Journal of Infrastructure Systems

Forum papers are thought-provoking opinion pieces or essays founded in fact, sometimes containing speculation, on a civil engineering topic of general interest and relevance to the readership of the journal. The views expressed in this Forum article do not necessarily reflect the views of ASCE or the Editorial Board of the journal.

This study reviewed the status quo of research on machine learning (ML), including deep learning, in the rail industry. This study conducted a scientometric analysis and critical review of 640 papers selected from 12,675 web-crawled papers. The extensive and complex networks of topics, researchers, and countries were analyzed using the Louvain method, a co-occurrence keyword analysis, a degree centrality analysis, and other network analysis methods. The results indicate that the majority of studies of ML in the rail industry focused on maintenance activities and traffic management, and mainly targeted rolling stock, rails, and passengers. Overhead contact systems, including catenaries, are a high-demand objective for ML-based maintenance. Although analyses of tunnels and stations remain rare, passenger flow prediction, station air quality estimation, shield tunneling performance improvement, and ground settlement are areas of high importance. Geographically, China, the US, and the United Kingdom lead ML studies in the rail industry, and the level of collaboration is higher among European countries than among countries on other continents. Future challenges include ensuring the security and stability of ML, along with considering novel mindsets, the black-box effect, improvements in ML techniques, and resource overload when introducing ML technologies.

Civil infrastructure systems, such as harbors and airports, support transportation between international communities. Inspection of facilities in a large area is labor-intensive and time-consuming work, and the result can sometimes be inaccurate due to humans’ limitations and inexperience. In this study, an autonomous aerial platform with a functional multipoint patrol module is developed to acquire images of inspected targets. It integrates a high-definition digital surface model (DSM) and a route-searching algorithm to optimize flying route planning. The collected unmanned aerial vehicle (UAV) images are then subjected to a matching technique that automatically detects exposure positions and corrects image distortions. Finally, target facilities are extracted from multitemporal images using object detection techniques so that the status of the inspected targets can be tracked and evaluated. Case studies in real-world settings have been conducted. As this proposed approach refines the efficiency and reliability of the facility inspection task, significant improvements are expected to deal with wide area monitoring and flexibility management.

Significant research has been conducted on the practice of bridge asset management (BAM). Although bridge site inspections yield observations on current conditions, their usefulness in a bridge asset management strategy requires a sound historical and statistical basis. Therefore, the present practice of bridge site inspection can only support a reactive mode in bridge asset life cycle management and is raising a critical need for a risk analysis methodology that will effectively support decision making in a forecasting mode. This research addressed that challenge by employing a series of relatively simple, well-established statistical techniques as a precursor to a forecasting methodology. A knowledge-based risk assessment method (RAM) is recommended for optimizing BAM of the New York City Department of Transportation (NYCDOT) bridge database. The limitations of the current practice of bridge condition rating are reactive in nature. Reactive approaches do not provide analytical tools for forecasting the aging effect on the time-dependent bridge infrastructure deterioration for the prioritization of BAM. The development of statistical data analysis can be done by using the aging effect on the corrosion rate as a health monitoring parameter of the girder. The results of this research provide bridge asset managers with a statistical data-driven analysis approach to the time-dependent aging effect as a decision support tool for preemptive prioritization and a prespecified vulnerability level in BAM planning. The infrastructure base selected for this study was 92 steel girder bridges with concrete decks (single-span/multispan) managed by NYCDOT, which were clustered into eight major age groups. There were multiple primary member degradation indicators, e.g., corrosion, section loss, cracks, and others. This research analyzed the corrosion rate of primary members. It illustrated the application of statistical data analysis for the development and application of RAM for forecasting the effect of aging on the corrosion rate as a structural degradation indicator for the preemptive planning of a BAM strategy.

Assessing the reliability and resiliency of transportation infrastructure is critical to improving the safety and sustainability of roadways. Such information, particularly when available on a network level helps transportation agencies identify vulnerable locations on their road network and make more informed decisions when managing infrastructure and when introducing design improvements. To that end, many design experts have promoted integrating the risk of failure of roadway design elements to satisfy road user demand (i.e., reliability measures) into the highway design process. In fact, previous work has established a link between design reliability and safety. Although exploring such relationships provides extremely valuable insights on the impacts of meeting or deviating from design requirements, it does not provide much information on where the most (and least) reliable road segments exist on a network. To overcome these critical shortcomings, this paper proposes the adoption of hot spot analysis and spatial interpolation to assess reliability of compliance with sight distance requirements on an aggregate network level. Light Detection and Ranging (LiDAR) data was first used to quantify available sight distance on 220 curved segments, and design reliability to sight distance requirements was then assessed at each location. Hot spot analysis and spatial interpolation using the inverse distance weighting method were then employed to identify regions of the Alberta road network where low or high design reliability existed or could be expected. The analysis revealed that the highest-risk regions on the Alberta highway network existed in areas of mountainous in the western region of the province as well as areas of rolling terrain in south Calgary. In contrast, it was found that curved road segments in the prairie region had a significantly higher design reliability to sight distance. The clustering of unreliable roadway segments in mountainous regions and regions of rolling terrain indicates that more effort is required to improve design reliability in those regions. Such important inferences are only possible when conducting an aggregate reliability assessment such as the one proposed in this paper.

