Journal of Computing in Civil Engineering

The ubiquity of smartphones has led to a variety of studies on how phone accelerometers can be used to assess road pavement quality. The majority of prior studies have emphasized supervised machine learning, assuming the ability to collect labeled data for model development. However, variances in vehicle dynamics and roadway quality, as well as the reliance on data labeling, limit the generalizability, scalability, and reproducibility of these approaches. Here, we propose an unsupervised learning framework that combines Pareto-optimized wavelet featurization and clustering. We first demonstrate the applicability of wavelets as features for dimensionally reduced accelerometer data. Wavelet featurization typically requires significant empirical tuning for optimal featurization. The presented framework automates tuning and selection of wavelet features based on subsequent clustering metrics. These metrics are related to inherent cluster characteristics such as intercluster variance and between-cluster variance. Rather than select a clustering method a priori, the proposed approach optimizes across a variety of clustering algorithms and hyperparameter configurations, again based on a set of clustering metrics. Experimental evaluation shows that the framework is able to detect road pavement distress and distinguish between classes of pavement defects, but that low-cost smartphone sensor data may not be as reliable in discriminating the more nuanced characteristics of pavement distress. This was most notable in the case of cracking, where the type and range of cracking severity caused undesirable cluster separation. The presented framework is general and cost-efficient and opens the way to further research on automatic pavement distress recognition from crowdsourced, low-cost data.

Building information modeling (BIM) object classification is a key step in supporting the full automation of architecture, engineering, and construction (AEC) domain tasks such as cost estimation and building code compliance checking. A machine learning approach is designated to address any classification task without requiring the domain knowledge to be explicitly or manually specified in detail. The success of machine learning, however, relies on the quality and suitability of input features. In order to support seamless interoperability of BIM applications, the authors have proposed invariant signatures that uniquely define each AEC object and capture their intrinsic properties. In this paper, the authors combine the use of invariant signatures together with machine learning approach to address BIM object classification. The developed invariant signatures include geometric signatures, locational signatures, and metadata signatures. To test the robustness of their use as machine learning features, the authors created a new BIM object data set with 1,900 AEC objects in five major categories of building elements, including beams, columns, footings, slabs, and walls. The data were manually annotated by independent annotators to ensure the quality. Among those AEC objects in the data set, 1,330 objects (70% of the data) were used as training/development data and 570 objects (30% of the data) were used as testing data. The authors extracted the predefined invariant signatures as features and tested the robustness of them in AEC object classification using different machine learning algorithms. The best-performing algorithm achieved 99.6% F1-measure in the testing data, which outperformed the state of the art (94.9% F1-measure). As a demonstration of the value of such object classification, a comparative experiment was conducted to take off quantities of walls from a student apartment complex, both using the state-of-the-art commercial software and using the object classification–based automation. Consistent results were found between these two quantity takeoff methods, whereas using object classification–based automation further saved time and manual efforts significantly (saved 98.1% of the loading and object selecting time). These results showed that the use of proposed invariant signatures and machine learning algorithms in BIM applications is promising.

The type and duration of construction workers’ activities are useful information for project management purposes. Therefore, several studies have used surveillance cameras and computer vision to automate the time-consuming process of manually gathering this information. However, the three-stage method they have adopted consisting of separate detection, tracking, and activity classification modules is not fully optimized. Additionally, the activity classification module is trained per-clip/segment on trimmed video clips and fails when applied to long untrimmed construction videos. This paper aims to (1) investigate the benefits of a fully optimized method such as you only watch once (YOWO) and a per-frame and per-worker annotated untrimmed data set over the previous approach for activity recognition of construction workers; (2) propose an improved version of YOWO, called YOWO53, to improve detection performance; (3) propose a semiautomatic data set annotation; (4) conduct a sensitivity analysis to compare the performance of YOWO, YOWO53, and the three-stage method; and (5) conduct a case study to compute the percentage of different workers’ activities. YOWO53 improves the detection recall of YOWO by up to 3%, and the classification accuracy of the three-stage method by 16.3%. Although YOWO53 has a lower inference speed, it is still sufficiently fast for productivity analysis.

System partitioning for effective simulation of civil infrastructure flow networks on parallel processors is a nontrivial problem. Arbitrary partitioning focused only on balancing processor workload can lead to a large interprocessor communication burden that limits parallel speedup. Thus, there is a need for intelligent partitioning algorithms that balance the estimated computational load while minimizing the number of connections between partitions. Graph theory provides widely used partitioning methods, but these are applicable to networks with power-law connectivity and where the computational workload is proportional to the number of system nodes—conditions that do not hold for finite-volume solution of water drainage networks (e.g., river systems, stormwater drainage systems). This paper presents the novel BIPquick algorithm, which is shown to be an effective approach to identifying network partitions with reduced connectivity for systems that are directed acyclic graphs (DAGs) and have a physical limit on the number of connections per network node. Novel developments include (1) a node-cut approach that allows a partitioning workload function to be exactly balanced in systems where the computational work is proportional to the link length between nodes, (2) a finite-pass approach to partitioning that ensures a partitioning solution in a known time, and (3) a new connectivity scaling metric that allows simple evaluation and comparison of different partitioning results. The BIPquick model is tested on a large river network with up to 10,000 partitions.

