A cost and expected consequence approach to planning and managing railroad transportation of hazardous materials

Although hazardous materials (hazmat) account for around 140 million tons of all railroad freight traffic in the US, it has not received much attention from academic researchers. This is surprising especially when one considers the volume of hazmat moved by railroads in both North America and Europe. In this paper we develop a bi-objective optimization model, where cost is determined based on the characteristics of railroad industry and the determination of transport risk incorporates the dynamics of railroad accident. The optimization model and the solution framework is used to solve a realistic-size problem instance based in south-east US, which is then analyzed to gain managerial insights. In addition, a risk-cost frontier depicting non-dominated solutions is developed, followed by conclusion.     

A bilevel flow model for hazmat transportation network design

In this work we consider the following hazmat transportation network design problem. A given set of hazmat shipments has to be shipped over a road transportation network in order to transport a given amount of hazardous materials from specific origin points to specific destination points, and we assume there are regional and local government authorities that want to regulate the hazmat transportations by imposing restrictions on the amount of hazmat traffic over the network links. In particular, the regional authority aims to minimize the total transport risk induced over the entire region in which the transportation network is embedded, while local authorities want the risk over their local jurisdictions to be the lowest possible, forcing the regional authority to assure also risk equity. We provide a linear bilevel programming formulation for this hazmat transportation network design problem that takes into account both total risk minimization and risk equity. We transform the bilevel model into a single-level mixed integer linear program by replacing the second level (follower) problem by its KKT conditions and by linearizing the complementary constraints, and then we solve the MIP problem with a commercial optimization solver. The optimal solution may not be stable, and we provide an approach for testing its stability and for evaluating the range of its solution values when it is not stable. Moreover, since the bilevel model is difficult to be solved optimally and its optimal solution may not be stable, we provide a heuristic algorithm for the bilevel model able to always find a stable solution. The proposed bilevel model and heuristic algorithm are experimented on real scenarios of an Italian regional network.     

A toll-based bi-level programming approach to managing hazardous materials shipments over an intermodal transportation network

This research proposes a bi-level bi-objective model to regulate the usage of rail intermodal terminals for hazardous materials (hazmat) shipments, where government imposes tolls to deter carriers from using certain terminals. The complexity of the resulting mathematical program motivates the development of a hybrid speed-constrained multi-objective particle swarm optimization algorithm, which is then integrated with CPLEX, to solve the model. Through a real problem instance based on the intermodal service chain of Norfolk Southern in US, the toll-setting model is examined and further compared with a regular network design approach, in which certain terminals are closed to hazmat containers. The computational results show that the toll-setting policy is more practical and efficient, and the two models can be combined as a two-stage strategy in long-term hazmat transportation regulations. Additional managerial insights are derived for different stakeholders.     

Scheduling railway freight delivery appointments using a bid price approach

This paper proposes a method for establishing aggressive but achievable delivery appointment times for railroad shipments, taking into account individual customer needs and forecasted available train capacity. The concept of scheduling appointment times is directly patterned after current motor carrier industry practice, so that customers can plan for rail or truck deliveries in the same way.A shipment routing problem is decomposed into a deterministic “dynamic car scheduling” (DCS) process for shipments already accepted and a stochastic “train segment pricing” (TSP) process for forecasting future demands which have not yet called in and for which delivery appointments have yet to be scheduled. Both are formulated as multi-commodity network flow (MCNF) problems, where each shipment is treated as a separate commodity. Gain coefficients represent recapture probabilities that a specific customer will accept a carrier’s service offer.A comparison with a widely used revenue management formulation is given. A Lagrangian heuristic for obtaining a primal solution is also described. The problem is solved within a 1% gap using the subgradient algorithm.     

Probabilistic analysis of the release of liquefied natural gas (LNG) tenders due to freight-train derailments

Liquefied natural gas (LNG) has emerged as a possible alternative fuel for freight railroads in the United States, due to the availability of cheap domestic natural gas and continued pursuit of environmental and energy sustainability. A safety concern regarding the deployment of LNG-powered trains is the risk of breaching the LNG tender car (a special type of hazardous materials car that stores fuel for adjacent locomotives) in a train accident. When a train is derailed, an LNG tender car might be derailed or damaged, causing a release and possible fire. This paper describes the first study that focuses on modeling the probability of an LNG tender car release incident due to a freight train derailment on a mainline. The model accounts for a number of factors such as FRA track class, method of operation, annual traffic density level, train length, the point of derailment, accident speed, the position(s) of the LNG tender(s) in a train, and LNG tender car design. The model can be applied to any specified route or network with LNG-fueled trains. The implementation of the model can be undertaken by the railroad industry to develop proactive risk management solutions when using LNG as an alternative railroad fuel.     

