Abstract Some of the most complex challenges in construction projects arise during the early stages of project delivery, where regulatory uncertainty and limited visibility of urban conditions often constrain feasibility assessments and delay coordination among project stakeholders. In this context, this study proposes a machine learning–based approach that enables the exploration, clustering, and classification of urban areas using structured and georeferenced data from construction permits issued by government agencies. A midsized Latin American city is used as a case study, drawing on official permit records collected over a 15-year period. Unsupervised learning techniques ( k -means) were applied to identify latent urban development patterns, followed by supervised models [random forest and extreme gradient boosting (XGBoost)] to classify construction zones based on functional and morphological features. The findings demonstrate how automated urban analysis generates consistent and interpretable indicators related to building density, location, and construction typology, which are directly relevant to early project delivery decisions. Beyond predictive performance, the study demonstrates that machine learning–powered urban regulatory intelligence can aid early project governance by decreasing information asymmetry, providing regulatory insights upfront, and fostering earlier alignment among project owners, designers, and contractors. These features are essential for collaborative delivery methods like integrated project delivery and lean project delivery, where early coordination, transparent information sharing, and a common understanding of constraints are crucial for ensuring project success.
Abstract Working in hot weather exposes construction workers to significant risks of heat-related strain, resulting in decreased labor productivity and increased occupational injuries and illnesses. The establishment of well-designed work–rest cycles serves as an effective strategy for alleviating these risks. This study aims to develop an optimal work–rest schedule for construction workers in hot weather that maximizes labor productivity while safeguarding worker health and well-being. Field studies were conducted at three construction sites in south China during the hot season (May–July 2024). The physiological data of 175 construction workers, along with meteorological data from these construction sites, were synchronously monitored and measured. Machine learning techniques were utilized to accurately estimate two key parameters essential for schedule design, in terms of maximum work duration (MWD) and rate of recovery (ROR). To reflect the variability and uncertainty in real-world construction environments, Monte Carlo simulations were employed to generate a wide range of potential worker environment scenarios. By systematically evaluating different work–rest patterns across these scenarios, the optimal schedule was identified, consisting of a 15-min break after 150 min of continuous work in the morning, and a 20-min break after 140 min of continuous work in the afternoon. This schedule achieves the maximum labor productivity while reducing the occurrence of heat-related incidents. The findings provide a scientific and practical work–rest schedule for the construction industry, offering a viable strategy for heat stress management that enhances both work efficiency and worker protection in heat conditions.
Abstract Construction fences can become unstable during high-wind events, particularly when attached objects such as signage, banners, or shade cloths increase the surface area exposed to wind. However, current inspection and preparedness practices often rely on manual observation, limited time and labor, and practitioners’ field experience, making it difficult to quickly assess fence stability and determine appropriate preparedness actions before high-wind events. To address this challenge, this study developed a vision-language model-based system that integrates prompt-based information extraction, computer vision-based fence condition assessment, engineering-based stability analysis, and large language model-based practical guideline generation. The system extracts key variables from user input, identifies fence panels and attached objects using instance segmentation, estimates wind-exposed surface areas, performs engineering-based stability analysis, and generates practical guidelines based on the stability analysis results. Technical evaluation showed that the instance segmentation model achieved an overall mAP50 of 0.866 in detecting fence-related objects. For practical guideline generation, the Construction Material and Incomplete and Temporary Structure 3-30E model with a temperature of 0.5 and top-p of 0.7 achieved the highest bidirectional encoder representations from transformers score F1 value of 0.853 among the tested configurations. In addition, a practitioner-based evaluation was conducted to examine the practical value of the proposed system from the perspective of potential users. The results indicated positive perceptions, particularly for practical guideline generation, system usefulness, and future adoption intention. This study contributes to automated inspection and proactive construction site management for severe weather preparedness by demonstrating how image-based condition assessment, engineering-informed stability evaluation, and natural-language decision support can be integrated into an artificial intelligence–based decision-support tool.
Abstract While the ubiquitous adoption of artificial intelligence (AI) monitoring in construction projects has substantiated significant gains in operational efficiency and safety, its backlash remains theoretically underdeveloped. In practice, AI is gradually replacing human supervisors to monitor workers’ operation trajectories and work status. However, existing research has not yet unpacked workers’ cognition and behavioral responses toward AI monitoring. Therefore, grounded in the combination of psychological reactance theory and social exchange theory, the current study investigates how AI monitoring triggers counterproductive work behaviors and the boundary role of perceived organizational support. Through a survey of 352 on-site workers across diverse engineering projects, our findings reveal that AI monitoring is positively correlated with both production and political deviance. This relationship is mediated by negative cognition and anger, while perceived organizational support serves as a critical moderator that mitigates these detrimental effects. Furthermore, combined with the two supplementary studies, this research further verifies that the AI monitoring context triggers reactance among on-site workers when they perceive threats to freedom. By uncovering these distinct underlying mechanisms, this study extends the application of psychological reactance theory to the research context of AI monitoring on on-site workers in construction projects and develops a context-specific explanation.
Abstract Digital technologies such as sensing systems, computer vision, and data analytics are increasingly proposed to support safety monitoring and inspection on construction sites. Despite rapid technological advances, their adoption in practice remains uneven. Existing review studies have primarily cataloged technological innovations but provided limited insight into how these tools align with onsite inspection workflows and managerial decision-making. This study addresses this gap by reframing digital safety monitoring and inspection as a managerial implementation challenge rather than a purely technical problem. A systematic literature review of 297 publications was conducted using the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) framework, combining bibliometric analysis with inductive content analysis. The results identified six domains of onsite safety monitoring and inspection activities and 17 technology-assisted tasks. Deployment readiness was evaluated across these domains based on reported validation settings, operational robustness, and integration feasibility with existing inspection practices. To explain the persistent research–practice gap, adoption barriers were synthesized using the technology–organization–environment (TOE) framework. The analysis showed that implementation constraints cluster around technical reliability, organizational readiness, and governance conditions. Workflow integration emerged as a key cross-cutting challenge linking technological capability with managerial feasibility. By organizing digital safety research around inspection activities rather than technologies, the study provides an activity-centered deployment readiness map and a management-oriented synthesis of adoption barriers. The findings offer guidance for researchers, contractors, and industry stakeholders seeking to develop and implement digital safety solutions that are operationally deployable and aligned with construction management practices.
