The application of a time-based consolidating policy in a freight train setting to achieve performance improvement for European infrastructure managers.

(1)

The application of a time-based

consolidating policy in a freight train

setting to achieve performance

improvement for European

infrastructure managers.

A simulation study

Geert-Jan Bakker

Groningen, 22th June 2015

Master Technology & Operations Management. Master International Business & Management. Supervisors:

dr. E. Ursavas dr. B.J.W. Pennink

Word count: 11642

Abstract

(2)

2

Preface.

Over the last couple months I had the interesting experience of carrying out a master thesis project for ProRail. Participation in this project has taught me a great deal about the railway industry and all the involved aspects. The experience was challenging and I feel grateful for the things that I could learn.

I would like to thank my lead supervisor dr. Ursavas and dr. Zhu for their input during the talks we had over the course of this project. The comments, feedback and recommendations received in our talks have contributed to the quality of my thesis.

Secondly, I would like to thank Rick Slot for making time to explain his complex model to me and provide me with many valuable insights. His unselfish support of my thesis has advanced my project for a great deal.

Lastly, I would like to thank my fellow students Jan-Eise Fokkema, Martin Pastoor and Ruben Schaafsma for the many valuable ideas and continuous support. Our talks were many times insightful and helped me stay focused during the course of this project.

June, 2015

(3)

3

1. Introduction.

Across many countries, train tracks have reached their limits of capacity utilization. This high utilization has the disadvantage that even a small delay can spread like an oil spill across the rail network and delay other trains for days. Events on the 22nd of January in the Netherlands have proven this. After a small failure in the computer system of the Dutch infrastructure manager, many train journeys have been delayed and even cancelled. It required a considerable amount of time to relocate the misplaced trains to their preferred positions (Spoorpro, 2015). Events like this happen all across Europe and delays become structural when utilization increases (Cule, Goethals, Tassenoy & Verboven, 2011; Milinković, Marković, Vesković, Ivić, & Pavlović, 2013; Marković, Milinković, Tikhonov, & Schonfeld, 2015). European infrastructure managers are searching for instruments to increase the railway capacity and cope with the ever increasing demand for cargo transport. Furthermore, upcoming railway infrastructure expansions will require the rerouting of freight transport. This puts extra pressure on existing infrastructure. For example, the construction of a third track on the Dutch Betuwelijn, a major European freight dedicated track, puts tremendous pressure on surrounding infrastructure. To cope with these problems, traditional methods have been exploited and researchers have concluded that new innovative approaches are essential (Middelkoop et al., 2012).

Capacity on the train tracks has to be allocated to passenger and cargo trains. In many instances cargo trains account for only a small percentage of the trains in the rail network (Meijer et al., 2012). However, cargo transport requests are hard to schedule due to their peculiar attributes. Not only are cargo trains slow and heavy, but due to high variability in cargo transportation demand many European infrastructure managers have difficulty allocating them to train paths. Moreover, more transport companies have emerged due to liberalization of the cargo market in Europe which has led to an even more complex situation for European rail infrastructure managers. Nonetheless, research on shipment consolidation in a variety of settings show promising findings which suggest that a more efficient utilization of train tracks might be possible (Geunes & Konur, 2012).

(6)

6

2012; Klincewicz, & Rosenwein, 1997; Kumar & Banerjee, 2012). Researchers have found performance enhancing outcomes of shipment consolidation in the setting of freight transport with trucks (Ulkü & Bookbinder, 2012). In a similar setting Qui & Huang (2013) present positive results in a study which investigates the consolidation of the freight in between industrial parks. Furthermore, Bernal, Blasco, Pellicer & González (2012) seek to optimize air cargo by implementing a consolidation strategy based on the destination of the freight, which reduces the costs of transportation significantly. Moreover, Wang & Cullinane (2014) investigate consolidation strategies for seaports and conclude that the importance of consolidation cannot longer be overlooked.

Infrastructure managers need to perform unraveling analyses of the infrastructure and freight demand relationship as a basis for their consolidation decision making (Wang & Cullinane, 2014). Freight trains with the same destinations can be consolidated which leads to economies of scale and therefore a significant cost reduction can be made (Klincewicz, & Rosenwein, 1997). Additionally, previous research suggests that infrastructure managers are also able to optimally use the freight train paths, increase robustness of the system, handle an increase in future demand and increase their punctuality (Milinković et al., 2013).

(7)

7

What would be the optimal consolidation time span for freight trains with regards to costs in the setting of the Brabant route and what is the effect of the optimal consolidation time span on travel times, utilization level and punctuality when also taking into account the effects of sensitivity and the creation of additional train paths?

Sub questions:

1. What is the optimal consolidation time span for freight trains with regard to costs? 2. How does this optimal consolidation time span influence the travel time of freight trains? 3. How does this optimal consolidation time span influence the utilization of freight train

paths?

4. How does the optimal consolidation time span influence punctuality of passenger trains? 5. How does an increase in demand influence the optimal consolidation time span?

6. How does the creation of more train paths influence the performance outcomes of the optimal consolidation time span?

According to Meijer et al., (2012) simulation modeling is an effective research method for analyzing complex sociotechnical infrastructures in general and to provide answers to specific questions in capacity management. Therefore, this paper uses an simulation approach. Historic datasets provided by an infrastructure manager are used to generate dynamic cargo transport demand, traffic intensity and the number of the available railway train paths in order to build a simulation model. The objective of this model would be to find an optimal consolidation time span that leads to optimal cost solution for freight trains. Therefore, an economic lot sizing model is developed to resolve the an optimal time-based consolidation policy. Furthermore, different performance measures are employed to ensure practical feasibility of the optimal solution.

(8)

8

The sixth chapter contains a discussion, implications and limitations section. The final chapter provides the conclusion.

2. Theory and background.

This chapter discusses several topics with regards to background and existing literature. Firstly, the problem description and some background information on freight trains is provided. Secondly, the literature with regards to the scheduling of freight trains is reviewed. The next section discusses the benefits of shipment consolidation policies employed in different settings. The fourth section discusses the specific consolidation policy employed in this paper. After that, the use of an economic lot sizing model in a freight train setting is discussed. The last section of this chapter presents the case of ProRail, the Dutch infrastructure manager.

2.1. Problem description.

To understand the problem at hand here, a thorough understanding of the characteristics of freight trains and the planning activities of infrastructure managers is necessary.