Spatial decision support system for hazardous material truck routing

Shipping hazardous material (hazmat) places the public at risk. People who live or work near roads commonly traveled by hazmat trucks endure the greatest risk. Careful selection of roads used for a hazmat shipment can reduce the population at risk. On the other hand, a least time route will often consist of urban interstate, thus placing many people in harms way. Route selection is therefore the process of resolving the conflict between population at risk and efficiency considerations. To assist in resolving this conflict, a working spatial decision support system (SDSS) called Hazmat Path is developed. The proposed hazmat routing SDSS overcomes three significant challenges, namely handling a realistic network, offering sophisticated route generating heuristics and functioning on a desktop personal computer. The paper discusses creative approaches to data manipulation, data and solution visualization, user interfaces, and optimization heuristics implemented in Hazmat Path to meet these challenges.     

Comparing rail fuel efficiency with truck and waterway

This paper presents a railroad energy efficiency model used to estimate the fuel economies for classes of trains transporting various commodities. Comparable procedures are used to estimate truck and waterway fuel consumption. The results show that coal unit trains are 4.5–5.0 times more energy efficient than movements in the largest trucks allowed in the eastern and western regions of the US, unit grain train movements in the central US are 4.6 times more fuel efficient, soda ash unit train and non-unit train shipments are 4.9 and 3.2 times more efficient, and ethanol unit train and non-unit train movements are 4.8 and 3.0 times more efficient. In terms of barge traffic, coal unit train and non-unit train are 1.3 and 0.9 times as energy efficient in the eastern US, grain unit train and non-unit train movements are 1.7 and 1.0 times more efficient from Minneapolis to the Gulf of Mexico, and grain unit train and non-unit train movements are 1.0 and 0.7 times more fuel efficient from the Upper Ohio River to the Gulf of Mexico.     

Disruption risk management in railroad networks: An optimization-based methodology and a case study

We propose an optimization-based methodology for recovery from random disruptions in service legs and train services in a railroad network. A network optimization model is solved for each service leg to evaluate a number of what-if scenarios. The solutions of these optimization problems are then used in a predictive model to identify the critical disruption factors and accordingly design a suitable mitigation strategy. A mitigation strategy, such as adding flexible or redundant capacity in the network, is an action that is deliberately taken by management in order to hedge against the cost and impact of disruption if it occurs. It is important that managers consider the trade-offs between the cost of mitigation strategy and the expected cost of disruption. The proposed methodology is applied to a case study built using the realistic infrastructure of a railroad network in the mid-west United States. The resulting analysis underscores the importance of accepting a slight increase in pre-disruption transportation costs, which in turn will enhance network resiliency by building dis-similar paths for train services, and by installing alternative links around critical service legs.     

Optimizing the freight train connection service network of a large-scale rail system

This paper presents a formulation and solution for the train connection services (TCSs) problem in a large-scale rail network in order to determine the optimal freight train services, the frequency of services, and the distribution of classification workload among yards. TCS problem is modeled as a bi-level programming problem. The upper-level is intended to find an optimal train connection service, and the lower-level is used for assigning each shipment to a sequence of train services and determining the frequency of services.Our model solves the TCS problem of the China railway system, which is one of the largest railway systems in the world. The system consists of 5544 stations, and over 520,000 shipments using this system for a year period. A subnetwork is defined with 127 yards having some minimum level of reclassification resources and 14,440 demands obtained by aggregating 520,000 shipments to the subnetwork. We apply a simulated annealing algorithm to the data for optimal computation after pre-processing and get an excellent result. Comparing our optimal solution with the existing plan result, there are improvements of about 20.8% in the total cost.     