Power infrastructure networks are susceptible to performance disruptions induced by natural or anthropogenic hazard events. For example, extreme weather events or cyberattacks can disrupt the functionality of multiple network components concurrently or sequentially, resulting in a chain of cascading failures throughout the network. Mitigating the impacts of such system-level cascading failures (systemic risks) requires analyzing the entire network considering the physics of its dynamic power flow. This study focuses on the draw-down phase of power infrastructure network resilience—assessing the power grid vulnerability and robustness, through simulating cascading failure propagations using a dynamic cascading failure physics-based model. The study develops and demonstrates the utility of a link vulnerability index to construct power transmission line vulnerability maps, as well as a node importance index for power (sub)station ranking according to the resulting cascading failure size. Overall, understanding the criticality of different network components provides stakeholders with the insights essential for building resilience and subsequently managing it within the context of power grids and supports policymakers and regulators in making informed decisions pertaining to the tolerable degree of systemic risk constrained by available resources.

Life-cycle structural assessment of existing bridges under aging and deterioration processes is of paramount importance for authorities managing road networks. In fact, in most developed countries many bridges and infrastructure facilities are approaching 50 years of lifetime or are even older, and a significant percentage of them are prone to be rated as structurally deficient. In this context, a quick and reliable estimation of the bridge condition is needed for the prioritization of maintenance and repair interventions and the optimal allocation of resources over large bridge stocks. In the last decades, computerized bridge management systems (BMSs) emerged worldwide to support decision makers in the definition of operational management policies. BMSs often incorporate soft computing tools, such as artificial neural networks (ANNs), which can provide reliable bridge assessments over time under uncertainty on the basis of limited data stored in bridge databases. This paper proposes an approach to the life-cycle assessment of deteriorating reinforced concrete (RC) bridges based on ANNs. Two-layer ANNs are formulated and trained for this purpose. The proposed approach is preliminarily applied to a time-variant structural capacity assessment of a RC bridge deck cross-section under chloride-induced corrosion to compare the results obtained with both full and limited input datasets. Finally, the methodology is applied to the prediction over time of the condition state of a group of existing RC bridges based on the information stored in the US National Bridge Inventory.

The ever-increasing investment gap for deteriorating infrastructure has necessitated the development of more effective asset management (AM) strategies. However, information asymmetry among AM stakeholder silos has been recognized as a key challenge in implementing effective AM strategies. The connectivity within the AM system introduces systemic risks (possibility of dependence-induced cascade failure) to the entire AM system operation when information asymmetry occurs. This study describes a toolbox to enable asset management stakeholders to assess such systemic risks through a network analytics approach. The network, representing the AM system, is examined through its centrality measures to identify the most critical subject areas within the AM system. These subject areas are subsequently paired with assets’ key performance indicators (KPIs). Within the developed toolbox, descriptive analytics provide transferrable KPI insights between stakeholders to reduce key asset information asymmetry. In parallel, predictive analytics forecast KPIs, ensuring stakeholder awareness of future asset performance to allow for appropriate preparation. Subsequently, prescriptive analytics employ heuristic-based optimization for optimal configuration of the AM network. The five tools presented are as follows: (1) dependence identification and network modeling; (2) network centrality analysis; (3) descriptive analytics of critical subject area paired KPI; (4) KPI-based predictive analytics; and (5) prescriptive analytics for optimal network configuration. The utility of the developed toolbox is demonstrated for Tools 1–3 using a real AM system network and KPIs associated with power transmission infrastructure outages. Based on the analyses, managerial insights are drawn to illustrate the usefulness of the developed approach in improving information asymmetry within the AM system, subsequently mitigating dependence-induced systemic risks.

