Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl
This report consists of ## pages and # appendices. It may only be reproduced literally and as a whole. For commercial purposes only with written authorization of Delft University of Technology. Requests for consult are only taken into consideration under the condition that the applicant denies all legal rights on liabilities concerning the contents of the advice.
Specialization: Production Engineering and Logistics Report number: 2014.TEL.7844
Title: Use of simulation for the analysis of railway systems
Author: Tim Jongerius
Title (in Dutch) Gebruik van simulatie voor de analyse van railsystemen
Assignment: literature
Confidential: no
Initiator (university): H.P.M. Veeke
Supervisor: H.P.M. Veeke
Mekelweg 2 2628 CD Delft the Netherlands Phone +31 (0)15-2782889 Fax +31 (0)15-2781397 www.mtt.tudelft.nl
Student: Tim Jongerius Assignment type: Literature
Supervisor (TUD): H.P.M. Veeke Creditpoints (EC): 10
Specialization: PEL
Report number: 2014.TEL.7844
Confidential: No
Subject: Use of simulation for the analysis of railway systems
Railway systems cope with complex planning problems that require analytical tools to support the planning and evaluation of railway processes, in order to be able to see the consequences of the choices that are made. The simulation of processes can aid in these analyses and is already being used by some railway companies. The assignment is to analyse a railway system by means of the Delft Systems Approach and to describe the current field of application of simulation models in such a system. Furthermore it is interesting to analyse what improvements could be made to the existing simulation models.
The report should comply with the guidelines of the section. Details can be found on the website. The professor,
3
Summary
Many passenger rail transportation systems exist, each with their own characteristics and size. These systems have requirements that originate from the passengers using the system and from the principle authorities. In order to determine and influence the performance of such a system means need to be found to evaluate the processes that are present. A railway system is a complex system, with many interrelated processes. Simulation models are
increasingly being used by railway companies to support and analyse these processes, since they model the behaviour of the system in high detail. To determine how, the main question that will be answered in this research is: “how does simulation support the design and evaluation of railway systems?” In order to answer this question a railway system will be analysed to determine where simulation can be used. Simulation methods are described, followed by a description of existing simulation software. Potential improvements and developments in research are finally discussed to create an outlook on railway simulation. A railway system can be described according to the Delft Systems Approach (Veeke et al, 2008) by three aspects; the flow of demand, passengers and resources (infrastructure, vehicles and personnel). These are handled by the processes of ‘planning services’, ‘transport’ and ‘assigning & maintain’ respectively. Interaction of these processes result in complex planning problems that require the assistance of software tools for design and evaluation. Several commercial simulation models exist that cover a large part of these processes. They all simulate infrastructure, vehicles, signalling systems and timetables to determine conflicts and vehicle delays. Often optimisation tools are integrated to further optimise this timetable. However, among other things, the simulation of passengers and drivers is not included in these existing models, which limits their ability to calculate passenger delays and their use for rescheduling purposes. Research describes algorithms and methods for simulating passengers and show some of the possibilities of including them in simulation models.
Although current simulation models already support planning and evaluation processes on an adequate level, current simulation models could still be improved to add more functionality and detail. For example by adding functionality that supports passenger and driver
behaviour. The accuracy of simulation models can also further be improved by supporting the collection of operational data.
4
Contents
- Summary………3
- Contents……… 4
- 1. Introduction……… 5
- 2. Analysis of a railway system……… 7
o 2.1 Requirements of a railway system……… 8
Passenger requirements……… 8
Ruling authority requirements……… 9
Internal requirements……… 10
o 2.2 Processes at a railway company……… 11
Planning services……… 12
Transporting passengers……… 16
Assigning resources……… 19
- 3. Railway simulation……… 23
o 3.1 Advantages and disadvantages of simulation……… 23
Advantages……… 23
Disadvantages……… 24
o 3.2 Methods of railway simulation……… 25
Macro-, meso- and microscopic simulation……… 25
Deterministic vs. stochastic……… 27
Synchronous vs. asynchronous……… 27
Continuous vs. discrete……… 27
- 4. Existing applications……… 28
Timetable construction and testing……… 29
Operations analysis……… 31
Rescheduling……… 32
- 5. Possible improvements and current developments in railway simulation…… 34
Rescheduling……… 34 Timetable creation……… 35 Operational process……… 35 General……… 36 - 6. Conclusions……… 37 - 7. References……… 38
5
1.