A bi-objective bi-level signal control policy for transport of hazardous materials in urban road networks

A bi-objective bi-level signal control optimization for hazardous material (hazmat) transport is considered to assess trade-offs between travel cost and environment impacts such as public risk exposure. A least maxi-sum risk model with explicit signal delay is presented to determine generalized travel cost for hazmat carriers. Since the bi-level signal control problem is generally a non-convex program, a bundle method using generalized gradients is proposed. A bounding strategy is developed to stabilize solutions of the bi-level program and reduce relative gaps between iterations. Numerical comparisons are made with other risk-averse models. The results indicate that the proposed bi-objective bi-level model becomes even amiable to signal control policy makers since provides flexible solutions whilst is acceptable to carriers since takes account of travel delay at signal-controlled junctions. Moreover, the trade-offs between public risk and generalized travel costs are empirically investigated among different risk models with a variety of weights. As a result, the proposed model consistently exhibits highly considerable advantage on mitigation of public risk whilst incurred less cost loss as compared to other alternatives.     

Enhancing intelligent agent collaboration for flow optimization of railroad traffic

Intelligent agents have successfully solved the train pathing problem on a small portion of railroad network [Tsen, 1995, Ph.D. Thesis, Carnegie Mellon University, USA]. As the railroad network grows, it is imperative that the agents collaborate to operate as efficiently as possible. In this paper, the authors demonstrate a collaboration protocol based on a conditional measure of agent effectiveness. Because agent effectiveness is not directly measurable, a suitable metric for agent effectiveness is introduced. Where typically agents run with uniform frequency, the collaboration protocol schedules the agents with a frequency proportional to their expected effectiveness. This protocol introduced a 10-fold improvement in the agent efficiency when tested with a simulation program on a portion of the Burlington Northern railroad.     

Regulating hazardous materials transportation by dual toll pricing

We investigate dual-toll setting as a policy tool to mitigate the risk of hazardous material (hazmat) shipment in road networks. We formulate the dual-toll problem as a bi-level program wherein the upper level aims at minimizing the risk, and the lower level explores the user equilibrium decision of the regular vehicles and hazmat carriers given the toll. When the upper level objective is to minimize the risk and all links are tollable, we decompose the formulation into first-stage and second-stage, and suggest a computational method to solve each stage. Our two-stage solution methodology guarantees nonnegative valid dual tolls regardless of the solution accuracy of the first-stage problem. We also consider a general dual-toll setting problem where the regulator rather wishes to minimize a combination of risk and the paid tolls and/or some links are untollable. To solve this truly bilevel problem, we provide heuristic algorithms that decompose the problem into subproblems each being solved by a line search. Case studies based on the Sioux Falls network illustrate the insights on the dual-toll policies.     

A multi-objective train scheduling model and solution

This paper develops a multi-objective optimization model for the passenger train-scheduling problem on a railroad network which includes single and multiple tracks, as well as multiple platforms with different train capacities. In this study, lowering the fuel consumption cost is the measure of satisfaction of the railway company and shortening the total passenger-time is being regarded as the passenger satisfaction criterion. The solution of the problem consists of two steps. First the Pareto frontier is determined using the -constraint method, and second, based on the obtained Pareto frontier detailed multi-objective optimization is performed using the distance-based method with three types of distances. Numerical examples are given to illustrate the model and solution methodology.     

Comparison of external costs of rail and truck freight transportation

In this article we estimate external costs for four representative types of freight trains. For each type of freight train, we estimate three general types of external costs and compare them with the private costs experienced by railroad companies. The general types of external costs include: accidents (fatalities, injuries, and property damage); emissions (air pollution and greenhouse gases); and noise. Resulting private and external costs are compared with those of freight trucking, estimated in an earlier article. Rail external costs are 0.24 cent to 0.25 cent (US) per ton-mile, well less than the 1.11 cent for freight trucking, but external costs for rail generally constitute a larger amount relative to private costs, 9.3–22.6%, than is the case for trucking, 13.2%.     

TransNIEMO: economic impact analysis using a model of consistent inter-regional economic and network equilibria

We describe a model that integrates a multiregional input–output (I–O) model of the USA (for 50 States and the District of Columbia) with the national highway network. Inter-state commodity shipments are placed on a congestible highway network. Simulations of major choke-point disruptions redirect traffic which increases the costs of some shipments. Increased costs show up in higher prices which help to determine a new I–O equilibrium. We find economic and network equilibria that are consistent. The simulations show only moderate economic impacts. We ascribe this to the resilience of the highway network. The model provides State-level detail on who bears the costs of the disruptions.     

