Bayesian Networks‐Based Probabilistic Safety Analysis for Railway Lines

Castillo, Enrique; Grande, Zacarías; Calviño, Aida

doi:10.1111/mice.12195

Cited by 41 publications

(26 citation statements)

References 21 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…BNs have been used to model single infrastructure networks such as inland waterway ports (Hosseini and Barker 2016), railway lines (Castillo and Grande 2016), highways (Grande et al 2017), and power (Tien and Der Kiureghian 2017) and water networks (Leu and Bui 2016). These studies did not consider interdependencies between different networks.…”

Section: Bayesian Networkmentioning

confidence: 99%

Framework for Probabilistic Vulnerability Analysis of Interdependent Infrastructure Systems

Applegate

Tien

2019

J. Comput. Civ. Eng.

View full text Add to dashboard Cite

Critical infrastructure systems are deteriorating and experiencing increased cascading failures. In this paper, we propose a new probabilistic framework for modeling interdependent infrastructure networks. As part of this framework, we present our modeling approach and the accompanying sets of algorithms that enable the computationally efficient probabilistic modeling of large infrastructure systems considering interdependencies between networks. The proposed method creates a representative Bayesian network (BN) of the system, which provides exact inferences over the network compared to simulation-based approximations. With the traditional computational limitations of BNs to tens of parent component nodes per child system node, our generalized framework enables computationally efficient BN modeling of systems of, for example, hundreds of component nodes, including component-level performance and interdependent connections across networks. Our approach requires only simple inputs based on the basic component characteristics of location, type, connectivity, and initial failure probabilities. We modeled three types of interdependencies across the system-service provision, geographic, and access for repair. We combined a minimum link set (MLS) formulation with the idea of supercomponents in order to reduce network complexity without any approximating assumptions. We present algorithms to efficiently identify MLSs and supercomponents, as well as to identify and remove any cyclic dependencies that arise across the network. Once the BN was constructed, we were able to perform exact inference analyses over a range of component state and hazard event scenarios in order to identify vulnerabilities across the network. The main novelty of the paper is to enable the probabilistic assessment of large, complex, interdependent infrastructure systems. We were able to consider performance from the component level and model hundreds of component nodes in a computationally efficient manner without approximating assumptions. We accounted for interdependencies between systems with exact inference results. The model can be used to investigate the potential for cascading failures and to prioritize critical components for repair, replacement, or reinforcement. We applied the proposed methodology and algorithms to the water distribution network in Atlanta, Georgia, and its dependencies with the power system. We validated the model using the results from a recent interdependent outage event.

show abstract

Section: Bayesian Networkmentioning

confidence: 99%

Framework for Probabilistic Vulnerability Analysis of Interdependent Infrastructure Systems

Applegate

Tien

2019

J. Comput. Civ. Eng.

View full text Add to dashboard Cite

show abstract

“…Following [71,72], we would like to consider how models developed for alert predictions can be interpreted. This section provides an example of using DT models for predicting patterns of interest.…”

Section: Interpretation Of Alert Patternsmentioning

confidence: 99%

Bayesian learning of models for estimating uncertainty in alert systems: Application to air traffic conflict avoidance

Schetinin

Jakaite

Krzanowski

2018

ICA

View full text Add to dashboard Cite

Alert systems detect critical events which can happen in the short term. Uncertainties in data and in the models used for detection cause alert errors. In the case of air traffic control systems such as Short-Term Conflict Alert (STCA), uncertainty increases errors in alerts of separation loss. Statistical methods that are based on analytical assumptions can provide biased estimates of uncertainties. More accurate analysis can be achieved by using Bayesian Model Averaging, which provides estimates of the posterior probability distribution of a prediction. We propose a new approach to estimate the prediction uncertainty, which is based on observations that the uncertainty can be quantified by variance of predicted outcomes. In our approach, predictions for which variances of posterior probabilities are above a given threshold are assigned to be uncertain. To verify our approach we calculate a probability of alert based on the extrapolation of closest point of approach. Using Heathrow airport flight data we found that alerts are often generated under different conditions, variations in which lead to alert detection errors. Achieving 82.1% accuracy of modelling the STCA system, which is a necessary condition for evaluating the uncertainty in prediction, we found that the proposed method is capable of reducing the uncertain component. Comparison with a bootstrap aggregation method has demonstrated a significant reduction of uncertainty in predictions. Realistic estimates of uncertainties will open up new approaches to improving the performance of alert systems.

show abstract

“…In railroads, safety is one of the most important topics and has attracted tremendous attention recently due to some reported accidents (Castillo et al., , b; Wang et al., ). Among all the causes of train accidents in the United States, track defects are one of the leading reasons.…”

Section: Introductionmentioning

confidence: 99%

A Customized Hybrid Approach to Infrastructure Maintenance Scheduling in Railroad Networks under Variable Productivities

Xie

Lei

Ouyang

2018

Computer aided Civil Eng

View full text Add to dashboard Cite

Railroads are maintained routinely by using various types of rail‐bound machines so as to achieve the longest possible rail life and reduce the safety risks associated with unanticipated rail failures. The rail maintenance routing and scheduling problem (RMRSP), which involves routing of multiple maintenance vehicles and scheduling of hundreds of maintenance jobs over a large‐scale network, is usually subject to various types of complex constraints and extremely difficult to solve. This article proposes a vehicle routing problem with time windows (VRPTW) formulation for RMRSP and develops a customized stepwise algorithm to solve the problem. A series of numerical experiments are conducted to demonstrate that the proposed algorithm works very effectively, significantly outperforming the state‐of‐the‐art commercial solver. The results of two real‐world instances from a Class I railroad company show that the proposed model and solution algorithm enable the expensive maintenance vehicles to achieve a higher level of utilization, that is, spending more time on working and less time on deadhead traveling.

show abstract

Bayesian Networks‐Based Probabilistic Safety Analysis for Railway Lines

Abstract: A Bayesian network model is developed, in which all the items or elements encountered when travelling a railway line, such as terrain

Cited by 41 publications

References 21 publications

Framework for Probabilistic Vulnerability Analysis of Interdependent Infrastructure Systems

Framework for Probabilistic Vulnerability Analysis of Interdependent Infrastructure Systems

Bayesian learning of models for estimating uncertainty in alert systems: Application to air traffic conflict avoidance

A Customized Hybrid Approach to Infrastructure Maintenance Scheduling in Railroad Networks under Variable Productivities

Contact Info

Product

Resources

About