Journal of Infrastructure Systems

Accurate recognition and location of pavement manholes are of great significance for pavement maintenance. This paper proposes an improved You only look once X (YOLOX) for automated detection of manholes on asphalt pavements. The proposed model improves the performance of the YOLOX model in two respects. First, the channel attention mechanism is introduced to enhance the model’s adaptive feature refinement; second, a microscale detection layer is deployed in the YOLOX model to extract more essential and distinct features. The experimental results are impressive, with the improved YOLOX achieving an F1 score and overall intersection-over-union of 98.14% and 91.61%, respectively, on 250 testing images, surpassing other state-of-the-art models such as YOLOv4, Faster R-CNN, EfficientDet, and the original YOLOX. To demonstrate robustness of the proposed model, the improved YOLOX is further applied to process manhole images taken randomly by a smartphone, which differ significantly from those acquired by a laser imaging system. It is found that the improved YOLOX can also yield similar detection efficiency in different scenes, which indicates the proposed model has a strong generalization ability. Particularly, the average frame per second (FPS) of the improved YOLOX is approximately 50.74 FPS using a modern graphic processing unit (GPU) device, implying the promising potential of the proposed model in supporting real-time automated detection of pavement manholes.

This paper delivers a detailed statistical assessment of pavement rehabilitation treatments by delivery methods via studying their performance in terms of pavement indicators (international roughness index, rutting depth, and pavement condition rating) and in terms of their pavement service lives. The data include 812 pavement segments that were rehabilitated under six commonly implemented treatments and through six commonly used delivery methods [with a focus on public–private partnerships (PPP)], which were let or completed in the United States between 1996 and 2011. The treatments include a two-course hot-mix asphalt (HMA) overlay with or without surface milling, concrete pavement restoration; three-course HMA overlay with or without surface milling; three-course HMA overlay with crack and seat of portland cement concrete pavement; 3R (resurfacing, restoration, and rehabilitation) and 4R (resurfacing, restoration, rehabilitation and reconstruction) overlay treatments; and 3R/4R pavement replacement treatments. The delivery methods/PPP types include performance-based contracting, cost-plus-time, incentives/disincentives, design-build and their derivatives, warranties, and lane rentals. To model and forecast pavement performance, a three-stage least-squares approach is employed. For the pavement service life analysis, the elapsed time until the pavement crosses a threshold is investigated, using random parameter hazard-based duration models. Separate models are estimated for each combination of delivery method/PPP type and rehabilitation type. The model estimation results show that several influential factors, such as traffic, weather, and pavement characteristics along with drainage condition affect pavement performance and pavement service life; these factors differ among pavement rehabilitation treatments and delivery methods/PPP types.

The lack of parking infrastructure continues to pose a problem for urban commuters, as the parking demand in most cities outstrips supply and significant driver time is expended searching for parking. The emergence of vehicle automation offers an opportunity to help mitigate this issue. In the autonomous vehicle (AV) era, it is expected that after dropping off its passengers at their destinations, the AV will park at a relatively inexpensive parking facility located outside the downtown area instead of the existing, higher-priced facilities in the central business district (CBD). This is expected to decrease CBD parking demand, ultimately leading to the possible decommissioning and repurposing of some existing parking infrastructure in the CBD and the construction of new infrastructure in the city’s outlying areas. What is needed, therefore, is a framework for city road agencies for decommissioning/relocating/locating and user pricing of parking infrastructure to serve human-driven vehicles (HDVs) and AVs. To address this issue, this study presents a bilevel framework. The road agency (at the upper level) seeks to: (1) minimize travelers’ cost systemwide; and (2) maximize monetary benefits of infrastructure decommissioning and parking fee revenue at the upper level. Travelers (at the lower level) seek to reduce their costs of travel in response to the road agency’s decisions made at the upper level. A hybridized solution approach (optimization heuristics and machine learning) is implemented for this mixed-integer nonlinear problem. The numerical experiments provided a number of insights regarding parking infrastructure location design and user pricing in the prospective AV era.

