Computing in Civil Engineering 2017: Smart Safety, Sustainability, and Resilience

Considering the significant proportion of manual operations in the construction industry, effective monitoring and control of manual activities through modeling and visualization can play a critical role in improving productivity and safety of projects. Furthermore, using motion capture data to visualize worker movements, as part of the virtual representation of a workplace, can be highly beneficial for applications pertaining to analysis of human motions. Thus, this study proposes a post-simulation visualization framework, which uses automated creation of motion capture data, to provide a reliable and realistic representation of manual operations. The approach has been implemented on a manual task, and the results confirm its effectiveness in enabling an easy to use modeling of manual activities for various visualization applications.

Productivity of craft workers in construction projects can be negatively impacted by ineffective communication of the project’s design and information. Traditional delivery of mechanical, electrical, and piping (MEP) designs rely on two-dimensional isometric drawings. Developments in three-dimensional (3D) computer aided design (CAD) and 3D printing have provided new format options for delivering engineering information, however providing them to crafts for use at the construction workface remains relatively rare. The objective of this research is to understand how eye gaze patterns of construction craft workers are influenced by information formats when completing complex spatial tasks. A series of field trials with MEP workers was conducted to examine the influence of information format on their eye gaze patterns in completing a scale model task assembly. Participants were provided eye tracking glasses with one of three information formats: traditional, two-dimensional (2D) isometric drawings; 2D isometric drawings supplemented with a three-dimensional (3D) image of the assembly; and 2D isometric drawings supplemented with a 3D physical model of the assembly. Eye gaze patterns were examined to understand if performance can be improved by alternative formats of information displays and/or training.

In the U.S, 2011, heating, ventilation, and air conditioning (HVAC) systems in buildings took nearly 20% of total energy consumption. To improve the efficiency of HVAC system and save energy, researchers have exerted great effort in developing models to optimize control schedule and strategies of HVAC systems. Such models not only can greatly reduce gross energy usage, but also maintain a comfort indoor environment. As one of the most important intrinsic factors for control models, however, real-time occupancy statistics has been rarely applied in previous studies, due to the difficulties in occupancy detection techniques including scalability, cost effectiveness and real-time consistency. Thus, the study discussed in this paper proposes an approach that utilizes processed probe request log dataset collected through Wi-Fi network for occupancy estimation and HVAC system control. The authors first discuss the requirements for the occupation information used in different optimization strategies for HVAC systems. In addition, occupancy detection approaches are explored and compared to illustrate that the WiFi-based approach fit the requirements of HVAC controls. Finally, WiFi-based occupancy detection and prediction approaches are investigated using a real-world case study, and an integrated framework for occupancy-based predictive HVAC optimization is proposed for future application.

When an emergency occurs in a building structure, real-time occupants’ location and their evacuation decisions are important for the overall evacuation time. It is difficult for the evacuees to perform reasonable behavior in the unknown environments during emergency situations. Communication support is of the utmost importance in such scenarios. The ideal communication support system consists of information collection system, information distribution system and social media etc. The information collection system is used to collect the information including evacuees’ location and situational awareness by evacuation officers; while the distribution system is to deliver real-time evacuation information to evacuees. Indoor positioning sensing and cloud computation technologies allow the evacuation officers to collect the real-time emergency information such as the evacuees’ locations and the calculated density. Provided with the situational awareness data, the evacuation officers in charge can make the corresponding real-time evacuation routes for the evacuees at different locations. This paper proposes a framework for real-time emergency evacuation management system enabling the interaction between evacuation officers and evacuees during emergency situations. The process for performing the pilot study has been introduced. The current system is limited by the power availability and internet accessibility in the physical environment. The REEMS system can be improved by considering the evacuees’ panic state during emergency and integrating the psychological acceptance extent for evacuation plan.

Construction projects are generally characterized by their evolving complexity where any given project would include-among other reasons-multiparty relationships and numerous transactions. Managing these relationships is often complicated further by the traditional contractual approaches, which diverge the focus of involved parties away from the project main goals and promote individualism among the associated parties. Hence, relational contracts were introduced in an attempt to align the goal of the different parties to enhance the multi-party interactions, communication, and collaboration. This study presents a simulated model-based comparative analysis between two of the main relational contracts standard forms in the U.S. and the U.K.; namely, (1) the Association of Consultant Architects (ACA) Project Partnering Contract PPC2000, and (2) the American Institute of Architects (AIA) Document C191–2009, Standard Form Multi-Party Agreement for Integrated Project Delivery. The performance of the various processes defined in these two forms was graphically and computationally analyzed using social network analysis (SNA). Employing SNA is an innovative approach to model and analyze the expected interactions among the involved parties under each contractual setting. In this study, UCINET 6 for Windows was used to analyze the network data for the change orders process according to the provisions of each contract form. Metrics such as adjacency, degree centrality, and closeness centrality were insightful in providing a deeper understanding of the expected level of collaboration and involvement among the involved parties at each step of the change order process. Meanwhile the processes in both forms promote direct communication among most of the key parties, the results of graphical and computational analysis favored PPC2000 form over C191 form in terms of the clarity of change orders procedures and in terms of the general results of connectivity and centrality. The results also demonstrated that PPC2000 placed significant emphasis on the client and its representative, while C191 placed more emphasis on the architect. This study builds a substantiated foundation for developing a more effective relational contract that capitalizes on the pros of the available contracts while mitigating their deficiencies. This should promote a more efficient and effective contracting environment.

Many researchers in the construction domain have explored various sensing technologies to acquire location awareness of construction assets. With the use of certain technologies, such as GPS, UWB, and RFID, previous research achieved significant progress for outdoor applications. However, unlike outdoor tracking, indoor environments pose considerable challenges because of the existence of a multitude of interactions among workers, equipment, and materials. These challenges appear in the form of signal interference such as signal degradation, occlusions, obstructions and multipath effects. To overcome signal interference, this study aims to develop and evaluate a probabilistic local search (PLS) algorithm using Bluetooth Low Energy (BLE). This study conducted a site test together with computer simulation to assess the performance of the developed algorithm with respect to accuracy, reliability and robustness. Although the test was conducted in a simple path, noisy data that are generated from the computer simulation reflect the challenges of signal interference, which are expected from occlusions and other objects that potentially introduce signal interference. Based on the test results, the PLS algorithm showed a significant improvement in accuracy, compared with two conventional positioning methods, especially when signal interference exists in the test environment.

The integration of life cycle assessment (LCA), and life cycle costing (LCC) of buildings into building information modelling (BIM) has been a trending research area in construction management. The current practice has involved the combined use of BIM tools (such as Autodesk Revit, and Navisworks) for 3D building modelling, building energy modelling (BEM) tools (such as EnergyPlus and DOE-2), and environmental impact assessment (EIA) tools (such as Athena Eco-calculator, Tally and SimaPro, etc.) to conduct LCA of buildings. One of the major limitations observed in this approach is the interoperability (data loss) issues, as well as the complexity arising from employing many tools especially at the early building design stage. This paper is a critique of conventional BIM-integrated LCA of buildings. It proposes a real-time BIM-integrated LCA/LCC that allows the designer to get immediate feedback (on energy, carbon, LCC, and exergy) at key stages of the design process. It concludes that an improved approach is needed for BIM-integrated LCA/LCC adoption at the early design stage, which is when most of the sustainability decisions are made.