Introduction
Rail transportation has been around for many years all over the world, becoming an important method of transporting goods and people. These rail systems have changed and developed into a number of forms we find today, ranging from lightweight trams used in urban areas for transportation of people to high speed heavy rail applications for
transportation of both freight and people between cities. The difference between both is mainly in the weight and speed of the vehicles used and the lay-out of the tracks (Vuchic, 2004). Trams are lighter vehicles than trains and use track sections that lay in the street; they share this space with road traffic and pedestrians, which makes them more susceptible to delays. The method for controlling movement authority through security (signalling) systems on these tracks is therefore also somewhat different from heavy rail infrastructure. Light rail systems are modern versions of trams that are able to brake and accelerate faster than trams. They often use the same street infrastructure as trams, but also run at higher speeds on dedicated tracks that are separated from road traffic and pedestrians. Light rail systems are often used to connect more distant areas within a city or neighbouring cities. Transportation capacities of these different types of vehicles vary from a higher capacity for heavy rail vehicles to smaller capacity for regional vehicles.
A rail system can be used for transporting freight and people. Both have comparable requirements like a desired speed and an efficient operation, but there are also some differences in terms of external demands. For example, the transportation of passengers requires some degree of comfort and ease of use, but also punctuality and safety are required to be of a higher level than for freight. As a result also the indicators from which the performance of the system is measured will slightly differ for the transportation of freight and passengers. For the sake of comparability the focus of this research will only be on the transportation of passengers by rail. There are however some performance indicators and applications that are also useable for freight transportation. The transportation of passengers by rail can be visualised by the following system, which is based on the method of the Delft Systems Approach (Veeke, Ottjes & Lodewijks, 2008):
6 Figure 1: Passenger transportation by rail
Rail operating companies focus on running one or more forms of rail transportation,
sometimes having to combine different types of vehicles on shared infrastructure, which can result in having to make complex decisions. This asks for analytical support during decision-making. Testing design choices and evaluating the performance of a rail system is more and more being done through the use of simulation. This technique allows to reproduce real processes in an abstract way, which enables the modelling and analysis of the behaviour of such a system.
The goal of this report is to get a global view of the simulation systems that are being used for planning, evaluation and decision-making processes in railway systems. Therefore the main question to be answered is:
‘How does simulation support the design and evaluation of railway systems?’
First the processes in a passenger railway system will be analysed and relevant requirements and performance indicators will be determined with the use of the Delft Systems Approach in Chapter 2. In the next chapter the advantages and disadvantages of using railway simulation are mentioned. Then a general description is given of the different methods that are used in rail simulation, in order to get some understanding of the way these systems work. In chapter 4 the ability of existing rail simulation software to test design changes and evaluate a passenger rail system based on the performance indicators is discussed. Furthermore the aim is to give insight into new rail simulation methods and software that are currently being researched in literature. The way in which they will contribute to a better or broader support for passenger railway processes will be pointed out in Chapter 5.
7
2.
Analysis of a railway system
To get insight into how a railway system performs it is not sufficient to look at a railway company as an isolated system. A national railway station for example can be the starting location for the journey of one passenger after he reaches it by foot, bike or car, but another passenger might use a tram or light rail vehicle to travel from his starting location to the railway station in order to continue his journey from there on. This depends on how the passenger plans his journey. Often these separate railway systems are connected at (national) railway stations. The company operating the trams or light rail vehicles can therefore have internal or external requirements regarding, for example, connectivity at this location. For this reason the scope of the analysis of a railway system will be set on a national level, in which also the interaction of (bigger and smaller) railway companies becomes visible. This way the total journey by regional and national rail of a passenger is taken into account. The system previously shown in figure 1 can then be viewed in more detail, as can be seen in figure 2.