We consider a budget-constrained rail network electrification problem with associated changes in costs of energy usage (via path gradient and curvature), operations, and long-term maintenance. In particular, we consider that freight flows on such a network form a user equilibrium. Interactions between electric and diesel trains on the same corridor are represented with nonseparable link performance functions that nevertheless have a symmetric Jacobian. This bilevel formulation is solved for the North American railroad network using a genetic algorithm (GA), incorporating domain-specific insights to reduce the number of solutions that must be considered. We analyze solution characteristics and decision-making implications. Results show that broad connectivity would be beneficial for the most impact. Increasing demand shifts electrified corridors toward the more populous east and gulf coasts, while increased operational costs results in the electrification of routes through mountainous terrains.

We consider the problem of setting tolls and capacities in transportation networks serving single origin–destination markets. The motivation is to understand the effects of network structure and of decentralized decision-making on the performance of such systems, as well as to identify opportunities for government intervention. Performance, in this case, refers to social welfare and profits, with the latter determining private firms’ incentives to participate in road operation. We first extend the Pareto efficiency analysis appearing in the literature for the single road case. Specifically, we consider a network operated by a single firm and show that aggregate Pareto-efficient capacity levels provide the same service quality as in the single road case and that any convex combination of social welfare–maximizing and profit-maximizing route tolls can be Pareto efficient, suggesting that such outcomes can arise naturally from negotiations between a government and a private operator. In the second part of the paper, we consider a market where, following bilateral negotiations with a government, self-interested, private operators set tolls and capacities independently. Again, we show that the optimal capacity for each link is set to achieve the same service level as that in the single road case. Further, we adapt the concept and show that decentralized Pareto-efficient tolling strategies are bounded. Characterization of Pareto-efficient tolling and capacity strategies for the two benchmark cases reveals when (government) interventions can lead to increased social welfare, operator profit, or both. As an example, we analyze simple stage networks and, among other results, show that, even under bilateral negotiations and decentralized decision-making, sufficient competition can lead to aggregate Pareto efficiency.

With an increasing emphasis on sustainable planning, highway decision makers need to shift their focus from developing a cost-effective highway program to developing a more comprehensive sustainable highway program. It includes incorporating the three components of sustainability, namely economics, the environment, and social equity. Studies in the literature have primarily focused on economic and environmental objectives. However, there are limited studies incorporating social equity in highway maintenance and rehabilitation (M&R) programming. Additionally, a review of the existing literature indicated that although studies have focused on single or dual objectives of sustainability, the existing literature failed to address all the three components of sustainability simultaneously. Accordingly, this study aimed to address this issue by proposing an M&R budget allocation model for developing a sustainable highway asset management plan that concurrently focuses on all the three objectives of sustainability. The proposed model assists decision makers with a range of potential decision alternatives that address multiple sustainability objectives. The proposed model was applied to a case-study example on a road network in Texas. The findings from the case study indicated that decision makers can implement sustainable M&R programs that can efficiently achieve social equity objectives without specifically compromising the environmental and economic objectives.

The controlled source electromagnetic (CSEM) geophysical method is used for a number of purposes, including exploration and characterization of oil and geothermal deposits. Here, based on idealized modeling scenarios, we explore an adaptation of the method to remotely inspect shallow buried pipelines and determine its suitability for monitoring corrosion because of the sensitivity of CSEM responses to electrical conductivity contrasts within the subsurface. We show that the electric field response from an in-line transmitter-receiver configuration provides more diagnostic information than the broadside configuration while both are likely to detect the pipeline response in the presence of ambient noise. The presence of a conductive pipeline is considerably easier to detect in sandy soils than in clays and there is a reduction in amplitude and response shape when the pipeline is resistive. In-line responses can diagnose the level of corrosion, although it is difficult to distinguish between a highly corroded pipeline and one with a break. In cases where a pipeline break is present, some data processing may be required to precisely locate the break.

A vast amount of investment in infrastructure is required to respond to short- and long-term social, technological, and environmental needs and developments. Because infrastructure systems are highly interconnected, much can be gained by considering these interdependencies when planning future investments. However, identifying opportunities that arise from these infrastructure interdependencies has mainly been neglected. So far, the risk perspective has dominated interdependency studies. Alternatively, this paper proposes an agent-based modeling approach supporting infrastructure decision makers (1) to reveal the effects of planned sector-specific investments on the performance of interdependent infrastructures; and (2) to identify situations around which cross-sectoral coordination and collaboration can be shaped. The selected modeling approach treats infrastructure as sociotechnical systems, incorporating sector-driven operational decisions and infrastructure demand changes. They included operational decisions that respond to temporary or longer-term demand-capacity mismatch to account for the flexibility in exploiting available systemwide capacities. This modeling approach results in a more realistic estimation of infrastructure performance and beneficial co-investment opportunities. A regional transportation infrastructure system in the Netherlands is used as a case to demonstrate the approach.