There is a clear and present desire from government authorities to actualize processes and procedures that place equity at the forefront of decision-making and improve supply chain resilience. Following a federal executive order, several agencies including the Federal Emergency Management Agency have instituted initiatives focused on equity in resilience planning, recognizing the variance in how different populations are impacted during disturbances. To actualize policy and programmatic priorities, decision makers need tools that are theoretically grounded yet computationally simplistic and adoptable for participants outside of academia. Thus, developing methodology that is deployable at scale within the framework of average agency capabilities remains critical. This research seeks to leverage network science, transportation theory, and social vulnerability analysis to explore the relationship between the exposure of counties to food supply disruptions through the road and highway transportation network and their community social vulnerability. This result highlights that in our food transportation network, targeted attacks are not more impactful than random removals when fewer nodes are removed. Our results also found that the most exposed counties are generally in the southeastern portion of the region. In our study area, a large part of the exposure is due to the high food-inflow demand. The framework presented is both theoretically grounded and computationally simplistic and presents a potential strategy for understanding social vulnerability alongside critical infrastructure resilience.

Abstract

This preliminary study presents the development of integrating a real-time mobile device with a machine vision algorithm to assess the retroreflectivity of the broken lane lines of in-service road marking. A stereo camera was used as the photometer, and the measuring vehicle’s headlights were used as the illumination system. The machine vision algorithm includes marking centroid determination, standard measuring condition control, illumination condition calculation, and luminance measurement. The test results show that the average absolute error percentage of 47 marking samples is 6.1%, with the highest and lowest accuracies of 99.9% and 85.7%, respectively. The left and right lane line (broken line) markings can be evaluated simultaneously in a single pass up to 100 kph. The hardware package, including a stereo camera, a camera support beam, cables, and a laptop computer, costs approximately USD 3,500, which is much lower than the cost of conventional and advanced fully automatic retroreflectometers. Moreover, the developed method is for general usage and can be easily modified and applied to camera sets and vehicle carriers of different specifications. Although the results are promising, the proposed method has some limitations. First, the accuracy decreases when the test section is rough with bumps and dips. Second, this algorithm is not ready for surveying solid lines. Furthermore, the current version can only be implemented under vehicle headlights. Future work can focus on improving the hardware and machine vision algorithm to overcome the challenges.

Practical Applications

This preliminary study aims to develop an integrated system for automatically surveying road markings’ retroreflectivity. In many countries and regions, road markings’ retroreflectivity is only examined during construction instead of being inspected periodically due to a lack of efficient equipment, budget, and techniques for surveying. Although some commercial instruments have shown promising results, no detailed information or discussion on the algorithm of reflective light acquisition and retroreflectivity computation has been reported. This study presents a detailed description about the developed machine vision algorithm that can detect and analyze the marking’s dry retroreflectivity in the nighttime. The hardware package, including a stereo camera, a camera support beam, cables, and a laptop computer, costs approximately USD 3,500, which is lower than the cost of conventional and advanced fully automatic retroreflectometers. Moreover, the developed method is for general usage and can be easily modified and applied to camera sets and vehicle carriers of different specifications. However, currently the algorithm is only for broken lane line under relatively dark environmental condition. Future studies are suggested to include more road marking types and reduce the effects of ambient light sources on the analyzed results.

The effects of global climate change are already observed in coastal communities, and one of the key impacts of climate change is sea level rise (SLR). Such increases in sea level are expected to have large impacts on transportation infrastructure and result in region-wide commute disruption. Studies and adaptation strategies in coastal areas will allow decision-makers to take better actions against SLR. The present paper uses the San Francisco Bay Area as a case study. The study quantifies the interregional traffic delay of commuters due to the inundation of shoreline protection segments of the Bay Area. The approach is based on the marginal effect of protection within San Francisco Bay where individual stretches of coastline known as operational landscape units (OLUs) are protected one at a time. In the study, we integrate detailed shoreline scenarios, coastal inundation modeling, and traffic disruption modeling. Important insights from the analysis are obtained: the protection of an OLU (and specifically San Rafael OLU) will produce an average commute time reduction of up to 17% in the neighboring areas or the neighborhoods located across the bay. Five OLUs along with four bridges were identified as critical for minimizing disruptions to commute times. This methodology may be used by decision-makers when proposing adaptation and protection strategies that account for both local and regional areas.