Scheduling is a mostly manual process, which is typically time consuming. To overcome this problem, researchers have developed various knowledge based, model based and case based representation approaches to automatically generate construction schedules. Despite the attention accorded by academics, the architectural, engineering and construction (AEC) industry has been reluctant to adopt automatic schedule generation. Several reasons exist for this, of which one lies in being able to robustly generate feasible schedules under different project contexts. These contexts include differing project layouts, and schedule milestones. This failure has eroded the confidence of industry practitioners in employing automatic schedule generation tools; practitioners prefer to rely on manual methods rather than trust automated methods in which they perceive as having little control over. This paper proposes a rule based approach which explicitly accounts for project context through a query mechanism. These rules are defined from construction requirements which represent pre-requisite conditions for construction processes to occur. This approach makes several research contributions: Firstly, a construction method model based on construction requirements is introduced to link product and process models. Secondly, a query language is utilized to enable project information to be embedded in the construction requirement specification. The approach is validated on a subway construction project against an actual expert.

Subsurface utility construction work often involves repositioning of, and working between, existing buried networks. While the amount of utilities in modern cities grows, excavation work becomes more prone to incidents. To prevent such incidents, excavation workers request existing 2D utility maps, use detection equipment and dig test trenches to validate their accuracy and completeness. Although test trenches are of significant importance to reveal information about subsurface conditions, the process of determining their location, number and size is not explicated by experts to date. This study therefore aimed to explicate the reasoning and logic behind the selection of utility test trenches, and to formalize this in a semantically-rich utility model. To this end, we conducted interviews with experienced excavator operators. We then derived heuristics and rules that the experts used to determine trench locations. Such rules related to, for example, the layout of the excavation site, and the type of utilities, and accuracy of available data. Based on these rules, we integrated various incomplete sources of data, and generated a 3D utility model that could generate several alternative construction situations. We used queries to identify the most suitable location for a test trench. The resulting answers to queries helped optimize the test trench selection process. Our prototype demonstrates that the identified rules (1) facilitate the generation of semantically rich 3D utility models, and (2) support test trench decision making.

In recent years, the building industry has been steadily increasing its use of building information modeling (BIM). This approach to design and construction leverages 3D virtual models that contain both physical and functional attributes in order to make better project decisions for the actual built facility. Despite the advances in BIM and 3D modeling, design communication between project participants still tends to rely primarily on 2D paper plans. This means that models are developed in 3D, reduced to 2D paper plans, and subsequently reinterpreted back into 3D mental models by project team members prior to being built. This research explores the use of mixed reality (MR) to enable design communication in order to allow electrical construction professionals to build conduit assemblies directly from BIM content. It addresses two research questions: Is it possible to use MR for design communication to enable construction professionals to assemble prefabricated electrical conduit? How does this mode of visualization influence the construction process performance? To address these questions, participants built two different conduit assemblies, once with paper communication and once with MR communication. The task was video recorded to allow the researchers to observe the participants’ process and measure their productivity when using each communication approach. The results suggest that MR did indeed prove to be a viable means for design communication. Furthermore, the results indicate that this mode of design communication enabled more accurate and faster construction of electrical conduit than was observed through 2D paper communication. Future work will aim to identify other use-cases where MR may offer value to project teams and determine best practices for leveraging this technology in practice.

Resurfacing is one of the most common highway projects in Georgia with large investment every year. The unit bidding price, proposed by contractors to fulfill the project, is one of a leading indicators to reflect the cost of department of transportation (DOT) in resurfacing projects. Compared to other cost index researches, the changing trend and big volatility of unit bidding price make the analysis more difficult, and thus the Pearson and Spearman correlation test turned out to be inapplicable here to find the relation of unit bidding price to other indicators. To fill this gap while focusing on the long term trend change, the research proposed a non-parametric analysis framework, which could first find the changing trend of unit bidding price in different segments and then find the relation of trend pattern among unit bidding price and other selected indicators. The results show that the national highway construction cost index (NHCCI) could be a useful indicator to reflect the trend change of unit bidding price.

Infrastructure planning and management teams need to promptly diagnose and resolve building problems when they occur. A common way to notice building problems is receiving service requests from occupants in terms of phone calls, emails, or specific information systems for facility management. Usually, facility management teams need to manually discern the request’s priority level, and decide necessary crews and tasks to solve the problem. However, decision making in response to the service requests is often problematic and delayed due to intensive labor demand. This paper demonstrates an effort to automatically predict the reaction type (regular or urgent) and the necessary crews through mining textual descriptions of service requests. Service request data were collected from a university’s facility management team. The dataset contains approximately 80,000 instances across two years. Multiple machine learning algorithms are employed to identify the optimal model to predict work priority and crew types. Findings suggest the model successfully and automatically determine critical patterns extracted from service request texts. The text-mining-based prediction model enables facility management to expedite the construction work process. The model is also scalable to other fields of construction management and facility management.

In a typical construction project, a significant amount of information is communicated to various stakeholders at different phases of the project lifecycle. The communication of this information tends to be informal and ad-hoc in the majority of the cases, which makes it more susceptible to loss of information or misinterpretation. Earthwork operations, which are one of the main operations of construction projects, also struggle with the challenge of effective information communication. There is an apparent shortcoming regarding the unified structure for data and information exchange in this domain. The existing models and ontologies do not address the explicit semantic representation of the earthwork operations. Accordingly, there is a need for a knowledge model to formalize the communication of information in an efficient manner. An ontological model can be used to organize the domain knowledge so that it can be utilized and reused by the stakeholders, e.g., project managers, designers, coordinators, equipment operators, etc. This paper purposes a framework to develop ontology for earthwork operations to support and enhance data exchange and communication among different stakeholders in the project. The main objectives of this paper are (1) to formalize the knowledge in the earthwork domain, and (2) to build an ontology that captures this formalization. The ultimate result of this ontology, which is demonstrated by means of a case study, is to facilitate the development of data standards that can be shared in actual projects to accelerate project execution.

Today, camera-equipped unmanned aerial vehicles (UAVs) are frequently used to capture large amount of images and videos on construction sites. Research has focused on methods that organize such data spatially and temporally, map them into 4D BIM, and use the resulting models for project monitoring applications. To guarantee quality in the data capturing process, this paper introduces a BIM-driven flight planning method. Specifically, a cloud-based 4D BIM environment is presented that highlights expected changes prior to capture. Based on a set of user-defined UAV waypoints that specify camera location and pose, a capture path is fit into the 4D BIM environment. The quality of the capture is then assessed per element using metrics of visibility (2D surface area associated with each element in each visual observation) and redundancy across visual observations. The effectiveness of the method and metrics are discussed using data from a pilot project.