Whereas passenger trains are scheduled far in advance and use a fixed schedule, the approach for freight trains is different. Departure and arrival time windows are less strict and routes sometimes have several intended intermediate stops. Several characteristics make freight trains harder to schedule (Cordeau, Toth & Vigo 1998), these are:

 Cargo trains can operate without fixed schedules.

 Special train paths need to be assigned to cargo trains in between the schedules of the passenger trains.

 Cargo trains are heavier and slower, therefore have longer breaking distances and accelerate slower.

 Cargo trains could stop (even multiple times) at shunting yards before reaching their destination.

(9)

9

every week, these can be scheduled accordingly. Nonetheless, almost 60% of the freight trains requests follow economic realities and requests are made on very short notice (ProRail, 2014). These ‘ad hoc requests’ arrive days or even hours before departure and are served on a ‘first come, first served’ basis (Meijer et al., 2012). This can lead to last minute changes and planning conflicts. Ultimately, the current state of affairs within infrastructure managers leads to a suboptimal use of capacity.

After the stepwise liberalization of the market, cargo transport has become more and more fragmented. This fragmentation leads to varying lengths of freight trains and therefore to fluctuating in capacity use. Even the scheduled trains do not always use the full capacity of their train path. According to Boysen, Fliedner, Jaehn & Pesch (2013) only 53% of the freight trains reach their destination with fewer than 30 minutes delay and the average delivery speed of a freight train is estimation to be between 10 km/h and 20km/h. These statistics show that there is significant space for improvement with respect to the planning of freight trains.

2.2. The scheduling of freight trains.

(10)

10

For a general overview of theory on scheduling and combinatorial problems arising in railroad operating systems this paper refers to the surveys performed by Bussieck, Winter, and Zimmermann (1997); Caprara, Fischetti, Toth, Vigo & Guida (1997); Cordeau, et al. (1998); Boysen et al. (2013) and Piu & Speranza (2014). This paper focusses on theory that discusses combinatorial train logistics and the consolidation of shipments.

A considerable amount of train logistics research is performed concerning the availability of railcars at the moment of departure, for example Fuegenschuh, Homfeld, Huck, Martin & Yuan (2008) present such a model. Furthermore, Brucker, Hurink & Rolfes (2003) investigate issues relating to the coupling and uncoupling of trains on stations and nodes. Furthermore, multiple studies also look into problem of assigning railcars and locomotives to a set of scheduled trains (Lingaya, Cordeau, Desaulniers, Desrosiers & Soumis, 2002). These researchers use an approach that allows to determine the individual scheduling of trains by considering a broad set of constraints to optimally schedule the position of a car in the train. The coupling and decoupling of individual cars is allowed at different destinations along the route. Moreover, Fioole, Kroone, Maróti & Schrijver (2006) describe a rolling stock problem that investigates the allocation of rolling stock to a certain timetable. The majority of the relevant train logistics research focuses on passenger trains. Little research is available that covers the combination or consolidation of freight trains shipments. However, academic literature has proposed consolidation methods in different settings.

2.3. The consolidation of shipments.

A rapidly growing body of literature emphasizes the benefits of shipments consolidation policies. Shipment consolidation policies can lead to effective cost reductions, the effective allocation of costs, the reduction of environmental costs, improvement of total system performance, shorter and more consistent travel times, a higher utilization and produce economic shipment deliveries (Çetinkaya, Tekin & Lee, 2008; Ülkü, 2012; Elomri, Ghaffari, Jemai & Dallery, 2013; Marklund, 2011).

(11)

11

improvements can be made in the freight train setting. Firstly, many freight trains have the same destinations and therefore economies of scale can be achieved when consolidating. This would lead to lower transportation costs (Klincewicz, & Rosenwein, 1997). Secondly, the consolidation of shipments will lead to greater efficiency within the planning department, because currently a large amount of transportation requests are planned ‘ad hoc’. Thirdly, consolidation programs allow for faster and consistent transit times and this would increase customer-service standards (Masters, 1980). Moreover, with faster transit times capital is tied up for a shorter time, fast deliveries might generate fast payments and speed up industry cash flows (Masters, 1980; Ülkü, 2009). Lastly, consolidating freight trains will lead to lower energy consumption per transported ton and therefore lead to a more environmentally friendly way of transportation (Ülkü, 2012).

Consolidation approaches found in the academic discourse can be differentiated into three categories: (i) time-based policies (Marklund, 2011), (ii) quantity based policies (Higginson & Bookbinder, 1995) or (iii) a hybrid (Ülkü, 2009). A time-based policy releases a consolidated freight load every period (T). A quantity based policy releases freight when an economic dispatch quantity (Q) is available. This quantity behaves like a threshold and releases all waiting stock at once. A hybrid consolidation policy releases shipments based upon the attainment of the earliest threshold.

Ülkü (2009) shows that a quantity based policy is the most cost effective with unit sized demand for arrivals using private carriage. However, in a freight train setting a quantity based policy might lead to very long waiting times, because demand is empirically distributed. Long waiting times are undesirable because infrastructure managers have strict contracts with the cargo companies and cannot easily discard those (Meijer et al., 2008). Furthermore, the very limited amount of shunting space in railway yards does compel infrastructure managers to let waiting freight trains leave swiftly (Netverklaring ProRail, 2013).

(12)

12

by the weight and length of the train (Meijer et al., 2012; Cordeau et al., 1998). The application of a pure time policy could lead to demand accumulation to such a level that it exceeds the length and weight restrictions. Therefore, a hybrid time-based policy model will be implemented to account for the logistic realities infrastructure managers have to deal with. The time policy is in place because when a certain amount of time has elapsed the infrastructure manager has to abide by their contract and transport the wagons even though that would mean that this train path is not fully utilized (Meijer et al., 2008). Simultaneously, fixed thresholds are set in place with to emulate the weight and length restrictions. Therefore, a train might leave regardless of the time policy. In order to determine the optimal consolidation time span, academic literature has proposed an economic lot sizing model.

2.4 Economic lot sizing model.

In any shipment consolidation problem, the most important question is how much to consolidate before shipping or how long to wait before shipping consolidated goods (Nguyen, Dessouky & Toriello, 2014). The microeconomics concerned with the choice of economic shipment size have been researched for a long time. In fact, the economic lot sizing problem which is the most elemental shipment size model, can be found in inventory theory and is introduced by Harris (1913). Wagner and Whitin (1958) originally proposed a dynamic economic lot sizing problem. Many extensions with increasing complexity and many different applications are added to literature (Nguyen et al., 2014; Combes, 2010). A survey of all the economic lot sizing applications and their heuristic solution approaches from 1988 up to 2009 can be found in Robinson, Narayanan & Sahin (2009).