Highway agencies are increasingly directed to broadly incorporate social equity goals into infrastructure planning. To better accommodate these goals in their decision making, it is essential for practitioners to assess the current level of social equity in road network performance. However, social equity in road network performance has primarily been assessed qualitatively in the existing literature, making objective assessments difficult. This work addresses this gap by proposing a quantitative approach to assessing social equity in a road network. This is achieved by formulating an assessment approach and exploring numerical measures to evaluate social equity for road network performance. The proposed approach is implemented on the street network of two cities in the United States: Seattle and New York. With the help of GIS-based visualization, the street network is studied in conjunction with the socioeconomic demographics of the city to gauge its current levels of social equity. The formulations and measures proposed in this study expand the use of quantitative metrics of social equity to support decision making. An objective evaluation of social equity levels can assist needs assessment and guide decision making.

This study used genetic algorithms as part of an optimization framework to directly minimize the expected impacts of road network disruption triggered by seismic events. This minimization is achieved by selecting an optimal set of bridges to retrofit to decrease their probability of being unavailable after an earthquake. We propose a genetic algorithm that outperforms other retrofitting techniques, such as ranking bridges by vulnerability or traffic importance. The proposed framework was demonstrated using the San Francisco road network as a testbed. This example showed that bridges selected by genetic algorithms are structurally vulnerable groups of bridges that act as corridors in the network. Additionally, this study evaluated and recommends domain reduction techniques and hyperparameter calibrations that can decrease the computational costs of this approach.

This paper presents a methodology to evaluate the direct and general equilibrium losses for a testbed community subjected to a megathrust earthquake and tsunami hazard originating in the Cascadia Subduction Zone. The testbed community studied consists of buildings for all residential and commercial sectors in the economy. A fragility analysis and functionality model are applied to estimate the direct damage and losses of these buildings at the parcel level. The process relies on Monte Carlo simulations (MCS) that propagate uncertainties from the hazards through to the damage and loss models. A computable general equilibrium (CGE) model is used to assess aggregated general equilibrium losses to the community. As an important mitigation strategy, seismic retrofit reduces the direct loss to building functionality and general equilibrium losses. Results show that the vulnerability of economic sectors depends on the hazard type, hazard intensity, economic zone, and building type, and the risks vary with the recurrence interval. The highest risks are associated with 500-year and 1,000-year mean recurrence intervals for joint seismic–tsunami hazards, respectively. Results from assessing different design alternatives show that whereas retrofitting all buildings to the highest code level considered results in the lowest losses, retrofitting only commercial buildings can be an efficient retrofit option for enhancing community resilience when controlling for costs. Last, a sensitivity analysis shows that losses and associated risks are sensitive to the definition of building functionality, which highlights the necessity for a common definition of building functionality when performing vulnerability analyses.

Abstract

Water main failure prediction models are a common tool in support of the asset management plans of utilities. This paper focuses on the development and evaluation of models that incorporate survival analysis in combination with machine learning algorithms. To understand the potential advantages of these models, five representative algorithms are applied to data from four different utilities. The Cox proportional hazards (Cox-PH) algorithm, a common survival analysis technique, is considered to provide a baseline measure of model performance. Other algorithms investigated include: extreme gradient boosting (XGBoost), random survival forest (RSF), neural multitask logistic regression (NMTLR), and XGBoost survival embeddings (XGBSE). XGBoost is a pure machine learning algorithm whereas the other algorithms embed survival analysis within a machine learning framework. Our findings suggest that the models that embed survival analysis within machine learning may predict water main failures more accurately than traditional survival analysis or machine learning models. All models are developed using data sets with high degrees of censorship and varying characteristics; thus, these approaches may be beneficial for a wide range of utilities, including those with limited break data. In particular, models based on the RSF algorithm are found to perform consistently well for the data sets investigated herein.

Practical Applications

To support decisions related to the repair or replacement of water mains, many utilities develop models that predict which water mains are most likely to break. Survival analysis (a branch of statistics) and machine learning models have been frequently applied in previous studies on water main failure prediction. However, when applied on their own, there are limitations to both of these types of approaches. More recently, several studies have focused on the potential for combinations of survival analysis and machine learning to overcome these limitations. In this study, three algorithms that combine machine learning and survival analysis are compared with a common survival analysis method and a machine learning algorithm. Two of the selected algorithms have not previously been applied in water main failure prediction, and this study exhibits their capabilities in this field. The algorithms are applied to data sets from four different utilities. Results from the analysis indicate that combinations of survival analysis and machine learning outperform a traditional survival analysis algorithm and a machine learning algorithm for the data sets included in this study.

