit Design 2012; 82(2): 28-33 Waimer F, La Magna R and Knippers J. Nature-inspired structural optimization of freeform shells. In: Cruz PJ (ed. Structures and architecture: concepts, applications and challenges, Guimaraes, Portugal, 24~26 July 2013, pp. 1177-1182, London: Taylor & Francis Group. Hildebrand T and Riiegsegger P, A new method for the model-independent assessment of thickness in three-dimen- sional images. J Microse 1997; 185(1): 67-75. ‘Naboni R and Kunic A. Design and additive manufacturing of lattice-based cellular solids at building scale. In: Proceedings of the XXI SIGraDi conference - Congreso Internacional de ta Soctedad Iberoamericana de Grafica Digital (ed. Herrera PC). Concepeién, Chile, 22-24 Novernbet 2017, pp, 369-375, Sio Paulo: Blucher. ‘Naboni R, Kuni A, Breseghello L, etal. Load-responsive cellular envelopes with additive manufacturing. J Facade Des Eng 2017; S(1): 37-49, Bendsoe MP and Kikuchi N. Generating optimal topologies in structural design using a homogenization method. Comput Methods App! Mech Eng 1988; 71(2): 197-224, Rozvany GIN, Zhou M and Birker T. Generalized shape optimization without homogenization. Struct Multidiscip 0 1992; 4(3-4): 250-252. ‘Lorensen WE and Cline HE. Marching cubes: a high resolution 3D surface construction algorithm. ACM SIGGRAPH Comp Graph 1987; 21(4): 163-169, hutps://www.