(13)

13

Mendiretta and Turnquist (1982) who developed a linear programming formulation. Furthermore, a real-life application of linear programming techniques to the daily distribution problem was presented by Markowicz and Turnquist (1990). The models are all solving problems concerning the appropriate empty wagon lot size at a certain train node. Even though economic lot sizing models are broadly used in literature to calculate the benefit of grouped shipments, the use of economic lot sizing models in freight train consolidation has been under-investigated in academic literature. The reason that the economic lot sizing model has been under-investigated in this setting is because the model has very specific data requirements. A complete set of data is required on the shipments by many variables. The weight, amount, volume, commodity nature, value and departure times (Combes, 2010).

In this paper, a novel design and solution approach is applied to derive performance insights on consolidation time span decisions of train track infrastructure managers. This approach is based on the premise that with the implementation of a consolidation time span performance will improve as compared to settings where no consolidation time span is implemented. At the same time, the cost of a train departure has to be balanced with the costs of holding a train which will result in an optimal consolidation time span.

(14)

14

paper includes the movements of the passenger trains. Passenger and freight trains in Europe share the same infrastructure and therefore a railway system has to be considered as one system to provide reasonable strategic prospects (Klug, Borndörfer, Fügenschuh, Schang, Schlechte & Schülldorf, 2013).

2.5. The case of ProRail.

For this paper, there has been cooperation with ProRail. ProRail is the infrastructure manager of the Dutch railways, it is a semi-governmental organization who is responsible for the Dutch railways. Some characteristics of the case have already been mentioned, but it is desired to provide a thorough characterization of the case selected in this paper.

The train corridor under investigation in this paper is the “Brabant route”. The Brabant route is befitting for this purpose because the corridor complies with several conditions. These conditions determine whether the case of the Brabant route is representative for the European railway network. Firstly, the corridor has a mix of freight and passenger trains. Secondly, the corridor has some major railway hubs with considerable amounts of train traffic. The presence of passenger train is important to measure interaction effects. Furthermore, the Brabant route is used to ship considerable amounts of freight, because it is one of the links connecting the Rotterdam harbor to major industrial areas in Germany. Furthermore, parallel to the Brabant route is the “Betuwelijn,” the main freight dedicated transport route to industrial areas in Germany. However, the Betuwelijn is expected to close down numerous times between 2015 and 2022 due to the construction of a third track. In that period, freight traffic would be redirected which would lead to increased freight train activities on the Brabant route.

(15)

15

(16)

16

In the next chapter, hypothesis are developed which include the determinants provided by ProRail and academic literature. The Brabant route and Betuwelijn are emulated in a simulation model to test those hypotheses. The Brabant route is highlighted in a map of the Netherlands in figure one.

(17)

17

3. Conceptualization and hypothesis.

When trains are consolidated they share the operational costs of the journey. The fixed costs of all involved stakeholders can be spread across a longer train. For example, personnel costs, setup costs and locomotive running costs. In different industrial settings this has been effectively illustrated by Klincewicz & Rosenwein (1997); Bernal et al (2012) and Islam et al( 2013) who describe consolidation strategies that lead to cost reductions in air cargo and freight truck settings because of the economies of scale. This leads to the following hypothesis:

H1. The optimal consolidation time span decreases the costs of freight trains.

Although, cost minimization is the most important performance indicator, several other important performance indicators are being discussed in literature and are employed by infrastructure managers. Çetinkaya, Tekin & Lee (2008) propose to look at a set of operational indicators, because when the optimal consolidation time span is evaluated with the use of several performance indicators the practical feasibility of the solution is enhanced. Therefore, this paper does not only focus on costs but also takes into account what the performance effects are with regards to time slot utilization, travel times and punctuality (Meijer et al, 2012; Bernal et al, 2012; and Islam, Olsen, & Daud Ahmed,2013). These performance indicators are also based on the determinants provided by ProRail. For infrastructure managers it is of great importance to allocate freight trains to train paths in such a manner that the least possible amount of train paths are used (Meijer et al., 2012). Theoretically, the optimal consolidation time span will provide the solution that leads to a significantly lower amount of used train paths when transporting the same amount of freight. This leads to the following hypothesis:

H2. The optimal consolidation time span will reduce train path utilization for freight trains.

(18)

18

before being transported to their destinations. It is expected that applying consolidation to railway cars based on their destination will lead to fewer trains in the system. Consolidation would decrease the likeliness that a train has to wait on a node and could ultimately lead to travel time reductions. Previous reasoning leads to the following hypothesis:

H3. The optimal consolidation time span reduces the travel time of freight trains.

Punctuality and reliability are vital for the quality of the provided service, especially in the case of passenger trains. This is because there is tremendous public pressure on passenger trains to arrive on time (Meijer et al, 2012). An increasing number of disruptive incidents in conjunction with an increasingly saturated railway capacity has led to a decline of reliability and punctuality of the railways (Goverde, 2005). Within ProRail punctuality of trains is of paramount importance. A punctuality level of 87% has been set by the Dutch Government with regards to the arrival times of passenger trains (Netverklarig ProRail, 2014). Punctuality relates to the amount of trains that arrive on time on their destinations. Trains are on time when they arrive within three minutes of the expected arrival time (Netverklaring ProRail, 2014). In this paper punctuality is expressed as the average lateness. Since lower utilization and lower travel times are associated with the optimal consolidation time span, this would mean the train track will be less saturated and therefore the arrival times of passenger trains will be more punctual. Therefore, based on previous reasoning the next hypothesis is:

H4. The optimal consolidation time span will increase the punctuality of the passenger trains.

It is expected that freight train demand will grow all across Europe (European Commission, 2014). The Brabant route is a European freight train corridor which transports a considerable amount of freight. Therefore, it is expected that the number of freight requests for the Brabant route will increase as well. As stated before, the optimal consolidation time span will have significant effect on the performance indicators mentioned in hypothesis one to four. Therefore it is expected that when saturation levels are even higher the effects of the first four hypothesis will be stronger in a setting with increased demand. The next hypothesis is:

(19)

19

ProRail is constantly searching for new approaches to deal with the high utilization on the Dutch railways. Recently, the Dutch government has approved the use of the ERTMS and reserved two and a halve billion Euro for this system. The implementation of this system reduces the headway between trains and enables up to 40% more capacity by creating more train paths without the need of investing in the construction of physical infrastructure (Kim et al., 2003). Theoretically, this increase in capacity would amplify the relationship between consolidation and performance, because more train paths becomes available for the same amount of trains. This leads to the following hypothesis:

H6. The creation of additional train paths will amplify the relationship between the optimal consolidation time span and the performance indicators.

