Modern scan-to-BIM approaches often use machine learning (ML) models to reconstruct BIM models from point clouds. Their performance relies heavily on a large amount of training data. Generating this data requires a significant amount of manual effort for outlier elimination and labelling. The approach uses a foundation model (FM) for the generation of point cloud training data. The presented generator enables the training of AI algorithms for the scan-to-BIM method without additional hardware. It also allows future researchers to generate data for a scan-to-BIM workflow.
Constructing digital models of firefighting assets is essential for informed decision-making in emergency response and management. However, current practices struggle to efficiently recognize and update these assets in building information models (BIM). This study proposes a framework integrating photogrammetric reconstruction and instance segmentation to enrich BIM. The framework involves creating an expanded firefighting asset dataset and leveraging panoramic images with supervised learning for scene reconstruction and asset segmentation. Real-world evaluation illustrates it performs satisfactorily in both asset recognition and positioning. The framework offers a practical solution for modeling digital twins to support various fire emergency applications.
The full carbon impact of the digitalisation of construction processes is somewhat neglected due to varying measurement approaches not fully tailored to design and construction processes. Utilising Jackson and Hodgkinson's data carbon ladder, our paper proposes a framework to measure the carbon emissions of digital design processes, from which the carbon impact of common project files is evaluated. Results demonstrate significant emissions come from data storage and transfer, highlighting the need for optimised digital workflows. Our framework increases awareness of the digital carbon footprint of construction projects and helps organisations adopt sustainable digital practices to align with Net Zero targets.
Digital twins have become transformative tools in design and operations, providing critical capabilities for real-time monitoring and management of building assets. However, creating high-quality digital building models required for the digital twinning of the built environment on a large scale remains challenging and requires significant human effort. This paper introduces an AI-based end-to-end automatic procedure for the creation of digital building models using point clouds and RGB images. The results demonstrate the effectiveness of the proposed method across multiple case studies, achieving an average accuracy of approximately 7 cm in estimating the parameters of the building's structural and opening elements.
Digital Twin Construction (DTC) is a data-centric approach to construction management, enabling real-time “Plan-Do-Check-Act” (PDCA) cycles for production control. Despite its potential to streamline processes and reduce waste, DTC remains untested in full-scale projects due to technical and practical barriers. To explore the feasibility of DTC, we therefore developed a controlled lab setup using a 1:25 scale model and automated monitoring to simulate DTC-driven planning. The setup enables experimentation across different levels of automation with the goal of testing feasibility and viability for real-time planning and control. This research contributes to understanding DTC’s potential to transform production control in construction.
The expanding body of literature on citizen empowerment in public decision-making underscores its significance in enhancing fairness, acceptance, and urban sustainability. This research first reviews existing literature on public participation and previously developed participation models. Next, it analyzes participatory design tools, evaluating their strengths and weaknesses with regard to their applications in urban regeneration actions. It then presents the functionalities of the participatory urban design tool MUST (Manage Urban Spaces Together), examining how its functionalities can be extended to support neighborhood regeneration and management in diverse urban contexts.
Automated design compliance checking has often overlooked graphical descriptions in regulatory clauses. This study investigates the use of multimodal large language models (MLLMs) to derive executable rules from both text and graphics. Using 23 accessibility regulation clause-graphic pairs with Prolog-based ground truth rules, MLLM outputs were evaluated under three input conditions: text-only (F1: 0.74), graphic-only (F1: 0.43), and combined (F1: 0.96). Results highlight the effectiveness of multimodal inputs, demonstrating MLLMs’ potential for accurate rule derivation in design compliance.
YOLOv11, a CNN-based object detection and instance segmentation algorithm, is used to automatically capture Masonry Quality Index (MQI) parameters for existing masonry structures. Training is performed using a suitable dataset for detecting bricks, and its hyperparameters are adjusted systematically for optimal accuracy. A workflow is proposed in which models are trained on the "MCrack1300" dataset and evaluated using orthomosaics of an unreinforced masonry building. Optimal hyperparameters are determined iteratively, and their impact on minimizing loss is compared. The proposed model captures block size distributions and staggering ratios associated with the construction quality of masonry walls.
The architecture, engineering, and construction supply chain is predominantly composed of small and medium enterprises (SMEs). Despite the growing adoption of digital tools such as Building Information Modeling (BIM), SMEs often face challenges streamlining these tools with existing enterprise resource planning (ERP) systems. This paper presents a study that worked with an SME to develop a "customized data environment", inspired by common data environment approaches, linking their BIM and ERP workflows. The paper describes the action research cycle used to develop the prototype. Insights into both the tool and the process evolution will help other SMEs facing similar integration challenges.