Camera-equipped UAVs are rapidly emerging as ubiquitous tools for capturing images/videos for monitoring construction sites. Currently, professionals use 2D maps for producing flight plans. These maps often depict the state of sites prior to construction and mainly allow for planar grid patterns for capturing overlapping visual data. To enable 3D capture and control percentage overlap among captured visual data necessary for photogrammetry, professionals need to manually operate the UAVs, a process which is time-consuming and requires dexterity. In this paper, a new client- system is presented to enable 3D path planning and visual data collection. It exploits prior knowledge of the BIM and assesses flight path for safety and quality requirements. The system includes a web-based user interface to plan and simulate the UAV missions, and a mobile application to execute the flights. To verify the system, planned flight missions are simulated and compared to executed flights. Results show that the system can achieve safe and quality capture while accommodating requirements for image-based 3D reconstruction and project controls applications.

A methodology for assessing and measuring project complexity can help project teams increase the likelihood of success and predictable project outcomes. This research used the principal component analysis (PCA) technique to create a satisfactory input for a potential project complexity predictive model that helps identify the levels of project complexity. In this research, 37 project complexity measures were statistically verified based on the data collected from 44 historical projects. However, the number of independent variables (37 complexity measures) was too large to generate a statistically stable and reliable model. Thus, the variable reduction process PCA was used to combine those significant complexity measures to create a required input for the project complexity predictive model. This PCA process resulted in a significantly smaller number of principal components functioning as the moderating variables. The process helped in generating a valid input for a numerically stable model while the subject observations were limited. The research helps enrich the theoretical basis in the field of project management by providing usable input for a potential project complexity model that can help scholars and practitioners assess a project’s complexity levels based on the applicably identified complexity measures.

Tower crane is one of the most versatile construction equipment that is needed in major projects. The determination of the location of the tower crane requires special consideration to its proximity to different facilities on site to optimize its performance. Several studies have been conducted on the optimization of single tower crane allocation using different techniques. The objective of this optimization problem is to minimize the total travel time used by the crane to deliver materials from supply to demand points. Mathematical and evolutionary algorithms (EAs) have been used in solving this problem. The algorithms used differ in their computational abilities, complexity, and performance. However, none of these models considered the application of ant colony optimization (ACO) although it has been proved to be one of the most powerful swarm-based algorithms used for solving combinatorial optimization problems. This paper: 1) presents the modeling and application of ACO in solving the single tower crane allocation problem; 2) shows the performance of ACO in comparison to other techniques used in the literature to solve this problem, 3) highlights some of the discrepancies identified in previous models, which make them confusing to the readers. The main contribution of this paper lies in providing the illustration needed for modeling and solving single tower crane allocation problem using ACO. This should prove very useful in solving large problems where mathematical techniques cannot be used.

Vision-based 2D human pose estimation provides a non-invasive and effort-saving means of extracting human motion data to facilitate an automated activity analysis of construction workers, such as unsafe behavior monitoring, ergonomic analysis, and productivity estimation. However, it continues to suffer from inaccuracies and a lack of robustness, particularly under a dynamic and cluttered environment like a construction site where occlusions are prevalent. To address these issues, the authors apply convolutional neural network (CNN) to human detection and pose estimation on sequential images from site conditions. Using the benchmark training datasets that do not include any images taken from the site, the result of 2D pose estimation in testing data shows that this approach achieves a high level of accuracy and robustness considering the presence of partial occlusion. The potential of this human pose estimation method a under dynamic and cluttered construction environment is demonstrated, and its further applications for a worker activity analysis are discussed.

This paper describes an analytical model along with a visualizing tool, based on the error propagation theory to quantify and visualize the uncertainty that is present in the position and pose of the excavator obtained from sensor-based measurements regardless of the type of sensors used. A technique is also presented to aid in the selection process of sensors (precision level) for tracking the excavator (pose and position) considering the tradeoff between the level of precision required in field operation on a specific task and the cost of sensors. A conceptual demonstration of field operation at a grading project is also given at the end; detailed workface level planning is made for a self-tracking excavator equipped with multiple sensors to track its bucket position.

Cognitive processes have been found to play a significant role in contributing to the human errors that lead to construction accidents. A better understanding of cognitive processes as they relate to construction will elucidate the more predictable varieties of human fallibility and allow for the creation of strategies to avoid said errors. To date, no published study has empirically investigated the relationship between working memory and the human errors that lead to accidents in construction. To address this gap, this study uses eye-movement metrics to examine how working-memory load impacts worker’s attentional allocation and the detection of potential hazards in construction. To determine whether changes in memory load influence scan patterns and awareness of hazards, 19 students were recruited to participate in an eye-tracking experiment in which working memory load was manipulated. The results demonstrated that as working memory load increases, the ability of participants to identify hazards decreases. In addition, the results demonstrate that there is a significant difference between the visual search strategies of people under different working-memory loads. This study provides a proof of concept that eye movements can be used as an indicator of variation in working-memory load among construction workers while they search for a hazard without any active interference with the subject.

Flow velocity is one of the key factors in hydrology. Considering workers on-site safety, non-contact methods have become the major approach for flow velocity measurement. Although particle image velocimetry (PIV) is a primary non-contact approach, which has been developed and improved since 1991, hydraulics experts are still required to be involved in the initial setting for PIV measurement. Without the initial setting executed by PIV experts, the uncertainty derived from surface flow that has varied features may cause inaccurate measurement. For these reasons, we proposed a surface flow measurement method that takes advantage of human visual awareness to measure flow velocity. People can mark the flow features on images using the measurement tool we developed to acquire flow velocity. We recorded two flow videos under different flow conditions and measured flow velocity using the developed method. The measurements established the feasibility of using the method.

Risk assessment based on imagery data is becoming popular in construction project management because cheap imaging devices can capture reality in real time. One challenge is that image-based safety risk identification heavily relies on the subjective image interpretation. Well-trained inspectors could be a limited resource to meet the inspection requirements for construction safety. “Crowdsourcing” is one of the ideal approaches collecting identified risks through online image interpretation games, and aggregate answers from game players into comprehensive and reliable risk recognition results. Unfortunately, inputs provided by nonprofessional players can distort the aggregated results. This paper presents a study showing that without knowing the actual risks on construction sites, human-image interaction behavior analyses of anonymous online image interpreters can overcome the influences of mixed reliable and unreliable answers from anonymous online players and acquire reliable risk evaluations of jobsite imageries. We conducted this study with the following steps: (1) train anonymous assessors in a few minutes on limited safety rules and have them assess some images for collecting crowd-based risk detection data, (2) collect assessment results using an online crowdsourcing platform, and (3) automatically aggregate risk assessments using a Bayesian network model into the risk detection results. Results from an online image-based elevator installation risk assessment experiment shows that the proposed method can overcome the limitation of the majority-based voting and achieve comparable results as experienced safety inspection professionals.