The conceptual model in figure 3 shows the relationships between the dependent and the independent variables. The plusses and minuses indicate in which way the optimal consolidation level and the creation of additional train paths are expected to influence the relationship.

(20)

20

4. Methodology.

This chapter discusses the hypothesis testing approach that is employed in this paper into detail. In the first section, the application of a simulation study is discussed, followed by the discussion about the collected data. The third section describes the simulation design. The next section provides the heuristics approach to finding the optimal consolidation time span. The fifth section discusses the simulation scenarios and the last section explains how performance is measured.

4.1. Simulation.

The consolidation of shipments for the Brabant route is challenging, because many different variables need to be incorporated. The case contains a large amount of complexity and incompleteness of data. Simulation is often used to deal with cases which are too complex for formal mathematical analysis (Karlsson, 2009).

Therefore, this paper uses a simulation study to measure the performance outcomes for ProRail. Much input data was not available and needs to be derived using educated assumptions. However, the model must be representative of the real Brabant route to such an extent that the hypothesis can be tested. The simulation model for this paper is built with Tecnomatix Plant Simulation from Siemens. License availability and knowledge of this simulation tool have affected the choice for this specific simulation program.

4.2. Data collection.

To build a representative model which is capable of testing the provided hypothesis, several types of data needed collecting. These types of data are detailed historic train data, detailed physical infrastructure data of the Brabant route, decision and policy rules, cost related data and context data. A considerable amount of data on the ProRail case was available from previous research conducted by Slot (2014).

(21)

21

Network statement of ProRail, this is publicly available data. The network statement contains detailed information in the number of junctions and measurements of the tracks (Netverklaring ProRail, 2013).

The data on policy and decision rules was available in work instruction documentation and safety regulations from different departments of ProRail. The standard headway time in between train is derived from the study of Kim, Lee & Ryu (2003). Other policies like the priority rule of passenger trains and first come first serve policy are also based on academic literature (Meijer et al., 2012).

The last type of data is data concerning the costs of running a train. Many costs of ProRail are publicly available in financial statements (Annual Report ProRail, 2014) or had to be derived with educated assumptions. A complete oversight of all the included operational cost data and how these costs are derived can be found in Appendix A. All relevant operational and variable costs are included in the model. Additionally, context data was derived from literature and documents provided by the infrastructure manager to fully understand the activities simulated in the model.

4.3. Simulation design.

The simulation design is the result of the research scope. The aspects relevant for the testing of the hypothesis are incorporated in the model. This section provides a short overview of the simulation design.

(22)

22

Two types of input data are included in the model, passenger trains and cargo trains. Passenger trains are scheduled based on historic data. The historic data is perceived as a fixed schedule over the year 2013. Passenger trains are added to include interaction effects in between passenger and cargo trains. Cargo trains are dynamically generated with regard to weight, length and destination based on distributions derived from historic data. All the distribution employed in the simulation model can be found in Appendix B. All trains are allotted different average velocities due to their peculiarities. Freight requests vary considerably for peak and non-peak hours. This means that a different dynamically generated datasets are used to model day and night demand. Night is between 22:01 and 06:00, Day is between 06:01 and 22:00. Furthermore, to emulate the activities of the different trains on junctions, different process times are allocated to the different types of trains.

4.3.1. The consolidation heuristic.

The simulation model described in the last section is highly complex. Therefore, a heuristics approach is chosen to decide whether a freight request is qualified for consolidation. The application of a heuristic can solve an economic lot sizing problem in a reasonable amount of time (Nguyen, Dessouky & Toriello 2014; Beraldi et al., 2006). Based on the peculiarities of the ProRail case and academic literature (Cordeau et al., 1998; Meijer et al., 2012; and Slot, 2014) five decision rules are developed.

(23)

23

span T is treated as a variable threshold because in this paper a time-based consolidation policy is investigated. Therefore, the time threshold T is considered the decision variable in this paper.

Secondly, When a train reaches a certain length, the train departs. When a train is too long it will be slower and cause more delays on the route (Cordeau et al., 1998). Therefore, when a train has reached the maximum length, it will depart regardless the consolidation time span (ProRail Netverklaring, 2014). The maximum length of a train is treated as a fixed threshold.

Thirdly, when a train reaches a certain weight, the train departs. Reasons for this are similar to the length constraint. When a train has a certain weight, it will be slower and cause too much delays on route (Cordeau et al., 1998). Therefore, when a train has reached the maximum weight, it will depart regardless the consolidation time span (ProRail Netverklaring, 2014). The weight of a train is treated as a fixed threshold.

(24)

24

A flowchart of all the steps that need to be completed before a consolidated train has reached its destination can be found in figure three. The flow chart depicts all the steps that have been discussed so far and are incorporated in the simulation model.

4.4. Simulation settings.

This section discusses reliability and validity issues of the simulation model. Firstly, the warm-up time of the model is discussed, followed by necessary number of replications that is needed to ensure reliability. The last part of this section discusses the validity of the simulation model.

4.4.1 Warm-up time of the model.

Before the model can be used the warm-up time has to be determined. The warm-up time of the model is the time it takes for the model to reach a steady state (Robinson, 2004). Welch’s method is employed with regard to the average travel time in the model for eight simulated replications of 15 days. The average travel time is chosen, because it will give a clear indication of when the model has reached the full utilization level. With a one day interval and a moving

(25)

25

average of three days, the Welch method determined that the warm up time of the model is one day. Therefore, the measurement of the performance indicators starts from the second day. Outcomes of the Welch method can be found in Appendix C.

4.4.2 Confidence interval method.

The confidence interval method (Robinson, 2004) is used to determine the number of replications for a single scenario that would lead to reliable results. This is because the dynamically generated freight train data can cause fluctuating performance outcomes. Every replication stands for one year of generated data on the Brabant route and has approximately 25.000 iterations of freight trains which run through the system. The confidence interval method (Robinson, 2004: 155-157) outcomes determine with 95% certainty that the number of runs larger than three should provide outcomes in which estimated error from mean is 5%. See Appendix D for outcomes. This means, when running four years of dynamically generated data reliable outcomes can be assured.