Figure 2: Regional and national transportation of passengers
Rail guided transportation of passengers is split up in four functions to represent a complete passenger journey by rail. A passenger here starts in a city and plans his journey, either by rail or by other means. He might use regional rail transportation (by tram or light rail) to travel towards his (intermediate) destination. If he has to travel to another city he will transfer to transportation by means of heavy rail at a national railway station. Once he arrives at the preferred station he might use regional rail transportation again to reach his final local destination.
8
2.1
Requirements of a railway system
As can be seen in figure 2 each of these transportation subsystems have their own
requirements and performance. These requirements come from both passengers and (local) stakeholders, as well as from the rail company that operates the system.
Passenger requirements
According to (van Hagen, Peek & Kieft, 2000) passengers rate the quality of public
transportation on the following indicators; safety (both of the infrastructure and of stations (social safety)), reliability, speed, ease-of-use, comfort and experience (visual layout,
neatness). These can be seen as base requirements that hold for all transportation functions in figure 2.
To determine how the rail system of figure 2 performs, it needs to be evaluated on the realisation of the abovementioned requirements. The passenger requirements should therefore be translated into measurable performance indicators that can be interpreted by the company and the environment. The performance indicators for passengers that can then be set at the current level of detail are:
Safety: This can be measured in the number of accidents and through expert ratings, but also the customer perception of safety is very important.
Customer satisfaction: To measure the perception of safety, ease-of-use, comfort and experience. This is often done through the use of a customer survey, in which they give a score on each of these indicators. It is also valuable to monitor the total passenger count. This gives a general indication of the customer satisfaction, although it is not sufficient to solely base conclusions on passenger counts alone.
Punctuality: This is used to quantify the reliability of the service and can be indicated by the percentage of trips that were on time at a stop or station. This is influenced by the quality of the timetable and by disturbances during operation.
Speed: The speed of the journey seen through the eyes of the passenger, this is not only the speed of the vehicle while driving, but the speed from the beginning to the end of the
journey. This is influenced by a number of things like the geographical location of the stops, the speed of the vehicle, the amount of delays, the regularity of trips (frequency of trips on a line) and the connectivity with other lines or modes of transportation at stops and stations.
9 Here a good connectivity means that it takes a relatively short time to walk to the next mode of transportation and to wait for the next vehicle.
Some of these performance indicators are quantitative whereas others are qualitative, which is why they require the use of different methods of evaluation.
Ruling authority requirements
The system in figure 2 shows a global picture of passenger transportation by rail as seen from a passenger perspective. However, as said before, a railway company does not only receive requirements from its passengers, but also from the ruling authorities. In public transportation much of the resources that are available are supplied by these authorities. In addition, changes that need to be made in the public domain will also have to be approved by them. This often means that some passenger requirements need to be adjusted in advance to match the available budget and limitations due to conflicting interests. It also requires a good communication and cooperation between the railway company and these authorities to find the most optimal solution for the passenger within these boundaries. Railway companies are often also bound to a performance contract with the government or local authorities. An example of this is mentioned by Vromans (2005), where he describes the performance contract of the NS (Dutch National Railways). From the government they receive the exclusive rights to operate trains on the core network of the Netherlands. In exchange they have to reach the performance goals; a minimal service frequency and a certain punctuality. These are, as mentioned before, determined by the authority on the basis of passenger requirements and possible constraints.
