Architectural intelligence最新文献

This paper investigates the development and optimization of artificial reefs through a new generative design method that integrates Computational Fluid Dynamics (CFD) with Bi-directional Evolutionary Structural Optimization (BESO). Since the 1950s, artificial reefs have been deployed to enhance marine ecosystems, and this study begins with a survey of existing reef designs. In response to limitations in current design approaches, we adopt a topology optimization strategy aimed at improving spatial allocation for polyphony expansion within reef structures. By coupling fluid-dynamic analysis with an iterative optimization loop, we evaluate the effectiveness of material exchange enabled by these artificial formations—an essential consideration given advanced manufacturing constraints and the need for rapid production of natural-like geometries. To extend artificial reef design into new possibilities, we propose a generative workflow in which reef morphology emerges from the interaction between CFD and BESO, iteratively removing and adding material in accordance with external loading conditions. The resulting reef is then compared with representative benchmarks from current artificial reef designs to assess material efficiency, structural performance, and geometric characteristics under complex underwater conditions.

Computer Numerical Control (CNC)-knitted textiles are flexible, lightweight, and highly customisable, which makes them promising materials for architectural and construction applications. In the context of tensile structures, both the final shape and the mechanical properties of knitted textiles can be controlled to follow a specific design intent. However, predicting their mechanical behaviour is challenging and currently requires experience, domain-specific knowledge, and extensive prototyping. Developing a computational method to design bespoke knitted textiles for a target geometry and behaviour is therefore essential. The proof-of-concept workflow introduced in this paper uses the Force Density Method (FDM) combined with gradient-based optimisation to compute force density distributions for a target geometry abstracted as a mesh. These force densities are discretised into domains and mapped to knit architectures with distinct deformation capacities, resulting in functionally graded textiles. The workflow is tested on a non-symmetric target geometry and evaluated through physical prototyping. The results highlight both the potential of the approach and the need for refined force density–knit architecture mapping and alternatives to prototyping. This computational method paves the way for material-informed form-finding, which can facilitate the integration of CNC-knitted textiles into architectural applications, such as flexible formwork.

Large old trees provide essential habitats for birds and many other species, yet they are rapidly disappearing from many landscapes. While artificial habitat structures have been trialled, their design rarely captures the morphological complexity of natural habitats. This limitation stems from both challenges in extracting relevant features from natural forms and the difficulty of developing cost-effective systems that can be reproduced at scale. This paper addresses this gap by presenting FloaTree, an experimental example of a human–machine design workflow to generate, optimise, and construct tensegrity structures derived from AI-generated visual abstractions of large trees. We developed a parametric workflow that translates such AI-generated polyline abstractions into X-module tensegrity configurations, refined through structural optimisation and represented via connectivity matrices. Iterative prototyping, from small-scale tests to an eight-module pavilion, validated the structural and constructability aspects of this workflow and culminated in the winning entry of the 2024 IASS “Design Competition and Exhibition of Innovative Lightweight Structures” in Zurich. The results demonstrate that tensegrity structures, typically confined to artistic installations or used with limitations as surrogates for other typologies, can be designed for packability, transport, and rapid low-tech assembly to enable their potential application in artificial habitat structures. The project also advances tensegrity design methods through a novel human–machine workflow and a visualisation technique based on connectivity matrices. It shows how the analogue and digital domains can co-exist in design workflows alongside emerging forms of human–AI collaboration. While ecological performance requires future field testing, the significance of this work lies in reframing tensegrity not only as an experimental artefact but as a transferable design framework integrating form abstraction, structural logic, and constructability, thereby suggesting broader applications for computationally optimised yet low-tech structures in disturbed landscapes.

Conventional 4D printed actuators often embed programmability in a single hygroscopic layer. As a result they bend mainly in one direction and recover slowly. In this paper we explore the integration of a myocardium inspired thermoplastic polyurethane (TPU) matrix between a rigid PLA base and a hygroscopic cellulose PLA active layer, forming a three material layered system that delivers rapid multi axis deformation without external power. Tile geometry and material assignment are generated in Rhino and Grasshopper and exported as G code, yielding lightweight modules suitable for large scale fabrication. Humidity cycling demonstrates three programmable motion modes (doming, slit opening, and hinge like rotation) obtained solely by adjusting the geometry of the TPU matrix. Tiles that incorporate the TPU matrix return more quickly to their initial flat shape and maintain a stable deformation amplitude within a given specimen over repeated cycles, because the matrix limits over curvature and prevents reverse bending in over dry conditions. By adding this strategically placed TPU matrix, the system converts simple bending elements into durable, zero energy actuators capable of complex and reversible transformations, offering a potential route toward self shaping facade elements for sustainable kinetic architecture.

Urban walkability is a critical determinant of health, safety, sustainability, and city life in general, yet most existing indices remain limited to amenity proximity and neglect the perceptual and morphological qualities that shape the walking experience. This paper proposes a Graph Machine Learning-centered multi-agent framework that integrates Large Language Models (LLMs), computational analysis, and human feedback to design and evaluate urban interventions. The framework positions Graph Machine Learning (GML) as the central predictive engine, capable of modeling network relationships and testing hypothetical scenarios, while LLMs act as perception interpreters that translate visual and textual information into experiential insights. Human agents validate and contextualize these results, ensuring alignment with policy and lived experience. A new Street Walkability Index (SWI) is introduced, combining traditional Walkscore metrics based on land-use and perception-derived data, to provide a multidimensional measure of walkability. Applied to Mexico City’s historic center, the system demonstrates improved predictive accuracy and interpretability compared to conventional models. Ablation studies confirm that integrating perceptual and topological features enhances performance, while intervention modeling shows the framework’s ability to simulate and evaluate interventions such as building massing and architectural program change. These results suggest that multi-agent GML systems offer a powerful decision-support approach for participatory urban evaluation, bridging data-based, perceptual, and human intelligences toward more equitable and actionable urban design strategies.