4.4.3 Model validation.

This paper is an empirical research that seeks to solve problems in a real life setting. However, Robinson and Bamsow (2010) state the exact replication of reality is not possible. Therefore, the model has to be built adequately enough to test the hypothesis provided.

To ensure model accuracy, the model has been extensively tested and reviewed. Outcomes are compared to general performance outcomes of ProRail. Furthermore, a number of talks with experts and stakeholders have been performed to validate the processes in the model. The simulation model was improved during the development and testing phase (Slot, 2014).

(26)

26

model validity. This means the model is assumed fit for testing purposes. The next section discusses the simulation scenarios that are performed to test the hypothesis.

4.5. Simulation scenarios.

The scope of this research is to investigate performance related effects of the optimal consolidation time span. Additionally, in hypothesis five, the effects of increased demand and the creation of additional train paths are investigated. Table one shows all the scenarios that are being tested in this paper.

Table 1. Oversight of simulated scenarios.

In order to generate comparison data the ‘base’ scenarios are simulated, which means no consolidation is applied. The base scenario is run for normal demand (S1) and for increased demand (S9). Hypothesis one to four will be tested by comparing costs, train path utilization, travel times and punctuality of scenarios two to eight to the outcomes of the base scenario. The optimal consolidation time is represented by the scenario with the lowest costs. For scenarios two to eight the decision variable will take values from one to seven hours.

(27)

27

Hypothesis six is concerned with the effects of creating more train paths on the costs, train path utilization, travel times and punctuality. This hypothesis will be tested by comparing the performance outcomes of scenario 17-20 with the optimal consolidation time span and the two base scenarios. In order to find the optimal consolidation time span an economic lot sizing model needs to be employed. The next section will discuss how the economic lot sizing model is implemented within the model and how performance indicators are measured.

4.6. Measuring performance outcomes.

In order to test the hypothesis, the performance indicators costs, travel times, utilization and punctuality have to be measured. This section describes how the performance indicators are incorporated in the simulation model.

4.6.1 The economic lot sizing model.

The most optimal consolidation time span relies on the costs that are related to the consolidation activities. When determining the optimal lot size for an economic order quantity the setup costs and the holding costs have to be balanced with each other. The following cost parameters are included in the model:

Index number per train Q Amount of wagons

n Total amount of trains in a year

t Actual consolidation time span of the train. t < T Actual consolidation time per train

T Threshold consolidation time span of the train. A Distance traveled in kilometer

Distance traveled per train O Overhead line

W Weight fee Weight per train U Usage fee H Holding fee

E Energy fee per ton S Setup costs per departure J Actual travel time

(28)

28

The setup costs (S) for every train are assumed to be the same. However, when the wagons of two or more trains are consolidated the setup costs are only applied once. The setup costs are fixed for every train to account for setup activities of the locomotive and personnel costs.

= 1,..,n (1)

The holding costs (H) are the costs related to every hour a train is waiting on a node to be consolidated. A fixed fee is applied for every hour a train is waiting to account for extra shunting activities.

= 1,..,n (2)

Furthermore, operational and variable costs are included in the economic lot sizing equation. These costs are included because the scope of this research is to the operational performance of the European railways. When consolidation is applied, the operational and variable costs will also be affected.

The usage fee is a fee based on the weight of the train, whenever a train falls into a higher weight class, a higher weight fee per km is allocated to that train. However, longer trains results in a higher fee but a lower fee per ton/km (ProRail Annual Report, 2013). The different weight classes and the corresponding fees can be found in Appendix E.

= 1,..,n (3)

The overhead line fee is a fee to cover the overhead costs of ProRail. This number is based on general estimations made by ProRail executives. A fixed fee per traveled kilometer is allocated to every train departing.

= 1,..,n (4)

A weight dependent energy fee is allocated to every traveled kilometer in order to account for the energy used during a train journey.

(29)

29

The last cost involved in consolidation activities is the personnel fee. The personnel fee account for the hourly wage of the train operator. An average wage is retrieved from the collective labour agreement of the Dutch railways (2013).

= 1,..,n (6)

The aim of the economic lot sizing model is to find the lowest cost per wagon (Q) for different consolidation time spans. All the separate cost equations above are incorporated in a total cost function per wagon for time span T. This leads to the following average cost function per wagon:

∑

∑ = 1,..,n (7)

This total cost function should be perceived as a performance indicator instead of a complete detailed cost calculation for all the activities of ProRail. However, the costs that are included do cover the most relevant costs. A complete oversight and explanation of the incorporated costs involved can be found in Appendix A.

4.6.2. Travel time, utilization and punctuality.

(30)

30

5. Results.

This chapter discusses the results. The results in this chapter will be presented in a pragmatic manner. Furthermore, chapter six will provide a more elaborate discussion of the results. The results will be presented for every hypothesis.

(31)

31

5.1. The results for hypothesis 1.

The first hypothesis is tested by comparing the total costs per wagon for all eight different simulated time spans with each other. The results of hypothesis five in which an increase in demand is emulated will be presented together with the outcomes of hypothesis one to four. The first hypothesis tested is:

H1. The optimal consolidation time span decreases the costs of freight trains.

Figure 4. The average cost per freight wagon expressed for all consolidation time spans.

Figure four shows that for normal demand the optimal consolidation time span is three hours (T = 3) In the setting of increased demand the optimal consolidation time span is four hours (T = 4). Outcomes are not distributed normally with α = 0.05 (Appendix F). Outcomes of the Mann Whitney U test indicate that the cost of optimal consolidation span (T = 3) are significantly different (P < 0.05) from the base scenario (T = 0) In the setting of increased demand comparing T = 4 to T = 0 returns P < ,001, which means that significant cost reductions can be made in both normal and increased demand. Based on these findings the null hypothesis is rejected and the H1 hypothesis is accepted. For an oversight of the results see Appendix G.

27,2 27,4 27,6 27,8 28 28,2 28,4 28,6 28,8 29 0 1 2 3 4 5 6 7 Co st p e r wagon

Consolidation time span (T)

Average cost per freight wagon

(32)

32

5.2. The results for hypothesis 2.

The second hypothesis is tested by comparing the train path utilization of the optimal consolidation time spans to the base scenarios. The second hypothesis tested is:

H2. The optimal consolidation time span will reduce train path utilization of the system.

Figure 5. The train path utilization by freight trains expressed for all consolidation time spans.

