An international, standardised product catalogue format representing parametric and variable product data, simplifying their exchange and integration, is to be developed. We map product data requirements in IFC files by leveraging lessons from existing national standards to enhance their parametric modelling capabilities. Using a case study, we evaluate the implementation of property value dependencies and valid combinations using a product catalogue example. Significant is the extension of data linking and value calculation in IFC, which integrates control structures. Of the three proposed implementation approaches, we favour the code-based solution to support the integration of parametrics in IFC for different applications.
Construction safety knowledge is often scattered across unstructured and semi-structured sources, complicating retrieval and reasoning. This paper introduces a novel Knowledge Graph Question Answering (KGQA) method leveraging Large Language Models (LLMs) for intelligent QA over safety hazard knowledge. The approach integrates an LLM-based assistant for natural language understanding (NLU), converting natural language queries (NLQs) into structured queries that retrieve information from a domain-specific safety KG. An NLQ dataset is constructed, and QA performance is benchmarked across GPT-4o, DeepSeek-v3, and Claude-3.5 to evaluate our method. Experimental results demonstrate that our method achieves high accuracy and efficiency in safety knowledge retrieval.
The paper presents a dataset for image-based damage detection in tunnels and compares multiple models in detecting these anomalies. It introduces first the dataset and explaines how it was collected and processed to label observable damages. It then benchmarks Mask RCNN, Cascade Mask RCNN and Mask2Former in the instance segmentation performance.
In this study, an approach to integrate concrete crack and spalling into BIM models using the open data format Industry Foundation Classes (IFC) will be introduced. After a learning-based crack and spalling detection algorithm has been developed, the creation of the BIM model will be demonstrated. The detection algorithm involves a transfer learning approach to minimise the need for annotated training data, while achieving a high prediction accuracy
Studio-based pedagogy, integral to design and construction education, emphasizes hands-on exploration, collaboration, and face-to-face interactions. The COVID-19 pandemic disrupted this model, shifting learning online and exacerbating workloads, pedagogical challenges, and community-building issues. This study investigates these effects through a survey of New Zealand educators, analyzed using inductive thematic methods. Findings highlight significant stress from tensions between traditional and online models, underscoring the urgent need for improved digital literacy, redesigned learning environments, and enhanced online community-building strategies. This research contributes insights into adapting studio-based pedagogy for resilient and effective online education.
Uncertainty analysis is a great forgotten method in systems development for construction applications based on artificial intelligence. Its capabilities to quantify and provide data regarding AI predictions and its consequences down data-driven decision-making processes are often overlooked in academic literature. This paper hopes to highlight the importance of such methods, providing two case studies where uncertainty analysis is performed following Bayesian approaches. The two case studies are computer vision applications for classification and localisation of elements within construction environments. These are taken as representative solutions that have been popular in the field.
Flood risk management in the Netherlands is transitioning to a decentralised model that actively involves municipalities and different stakeholders, such as property owners and citizens. This research addresses the technical challenges of integrating semantic 3D city models, Building Information Modelling (BIM), and Geographic Information Systems (GIS) for improved flood risk management. A web-based platform was developed using Django, integrating cadastral data, BIM, and semantic annotations. The system incorporates Cesium-based 3D and IFC viewers to visualise both urban environments and building-level details. This study contributes to participatory flood management by enhancing stakeholder engagement, preserving semantic richness and providing a scalable framework.
The industrialized construction promises to transform the industry by performing activities in shop floors to improve operational efficiency. However, the lack of integration between practical experience and real-time data limits more efficient operations in the shop floor. This study addresses this limitation by combining design parameters from BIM models with real-time production data collected via RFID in a semi-automated shop floor. By including context-based data in machine learning models, cycle time prediction accuracy was improved, compared to models containing only design-based data. highlighting how data granularity optimizes operational planning and production processes in Industrialized Construction.
Human variability challenges the use of robotics in construction. While research often focuses on robots adapting to humans, reducing human behavior variability is less explored. This paper presents a method using time-expanded graphs to identify safe pedestrian paths during human-robot interaction (HRI). Simulating hazardous energy in a virtual environment provides the safest routes for construction workers in dynamic hazardous environments. Forwarding these paths to pedestrian workers guides them through safe paths. User studies show this approach reduces hazardous exposure and increases behavior predictability, improving safety in HRI. This research underscores that providing safe worker routes enhances safety during HRI.
Building design requires various disciplines and spatial relationships, which can lead to errors. This study introduces a comprehensive conceptual framework for automated BIM compliance checking. It involves the creation of knowledge graphs by formulating ontologies for building regulations and developing models for semantic role annotation. Data extraction pipelines are established using the Dynamo module within Revit to gather pertinent information from BIM models. Compliance checking logic is articulated using graphs to match the extracted knowledge from building standards with the information in BIM models. The practicality of this automated compliance-checking framework was tested using BIM models from two actual projects.