The described performance indicators are generally found in literature about railway systems. However, there are more requirements and performance indicators that are of interest for both the involved authorities and the railway company. In some cases strict performance demands are set and the way these are reached is not of interest, like in the NS case, but for smaller railway companies, for example local tramway companies, the influence of the authorities is often larger. Indicators on the quality of internal processes of the
company are then required. Even if these are not required, they are still of great interest to the company itself in order to analyse its performance. Performance indicators for controlling processes are presented as part of process criteria in the Delft Systems Approach (Veeke et al., 2008):
10 Productivity: The way the available resources are used by the railway company is of interest to the authorities, since they need to know if the realised output corresponds to what was expected based on the resources that were made available. If this is not the case it might be necessary to change the means that were used. In the Delft Systems Approach this is called the productivity, which is one of the main process criteria for designing a system. It gives an indication of how well the processes in a railway company perform.
Related to the productivity are the criteria of efficiency and effectiveness. Both influence the productivity level of the system and therefore need to be analysed by the company.
Effectiveness: The effectiveness indicates whether the realised service level is equal to or near the agreed upon service level, given a certain amount of resources. This shows if the provided transportation capacity actually realises the goals that were set. If the goals that were set are not reached by the current output then an analysis needs to be done to determine the cause.
Efficiency: The efficiency is seen as the degree to which the amount of resources that were used corresponds to the standard amount of resources used for realising a certain process. To decrease costs the company should therefore efficiently use the available resources. This translates into making sure that the same transportation service level can be obtained with fewer funds, vehicles and personnel.
Flexibility: Another main process criterion that is relevant for the authorities and that serves as a measure of performance is the flexibility. This is the way in which the railway company is able to respond to changing demand from the authorities or the environment. In a railway system it is important to adjust the capacity of each line to the changing demand and constraints for transportation throughout the network, this requires a certain level of flexibility in the process of adjusting a timetable. Apart from adjusting the capacity it is also important to quickly react on both planned and unplanned large disturbances that require action beyond the regular control mechanisms of the daily service (see next paragraph), in order to guarantee a reliable supply of transportation. Both are of importance for obtaining and keeping satisfied customers and principals.
11 Internal requirements
To be able to realise these external requirements there is also the internal requirement of control that needs to be met:
Control: This requires the processes in a railway system to be controlled, which means that the processes need to be evaluated regularly according to protocol and if deviations from the desired situation occur they need to be controlled to return to the desired state. This is especially important for coping with disturbances in the current timetable, in order to return the daily service to normal. The level of control over internal processes influences the external performance indicators.
2.2
Process at a railway company
The system as described in the previous section represents the core aspect of a railway system, that of the transportation of passengers. However, as said before, besides the flow of passengers a railway system has to deal with more than only this aspect. Throughout literature different aspects can be found. For example, Marinov et al. (2012) describes the aspects of static (e.g. infrastructure) and dynamic (e.g. trains) resources, demands for services, patterns, traffic rules and priorities and different operating forms (transportation with different train types). In this research the aspects are defined, according to the
PROPER-model of the Delft Systems Approach (Veeke et al., 2008), as the flow of resources (both static and dynamic) like trains, personnel and infrastructure, the flow of the total demand for transportation (order flow) and the flow of passengers (product flow). This best describes the different flows and their related processes, including the interaction between these aspects. Patterns, traffic rules and priorities are included through the requirements and standards that are set for the processes of these aspects. The three aspects and their mutual relations are represented in the PROPER-model of figure 3, which represents a single (either regional or national) railway system. The interactions between these aspects need to be evaluated to understand how the abovementioned requirements are internally processed and where analyses for design and performance evaluation of the system are required.
The model of figure 3 directly shows the complexity of a railway system. It is a continuous puzzle to create an optimal transportation service, in which timetables and the assignment of trains, personnel and infrastructure need to be mutually tuned while the input of
12 requirements, transportation demand and availability of infrastructure, trains and personnel continuously changes. To use the system requirements for controlling internal processes in the three aspects they must first be translated into standards in order to be measurable. By zooming in on the aspects the underlying processes can be described.
