The benefits of a digital twin city have been assessed based on real-time data collected from preinstalled Internet of Things (IoT) sensors (e.g., traffic, energy use, air pollution, water quality) for managing the complex systems of cities, but the sensor-based reality information is likely insufficient to provide dynamic spatiotemporal information about physical vulnerabilities. Understanding cities’ current states of physical vulnerability can support city decision makers in analyzing associated potential risk in urban areas for data-driven infrastructure management in extreme weather events. As a step toward creating a digital twin city for effective risk-informed decision-making, this paper proposes a new framework to bring crowdsourced visual data-based reality information into a three-dimensional (3D) virtual city for a model update with interactive and immersive visualization. Unstructured visual data are collected from participatory sensing and analyzed to estimate the geospatial information of vulnerable objects in the distance representing physical vulnerability in cities. The crowdsourced visual data–based reality information of physical vulnerability in a given region is then integrated with a 3D virtual city model, and the updated 3D city model is fed into a computer-aided virtual environment (CAVE) for immersive visualization to enable users to navigate the intersection of reality and virtuality. To test the proposed framework, case studies were conducted on Houston. The outcomes demonstrate that the proposed method has the potential to make the virtual city model live in terms of local vulnerability. The digital twin city building on crowdsourced visual data is expected to contribute to risk-informed decision-making for infrastructure management in cities and help analyze various what-if scenarios in disaster situations with increased visibility of hazard and city interactions.

Defining, measuring, and dealing with contractual risks are crucial for successful construction projects because the contractual risks can lead to serious claims and disputes. In general, construction participants make a stipulation regarding their roles and responsibilities by contracting in order to prevent such claims and disputes. A common practice for preparing construction contracts is to modify the standard contract forms to reflect the interests of the given project from the owner’s perspective. In this process, however, favorable clauses that may be beneficial to the contractor are often modified or even removed, causing significant potential risks to the contractor. Therefore, an in-depth review of contract terms and conditions is required to avoid future risks. This study presents a new proactive risk assessment model to identify missing contractor-friendly clauses in the owner’s modified contract conditions from the contractor’s point of view. A case study is used to demonstrate the proposed framework, and real-world project cases were analyzed to understand what type of contractor-friendly clauses would likely be omitted in the owner’s modified contract. In this study, the developed model builds on rule-based natural-language processing (NLP) to analyze unstructured text data through preprocessing, syntactic analysis, and semantic analysis. The proposed data-driven risk assessment model is expected to reduce the extent of human errors by (1) identifying potential contractual risks that could arise disputes; and (2) supporting to develop an appropriate response strategy for the given risks.

As one of the representative parameters for human energy metabolism, the metabolic rate has been considered as the significant factor for occupants’ thermal comfort analyses. Despite the importance of metabolic rate as a predictor of thermal comfort modelling, prior works rely on uncertain metabolic rate estimation without considering actual activity variations while occupying a building. This study aims at identifying the effect of metabolic rate on the thermal comfort models by proposing a robust data-driven personalized model in consideration of human activity variations. To investigate heterogeneous thermal state of occupants, wearable sensors and machine learning algorithms were used to continuously monitor and analyse individual physiological signals, activity-based metabolic rates and environmental indices. Field experiments were conducted with 10 subjects in a campus building in the US, and the results showed that predictive models considering metabolic rate yield advanced performance of up to 8.5%, implying that activity-based metabolic rates provide better understanding of personal thermal comfort. This paper quantitatively validates the effectiveness of reflecting metabolic rate based on human activity variations into personal thermal comfort modelling, which provides an insight into how to better model personal thermal comfort of occupants in real-life settings.