Public–private partnerships (PPPs) are the primary means of delivering water projects in developing countries. In China, most water PPP projects remain in the construction and operation stages, with few having passed to the transfer stage. The lack of experience in managing transfer risks is itself an impediment to the further development of China’s PPPs. Thus, this study attempts to address two research concerns: (1) identifying transfer risks and risk factors, and (2) dynamically evaluating those transfer risks. Literature review, expert interviews, and process modeling via Integrated Computer-Aided Manufacturing Definition Method 0 (IDEF0) were utilized to identify and assess transfer risks with risk factors. The time-varying nature of transfer risks was evaluated using a system dynamics method. Eleven transfer risks and their value accumulation curves were identified. Value growth of legal and residual value risks remains steady over time. The risk values of contract, political, overhaul, information, and employee risks pose severe initial threats, but reduce on project transfer. However, credit, data transfer, after-transfer operational, and succession operator risks tend to escalate over time. The proposed model allows for a dynamic evaluation of transfer risks, and in so doing, improves the efficiency of PPP project transfer risk management.

The need for healthcare facilities has increased with aging populations. Public–private partnerships (PPPs) have a proven record of accomplishment in supporting the financial viability of various types of infrastructure projects. Thus, globally, governments recently applied the PPP model to characteristic health town (CHCT) projects. However, previous studies focused primarily on the success factors of PPP projects, whereas studies of the factors in CHCT-PPP projects’ operational performance have been limited. This study identified the critical factors in a CHCT-PPP project’s operational performance. First, a literature review and expert interviews identified 24 factors in a CHCT-PPP project’s operational performance. Second, exploratory factor analysis (EFA) and confirmatory factor analysis (CFA) were performed to determine the common factors in the CHCT-PPP project’s operational performance. To establish the final factor system, 264 effective questionnaires were collected to investigate stakeholders’ views on the relative importance of different factors. Finally, importance–performance analysis (IPA) was performed to identify the four key factors in the CHCT-PPP project’s operation performance: customer satisfaction, government satisfaction, value-for-money index, and healthcare-product development. This study can serve as a reference for identifying the key factors in a CHCT-PPP project’s operational performance and promoting the development of CHCT-PPP.

Abstract

The global supply chain depends on the efficient transfer of containers between cargo ships and ground transport. To ensure reliable operations, equipment at ports requires predictive maintenance, with minimal downtime, at scheduled times. One common piece of equipment, rail mounted automated gantry (RMG) automatic stacking cranes (ASC), can be outfitted with commodity acceleration sensors to better monitor the structural health of the crane and the rails on which the crane runs. The ratio of the rolling standard deviation of the crane vibration at speed versus at rest is a metric that can separate the cranes and rails into those that are in need of detailed inspection and repair, and those that are in good health. This method produced zero false negative, and an even split of true positives and false positives. The cost of a false positive is orders of magnitude less than the cost of a false negative, so this technique should help bend the cost curve down on RMG maintenance.

Practical Applications

Industrial RMG ASCs and the rails on which they run are difficult and expensive to survey, but abnormalities can cause damage to the cranes that will be much more expensive and difficult to fix. We are developing a method using low cost computing in the form of Raspberry Pis and commodity acceleration sensors to detect abnormalities without disturbing the use of the crane for a survey. One limitation of low cost sensors is that they produce a signal with a relatively high amount of noise. However, the signal is usable. We quantify the noise level as the rolling standard deviation (RSD) of the raw signal from the smoothed signal, which can be calculated for the crane at rest and at different speeds. The ratio of resting RSD to moving RSD is used to classify a rail as needing attention or as nominal. That ratio can be compared between two cranes that are operating on the same set of rails to determine whether the abnormalities are caused by the crane or by the rail. Utilizing this metric to prioritize track repairs such as tamping or grinding, or crane repairs such as wheel replacement or bogie alignments can significantly reduce the maintenance cost for the overall crane systems.