Cost estimation and as-built documentation on construction sites require extensive manual work. Combining LiDAR with computer vision offers an effective solution for 5D BIM applications. Therefore, two deep neural networks are trained on a database of over 1000 images of radiators, as example of MEP components. Transfer learning has been applied on SOLOv2 for instance segmentation and on YOLOx for object detection. Given the variety of the image database, the generated models achieve satisfying performance, suitable for the intended applications. The developed models will be used in a scan-to-BIM workflow, enabling the semi-automatic capture of as-built information.
Previous expert systems for excavation support system (ESS) recommendations often overlooked lateral supports and water-barrier methods, limiting their applicability to small projects with a single retaining wall type. These methods were unsuitable for large-scale projects, where varying site conditions require different ESS configurations for each excavation side. This study proposes an ESS design assistant tailored for large construction projects. While LLMs like GPT-4.0 and DeepSeek show promise, they struggle with practical ESS recommendations. To address this, we developed a multi-step machine learning-based ESS recommendation system, achieving weighted F1 scores of 0.951 (retaining walls), 0.981 (water-barrier applications), and 0.831 (lateral supports).
This study aims to enhance community safety and resilience by assessing post-earthquake risks using Google Street View images for 3D data acquisition in the Nurdagı district, Gaziantep. A total of 5548 images were collected via 360° and 180° scans, then processed using the ’Structure from Motion’ algorithm to construct a 3D point cloud, generate depth maps, and create surface meshes. Texture mapping was applied for successful 3D modeling of structures. This innovative approach utilizes Google Street View images as an alternative to traditional 3D urban modeling, offering detailed data resources and significant contributions to urban planning, architectural conservation, and structural engineering.
Digital twins are increasingly applied in the construction industry. Utilizing real-time, structured data from digital twins for quality control remains a critical challenge. This paper defines the evolution mechanisms of defect triggers through a literature review. The quality linked-data is designed by using Resource Description Framework and related technical concepts to transform multi-source heterogeneous data into structured linked-data. A workflow with process patterns for defect trigger identification is developed to invoke the quality linked-data. The framework is validated with the reinforcement cage length deviation in cast-in-place piles. This work provides theoretical insights for automated defect trigger identification with digital twins.
Determining the number and positions of individuals inside a building is critical for effective data management and rescue operations during building disasters. Although recent vision-based methods have been developed for people counting and tracking, they often struggle to process video data from non-overlapping zones. To overcome these challenges, this study proposes a novel multi-camera analysis methodology that integrates a camera network with Re-Identification (Re-ID) models to manage personnel information. Experimental results demonstrate that the system provides precise zone-specific counts and locations of individuals. This approach can significantly improve emergency response by enabling strategic resource allocation and prioritized access.
Recent years have seen a surge in development of formal ontologies within the domain of built environment. Yet, experts like novices face difficulties in locating relevant ontologies, hampering the use of ontologies. Besides, these ontologies demonstrate a range of development maturity. This highlights a gap: the need for a continuously updated repository to help users in discovering, evaluating and (re)using ontologies. This paper presents the Ontology Lookup Service for the Built Environment: BE-OLS, which allows to easily find and consult existing ontologies. It also enables researchers to find gaps or redundancies to help bring some order to the current state of the field.
The EU's energy targets highlight the importance of retrofitting older buildings to reduce carbon emissions. However, many rural properties remain in the lowest energy rating categories, complicating retrofitting efforts. Urban buildings dominate Energy Performance Certificates (EPC) models, while rural structures require tailored approaches due to their diversity and lower energy performance. This research compares machine and deep learning models to address gaps in predictive accuracy and scalability in retrofitting simulations. The methodology predicts EPC ratings based on renovation policies and improves regional segmentation and archetype classifications. These strategies offer insights for rural residential buildings aligned with EU energy efficiency standards.
Digitalization is transforming the construction industry, yet gaps persist between industry needs and educational training. Current curricula emphasize broad digital literacy but lack structured training in AI, BIM, and automation, leaving students underprepared for digital workflows. This study applies a mixed-methods approach, combining surveys and interviews with students and educators, to assess digital skill gaps. Findings highlight mismatches between student expectations, educational priorities, and industry demands. The study underscores the need for interdisciplinary collaboration, structured AI training, and closer industry-education alignment. Recommendations include curriculum enhancements and practical exposure to digital tools to better equip future construction professionals.
Technological advancements and methods for tracking CO₂ within the construction industry have enhanced emissions monitoring. However, traditional CO₂ data collection methods often rely on manual processes that lack real-time capabilities, limiting their effectiveness in offering meaningful guidance for reducing emissions. This study employs a literature review and expert opinions to examine current practices and advancements in CO₂ tracking for the construction sector, assessing their effectiveness and potential for widespread industry adoption based on European Union regulations. Ultimately, this work contributes to the advancement of CO₂ monitoring by providing recommendations that meet regulatory requirements and stakeholder demands for sustainable construction practices.
A framework for monitoring HVAC data streaming and detecting anomalies is presented. Latent Dirichlet Allocation identifies three states corresponding to heating, cooling, and baseline system behaviors, while the Dirichlet distribution detects proportion anomalies independent of total data volume. Using 34 months of sensor data from an academic-residential building, the framework reveals seasonal and operational trends, with notable anomalies linked to maintenance events. This method enables facility managers to monitor system states and diagnose deviations efficiently but requires sufficient historical data and expert state interpretation. This low-complexity approach provides a practical tool for real-time HVAC monitoring in dynamic environments.