Planning services
In the function of planning services the actual timetable is made based on the demand for transportation. Therefore this demand first needs to be measured, often through ticket sales, and is further predicted based on experience and historical data. Then the required
transportation capacity is determined from this demand over time, after which the current routing of different lines on the infrastructure is evaluated. According to Huisman et al. (2005) this is based on the available infrastructure, sufficient coverage of all stops and the interconnection of different stops and lines. He says a trade-off needs to be made between long lines which guarantee a direct connection and the stability of the system with respect to delay propagation. He also mentions that long lines introduce an inefficient rolling stock circulation due to fluctuations in demand for transportation along the line. The route choice for connecting stops also influences the passenger requirement of speed in another way, since the journey of one stop to another often contains one specific route that is the fastest. Improving this route can speed up the journey between these stops. The influence of the demand on the routing and travel time is therefore hard to predict and is preferred to be analysed before it is adjusted. Most standard routes stay the same over longer periods of time, but maintenance tasks on parts of the infrastructure can regularly cause temporary route changes.
When the routes are defined, a timetable can be made which contains all the trips including the halting times at stops. According to Pachl (2008) scheduling through timetables has the following use:
- It coordinates the train paths (trips) in the planning process for optimum use of the infrastructure.
- It ensures the predictability of train traffic.
- It produces timetable data for passenger information.
13 Figure 3: Aspect model of Railway Company
The process of making a timetable can be a complex puzzle in which the infrastructure capacity, the routes, the demand for transportation and the availability of vehicles greatly influence the design. Many tramway companies control the infrastructure which they use and are therefore able to completely make their own timetables. On a national level however the infrastructure is often operated by a separate infrastructure operator, for example in the Netherlands (Vromans, 2005). Train paths are sold to multiple companies, which is why a train operating company has to make requests for certain trips and for some routes the timetable becomes dependent on this commercial interface. When all trips for the vehicles are determined they can be fitted with crew duties, which form the crew schedules. These need to be tuned to the available personnel and comply with the collective employment agreements. This process of making these schedules is shown in figure 4.
14 The control mechanism in figure 4 corresponds to process control as described by Veeke et al. in the Delft Systems Approach. It uses the measurable standards that are created from the requirements to control the process of making timetables and crew schedules. The requirements of productivity, efficiency, effectiveness and flexibility hold for the entire
aspectsystem as seen in the figure. Efficiency and productivity can be translated to standards that make sure that in the planning phase the available capacity of infrastructure, vehicles and personnel is allocated in an optimal way. This means that the demand for transportation at each stop should be satisfied by using as much as possible of the capacity of each vehicle, thereby minimising the total number of required vehicles. Without further requirements, this would mean that the train would only depart if is completely full. Here another standard has to be taken into account, which is the frequency of trains on a route. This standard is translated from the requirement of speed and comfort, since a passenger wants to be able to leave at any moment and not have to wait for too long for the next train. A certain interval between trains must therefore be offered on each route. An optimum has to be found between these two standards. For the productivity of the trips the trade-off is made in route setting between the direct connection of as much stops and lines as possible and the travel time on the line. All these parts of the process of making a timetable are interrelated and therefore changes in any stage of the process have an effect on the other stages.
15 This means it is an iterative process in which all standards that were set need to be analysed at the same time if an optimal solution is preferred. Otherwise priorities have to be set on a number of standards in order to guarantee that these are first met. The process of making a timetable is therefore very well suited for the use of design and evaluation tools that support this iterative nature.
