Accurate diagnosis of cracks in concrete structures requires integrating data-driven perception with mechanism-based knowledge. This paper presents a knowledge-infused framework integrating crack perception, knowledge-guided reasoning, and human-in-the-loop validation, implemented as an engineer-friendly interactive system for interpretable assessment. A domain-specific knowledge graph containing 530 nodes and 716 edges is constructed to represent crack characteristics, deterioration mechanisms, and maintenance strategies, and Retrieval-Augmented Generation (RAG) is integrated to enhance standards-guided reasoning and improve code compliance. The crack detection and quantification module achieved high accuracy (R2 = 0.976, MAE = 0.505 mm) with an average processing time of 4.6 s per image. The framework is completed within 164.6 s per case. Under the representative bridge inspection scenario considered in this study, the proposed framework is estimated to reduce inspection labor cost and time by 78.3% and 68.7%, respectively, compared with manual inspection. The framework provides an efficient, transparent, and standards-guided solution for intelligent infrastructure maintenance.
Automating the encoding of building codes into formal representations is a critical step for compliance checking, yet it is often hindered by the linguistic complexity of regulations. This paper develops a two-stage pipeline that utilises Large Language Models (LLMs) to generate LegalRuleML (LRML) through a Structured Intermediate Representation (SIR). Using a dataset derived from the New Zealand Building Code (NZBC), the paper demonstrates that the SIR-based approach outperforms direct translation, with four fine-tuned models achieving F1 scores of 73.29%–77.23%, representing improvements of up to 2.40% over the current state-of-the-art method. Despite these gains, a large gap remains compared to expert-crafted SIRs (80%+). Analysis reveals that LLMs still struggle with logical structural integrity and that standard cross-entropy loss poorly aligns with formalisation goals. These findings highlight the potential of LLM-generated SIRs while underscoring the necessity of task classification and logic-driven verification in automated rule encoding systems.
Although various technologies have been applied to automated proximity monitoring and warning (PMW) to prevent struck-by accidents and enhance construction safety, comprehensive reviews of this topic remain scarce. Therefore, this paper conducts a systematic review based on 97 PMW-related articles published between 2010 and 2024. The bibliometric analysis reveals the technical roadmap over time and influential research networks. These articles are then classified and summarized from four aspects: factors affecting PMW, hazard level definition and determination, proximity perception, and alarm mechanism. Current limitations of PMW include low accuracy, high sensitivity to obstacles, and poor real-time performance. Future directions thus may focus on end-to-end three-dimensional (3D) object/pose detection, real-time 3D reconstruction and local updating, and real-time multimodal fusion. This review synthesizes the current status, limitations, and future trends of PMW through a new taxonomy, providing insights to advance the development of automated PMW in construction.
Accurate point cloud classification of bridge components is essential for Scan-to-BIM and semantic model reconstruction. However, existing learning-based methods rely mainly on data-driven feature learning and fail to capture bridge-specific structural characteristics. This paper, therefore, proposes a structural prior knowledge-guided method that incorporates bridge component distribution characteristics into point cloud classification. It combines (i) a bridge-wise batching strategy to preserve the component relationships within the same bridge with (ii) three bridge-specific structural prior knowledge constraints, namely elevation hierarchy, spatial adjacency, and projection overlap. Validated on 34 RC girder bridges, the proposed method consistently improved classification accuracy and convergence stability across different backbones and training-data settings, mitigated the synthetic-to-real domain gap, and achieved up to 99.1% OA and 98.3% mAcc with mixed synthetic-and-real training. These results highlight the value of incorporating scene-level structural prior knowledge into the processing of infrastructure point clouds.
As circular economy principles become increasingly embedded in regulations, digital passports are being recognised as key enablers of circular construction. Although several digital passport concepts have been proposed, their implementation remains constrained by the lack of standardised, hierarchically structured data requirements. This paper presents a computational workflow that combines semantic topic modelling and unsupervised clustering to derive a common digital passport taxonomy from the state-of-the-art literature. To address inconsistencies in property names across data requirements, large language models are used to generate contextual property descriptions from the literature. Expert evaluation aligns moderately with cluster cohesion and diverges on hierarchical fit, indicating that statistically optimal taxonomies require manual moderation. The workflow can be used to semi-automatically generate and moderate taxonomies that structure disparate data requirements. This application can assist construction professionals in developing standardised digital passport data templates aligned with evolving regulations.