Figure five shows that for normal demand an improvement of 8.4% and for increased demand an improvement of 8.39% can be made. Outcomes are normally distributed with α = 0.05 (Appendix F). Outcomes of the independent t-test indicate that number of train paths utilized for optimal consolidation span (T = 3) is significantly different from a non-consolidation (T = 0) setting. P = ,047 which is below α = 0.05. The increased demand scenarios return P = ,034. This means that a significant train path utilization reduction can be made in the setting of normal and increased demand. Based on these findings the null hypothesis is rejected and the H1 hypothesis is accepted in both normal and increased demand. For an oversight of the results see Appendix G. 22500 23000 23500 24000 24500 25000 25500 26000 26500 27000 0 1 2 3 4 5 6 7 N u m b e r o f tr ai n p ath s u til ize d

(33)

33

5.3. The results for hypothesis 3.

The third hypothesis is tested by comparing the average travel times per wagon of the optimal and base scenarios with each other. The third hypothesis tested is:

H3. The optimal consolidation time span reduces the travel times of freight trains.

Figure 6. The average travel times for freight trains expressed for all consolidation time spans.

Figure six shows that for normal demand an improvement of 6,98% can be made and for increased demand an improvement of 8.49% can be made. Outcomes are not normally distributed with α = 0.05 (Appendix F). Outcomes of the Mann Whitney U test indicate that average travel times for optimal consolidation span (T = 3) are not significantly different from a non-consolidation (T = 0) setting with P = ,148. The increased demand scenario returns P = ,471 which means that no significant travel time reductions can be achieved by consolidation. Based on these findings the null hypothesis cannot be rejected. For an oversight of the results see Appendix G. 3750 3800 3850 3900 3950 4000 4050 4100 4150 4200 4250 0 1 2 3 4 5 6 7 Tr av e ltim e in se co n d s

Consolidatino time span (T)

(34)

34

5.4. The results for hypothesis 4.

The fourth hypothesis is tested by comparing the punctuality of passenger trains of the optimal consolidation scenario and the base scenario. The fourth hypothesis tested is:

H4. The optimal consolidation time span will increase the punctuality of the passenger trains.

Figure seven shows that for normal demand an improvement of 0,48% can be made and for increased demand an improvement of 0.60% can be made. Outcomes are not normally distributed with α = 0.05 (Appendix F). Outcomes of the Mann Whitney U test indicate that punctuality percentage for optimal consolidation span (T = 3) are significantly different from a non-consolidation setting (T = 0). The P-value is ,039 which is below α = 0.05. The increased demand scenario returns P = ,244 which means that no significant punctuality reductions can be achieved by consolidation. Based on these findings the null hypothesis is rejected in the normal demand setting, however it cannot be rejected in the increased demand setting. For an oversight of the results see Appendix G.

86,60% 86,70% 86,80% 86,90% 87,00% 87,10% 87,20% 87,30% 87,40% 87,50% 87,60% 87,70% 0 2 4 6 8 Pu n ctu al ity p e rc e n tage

Punctuality percentage for passenger trains

Increased demand Normal demand

(35)

35

5.5. The results for hypothesis 6.

The last hypothesis is tested by comparing performance outcomes of a setting with consolidation to a setting with consolidation and the implementation of ERTMS. A total of four comparisons is made. Performance will be compared in a setting with normal demand and no consolidation (T = 0), a setting with normal demand and optimal consolidation (T = 3), a setting with increased demand and no consolidation (T = 0) and a setting with increased demand and optimal consolidation (T = 4). The last hypothesis is:

H6. The creation of additional train paths by implementing an ERTMS will amplify the relationship between the optimal consolidation time span and the costs, utilization, travel times and punctuality.

Figure eight shows the outcomes of the first performance indicator tested for hypothesis six. This is the average cost per wagon. Since outcomes are not normally distributed, the Mann Whitney U test is employed to compare the output data. The P-value for normal demand (T = 0) the P = ,214, for normal demand and optimal consolidation (T = 3) this is P = ,672, for increased demand (T = 0) the P-value is ,876 and for increased demand with optimal consolidation (T = 4)

26,5 27,5 28,5 29,5

Normal Demand T = 0 Normal Demand T = 3 Increased Demand T = 0 Increased Demand T = 4

Co st p e r wagon Scenarios

Average cost per freight wagon

Condolidation + ERTMS Consolidation

(36)

36

the P = ,069 (Appendix H). In none of the scenario a significant cost reduction is found. This means the null hypothesis concerning costs cannot be rejected.

Figure nine shows train path utilization outcomes for hypothesis six. Since outcomes are normally distributed, the independent t-test is employed to compare the output data. For normal demand (T = 0) the P-value is ,002, for normal demand and optimal consolidation (T = 3) this is P = ,110, for increased demand (T = 0) the P = ,003 and for increased demand with optimal consolidation (T = 4) the P = ,526 (Appendix H). This means that only when consolidation is not applied, the application of an ERTMS leads to a significant decrease in used train paths. Therefore, the null hypothesis cannot be rejected.

21000 22000 23000 24000 25000 26000 27000 28000

Normal Demand T = 0 Normal Demand T = 3 Increased Demand T = 0 Increased Demand T = 4

Nu m b er of t rai n spath s u sed Scenarios

Train path utilization

Consolidation + ERTMS Consolidation

(37)

37

Figure ten shows the travel times output data for this hypothesis. Since outcomes are not normally distributed, the Mann Whitney U test is employed to compare the output data. For normal demand (T = 0) P < ,001, for normal demand optimal consolidation (T = 3) this is P < ,001, for increased demand (T = 0) the P < ,001 and for increased demand with optimal consolidation (T = 4) the P < ,001 (Appendix H). A significant improvement is found in all four comparisons. Therefore the null hypothesis can be rejected with regard to travel times and hypothesis six can be accepted.

3200 3300 3400 3500 3600 3700 3800 3900 4000 4100 4200 4300

Normal Demand T = 0 Normal Demand T = 3 Increased Demand T = 0 Increased Demand T = 4 Tr av e l tim e in se co n d s Scenarios

Travel times

Consolidation + ERTMS Consolidation

(38)

38

Figure 11 shows the punctuality output data for passenger trains. Since outcomes are not normally distributed, the Mann Whitney U test is employed compare the output data. For normal demand (T = 0) P < ,001, for normal demand and optimal consolidation (T = 3) P < ,001, for increased demand (T = 0) P < ,001 and for increased demand with optimal consolidation (T = 4) P < ,001 (Appendix H). Significant improvement is found in all four comparisons. Therefore the null hypothesis can be rejected with regard to punctuality of passenger trains and hypothesis six can be accepted.