According to Vromans (2005) important performance indicators for railway timetables are robustness, reliability, stability and punctuality. The reliability was mentioned earlier as a passenger requirement, for which the punctuality can be used as a performance indicator. Both concepts are often found in literature. This punctuality requires some measuring points in the system, usually at stops or stations. Vromans sees the robustness of a railway system as the influenceability of the system by disturbances. A railway system that is not robust will make small disturbances cause large delays that propagate quickly throughout the system. Since the passengers are only interested in the punctuality of their trains, robustness can here be seen as a concept that influences the punctuality. The stability of a railway system is described as a measure for the time and effort that is needed to return to normal operations after a disturbance. An instable railway system will have irregular traffic for a long time. This means that in order to achieve a good punctuality level as an external requirement, the stability and robustness of the timetable also need to be of a good level. Stability and robustness are improved by adding recovery time to the timetable (Pachl, 2008). Analytical tools can be helpful here in testing the robustness and stability of a timetable and to see how these are affected by changes in the recovery times and in the rest of the system. Very closely related to the punctuality is the total amount of delays that occurred, which is relevant for the requirement of speed. The punctuality is a measure for the total delay of individual vehicles that has occurred, but the resulting delay that passengers might have can be even bigger as a result of missing the connecting train. It is therefore important to
distinguish these two concepts and to analyse the consequences of vehicle delays for the total delay of passengers. When the total journey of the passenger can be monitored, not only the total delay becomes apparent, but it is also possible to monitor the individual travelled distance. From this, the total revenue that is generated by passengers can be calculated and consequences of changes made in the system for this total revenue becomes visible.
The planning of personnel is a similar process to the creation of timetables. The crew
16 the timetable, but also on the collective labour agreement of the company and on the
availability of personnel. These last two both give quantitative standards that are directly useable for planning crew duties. The collective labour agreement gives boundaries on the length and content of the duties, which limits the flexibility of this planning process. The availability of personnel is a matter of considering to invest in hiring more personnel if there is a shortage, or repositioning or firing personnel if there is a surplus. Therefore an optimal allocation of personnel is required, in order to minimize costs. This means that crew duties need to be created in the most efficient way, while the conditions of the collective
agreement are satisfied. A simple analysis of the produced schedules quickly shows if they comply to these standards, but to efficiently allocate personnel within these boundaries is a more difficult task that requires the support of analytical tools. Rescheduling of services might also influence the driver schedule, due to which duties might end up outside of the limits of the agreements. The consequences of rescheduling for the crew duties therefore need to be known.
The effectiveness of the planning process is measured by determining how well the realised theoretical transportation capacity and service level satisfies the goals that were set.
Standards are made from which the current planning process can be controlled and evaluated. These standards include the quantity and quality of the offered capacity and service. The required service quality contains the requirements that are set by the authorities, which in turn represent the adjusted passenger requirements. The produced schedules need to be analysed to determine the theoretically realised capacity and service level to be able to communicate this with the authorities.
Transporting passengers
The main aspect of a railway system, the process of transporting passengers, was shown globally as part of a combined national and regional railway system in figure 2. When looking at this aspect in a single railway network it can be drawn as in figure 5, which is actually one of the transportation sub-processes of figure 2 (apart from the added process of trip
planning, which is also added for completeness of the model). Here the control function is also present to make sure the standards are used to control the process. In this process the schedules and resources like infrastructure, vehicles and personnel come together to fulfil the main function of the system; transporting passengers.
17 Many of the passenger requirements apply here, since it is the process that directly interacts with the passenger. Safety is a basic requirement that needs to be fulfilled and should be guaranteed here.