With the growth of interest in the relationships between neighborhoods and physical activity, public health, and pedestrian safety, various efforts have been made to identify and assess physical disorders (PD) (e.g. garbage, litter, broken glass, damaged/disconnected sidewalk/driveway, shrubs on the sidewalk, graffiti, and strong odors) in the neighborhoods. Traditional assessment approaches mainly rely on trained inspectors, which require significant cost and time, and most of these methods lack reliability as they depend on subjective judgment to determine the conditions of the physical environment. In this context, this study explores the potential for using crowd-sourced physiological sensory data from pedestrians (e.g., gait and heart rate data) to identify and assess PDs in the neighborhoods. Though previous studies indicate that humans present diverse physiological responses in their interactions with their environment, there is a lack of empirical research exploring the physiological effects of the neighborhoods on walking activity. In an experiment conducted in a neighborhood of Lincoln, Nebraska, participants were asked to walk on a pre-defined path consisting of sixty measurement units. Three types of data (wearable inertial measurement units (IMU), photoplethysmography (PPG), and global positioning system (GPS)) were collected for each participant in order to examine the relationship between their physiological responses and the existing PDs they encounter. The results of this test indicate that the average values of physiological responses of participants in locations with PDs is significantly different than the average values of physiological responses of participants in locations without PDs. The results of this study will help improve the detection of PDs, thereby enhancing walkability.

Many studies indicate that energy consumption in commercial buildings is highly related to occupants’ energy-use behaviors, and improving these behaviors is regarded as the most cost-effective approach toward enhancing commercial building energy efficiency. Effective behavior-interventions rely on the availability of occupant-specific energy-use information, which is extremely expensive to capture with existing intrusive load monitoring (ILM) technologies. On the other hand, non-intrusive load monitoring (NILM) approaches have proven cost effective for monitoring appliance-specific energy consumption. In order to extend the concept of NILM to occupant energy-use monitoring in commercial buildings, this paper examines the importance of two occupancy-related energy-use variables—delay interval and energy-load variation—in identifying occupant-specific energy-use information. The results from implementing a k-Nearest Neighbors classifier into aggregate energy consumption data collected over the course of one month from a small office space reveal that these variables are effective in developing sophisticated NILM-based approaches for obtaining occupant energy consumption information. By providing this information at minimal cost, such approaches could make a great contribution to behavior-related energy research.

Knowledge workers are the greatest asset of any corporation. Environmental factors like air quality and temperature impact knowledge workers’ well-being, and influence concentration and motivation to work. Built environment research focuses on occupant comfort which typically prioritizes tangible factors, such as temperature, over imperceptible yet impactful factors like indoor air carbon dioxide (CO₂) concentration levels. Well-being offers a comprehensive perspective that includes comfort, psycho-physiological, social and personal needs. The accuracy of current models to predict building performance and occupant well-being is compromised by assumptions and simplifications leading to discrepancies between model results and measured performance. We present a case study to understand the CO₂ concentration variability due to occupant biometric and psycho-physiological states. The results reflect the lack of adjustment of the building systems to CO₂ concentration variability in response to the diverse activities in the spaces, and the lack of occupant awareness to the impact of the increased CO₂ concentration levels on knowledge work productivity. We present a comprehensive framework to model the dynamic interaction between building performance and occupant well-being.

Well-designed automated lighting control can take advantage of the natural light and save energy. Integration with a window shading system with blinds could enable a strategy that regulates the amount of natural light while maintaining occupants’ visual comfort. Previous studies have developed various automated lighting and blind control systems. Most of the works directly take light level and glare index listed in lighting guidelines or previous studies as the control set point and the criteria to evaluate visual comfort. However, users have subjective preferences for lighting level, visual comfort, and moreover, for outside views and privacy. That is, the consideration for each user’s need is required. The objective of this work is to develop an occupant-centered integrated lighting and blind control system using machine learning, sensor network, and control. For the implementation of the system, pyranometer, photometer, thermocouples, camera, and occupancy sensors are utilized to receive the environmental conditions, indoor and outdoor. The Arduino board is adapted for hardware control. By capturing occupants’ interaction with the blind and lighting system and by learning the relationships between occupants’ behavior and the environmental conditions, we build a preference profile of lighting and shading for each individual, and provide the feedback to regulate the control system. Through this work, an occupant can have a personalized optimal adjustment for the lighting and blind control based on her/his personal preferences. By considering personal preferences, the system can provide not only a user friendly but also an environmental friendly approach to control lights and blinds.

This paper develops a stochastic model to address uncertain (non-deterministic) water demand and supply that supports decisions regarding which water treatment systems to select. The optimal decision is the one that meets the water demand and achieves the minimum energy use or cost. The model constitutes the first stage of a broader research plan that aims to understand the dynamics among the water, energy, and food nexus and applies to a closed-loop, controlled environment system. The optimal values of decision variables are calculated for two different climates; one is relatively cloudy (with an average solar radiation of 3.6 kWh/m²/day) and humid (with an average rainfall of 1,000 mm/year) and the other is sunny (with an average solar radiation of 5.6 kWh/m²/day) and arid (with an average rainfall of 100 mm/year). The results of the study provide a basis for recommendations regarding actions to improve technological problems and processing approaches to optimize water treatment with minimal costs and energy inputs.

Majority of building energy consumption is used to heat and cool enclosed spaces. An innovative ultra-high-performance fiber-reinforced concrete (UHP-FRC) façade system provides the vital opportunity to reduce this energy consumption by utilizing UHP-FRC’s high structural strength hence more space for insulation. The building energy consumption and consequently the energy savings attributed to the proposed façade system are subject to uncertainties, such as variations in wall thermo-physical properties and building occupancy. The objective of this research is to assess the probabilistic energy performance of the proposed UHP-FRC façade system in comparison with conventional sandwich panel façade system. EnergyPlus was used to conduct whole building energy simulation. Three Department of Energy (DOE) prototype buildings (high-rise building, mid-rise building, and hospital) were used to represent the building context. The assessment was conducted for three cities with different climates: Chicago, Fairbanks, and El Paso. Probabilistic models were created to represent uncertainties of critical input parameters of the simulation model, such as occupancy level. We randomly sampled from the probabilistic models of the input parameters and propagated the sampled values through the simulation model using Monte Carlo simulation to create a pool of randomly generated buildings. The results show that the average of the annual energy consumptions of the randomly generated DOE prototype buildings with the UHP-FRC panels is less than the average of the annual energy consumptions of the buildings with the conventional panels for all scenarios. Moreover, the T-test (parametric) and sign test (non-parametric) also show that we can accept the alternative hypothesis that the average of the annual energy consumptions of the randomly generated DOE prototype buildings with the UHP-FRC panels is less than the average of the annual energy consumptions of the buildings with the conventional panels for all but one scenario (High-rise building in El Paso).