6. Discussion.

This chapter will start with a results discussion section. The second section discusses the implications, limitations and distinguishes further research opportunities.

6.1. Wagon costs for freight trains.

Results indicate that the optimal consolidation time span by normal demand is three hours. An average reduction in wagon cost of 5,29% is achieved in the normal demand setting and a reduction of 6,4% in an increased demand setting. However, after a demand increase the average wagon price is the lowest at four hours. Results indicate that when demand increases, wagon costs decrease disproportionally. This can be explained by the larger amount of trains that is available for consolidation. However, no significant cost reductions can be made when

Figure 11. Four comparisons based on passenger train punctuality.

84,00% 85,00% 86,00% 87,00% 88,00% 89,00% 90,00% 91,00% 92,00%

Normal Demand T = 0 Normal Demand T = 3Increased Demand T = 0 Increased Demand T = 4 Pu n ctu al ity r ate Scenarios

Punctuality passenger trains

(39)

39

implementing ERTMS. An explanation could be that the improvements made in travel time are being overturned by a larger amount of setup costs. This is caused by an increased availability of train paths. Therefore, trains have earlier possibilities to leave when the consolidation time span has elapsed.

6.1.1. Utilization of train paths.

The optimal consolidation level in a normal demand setting decreases the train path utilization with 8,4%. In a setting with increased demand the number of train paths utilized decreases with 8,39%. This result can be seen as a very conservative estimation because this model does exclude wagons with dangerous contents and only allows the consolidation on Rotterdam and Venlo. It is plausible that additional train path reductions can be made when incorporating these factors in the model.

The output data provides a valuable insight with regards to the application of the ERTMS. The creation of additional train paths does have a significant impact in all four comparisons. However, the amount of utilized train paths is significantly higher instead of lower. A closer look at the data offers the explanation that with an increase in capacity of up to 40% (Kim, Lee & Ryu, 2003) the absolute number of utilized train paths might increase but the percentage of utilized train paths might decrease. This means that better performance can be achieved even though the number of utilized time slots in increased. This is caused by an increase in total capacity. This reasoning is confirmed by the finding on travel times of freight trains and punctuality of passenger trains. These two performance indicators increase significantly when implementing ERTMS.

(40)

40

strong indication of the improvements that can be made with the implementation of a time-based consolidation policy.

6.1.2. Travel times of freight trains.

No significant improvement is found for travel times of cargo trains when implementing a time-based consolidation policy. Even though time-time-based consolidation can lead to a lower amount of shipments (Marklund, 2011), cargo trains seem not to benefit significantly from the additional capacity. This can be explained by the prioritizing of passenger trains. Often freight trains wait several times on shunting yards before reaching their destination, therefore they stay long in the system (Boysen, et al., 2013). The unimpressive result with regards to travel times for freight trains question whether time based consolidation is a valuable approach when travel time reductions for freight trains are sought.

Nonetheless, travel time length decreases greatly when the amount of train paths is increased by implementing ERTMS and the consolidation approach simultaneously. In normal demand with no consolidation a reduction in travel times of 8.18% can be achieved, in the optimal consolidation time span a reduction of 7,62% can be made, in increased demand with no consolidation 9,94% reduction can be made and in the optimal consolidation time span (T = 4) with increased demand a reduction of 6,88% can be achieved. These are valuable findings because the current system already has a very high utilization. The implementation of ERTMS can be considered as an inventive capacity approaches are necessary to improve performance of the saturated railways, of which the necessity is discussed by Middelkoop et al. (2012). An explanation of why the reduction in travel times is lower in the increased demand scenario is related to the degree of saturation. This finding is in line with findings of Cordeau et al. (1998) who state that longer travel times are normal in high utilization settings.

6.1.3 Punctuality of passenger trains.

(41)

41

all simulated scenarios. This can be explained by the priority rule for passenger trains that is employed by infrastructure managers.

Currently, ProRail has a 87% target on punctuality (Netverklaring ProRail, 2014). ProRail considers a train that arrives within three minutes of expected arrival time to be on time. The punctuality results show that significant improvements with regards to passenger train punctuality can be made by applying the consolidation time span in normal demand. However, with ever growing demand and upcoming construction activities the implementation of ERTMS is recommendable.

6.2. Implications, future research and limitations.

(42)

42

This reasoning leads to more research directions which are valuable to consider. Infrastructure managers should investigate the possibility of a request infrastructure that makes the demand requests and the capacity more transparent for all the parties involved. The use of such a system could be made more desirable when dynamic pricing is integrated in the system. A practice that is successful in use within the airline industry (McAfee & Te Velde, 2006). This would enable ProRail to consolidate freight with the same destination in the optimal consolidation time span that is provided by the findings of this paper. This is especially of interest for infrastructure managers within Europe because it is not legally possible for them to force cargo companies to consolidate.

When implementing a dynamic pricing policy, prices should incentivize consolidation. The current weight fees as used by ProRail (2014) do not have an incentivizing effect on the implementation of a time-based consolidation policy (Appendix E). When a train has a weight that is close to the upper threshold of a weight fee, it is not beneficial to consolidate. A small increase in weight would results in reaching the lower bound of a higher threshold and result in a higher cost per wagon. The implementation of a flexible weight fee that incentivizes cargo companies to consolidate their freight could be interesting. Meijer et al (2012) have already done exploratory investigations into this topic and has explored the implementations of price mechanisms and transparency in a railway setting. However, further investigations are necessary into the implementation of price mechanism and transparency together with a time based consolidation policy. Especially because Ülkü and Bookbinder (2012) have proven the benefits of periodic pricing schemes and consolidation in the setting of a private trucking company.

(43)

43

stays consolidated over longer distances. Further implications and additions of this paper to the field of International Business and management are discussed in Appendix I.

6.2.1 Limitations.

Several limitations need mentioning. Firstly, only one way train tracks are simulated. In reality, tracks can be used both ways when demand in one direction transcends the demand for a different direction. Another limitation is that this model does assume a perfect information supply. In reality this information is only partially available to ProRail and arrives at many different moments in time. Furthermore, punctuality of freight trains is not included in the model because the model is not fit to simulate a planning department that selects the best train paths for departing trains. Therefore, the punctuality of freight trains cannot be calculated, because every different freight train path has a different expected time of arrival. This is because oftentimes freight trains have to wait at shunting yards before continuing their journey. Moreover, the relevant cost with regard to consolidation is included including an increase in shunting movements. However, no consideration is given to the specific costs of renting rolling stock. When more wagons are consolidated, savings can be made by accommodating the handling of rolling stock within the simulation model.

