In interior space planning, the furnishing stage usually entails manual iterative processes, including meeting design objectives, incorporating professional input, and optimizing design performance. Machine learning has the potential to automate and improve interior design processes while maintaining creativity and quality. The aim of this study was to develop a furnishing method that leverages machine learning as a means for enhancing design processes. A secondary aim was to develop a set of evaluation metrics for assessing the quality of the results generated from such methods, enabling comparisons between the performance of different models. To achieve these aims, floor plans were tagged and assembled into a comprehensive dataset that was then employed for training and evaluating three conditional generative adversarial network models (pix2pix, BicycleGAN, and SPADE) to generate furniture layouts within given room boundaries. Post-processing methods for improving the generated results were also developed. Finally, evaluation criteria that combine measures of architectural design with standard computer vision parameters were devised. Visual architectural analyses of the results confirm that the generated rooms adhere to accepted architectural standards. The numerical results indicate that BicycleGAN outperformed the two other models. Moreover, the overall results demonstrate a machine-learning workflow that can be used to augment existing interior design processes.

The application of machine learning (ML) for the automatic classification of building elements is a powerful technique for ensuring information integrity in building information models (BIMs). Previous work has demonstrated the favorable performance of such models on classification tasks using geometric information. This research explores the hypothesis that incorporating contextual information into the ML models can improve classification accuracy. To test this, we created a graph data structure where each building element is represented as a node assigned with basic geometric information. The connections between the graph nodes (edges) represent the immediate neighbors of that node, capturing the contextual information expressed in the BIM model. We devised a process for extracting graphs from BIM files and used it to construct a graph dataset of over 42,000 building elements and used the data to train several types of ML models. We compared the classification results of models that rely only on geometry, to graph neural networks (GNNs) that leverage contextual information. This work demonstrates that graph-based models for building element classification generally outperform classic ML models. Furthermore, dividing the graphs that represent complete buildings into smaller subgraphs further improves classification accuracy. These results underscore the potential of leveraging contextual information via graphs for advancing ML capabilities in the BIM environment.

This paper presents an assessment and comparison of the effects of static and kinetic external shading elements on the dynamic measurement of daylighting. For this purpose, we used a method and parametric tool developed previously for the design and analysis of external shading elements in buildings. The proposed approach was used to compare static and dynamic movement scenarios for achieving optimal internal adjusted useful daylight illuminances (AUDI). The current paper presents the results of a methodical analysis, which compared various types of louvers in static and dynamic operation scenarios for a typical office in a Mediterranean climate. The results show that dynamically adjusted louvers perform notably better than fixed or seasonally adjusted modes of operation. The results show that dynamic operation scenarios can increase the AUDI by up to 51%. The results also show that in some conditions the existing rules of thumb fail to predict the correct design approach to louver geometry and that the use of rules of thumb in architectural daylight design needs to be revaluated.

Recent advancements in Machine Learning (ML) have significantly enhanced the capability to generate 3D objects from textual descriptions, offering significant potential for design and fabrication workflows. However, these models typically fail to meet practical requirements like printability or manufacturability. In addition, current approaches often cannot accurately control the dimensions and the interrelations of elements within the generated 3D models. This presents a major challenge in utilizing ML-generated designs in real-world applications. To address this gap, we introduce a novel method for translating natural language descriptions into parametric 3D objects using Large Language Models (LLMs). Our approach employs multiple agents, each agent is LLM pre-trained for a specific task. The first agent deconstructs textual prompts into design elements and describes their geometry and spatial relations. The second agent translates the description into code using the Rhino.Geometry coding library in the Rhino3D-Grasshopper modeling environment. A final agent reassembles the models and adds parametric control interfaces, enabling customizable outputs. In this paper, we describe the method’s architecture and the training methodologies, including in-context learning and fine-tuning. The results demonstrate that the suggested method successfully generates code for variations of familiar objects. However, it still struggles with more complex designs that deviate significantly from the training data. We outline future directions for improvement, including expanding the training dataset and exploring advanced LLM models. This work is a step towards making 3D modeling accessible to a broader audience, using everyday language to simplify design processes.

The construction industry is a major contributor to global CO₂ emissions, necessitating sustainable alternatives for building materials. Additive manufacturing (AM) using wood-based composites offers an eco-friendly solution for thermal insulation applications. This study explores the thermal and mechanical properties of wood powder–carboxymethyl cellulose composites fabricated via liquid deposition modeling (LDM). Six formulations incorporating industrial wood waste from beech and oak, with varied particle sizes, were developed to evaluate their extrudability, structural stability, and insulation efficiency. Material characterization included thermal conductivity testing via the transient plane source method and compressive strength assessment following ISO standards. Results indicate that particle size and wood species significantly influence material properties. Finer wood particles yielded higher compressive strength, whereas coarser particles exhibited lower conductivity, enhancing insulation performance. The best-performing formulation (B2: beech wood, medium particle size) demonstrated a balanced thermal conductivity of 0.188 W/m·K and compressive strength of 3 MPa. To assess large-scale buildability, a 3D-printed block component (200 × 350 × 220 mm) was fabricated. A refined formulation with reduced water improved print stability, demonstrating the viability of LDM for producing rigid, lightweight insulation blocks. This research establishes a foundational understanding of AM wood composites for thermal insulation, offering insights into material formulation, printability, and structural behavior. The findings underscore the potential of bio-based AM in sustainable construction, paving the way for scalable applications of wood waste in energy-efficient building systems. Future work will focus on optimizing binder composition, refining printing strategies, and exploring reinforcement techniques to enhance mechanical properties while maintaining thermal efficiency.

This paper presents a framework for an analysis method and computational tool, which evaluates the fabrication parameters of complex geometries. The suggested method predicts the feasibility, material use, and machining time required for fabricating the moulds for these geometries. It achieves this by interrogating geometric properties instead of the traditional machining simulations. Using the algorithms developed in this research, the method can provide real time evaluation of computer-controlled mould fabrication techniques such as cutting and assembling sheet materials, multi-axis milling of volumetric material, and robotic hot wire cutting. In the paper, we describe the mathematical basis of the suggested method. We demonstrate how the method provides real-time visual feedback for designers and allows them to adjust their design according to fabrication constraints in the early design stages. Using architectural case studies, we show how the analysis results provide precise cost estimates and help minimize fabrication resources in manual or automatic fabrication optimization processes.