Widespread damage of water supply systems during recent earthquakes clearly shows the need for seismic planning. However, water supply seismic planning is subject to uncertainties in location, magnitude, and resulting damage of earthquakes. This problem is further complicated by complex topology of water supply systems along with the use of different materials, joint characteristics, pipe diameters and soil corrosivities. The objective of this paper is to identify critical links of water supply networks subject to various earthquakes and find optimum renewal decision given resource constraints. The methodology comprises of four interconnected components: (1) repair rate probabilistic modeling; (2) Monte Carlo simulation; (3) hydraulic damage modelling; and (4) resource allocation optimization. The first component calculates repair rate for each pipe in the network based on empirical fragility curves. Empirical fragility curves depend on the pipes’ location, material, diameter, joint property and soil corrosivity. Monte Carlo simulation generates probabilistic damages (i.e., leaks and breaks) in the pipe network. The hydraulic model calculates serviceability index considering simulated damages. Resource allocation optimization model uses genetic algorithm to find the optimum renewal decision to maximize serviceability index given resource constraints. The proposed model was validated using a water supply network. The network consists of 117 pipes and 92 junctions. The results show that the proposed methodology outperforms the latest proposed methodology in the literature to identify critical links in a water network. The network serviceability index is used as the measure to compare the results with the latest proposed methodology in the literature.

Building occupant behavior impacts energy consumption but this is often not adequately captured when monitoring energy use. It is important to provide indoor environmental conditions that are well suited to occupant needs and promote their well-being while minimizing energy wastage. Machine learning methods have been employed to predict energy consumption but in the absence of adequate occupant-related monitoring data, it is challenging to establish the right inputs and obtain predictions that reflect occupant characteristics (behavior and preferences). This paper presents the approaches employed to capture occupant behavior-related energy data using a combination of sensors, metering devices, and surveys to continuously track space conditions, occupant behavior, and energy consumption over a period of time. Two case study buildings are selected for this study, some of the challenges with energy monitoring are identified, and the need for interoperability among different sensors is discussed.

Given ambitious goals set by cities to reduce carbon emissions by as much as 80 percent by 2050, rapid progress is needed in improving building energy efficiency. This paper provides the first rigorous empirical examination of the effects of energy disclosure policies on building energy consumption over time by analyzing matched energy data from 2010 to 2015. These data, from NYC’s Local Law 84 benchmarking ordinance and the NYC PLUTO land use database, contain six years of detailed information on energy use, occupancy and use characteristics, and other property attributes for over 15,000 large buildings, resulting in a total feature space of more than 10,000,000 data points. Time series analysis and clustering techniques are used to identify common attributes of buildings with similar observed performance over time. The counterfactual scenario, to control for energy use changes in buildings not subject to mandatory disclosure, is established through the comparison of the normalized changes over time in energy use from buildings subject to the reporting mandate with a sample of matched peer buildings that voluntarily reported energy data, derived from a sample of 3,000 buildings collected through the Urban Land Institute’s Greenprint Center. Our findings reveal that buildings subject to mandatory disclosure have reduced their energy consumption more than those voluntarily reporting data, although more extensive analysis is needed to identify causal links.

The objective of this study was to create an integrated simulation framework to examine the dynamics that affect the long-term resilience of infrastructure systems. The proposed framework captures three phenomena that shape the dynamics of infrastructure systems: (1) deterioration of physical networks; (2) renewal and adaptation decision-making processes of agencies; and (3) chronic and acute stressors. Accordingly, a complex systems approach was adopted to examine the long-term resilience of infrastructure based on understanding of performance regimes as well as critical points at which shifts in the performance regime of infrastructure occur under chronic and acute stressors. The application of the proposed framework is demonstrated in a case study of a water distribution network where the system resilience was examined over a 100-year horizon under stressors such as aging infrastructure, population change, and funding constraints. The results identified the effects of internal dynamics and external stressors on the resilience landscape of infrastructure systems. The proposed framework enables building complex system-based theories for a more holistic understanding of infrastructure resilience.

The National Infrastructure Advisory Council identifies four lifeline infrastructure networks, specifically energy, communication, water and transportation, which underpin the critical functions of business, government and communities. Acute disturbances can cause cascading failures in interdependent network-of-networks resulting in severe loss of essential functionality and devastation at community, urban and regional scales. Quantitative recovery strategies, designed in anticipation of failure, inform timely, reliable and on-budget restoration. Post-hazards learning at recovery time helps develop anti-fragility, which is the ability of systems to attain a higher level of functionality and/or a state of greater robustness than the pre-disaster state. We propose an analytical resilience framework, which considers both failure and recovery processes of networked systems integrating complex networks science. A case study with real and simulated data showcases the potential value of our approach, while motivating systematic evaluation strategies and the need for gathering, benchmarking, and widely disseminating critical infrastructure information.

Residential buildings account for more than 20 percent of total energy usage in the U.S. in the past decade. Reduction of household energy consumption has environmental and economic impacts. As the first step, building scientists and construction engineers have endeavored efforts to obtain an accurate energy use prediction; however, few have focused on the relationship between construction cost and energy use. This paper investigates the associations between detailed construction cost takeoffs and actual energy uses in multifamily green buildings. The researchers employ advanced machine learning analytics to model the correlations between construction cost and energy use data which were collected from multifamily residential units. Findings identify the cost divisions in the construction stage that significantly correlates the energy use in the operational stage. The model allows developers to predict the energy consumption based on the construction costs, and enables them to adjust the investment strategies to amplify the energy efficiency of green building technologies.

Transitioning from centralized distribution to decentralized energy supply that is more adaptable to local patterns of energy demand is rapidly spreading in urban areas. However, urbanization is creating a significant increase in the amount, diversity, and complexity of human activities, all of which have a substantial impact on patterns of energy demand. Despite the increasing uncertainty on the demand side of energy, the effects of location-specific attributes associated with changing patterns of human activities are often neglected in classifying patterns of energy demand. Here, we present an urban-level spatially constrained clustering approach for classifying energy demand across residential and commercial buildings driven by the activity patterns of urban population as indicated by human mobility. Using a full month of 1,112,019 individual positional records from an online social networking platform (Twitter), we examined spatial fluctuations in energy demand in the City of Chicago across 801 spatial divisions. This approach is a foundational step toward contextualizing decentralization of energy supply by incorporating energy and mobility imprints of large populations in urban areas.

Buildings consume a tremendous amount of the total use of secondary energy resulting in a considerable impact on the environment. Therefore, it is necessary to reduce their energy consumption by improving the design of new buildings or renovating existing buildings. However, renovating building envelopes and energy systems to lessen energy losses is usually expensive and has a long payback period. Despite the significant contribution of the research about optimizing energy consumption, there is limited research focusing on the renovation of existing buildings to minimize their environmental impact using life cycle assessment (LCA). This study aims to develop a methodology to optimize the energy performance of existing buildings by selecting the optimal renovation strategies considering LCA. Different scenarios can be combined in a building renovation strategy to improve energy efficiency. Each scenario considers several factors including improvement of the building envelopes, heating, ventilation, and air conditioning (HVAC) systems, and local energy generation systems. However, some of these scenarios could be inconsistent and should be eliminated. Another consideration in this research is the appropriate coupling of renovation scenarios. For example, the HVAC system must be redesigned when renovating the building envelope to consider the reduced energy demand and to avoid undesirable side effects. A genetic algorithm (GA) coupled with an energy simulation tool is used for simultaneously minimizing the energy consumption and the LCA of the building. The simulation tool is used to calculate the energy consumption for each potential solution representing one renovation scenario. The data of the building characteristics are extracted from the building information model (BIM). The feasibility of the proposed method is demonstrated using a case study focusing on the buildings of Concordia University.