With the emergence of the smart city, there is a growing need for scalable methods that sense how humans interact and use infrastructure in order to model social behaviors relevant to designing sustainable and resilient built environments. Cyber-physical system (CPS) frameworks used to monitor and automate infrastructure systems in smart cities can be extended to sense people to better understand how they use infrastructure systems including social infrastructure (e.g., parks, markets). This paper adopts convolutional neural network (CNN) architectures to automate the detection and spatiotemporal mapping of people using camera data to form a cyber-physical-social system (CPSS) for smart cities. The Mask region based convolutional neural network (R-CNN) detector was adopted and tailored to identify and segment human subjects in real time using camera images with an average speed of 7 frames per second. The Mask R-CNN framework was trained end to end using the Objects in Public Open Spaces (OPOS) image data set that includes classified segmentations of people in public spaces. A two-dimensional/three-dimensional (2D-3D) lifting algorithm based on a monocular camera calibration model was also employed to accurately position detected people in space. Finally, a Hungarian assignment algorithm based on association metrics extracted from detected people was used to assign people to spatiotemporal trajectories. To demonstrate the proposed framework, this study used the Detroit riverfront parks to study how people utilize community parks, which are a form of social infrastructure. The Mask R-CNN detector is proven precise in detecting and classifying the behavior of people in parks with mean average precision well above 85% for all class types defined in the OPOS library. The framework is also shown to be effective in spatially mapping the various uses of park furnishings, leading to better management of parks.

The COVID-19 pandemic has impacted how the construction industry operates around the world. To fight the risk of transmission, new health, safety, and environmental (HSE) protocols have been put in place. Among these protocols are social distancing and limiting the number of workers per area, where social distancing acts as a so-called protective bubble for each worker. Contractors are now required to attempt to achieve (and be prepared to keep) social distancing among their workers whenever needed and possible. Otherwise, they could be forced to halt operations due to having an unsafe environment. Accordingly, construction plans, and corresponding workspace assignments, should be revised in a four-dimensional (4D) environment to ensure fulfillment. Even after the end of this pandemic, the new HSE awareness achieved during this experiment is expected to reshape the so-called new normal of construction. Therefore, this paper presents a novel workspace simulation and management solution comprising a theoretical framework and a semiautomated tool to incorporate physical distancing during 4D planning. The semiautomated tool creates a 4D building information model, loaded with workspaces and social distance bubbles as stochastic variables, and utilizes Monte Carlo simulation to model uncertainties occurring onsite. The uncertainties considered are both temporal and spatial, i.e., changes in productivity and workspace sizes, respectively. This tool surpasses existing workspace management solutions in that (1) it has a schedule generation module to recompute schedule projections based on temporal uncertainties, (2) its workspace generation module can automatically create physical distance buffers around selected workspaces, as per site conditions, (3) its 4D simulation can realistically mimic the work progress on the site, and (4) its 4D clash detection module can smartly detect and report both soft and hard operational clashes. Additionally, the proposed analytics target three levels of clash resolution: site, workspace, and activity level. The framework and developed tool were tested against a residential building case study. Over the course of 155 days, 26 activities with 257 workspace assignments were examined. The proposed solution was able to capture the critical schedule duration (21 out of 155 days), the impactful 4D clashes (44 out of 2,900), and the activities involved in the most sever clashes (5 out of 26). Hence, the proposed method and the developed software tool will help planners/construction managers understand the space requirements for construction operations considering social distancing and other required safety buffering, identify critical spatiotemporal zones, and suggest resolution strategies for the resulting clashes based on the analytics.

Smart devices for structural health monitoring provide edge computing capabilities to reduce wireless transmission and, thus, power consumption. Although effective algorithms have been proposed in the last few decades, traditional microcontrollers require heavy data flow between the memory and the central processing unit that involves a considerable fraction of the total energy consumption. Phase change memory has recently emerged as an attractive solution in the field of resistive nonvolatile memory for analog in-memory computing, which is a valid approach to avoid data being conveyed among distinct elaboration units. However, it has never been envisaged in structural health monitoring applications. As this technology is still in an embryonic state, several challenges related to nonlinearities and nonidealities of the memory elements and the energy expenditure related to the memory reprogramming process may undermine its usage. In this paper, the application of a novel identification approach for civil infrastructures is investigated using phase change memories. The main computational core of the presented algorithm, consisting of one-dimensional convolutions, is particularly suitable for implementations involving analog in-memory computing, thus showing the great potential of this technology for structural health monitoring applications. The test unit is an embedded phase change memory provided by STMicroelectronics and designed in 90-nm smart power bipolar complementary metal-oxide-semiconductor (CMOS)-double-diffused metal-oxide-semiconductor (DMOS) technology with a Ge-rich Ge-Sb-Te alloy for automotive applications. Experimental results obtained for a viaduct of an Italian motorway support the efficacy of the method. Moreover, the influence of nonidealities on the outcomes of damage identification based on both dynamic and quasi-static structural parameters is examined.