Integrated operations planning and revenue management for rail freight transportation

In the rail industry, profit maximization relies heavily on the integration of logistics activities with an improved management of revenues. The operational policies chosen by the carrier have an important impact on the network yield and thus on global profitability. This paper bridges the gap between railroad operations planning and revenue management. We propose a new bilevel mathematical formulation which encompasses pricing decisions and network planning policies such as car blocking and routing as well as train make-up and scheduling. An exact solution approach based on a mixed integer formulation adapted to the problem structure is presented, and computational results are reported on randomly generated instances.     

Reinforcement learning approach for train rescheduling on a single-track railway

Optimal rail network infrastructure and rolling stock utilization can be achieved with use of different scheduling tools by extensive planning a long time before actual operations. The initial train timetable takes into account possible smaller disturbances, which can be compensated within the schedule. Bigger disruptions, such as accidents, rolling stock breakdown, prolonged passenger boarding, and changed speed limit cause delays that require train rescheduling. In this paper, we introduce a train rescheduling method based on reinforcement learning, and more specifically, Q-learning. We present here the Q-learning principles for train rescheduling, which consist of a learning agent and its actions, environment and its states, as well as rewards. The use of the proposed approach is first illustrated on a simple rescheduling problem comprising a single-lane track with three trains. The evaluation of the approach is performed on extensive set of experiments carried out on a real-world railway network in Slovenia. The empirical results show that Q-learning lead to rescheduling solutions that are at least equivalent and often superior to those of several basic rescheduling methods that do not rely on learning agents. The solutions are learned within reasonable computational time, a crucial factor for real-time applications.     

Intermodal and international freight network modeling

The authors describe the development and application of a single, integrated digital representation of a multimodal and transcontinental freight transportation network. The network was constructed to support the simulation of some five million origin to destination freight shipments reported as part of the 1997 United States Commodity Flow Survey. The paper focuses on the routing of the tens of thousands of intermodal freight movements reported in this survey. Routings involve different combinations of truck, rail and water transportation. Geographic information systems (GIS) technology was invaluable in the cost-effective construction and maintenance of this network and in the subsequent validation of mode sequences and route selections. However, computationally efficient routing of intermodal freight shipments was found to be most efficiently accomplished outside the GIS. Selection of appropriate intermodal routes required procedures for linking freight origins and destinations to the transportation network, procedures for modeling intermodal terminal transfers and inter-carrier interlining practices, and a procedure for generating multimodal impedance functions to reflect the relative costs of alternative, survey reported mode sequences.     

Inhalable railroad particles at ground level and subterranean stations - Physical and chemical properties and relation to train traffic

The paper compares PM10 concentrations in railroad environments with EU air quality standards and characterizes particle concentrations and particle properties in relation to train traffic. The results show that PM10 concentrations in ground-level railroad environments do not exceed the EU directive 24-h limit value, while on the platforms of the two subterranean stations PM10 concentrations were far above the outdoor limit value. Diurnal and weekly patterns in PM10 concentration could be identified, co-varying with train traffic intensity. The particle mass size distribution peaked at around 2-6 μm. The elemental composition of PM10 at the subterranean stations was dominated by Fe.     

Improving rail network velocity: A machine learning approach to predictive maintenance

Rail network velocity is defined as system-wide average speed of line-haul movement between terminals. To accommodate increased service demand and load on rail networks, increase in network velocity, without compromising safety, is required. Among many determinants of overall network velocity, a key driver is service interruption, including lowered operating speed due to track/train condition and delays caused by derailments. Railroads have put significant infrastructure and inspection programs in place to avoid service interruptions. One of the key measures is an extensive network of wayside mechanical condition detectors (temperature, strain, vision, infrared, weight, impact, etc.) that monitor the rolling-stock as it passes by. The detectors are designed to alert for conditions that either violate regulations set by governmental rail safety agencies or deteriorating rolling-stock conditions as determined by the railroad.Using huge volumes of historical detector data, in combination with failure data, maintenance action data, inspection schedule data, train type data and weather data, we are exploring several analytical approaches including, correlation analysis, causal analysis, time series analysis and machine learning techniques to automatically learn rules and build failure prediction models. These models will be applied against both historical and real-time data to predict conditions leading to failure in the future, thus avoiding service interruptions and increasing network velocity. Additionally, the analytics and models can also be used for detecting root cause of several failure modes and wear rate of components, which, while do not directly address network velocity, can be proactively used by maintenance organizations to optimize trade-offs related to maintenance schedule, costs and shop capacity. As part of our effort, we explore several avenues to machine learning techniques including distributed learning and hierarchical analytical approaches.