Over the past two decades, the construction industry has experienced a lot of positive changes in Saudi Arabia. Governments and lawmakers have come up with various regulations that govern the construction industry, such as the Saudi Building Code. These regulations have been effective in creating legal boundaries and rules for contractors and developers to abide by. However, critics and environmentalists suggest that contractors still have undertaken various activities that are of great harm to the environment. For example, wasted materials left in the surrounding area of construction job sites are a great cause of environmental pollution. Such pressure has led to the pursuit of a new trend in the construction industry known as sustainable construction. Sustainable construction entails the use of processes and materials that have little or no negative impact on the environment, and the construction of tall and complex structures that are important to accommodate the ever-growing population. In spite of the universal agreed guidelines for sustainable construction, detailed implementation methods depend on local contexts. This paper presents an initial effort in creating an ontology that describes the potential sustainable construction strategies in Saudi Arabia, to create a guidance for improving the energy efficiency of buildings in Saudi Arabia. The ontology is expected to (1) help create construction job sites that are not only environmental friendly, but also better serve the interests of all stakeholders; and (2) help reduce energy usage in buildings.

An important step toward harnessing urban wind energy is to determine the local wind levels. Local weather services commonly report regional wind levels, which don’t account for the urban morphologies. Urban local wind patterns can deviate significantly from regional wind levels. In this research, a semi-analytical methodology is employed to calculate local wind levels in New York City using regional wind levels. The model up-scales the regional wind level to the urban boundary layer; it then downscales the wind level to the average building height of urban localities through two downscaling steps. The analysis is implemented in a geographic information system (GIS) framework. Detailed information regarding building geometries in Manhattan is used to calculate aerodynamic parameters of the city localities, which is used as input for the semi-empirical method. The results are presented as localization wind index map of the city, which can be used to calculate local wind level at any location in the city as a function of the regional wind level. The local wind index can be used to identify the urban localities with higher local wind index, which are more suited for investment to harness the wind energy on top of high rises.

Studies have shown that the human-centered control of heat, ventilation and air conditioning systems could enhance their performance. Respiration is one of the principle mechanisms that human body uses to exchange heat with the ambient environment. This study evaluates the feasibility of using respiration to represent occupants’ thermal comfort using a non-intrusive Doppler radar sensing (DRS) system. The DRS system non-intrusively measures the movement of the target by detecting the changes of the returned signal. We identify human subjects’ pulmonary activities using the movement of chest and abdomen area by a DRS system under the low and high indoor temperature. For five human subjects out of six, an increase in respiration intensity in a higher temperature was observed. The findings demonstrate the feasibility of such approach in quantification of human body thermoregulation. Therefore, this study contributes to measuring real-time respiration as an index of occupants’ thermoregulation state to enable smart thermal control in building operations.

A numerical tsunami model that can incorporate the constructed environment onshore is difficult to build either with a two-dimensional (2D) approach or a three-dimensional (3D) approach. A depth-integrated 2D approach assumes independence of the z-coordinate (in the vertical direction), thus can be oversimplified when modelling complex and variable flow. 3D models require very fine meshing near coastal structures, which dramatically increases the computational cost and makes modeling of some problems impractical. In this study, an experiment modeled after a 500-year Cascadia subduction zone (CSZ) tsunami inundation with a constructed environment is modeled and analyzed. Comparisons are made between a 2D model and a 3D model. Modeling reasonably selected subdomains is shown to alleviate the computational cost of the 3D model. The 2D model does not accurately capture the important details of the flow near initial impact due to the transiency and large vertical variation of the flow. The 3D model does a better job at a much higher cost of computational resources.

The sound recognition and identification technology, which has been significantly developed, is the one of the most promising methods to monitor construction operations and measure their productivity. However, one current challenge resides in identifying event types and working hours from the analysis of overlapped construction working and equipment sounds. To facilitate real-time process monitoring of dynamic construction environment, this paper proposes a frequency-domain-based approach to identifying event detection of overlapped sounds. The frequency-based method can identify response frequencies and their pattern from each type of construction tasks. The experimental validation, which measured frequency patterns of construction working and equipment sounds shows that types of sounds can be detectable based on time history with the aid of spectrogram. Furthermore, hidden Markov model (HMM) is adopted to accurately classify types of construction sounds. The construction sound detection technology will be an imperative asset for diverse disciplines to develop an innovative technology that captures, analyses, and visualizes sounds of construction work and equipment in a dynamic and vulnerable construction site.

Appealing visual design and economic optimization are traditionally two conflicting aspects of good architecture and engineering. An emerging way of improving construction efficiency is the use of highly sophisticated precast elements. While prefabricated elements dramatically improve the efficiency of the construction process, they introduce another limitation on the architectural design. The approach in avoiding this limitation is a combined strategy: In a first step, the architect should generate designs without any precast elements in mind. Within a second step the generic design has to be transformed in a design using a predefined range of precast elements. This process has to consider many constraints and follow several given objectives. Finally, the stated situation is an NP-complete optimization problem. In a three-dimensional BIM based environment this optimization task can be ideally supported by computer aided optimization strategies. The presented solution shows a comprehensive concept and its implementation in a software module which uses genetic optimization for selection of appropriate precast elements.

Over the past two decades, highway agencies have been using geographic information systems (GIS) to store, visualize, and analyze highway-related issues. The fusion of projects information residing in disparate highway data sources can yield useful information to support highway maintenance and rehabilitation (M&R) planning. Through a case study, this study presents applications for projects data fusion in a GIS, identifies data integration challenges, and proposes recommendations on data management practices. Potential applications for integrating highway projects information in GIS: including spatio-temporal conflict analyses and pro-active scheduling of M&R projects can lead to the obviation of costly planning errors. Furthermore, the study identifies data integration issues including data granularity, missing data, semantic interoperability, and discordant use formats. Based on these challenges, this paper proposes the development of data “validation rules,” and pro-active data entry efforts to improve projects data fusion in GIS. The insights from this study can provide directions for future research in pushing the frontiers of data integration of projects data in GIS and its subsequent applications for M&R project planning.

Construction machines and devices often generate distinct sound patterns while performing different tasks, making it possible to extract useful information about jobsites by placing microphones and recording and processing the generated audio files. This paper presents the results of current studies conducted by the authors on the necessary hardware and software for audio modeling of construction jobsites. As the first step, an audio-based system for recognizing activities of construction equipment has been devised. The presented system includes a de-noising algorithm for enhancing the quality of audio files as well as a short-time Fourier transform (STFT) and support vector machines (SVM) for classifying various activities. In the second step, three types of audio recorders (off-the-shelf microphones, contact microphones, and multichannel microphone arrays) and two types of installation settings (microphones mounted on board vs. installed on the job site) have been selected and several experiments were conducted to optimize the hardware settings, tune the algorithmic parameters, and evaluate the different approaches. The results show that for several different types of machines, the accuracy of the audio-based activity recognition system can exceed 85%.

