Architectural intelligence最新文献

This research focuses on the design, fabrication, and structural and embodied carbon analysis of the world’s first topologically optimised multi-metal I-beam. Specifically, the beam under study is a European Parallel I-beam with a nominal height of 100 mm (commonly referred to as ‘IPE-100’), and the materials used are mild steel and tool steel. Topology Optimisation (TO) is performed using Altair’s OptiStruct software package, applying the Solid Isotropic Material with Penalty (SIMP) method. The multi-metal beam is fabricated using 3D printing, specifically Laser Metal Deposition (LMD), with a dual built-in metal wire feeder attached to a robotic arm. The beam is analysed both environmentally and structurally — the former focusing on an embodied carbon assessment of material extraction and component manufacturing, and the latter on four-point structural load testing. The fabrication method and analysis results are compared with those of the standard IPE-100 beam currently used in construction. Environmentally, the Multi-Material Topologically Optimised (MMTO) beam’s reduced mass results in lower carbon emissions compared with the standard IPE-100; however, due to the high emissions associated with its fabrication process, its overall carbon footprint is higher. Structurally, the MMTO beam can withstand a higher machine load than the standard IPE-100 before undergoing plastic deformation. This research is the result of an international, multidisciplinary collaboration between academia and industry across the United Kingdom, Germany, and Spain.

Architectural design relies heavily on rich and multimodal knowledge—including text descriptions, detailed tables, and complex visual information—to inform creative and technical decision-making. However, effectively retrieving and generating meaningful insights from such diverse data sources remains challenging. In this study, we propose Multimodal Retrieval-Augmented Generation (M-RAG), a multimodal retrieval-augmented generation framework that integrates a Multimodal Large Language Model (MLLM) with a unified knowledge base via a shared semantic embedding space, dual-mode table decomposition (text and image), and adaptive query handling to enable traceable cross-modal retrieval and augmented answer generation. We use a multimodal embedding model to project text, table-rendered images, and diagrams into a common space. Quantitative evaluation on the dataset employed in the study reports average text–image similarity of 0.55 ± 0.07; we use an operational threshold θ = 0.5 to flag likely out-of-domain queries and trigger clarification, reducing low-relevance returns. We evaluate the M-RAG framework across multiple architectural scenarios—including reducing technical hallucinations in structural-engineering queries by returning supporting diagrams, enabling special-technical comparison reasoning, handling floorplans and sections, supporting regulation and compliance lookups, retrieving sustainable concept design images and explanations, facilitating material queries, and demonstrating depth-first and breadth-first exploration examples—showing its practical usefulness. By quantifying embedding alignment, retrieval performance, and an operational relevance threshold, M-RAG delivers a more reliable and traceable multimodal retrieval and generation capability for building engineering tasks.

Most 4D-printed hygromorphic mechanisms using fibre-polymer composite filaments are thin laminated bilayer composite structures, which are printed flat on the 3D printer bed and shape-change into single or double-curvature geometries in response to moisture uptake. The tendency towards thin laminated 2D-like bilayers might be traced to early research on shape-change laminates, such as bimetals, or to the vertical layering sequencing of most additive manufacturing methods. Recently, actuators with non-planar geometries have been achieved through complex rotational 3D printing, 5 or 6-axis printers, or multi-step repositioning of parts, but these methods require bespoke equipment or intensive post-processing, which greatly limits their application to the ubiquitous 3-axis gantry system. In this paper, we introduce a strategy for vertically printing non-planar shape-changing hygromorphic structures that requires only one print session and minimal post-processing. We demonstrate the novel capability of the presented strategy by creating actuators with double-curvature geometries based on the biological models of the Bauhinia variegata seed pod and the Lilium ‘Casa Blanca’ tepals. This novel approach reduces the complexity of manufacturing intricate 4D-printed actuators that have doubly curved geometry. The possibilities in 4D-printed shape parameters are expanded through this method, considering that dimensional curvature and form can now be more easily introduced to structures previously limited by planar printing.

Different users have unique housing needs; however, in the current residential design process, communication based on two‑dimensional architectural drawings requires a certain level of technical literacy; non‑professional users often find the interaction modalities difficult to use and are therefore unable to effectively express their design preferences. This study addresses this issue by introducing virtual reality (VR) into the participatory design process, developing VR-Aided Design as a solution. This system utilizes speech input and ray casting to enhance user interaction. To validate the feasibility of the proposed method, we applied it to real-life residential design cases and conducted thorough testing of the system. The communication functions of SketchUp, VR Sketch, and VR-Aid Design were compared and analyzed using protocol analysis. The results of the study demonstrate that the use of VR in the design process improves interaction, communication, and understanding, leading to increased user engagement with the design. Results showed the system effectively increased engagement in the design process and enhanced users’ perception and understanding. Focus interviews confirmed the system enabled easy design adjustments.