7. Conclusion.

All elements of this paper build up to answer the research question that is introduced in chapter one:

RQ: What would be the optimal consolidation time span for freight trains with regards to costs in the setting of the Brabant route and what is the effect of the optimal consolidation time span on travel times, utilization level and punctuality when also taking into account the effects of sensitivity and the creation of additional train paths?

(44)

44

the length of the consolidation search span that leads to the lowest costs. This lowest point is determined by the implementation of an economic lot sizing model.

Results suggest that the optimal consolidation level leads to significantly lower costs for freight trains when a time-based consolidation policy is applied. Furthermore, significant improvements in punctuality of passenger trains can be made when consolidation is applied in a setting with normal demand. Moreover, the utilization of train paths is significantly improved. This means that when consolidation is applied a significantly lower amount of train paths is utilized in all tested settings. The increase in capacity can facilitate future growth and help infrastructure managers to deal with disruptions. However, outcomes with regard to travel times of cargo train paths are unimpressive. No significant reductions could be found. The benefits of reducing the amount of utilized train paths seem to be disproportionally allotted to passenger trains, since they have priority in the system. Moreover, costs reductions and utilization are further improved in the setting of increased demand. However, travel times for freight trains and punctuality of passenger trains are deteriorated because the system has a higher saturation rate.

The creation of additional train paths does provide no significant cost reductions. However, significant improvements can be made with regards to utilization and travel times of cargo trains. Furthermore, the punctuality levels of passenger trains can be improved. It is commendable to combine the application of an optimal consolidation time span and ERTMS. This combination will enable European infrastructure managers to improve performance with regards to costs, utilization, travel times and punctuality. Significant improved outcomes in the setting of increased demand show the robustness of the application of the optimal consolidation time span and ERTMS.

(45)

45

policy. However, due to their unique capacity as market makers, European infrastructure managers are able to place incentivizing instruments in place to motivate cargo companies into consolidating their freight. Several options such as information transparency and flexible price strategies are put forward.

The findings of this paper provide strong evidence for the application of consolidation and ERTMS. Additional research is necessary to extend the findings. More research is needed into the costs of rolling stock, the willingness of cargo companies to cooperate and the exact movements of consolidating trains on the shunting yards.

(46)

46

References.

Avi-Itzhak, B., Benn, B. A., & Powell, B. A. (1967). Car pool systems in railroad transportation: mathematical models. Management Science, 13(9), 694-711.

Beraldi, P., Ghiani, G., Grieco, A., & Guerriero, E. (2006). Fix and relax heuristic for a stochastic lot-sizing problem. Computational optimization and applications, 33(2-3), 303-318.

Bernal, M. P., Blasco, S. V., Pellicer, E. L., & González, R. S. (2012). Optimization of the air cargo supply chain. Journal of Airline and Airport Management, 2(2), 101-123.

Boysen, N., Fliedner, M., Jaehn, F., & Pesch, E. (2013). A survey on container processing in railway yards. Transportation Science, 47(3), 312-329.

Brucker, P., Hurink, J., & Rolfes, T. (2003). Routing of railway carriages. Journal of global

optimization, 27(2-3), 313-332.

Bussieck, M. R., Winter, T., & Zimmermann, U. T. (1997). Discrete optimization in public rail transport. Mathematical programming, 79(1-3), 415-444.

Cao, M., & Zhang, Q. (2010). Supply chain collaborative advantage: a firm’s perspective. International Journal of Production Economics, 128(1), 358-367.

Caprara, A., Fischetti, M., Toth, P., Vigo, D., & Guida, P. L. (1997). Algorithms for railway crew management. Mathematical programming, 79(1-3), 125-141.

Çetinkaya, S. (2004) Coordination of inventory and shipment consolidation decisions: a review of premises, models, and justification, in Applications of Supply Chain Management and E-Commerce Research in Industry, Geunes, J., Akcali, E., Pardalos, P.M., Romeijn, H.E., Shen, Z.-J. (eds.), Springer, New York, pp. 1–48.

Çetinkaya, S., Tekin, E., & Lee, C. Y. (2008). A stochastic model for joint inventory and outbound shipment decisions. IEE Transactions, 40(3), 324-340.

(47)

47

Cordeau, J. F., Toth, P., & Vigo, D. (1998). A survey of optimization models for train routing and scheduling. Transportation science, 32(4), 380-404

Cule, B., Goethals, B., Tassenoy, S., & Verboven, S. (2011). Mining train delays. In Advances in Intelligent Data Analysis X (pp. 113-124). Springer Berlin Heidelberg.

De Jong, G., Vierth, I., Tavasszy, L., & Ben-Akiva, M. (2013). Recent developments in national and

international freight transport models within Europe. Transportation, 40(2), 347-371.

Elomri, A., Ghaffari, A., Jemai, Z., & Dallery, Y. (2013). Cost allocation in a full truckload shipment consolidation game. International Journal of Inventory Research, 2(1), 127-143.

European commission (2014) REPORT FROM THE COMMISSION TO THE COUNCIL AND THE EUROPEAN PARLIAMENT Fourth report on monitoring development of the rail market. Retrieved March 4, 2015, from:

http://ec.europa.eu/transport/modes/rail/market/market_monitoring_ en.html

Fioole, P. J., Kroon, L., Maróti, G., & Schrijver, A. (2006). A rolling stock circulation model for combining and splitting of passenger trains. European Journal of Operational Research, 174(2), 1281-1297.

Fuegenschuh, A., Homfeld, H., Huck, A., Martin, A., & Yuan, Z. (2008). Scheduling locomotives and car transfers in freight transport. Transportation Science, 42(4), 478-491.

Geunes, J., & Konur, D. (2012). Impacts of Efficient Transportation Capacity Utilization via Multi-Product Consolidation on Transportation Network Usage and Congestion (No. CMS Project No. 2010-018).

Goverde, R. M. (2005). Punctuality of railway operations and timetable stability analysis. TU Delft, Delft University of Technology.

Graudina, V., & Grundspenkis, J. (2005, June). Technologies and multi-agent system architectures for transportation and logistics support: An overview. In International

The application of a time-based consolidating policy in a freight train setting to achieve performance improvement for European infrastructure managers.