Virtual design and construction (VDC) is the use of multidisciplinary models for improving project performances, encompassing building information modeling (BIM), process optimization, and collaboration management. Active research in the field of VDC has resulted in the development of VDC techniques that can potentially improve the outcome of architecture, engineering, and construction (AEC) projects. However, not every VDC technique is equally suited for every project, and different features of the techniques work together towards the project goals in different ways and magnitudes. While these features of VDC are widely discussed and, in some studies, tracked and measured, there has been limited research that relates the measurements of the features to the achievement of project goals, which VDC ultimately aims to achieve. Initiating a study on this topic would (1) further our understanding of the relationship between VDC features and project goals, (2) help develop models to predict performances based on VDC features, and (3) inform both researchers and project stakeholders the VDC investments that would yield maximal project performance improvements. This research studied the combined effects of VDC techniques on real-life AEC construction projects. The authors predicted various aspects of VDC-related project performances based on 50 VDC features in 108 worldwide project cases. Using artificial neural networks (ANNs), this research achieves a prediction accuracy of up to 78.2% under 10-fold cross-validation. This prediction accuracy shows the relevance of VDC features proposed by our previous research; the predictive model could also help AEC practitioners and researchers estimate project performance in the early stage and adopt appropriate VDC implementations to better achieve project goals.

Traditional paper-based methods for construction activity monitoring are inefficient and error-prone. In this paper, an automated approach is proposed for monitoring the progress of construction activities by tracking major construction equipment and bulk material, using sensor-based technologies. Data obtained from sensor-based technologies will be fused to determine the completed steps of the activities, and consequently to identify the overall progress of the activities. Also, human-generated data, such as daily site reports, will be used to support the sensor data. A rule-based approach is developed for fusion of the collected data. This paper describes the initial field tests that were performed to monitor the rebar installation activity at a pilot construction site. To collect data about the ongoing rebar installation activity, the embedded sensors (e.g., position and load sensors) of a tower crane anti-collision system, and the site reports and quantity takeoffs were used. Initial results show that the developed approach can estimate the progress of rebar installation activity with 98% accuracy.

Traditional earthwork projects rely heavily on project managers and superintendents to generate the truck hauling plan based on experiences. Though various linear programming methods have been proposed to produce cost effective plans specifying each job with source, destination, and volume of material to be transported, how to sequence these haul jobs tends to be neglected. As such, temporal and spatial conflicts between jobs are commonplace in resulting construction plans. On the other hand, establishing linear equations requires considerable expertise and time. This could hamper the updating of the project plan due to the rapidly changing situation on the earthmoving site. In this paper, a fully automated flow network based project planning method is applied to: 1) optimize the earthwork operation; 2) define conflict-free haul jobs; and 3) produce an activity-on-node (AON) network model. Further automation can be achieved through integration with resource constrained scheduling. The clearly defined haul jobs and the AON network will serve as input to the computer platform of simplified discrete event simulation approach (SDESA) for schedule optimization subject to resource availability limits. The complete automation approach is applied to a rough grading case based on a real-world project in northern Alberta, Canada. The case study shows the proposed approach is intuitive to communicate and implement in the field. A comparison is also done between the plan generated by automation and the plan proposed by the superintendent. The result shows considerable savings on both haul effort and total cost.

There has been a need for a systematic risk analysis method for struck-by-equipment hazard that considers interactions and interrelationships among entities (equipment and workers-on-foot) on a job site and also analyzes safety risk at multiple levels. This paper presents a comprehensive spatio-temporal network-based model, developed for dynamic risk analysis pertinent to struck-by-equipment hazard. In the developed model, the dynamic interactions and interrelationships among entities presented on a job site are quantified and conceptualized in a weighted network. The network-based model is capable of conducting safety risk analysis at both entity and network levels in a real-time and proactive manner. Three safety leading indicators, including relative risk score (based on algebraic connectivity), degree centrality, and eigenvector centrality, are adopted to identify different aspects of the dynamic risk of individual entities and job sites. Three simulated job sites were used to illustrate the network-based model for risk analysis. The results show that while the risk level of each job site can be identified, the entities bearing different aspects of risk can also be detected. The over-time risk evolution at both entity and network levels were tracked to gain a full understanding of the dynamic risk on job sites.

Being struck by equipment, such as a heavy machine and vehicle, is one of the leading causes of occupational fatalities in an outdoor construction project. As demonstrated through previous sensor-based research efforts, the proximity detection of workers and active equipment is useful in identifying potential hazards at construction sites. However, attaching sensors to entities can be burdensome for both contractors and workers. Additionally, the measurable ranges of proximity sensors may not be consistent because of the vulnerability to ambient conditions. Against this backdrop, we propose the use of a drone with computer vision techniques for on-site proximity detection. Specifically, the potential for the visualization of a struck-by hazard on air-view image frames is investigated in this paper. The drone can fly over a construction site so that moving entities are persistently captured with a drone-mounted camera while being tracked by a computer vision-based tracking algorithm (i.e., mean shift embedded particle filter). In addition, single view geometry-based rectification is adopted with the known geometry information of a reference object, to reflect a real scene scale and dynamic distortions derived from the drone’s altitude and location. Based on the rectified proximity within the image plane, a struck-by hazard zone is defined and visualized in the video, which provides intuitive feedback on whether a worker of interest is within the potentially hazardous area or not. To validate the presented method in terms of accuracy and applicability, a site video from a wastewater treatment plant renovation project was collected. The preliminary test results showed that the defined hazard zone on the drone’s image plane can be rectified to reflect reality on the ground plane. The proposed drone-based on-site proximity detection potentially may be used to identify the risk of workers being struck by equipment in an effective way.

Gazetteers, also known as entity dictionaries, can be applied to support many information extraction (IE) applications such as named entity recognition (NER). However, gazetteers are not always available, because they require not only domain knowledge but also human effort during development. Existing gazetteers are also mostly general in nature; they are limited to providing common types of entities such as locations, organizations, person names, etc. These common types of entities cannot accurately reflect the semantics of specific domain knowledge, and their applicability is therefore limited. A useful gazetteer, on the other hand, must be able to indicate the important types of entities (i.e., classes of concepts) and must also contain a sufficient number of entities. This creates the need for domain-specific gazetteers, especially when involving specialized IE tasks. In this paper, the authors take construction safety management as the target domain and propose a semi-automated approach to develop a construction safety gazetteer, aiming to eventually support IE applications for constructions safety management. The proposed approach consists of three steps: (1) applying natural language processing (NLP) techniques to extract important phrases (not limited to single terms, but including bigrams and trigrams) from text resources, (2) defining important types of entities as entity classes for the construction safety domain, and (3) assigning the extracted phrases to the predefined entity classes. The authors also discuss how the proposed methodology could be affordable for domain experts, and the possible scenarios for IE applications for supporting construction safety management.