Driven by technological innovation and emission standards, the actual emission level of road motor vehicles is far lower than the design value of the current standard, so it is necessary to update the outdated parameters in the standard for research. The estimated emission factors were obtained based on the updated statistical data of Chinese vehicle models and the comprehensive baseline emission factors, and the inter-annual variation was found to be far more than 2% of the standard setting, and a grey prediction model was established based on the deduced emission factor sequence. The predictive emission reduction rates of gasoline vehicle CO, diesel vehicle NO_x and PM_2.5 are 14.5%, 4.7% and 10.1%, respectively. Field tests were carried out in the Yanglin extra-long tunnel. Through multiple regression analysis, the emission factors of gasoline vehicle CO, diesel vehicle NO_x and diesel vehicle PM_2.5 are 0.79 g/ (km·veh), 5.64 g/ (km·veh) and 0.149 g/ (km·veh), respectively. The measured emission factor results are significantly lower than those in the China Guidelines 2014, with a difference of more than 80%, much closer to the levels reported by PIARC. It is proved that the long base year and unreasonable reduction rate selection are the key reasons for the deviation from the actual standard emission factor. Combined with the measured data and predicted model data, this paper proposes to update the baseline emission factors and reduction rates of ventilation design, taking 2020 as the base year, the baseline emission factors for CO, NO_x and PM_2.5 are 0.46 g/ (km·veh), 4.72 g/ (km·veh) and 0.027 g/ (km·veh), respectively, and the emission reduction rates are 12%, 4.7% and 10.1%, respectively. Through case calculation, the air demand after the parameter update was 70% lower than that before the update. The results can provide reference for tunnel ventilation design in China.

Free-form space frames enable striking architectural designs but often require complex 3D nodes, leading to high construction costs. This paper presents a cost-effective solution using the curved-crease origami-inspired approach to achieve flat-foldable nodes. The proposed node features a self-locking mechanism that is activated upon reaching its 3D target state and is further reinforced by member insertion, preventing unfolding. Despite strict planarity and straightness design constraints, the proposed node can achieve vertex valences of two, three, and four, enabling diverse planar and non-planar configurations, demonstrated through a series of numerical and physical examples. Additionally, an optimization tool is developed to verify and modify input meshes for constructability. Results show that the proposed node can effectively realize double-ruled surfaces that naturally satisfy design constraints and support double-layer systems for free-form realizations. Beyond space frames, the node’s applications extend to tree-like structures, with key challenges and future research directions discussed.

The dispersion of dust during vehicle unloading in semi-open industrial buildings poses environmental and health risks. Air curtain is normally used as one of the effective dust removal technologies, while its performance under natural wind around an isolated semi-open building has been rarely reported. This study aims to evaluate the potential of air curtains as a solution for this issue. To validate the used turbulence model and numerical methods, comparisons are made with reported wind tunnel experimental results. The isolated semi-open industrial building is subsequently established to assess the impact of air curtains at varying jet velocities (6, 8, 10, 12, 14, 16 and 18 m/s) and jet angles (0°, 5°, 10° and 15°) on pollutant leakage. Results show that as the air curtain flow speed increases from 6 m/s to 14 m/s, the control efficiency rises from -42.1% to 98.1%. Meanwhile, further increases of flow speed show a slight decrease in efficiency, down to 97.18%. This study provides preliminary insights into the application of air curtains in semi-open industrial buildings.

To overcome the limitations of Artificial Intelligence (AI) in the field of architectural design, particularly regarding issues of interoperability, domain-specific knowledge and interdisciplinary, we propose an innovative multimodal AI platform—FUGenerator. Within this framework, we designed a multimodal knowledge graph, a multimodal algorithm library and a traceable workflow. Additionally, by implementing advanced AI technologies such as Natural Language Processing (NLP), image processing, and 3D model generation, the platform is capable of processing diverse formats of input data and generating preliminary design proposals that are accurate and personalized design solutions. During the application experiment, students used the platform in their fourth-year undergraduate design projects. The outcomes demonstrated its effectiveness in not only generating diverse design alternatives based on various design requirements, but also significantly improving design efficiency and flexibility in different scenarios. With further optimization and expansion, the platform can become the multimodal intelligent support tool throughout the entire design and construction process, driving the digitalization and intelligent evolution of architectural practice.