As urbanization accelerates and population density rises, noise pollution increasingly undermines residents’ quality of life. While the link between noise distribution and urban morphology has been widely established, most studies focus on two-dimensional noise patterns and overlook its continuous three-dimensional variation. This study proposes a method that integrates urban section analysis with noise simulation. Taking typical residential forms in Nanjing as case studies, we treat profiles as core analytical units and incorporate noise simulation data from CadnaA to generate section-based maps that illustrate noise propagation paths and intensity variations. Based on a 50 threshold, high-noise and quiet areas are identified, and an area density index is introduced to enable quantitative comparison across residential morphologies. The results reveal that building form and layout significantly influence noise propagation and attenuation. Compact low-rise layouts foster stable quiet areas, whereas high-rise and wide-spaced layouts often create noise corridors that channel noise deeper into the site. Across cases, a consistent spatial pattern emerges: noise attenuates from the perimeter inward; the first building row provides substantial shielding; and noise troughs frequently occur at interfaces between buildings and open spaces. This section- based framework advances the analysis of coupling between urban morphology and acoustic environments and offers a rigorous basis for optimizing residential layouts and guiding urban acoustic environment management.

The exploration of quantitative approaches to urban fabric, grounded in urban morphology theory, provides digital tools for accurately characterizing its physical attributes. Existing quantitative methods, however, have primarily focused on general morphological features such as dimension, shape, and density, leaving a gap in the development of methodologies targeting building types and their compositional relationships. In this study, two historical districts in China—Datong and Nanjing—are examined as case studies. By applying graph theory, we quantify the compositional patterns of urban fabric, defining plots, courtyards, and building typological units as nodes. This approach reveals the structural and typological differences between the two cases. The results demonstrate that integrating quantitative metrics such as building width, depth, and courtyard configuration with graph-theoretic indicators—including node degree and depth values—effectively uncovers distinct organizational patterns within the fabric. Furthermore, metric intervals derived from these indicators provide a means to differentiate the morphological characteristics of each district. This study highlights the potential of combining morphological analysis with graph theory to enhance the understanding of the complexity and diversity of urban fabric in historical contexts.

Current timber construction predominantly employs solid mass‐timber panels, such as cross‐laminated timber. While these solid construction systems simplify assembly processes, they consume large volumes of material. The potential for targeted, efficient material placement offered by additive manufacturing (AM) has yet to be fully realized in the context of solid wood, largely due to its anisotropy, variable material quality, and the difficulty of incorporating its continuous fibers into printable mixes. Wood is therefore usually ground into particles within varying binder mixtures. We developed a method that uses automated lamination of veneer filaments using PUR adhesives, which has the advantage of keeping intact and continuous natural long wood fibers. This study introduces a workflow that integrates finite‐element (FE) modeling and optimized principal stress‐line (PSL) for form finding of highly material-efficient structures. With FE analysis, principal stress trajectories are identified, filtered, and translated into deposition paths. A custom robotic end‐effector was developed to deposit stress direction-aligned linear wood filaments. We developed slab components that consist of a thin timber sheet that acts as a printing surface, as a functional outer layer, and out of 3D-printed reinforcement ribs. This approach is benchmarked with small‐scale fused deposition modeling (FDM) polymer specimens against a large‐scale series of robotically produced specimens using printed beech veneer filament as reinforcement ribs. This paper evaluates the method’s feasibility and structural behavior, highlighting its potential to advance renewable‐material AM and foster more sustainable, resource‐efficient construction practices.

Land use/cover change (LUCC) is recognized as one of the key drivers of global warming and extreme weather events. Most current articles on urban LUCC focus on megacities or regional urban clusters, while fewer studies have been conducted on medium-sized cities growing steadily and rapidly. Wuxi, the center of the Yangtze River Delta (YRD) region in China, has been experiencing rapid and sustained economic growth since 1981, with its per capita GDP approaching the level of moderately developed cities worldwide. By analyzing the characteristics of land use type, temperature, and precipitation in the sample of this typical rapidly urbanizing medium-sized city, we can reveal the quantitative relationship between the impacts of different land use cover types on the natural factors. It also provides new data and analytical perspectives to understand how land use/cover changes in urbanization affect climate change. Based on the remote sensing and meteorological data of Wuxi City from 1990 to 2020, we quantitatively calculate the single land use dynamic index, land use intensity index, and land use transfer matrix. In addition, we dynamically analyze the spatial and temporal trends of LUCC and natural factors by using the center of gravity model and standard deviation ellipse model to find out whether there is any correlation and consistency between them. Based on this, the gravity center model is combined with statistical methods to obtain the equilibrium equations composed of the center of gravity shifts of different land use types and natural factors, to clarify the quantitative influence of different land use types on the changes of various natural factors. The study shows that cultivated and construction land areas changed drastically from 1990 to 2020. The simulation results of the center of gravity model indicate that all types of land use correlate with climatic factors, and that water area has the greatest positive influence on precipitation. Grassland has the greatest positive influence on temperature.