One area of railway infrastructure management that can benefit greatly from digitalization is the determination of optimal intervention programs, i.e., when, where, and which type of interventions are to be executed. The potential benefit is considerable because of the large variety of assets required for the infrastructure to function as intended, the interconnectedness of the assets, the extensive number of different types of possible interventions, and the wide range of service measures to consider when deciding between different intervention programs—all of which are difficult, if not impossible, to consider qualitatively. In this paper, a network flow optimization model is presented that determines the optimal intervention program considering different types of assets, interventions and service measures to execute the interventions, the dependencies between interventions, and the relation between interventions and service in the short and long term. The model is developed and used to determine the intervention program that maximizes the net benefit for a 17-km railway line over a 12-year planning period, divided into three four-year blocks. The example demonstrates that the model can be used to determine optimal intervention programs on real-world railway networks, taking into consideration the intervention costs and relevant measures of service, the interrelationships between the different assets, and multiple time periods. It also demonstrates that the model is a powerful management tool for leveraging the digitalization of railway infrastructure.

This study introduces a framework for quantifying the time-evolving functionality and, consequently, the resilience of rail-truck intermodal freight transportation networks subjected to regional disruptions from events like natural hazards. The proposed framework leverages publicly available datasets and probabilistic models for estimating the damage and functionality of the components of the intermodal network. New restoration models linking physical damage to functionality have been proposed in this study, improving upon previous literature in this area. Furthermore, the framework for quantifying the resilience of intermodal networks is posed to consider not only the functionality of constituent highway and railway networks and their components but also that of intermodal terminals in its formulation. The framework enables the estimation of network throughput and functionality metrics at various instances during the recovery timeline and can be extended to assess the broader economic impacts of freight disruption on the community.

This paper presents an availability analysis of a multistate infrastructure in the case of regular inspections and renewals, given the assumption that an infrastructure is renewed after exceeding a certain threshold by the infrastructure risk function. In the latter case, the procedure of a condition-based maintenance strategy is proposed. A wind farm infrastructure located in the North Sea is considered a case study. The identification and prediction of its reliability are performed, distinguishing three reliability states and assuming that its components have Weibull reliability functions. This paper takes into account the multistate process of infrastructure degradation and includes infrastructure inspections and repairs, and applies the multistate approach to a maintenance and availability analysis of a wind farm infrastructure using its reliability indicators. In its reliability and availability analysis, three periods of “early,” “random,” and “wear-out” failures are distinguished. The results of a comparative analysis of wind farm infrastructure assuming various reliability structures that correspond to its output power differ significantly. Thus, determining at what minimum number of operating wind turbines the infrastructure ensures sufficient efficiency is crucial for analyzing its reliability and availability.

Water pipeline infrastructures are critical for the delivery of lifeline services; however, these aging systems are experiencing increasing breakage rates. To assist utilities in identifying the most vulnerable assets, sustained research efforts have been made in developing machine learning models to accurately predict future failures. The performance of these methods heavily depends on the quantity of reliable data, while most utilities only have limited records of historical pipe breaks. To overcome the limitation of data availability, this article presents a case study exploring the performance of machine learning methods for predicting future failures when system information from multiple utilities is combined. Six utilities are considered, for which predictive models are trained and evaluated in several scenarios, (1) using data from only a single reference system, (2) all systems combined, and (3) a bootstrapped sample of multiple systems to match the pipe material distribution of the reference system. Empirical results suggest that variance controlling algorithms, such as random forests, are less sensitive to the availability of data, and that introducing information from third-party sources only leads to marginal changes in performance. Overall, the amount of break records from the reference system itself has the largest influence on accuracy, suggesting that utilities must keep reliable historical break data to maximize the power of predictive modeling for their asset management programs.

Public-private partnerships (PPPs) are adopted worldwide to provide public infrastructure yet may present problems due to the multiple uncertainties embedded in this class of projects. Therefore, in order to attract private investment, some form of mitigation of risks may be required. These risk mitigation mechanisms may take many forms and place contingent liabilities and financial burdens on the government budget that must be adequately priced. This article analyzes the case of the Salvador–Itaparica Bridge system concession project under the real options approach and compares the results with publicly available government reports on the project. Our results indicate that the expected cost of the State of Bahia government’s contingent liabilities is undervalued by a factor of five and that the guarantees provided were excessive as they resulted in the total elimination of demand risk to the concessionaire. This article contributes to a better understanding of the effects of risk-mitigating mechanisms for private investment in infrastructure projects and the importance of adequately calibrating and pricing these contingent government liabilities.