Because large-scale building projects are designed through multiprofessional collaboration and the design changes are unavoidable and frequent, it is essential to efficiently detect the changes between submodels in different versions for building projects. However, the traditional methods may have lower efficiency problems, such as time-consuming and inaccurate results via manual operation and visual inspection. To address these issues, this paper proposes a submodel comparison method based on Industry Foundation Class (IFC) file contents, called as IFC-submodel content comparison (SCC). First, all instances in two IFC files were, respectively, preprocessed and stored in a unified format. Next, all instances related with the required demand were extracted to form submodels. Then, the instances of the two submodels were all checked to remove redundancies, respectively, and each remaining instance was assigned at least one signature, which is a string consisting of feature information. Finally, these instances in the two submodels were compared by their signatures in a breadth-first search manner, and the matched information and unmatched information were extracted. The experimental results showed that the proposed IFC-SCC method is superior compared with the previous methods. Compared with the globally unique identifier (GUID)-based comparison method, the IFC-SCC method generated more accurate matched information, and obtained an improvement in time efficiency by 16.85%. Compared with the flat-based comparison method, the IFC-SCC method had an improvement of 48.35% on the time efficiency on average and removed the redundancies.

Bridges experience regular inspections from transport agencies/authorities periodically. However, current approaches for the storage and visualization of inspection-related data are often restricted to spreadsheets and inspection reports and cannot sufficiently support open building information modeling (BIM) processes. This research proposes an industry foundation classes (IFC)-based method to document and represent inspection-related information in bridge BIM models. Specifically, the geometry of defects, including spatial placement and shape representation, was modeled parametrically. Interrelationships between defects and other IFC entities (e.g., bridge elements, root causes, and maintenance actions) were represented. The proposed method was validated on a concrete highway bridge, with two scenarios designed to illustrate its capability to support cause diagnosis and maintenance decision making. The interoperability of the proposed method across different BIM tools was evaluated as well. This work demonstrates the parametric-driven representation of defect information based on the latest IFC, facilitating an integrated BIM environment for the lifecycle management of civil assets.

Disorientation has been identified as one of the leading causes of firefighter injuries and fatalities. In search and rescue tasks, firefighters have to build up a general understanding of an unfamiliar space with limited time and limited information at hand. These two factors together lead to firefighters’ insufficient development of spatial knowledge that could induce firefighters’ disorientation. Therefore, there is a pressing need to better understand how and why different types of spatial information, including landmark, route, and survey information, affect the development of firefighters’ spatial memory of unfamiliar environments. This study examined functional brain connectivity associated with different forms of spatial information that regulate firefighters’ wayfinding performance. We conducted a virtual reality (VR) experiment to investigate the brain-based connectivity changes during wayfinding tasks. To achieve the research goal, we recruited 40 firefighters from the Bryan Fire Department in College Station, Texas. The Firefighters were requested to find three missing victims in a VR office maze, given different wayfinding information, including landmarks, routes, and maps. Functional near-infrared spectroscopy (fNIRS) was used to analyze firefighters’ cerebral hemodynamic responses associated with neural activation and connectivity patterns. The results showed that the firefighters had better task performance when they used route or survey spatial information than when they used other information. Survey information (maps) led to higher neural activation and less-desired neural connectivity patterns than did landmark and route information. Although maps provided complete spatial information about the building layouts in the experiment, they also burdened firefighters’ cognitive processes and required additional effort in memory retention and path planning. Thus, survey information did not result in better performance as suggested by previous studies. The findings are expected to help fire departments design better training protocols and inspire the design of cognition-driven personalized wayfinding systems for firefighters.

The relationship between the demographical characteristics of building occupants and their perception of indoor comfort is increasingly being studied. However, the added value from accounting for such characteristics when modeling and predicting occupants’ perceptions remains unclear. An incremental machine learning (ML) modeling and analysis approach is proposed to quantify the influence of four demographical factors (gender, age, nationality, and time lived in the environment) on occupants’ perceptions of their indoor environment conditions. A three-step methodology is presented: (1) data collection through sensors and a questionnaire administered on 206 occupants of academic and office buildings in Abu Dhabi, UAE, (2) development of ML models (i.e., support vector machine, random forest, and gradient boosting) to predict occupants’ perceptions under different scenarios of demographical representation (i.e., from no representation to all demographical parameters included), and (3) analysis of the impact of demographical parameters’ inclusion on the performance of the ML models in terms of predictive accuracy, F1-scores, and computing time. Results confirm that including demographical variables could increase prediction accuracy and F1-scores by approximately 19% and 56%, respectively. However, in some instances, the inclusion of these variables reduced model performance while increasing computing time by as much as 50%. A detailed discussion is presented on the comparative performance of the different tested ML algorithms and the need to strike a balance between increasing model complexity and computational costs.