There is an increased interest in using text mining techniques to automatically classify text-based accident descriptions in industries such as aviation, medical and construction. The ability to automatically classify accident narratives enables large text databases to be analyzed to provide insights on accidents and near misses. Several machine learning and text mining approaches such as support vector machines (SVM), naïve Bayes, and neural networks have been adopted in the literature in order to classify accident narratives. In the recent years, ensemble approach has gained popularity among machine learning applications due to the fact that, ensemble approach combines multiple machine learning algorithms into a much stronger learning algorithm yielding better results. Thus, this study evaluates the effectiveness of an ensemble approach which often performs better than a single learning algorithm, using popular machine learning algorithms: support vector machine, decision tree, linear regression, k nearest neighbor, naïve Bayes and neural network. Analyzing the accident narratives reported in the construction safety data yields useful knowledge that can be used to improve the understanding of what went wrong in the past and take necessary precautionary measures to prevent future accidents. One thousand accident narratives obtained from the US OSHA website are used to facilitate this study. Uni-gram tokenization, tf-idf document term matrix representation along with 11 class labels are used to perform the research approach. The precision of ensemble model ranged from 0.6 to 1.0; recall ranged from 0.18 to 0.96, and F1 score was 0.14 to 0.96 across the 11 labels of accident types. The highest average F1 score reported herein was 0.69.

Psychological states are known to be related to human behaviors, and investigating the relationship between psychological states and unsafe behaviors of workers may shed light on how to better prevent accidents on the construction jobsites. One challenge of such investigation is that it is difficult to track workers’ behaviors and psychological states on a construction job site, because (1) psychological states are difficult to be monitored timely and accurately on site due to the lack of non-interruptive monitoring methods and workers’ resistance to be monitored; and (2) monitoring unsafe behaviors usually requires extra on-site staff, which may lead to more safety risk. To address this problem, a virtual environment-based experimental method is proposed. The collected data was analyzed to reveal the relationship between workers’ psychological states and unsafe behaviors, which can help improve safety training and safety management on a construction jobsite by providing the evidence of emotion’s influence on unsafe behaviors and to what extent those emotions can affect behavior.

The unscripted nature of construction jobs puts equipment and workers in close proximity of one another which can lead to near-miss situations or contact collisions. Previous research has primarily focused on location-aware methods to improve construction safety from a technological perspective, but has fallen short in investigating the role of individuals in creating or avoiding safety accidents. This paper presents a preemptive safety framework that incorporates not only positional data but more importantly, workers’ perception of and attitude toward safety. First, a comprehensive review of previous work related to construction safety, resource tracking, trajectory prediction, and risk behavior is presented. Next, the designed methodology is explained and tested which incorporates worker’s risk index (a measure of affinity for or aversion to risky behavior near hazards) with conventional trajectory modeling in order to better predict the worker’s future position, and detect potential contact collisions in advance. In particular, we examine the reliability of polynomial regression and hidden Markov model (HMM) in predicting a worker’s position given his/her positional data collected by a smartphone’s GPS sensor. Next, HMM prediction is further modified by considering the worker’s past walk trajectory near hazard zones. Preliminary results demonstrate that the developed motion trajectory prediction can reliably detect unsafe movements and near future collision events. The outcome of this research contributes to improving current practices of site safety monitoring.

For a construction project to be on schedule and within budget, project managers must constantly monitor work progress, identify deviations from plans, and design a more efficient workplace. This often requires meticulous attention to how field tasks are conducted by workers over time. From a worker’s perspective, to be more productive may translate into carrying out tasks that exceed one’s natural physical limits. The sustained physical labor will result in work-related musculoskeletal disorders (WMSDs) which can adversely affect the project budget, schedule, and productivity. To prevent WMSDs, health and safety organizations have established rules and regulations which limit the duration and frequency of labor-intensive activities. Proper implementation of these standards requires that worker’s physical motions are constantly tracked. Monitoring workers’ activities serves a two-fold purpose: it provides information about the work progress, as well as help quantify the ergonomic risks associated with those activities. The complexity of activities, large number of workers, and psychological and physical barriers in the workplace make manual data collection very challenging. In this paper, the authors present and validate a model for remotely and unobtrusively monitor worker’s activities. The model deploys built-in smartphone sensors and machine learning algorithms to recognize worker’s activities in the field, and extract duration and frequency information, which will be ultimately used to evaluate productivity as well as ergonomic risks associated with each activity.

Crane operations can be safety hazards to their surroundings on job sites, particularly during blind lifts, when crane and its payload movement is affected due to the reduced visual field of the crane operator. The increased availability of high-definition cameras mounted on unmanned aerial vehicles (UAVs) helps to address this issue as it allows the crane and its surrounding resources, such as workers and their activities, to be monitored in real time. In this paper, we focus on the task of localizing cranes in images obtained from these flying cameras. Our aim is to prove that state-of-the-art object detectors can localize cranes in visual feeds accurately and in real time. To this end, we apply a state-of-the-art object detector to a dataset of crane images captured from a real-world construction site. Preliminary results show that this object detector is effective at pinpointing crane locations in real-time and could be integrated into end-to-end machine vision applications such as payload tracking or 3D crane pose estimation.

Hazard recognition is the first step in mitigating safety incidents on a construction site. The unprecedented growth in jobsite visual data provides a unique opportunity to leverage computer vision systems to handle various aspects of safety inspections. Such systems can be widely deployed on jobsites for safety monitoring and safety performance measurement. They also provide a semantic-rich object-oriented way to represent safety hazards to as-built BIM and VR-based safety training. To achieve this goal, one needs to properly introduce safety domain knowledge in the design of computer vision systems. Towards this end, this paper presents the design of a language-image framework that aims at understanding and detecting semantic roles of activities mentioned in safety rules. We define this task as visual safety checking. The framework, which is based on visual semantic role labeling, detects visual grounding of ‘verbs’ described in safety rules, such ‘verbs’ represent states or processes of visual activities. This framework includes (1) semantic parsing of safety rules, (2) training construction object detectors, and (3) training semantic role detectors. We present preliminary results in semantic safety rule parsing and the construction of a new visual safety dataset based on parsed construction objects. Preliminary results are presented from implementing two state-of-the-art object detectors on this dataset and the benefits and the limitations are discussed in details.

Hazard identification in construction typically relies on visual inspection. Visual inspection, however, is very time-consuming and error-prone due to the dynamic nature of the construction workplace and the limitations of human perception. Our previous study introduced a new approach, one that infers the existence of hazards by analyzing gait abnormality patterns of workers based on information from wearable inertial measurement units (IMU). Previous studies have not integrated such analysis with the two-dimensional spatial information of each gait cycle, which is essential to localize safety hazards in a typical construction workplace. To this end, this study used two-dimensional spatial information captured from ultra-wideband (UWB) localization technique and analyzed the spatial pattern of workers’ gait cycles in typical construction settings. An experiment simulating an indoor brick installation space was conducted with different types of hazardous conditions (e.g., obstacles and slippery surfaces) to demonstrate the feasibility of the proposed approach. The experiment results show that proposed approach is able to identify hazards and their locations during indoor brick installation work. This result highlights the potential of revealing latent fall hazards in an efficient and semi-automated way, hazards which may be difficult to recognize purely through visual inspection.