Figure 5: Aspect of the transportation of passengers
For this aspect this is done by setting standards for the driving behaviour of train drivers, like local speed limits and instructions for halting at stops and handling failures, in order to prevent accidents. Part of the control function in this process is performed by the traffic control centre. Safety standards should be set for the controlling of train operations by this control centre. This will make sure that no conflicting movements will occur and that when an accident happens or a part of the infrastructure becomes unavailable, part of the
operations are stopped or rerouted. This requires the development of protocols that include the actions that need to be taken in different kind of scenarios. This can only be done for the most basic scenarios, since a combination of parts of unavailable infrastructure can create an infinite amount of different scenarios. Apart from safety these decisions also influence the routes and delays of the vehicles and therefore the requirements of reliability and speed. The total delay that results from the control alternatives therefore needs to be taken into account when making a decision. Whenever a vehicle is delayed or rerouted this will also influence the schedule of the train driver. Drivers and vehicles follow different schedules, in which a train driver can even drive multiple vehicles in one duty. This means that changing
18 the vehicle schedule also influences the driver schedule. It is also the task of the control centre to make sure the drivers end up in the right place at the right time to be able to continue their duties. The total consequences of rescheduling are therefore difficult to oversee. According to Marinov et al. (2012) ‘dealing with various tasks with complex interactions in an evolving dynamic environment requires adopting a systems approach for efficient and effective train operations’, which was said in relation to train traffic controller tasks. They further describe the complexity of train traffic control. To support traffic controllers in making these complex decisions they need to be able to quickly see the consequences of the different alternatives, for which the use of a computer tool is almost required.
In the wait-for-vehicle process a passenger spends his time waiting at a station or stop. At these locations the social safety should be ensured by creating a safe and comfortable environment. Often the perception of safety may even play an equally important role as the actual safety in this situation, so creating an environment that looks safe helps to improve the performance on safety (van Hagen, Peek & Kieft, 2000).
Reliability can be measured and controlled in this process by monitoring the driving and halting times of the vehicles in operation. This is already done often by railway companies, since these data are used for communicating real-time travel information, as seen in figure 5. This is also used to determine the realised punctuality.
Performance indicators like ease-of-use, comfort and experience are influenced by qualitative measures that are taken. These are improvements on facilities like the interior of the vehicles or stations or the placement of adequate way finding. These qualitative measures, however, are not suited for analyses with quantitative methods like simulation and will therefore not be discussed further.
The speed of the journey is for a great part determined by the outcome of the other two aspects; the limitations of the speed on parts of the infrastructure and of the vehicles together with the chosen routes determine the pure driving time. Disturbances can
negatively influence this driving time. Although the speed limitations that are set for safety or nuisance reasons often determine the actual driving speed, the behaviour of the driver still influences a big part of this speed. He might choose to ignore the speed limits if the security system allows him to and drive faster, or he might drive slower than the allowed
19 speed. Especially on tramway systems the security system often gives drivers a greater flexibility by not intervening in the speed of the vehicle and this causes bigger fluctuations in realised running times. Also the amount of passengers at a stop that need to board the vehicle can influence the halting time of the vehicle. A timetable has built-in recovery times at stops or end-points to compensate for some of these delays, but this can only cope with a small amount. How these factors exactly influence the total transportation time needs to be analysed before the timetable is used. Since driver behaviour and passenger counts can vary significantly, multiple scenarios need to be taken into account.
As a consequence of the speed profiles and total distance that is driven the energy usage changes. Albrecht (2008) says that also for railways the energy efficiency of transportation becomes more and more important in times of climate change and growing population. He presents algorithms to minimise mechanical energy consumption during train operation. These include the use of spare recovery time to influence the driving style. Also the maximum power load at energy distribution stations along the network needs to be considered when determining train frequencies. This energy usage per section of the network needs to be determined to make sure the network is able to process the required capacity and to be able to further minimise energy usage. This contributes to the efficiency of the system.
From the output of this passenger aspect the actual productivity, effectiveness and flexibility can be determined. The productivity and effectiveness can then be compared to the
theoretical values obtained from the planning aspect to compare the theoretical planned service to the realised service. This gives valuable information on the timetable, internal processes and external influencing factors. The flexibility can be determined from the way in which the railway company was able to react to planned and unplanned disturbances, changes in demand over a time period, usually a year, and changing requirements from the authorities.
Assigning resources
The last aspect of figure 3, the aspect of assigning and maintaining resources, seeks to make sure that the transportation process is supplied with the required infrastructure, vehicles and personnel. This is done in cooperation with the planning department; since there is not an infinite amount of vehicles and personnel available, the planned transportation capacity
20 should be adjusted to the availability of resources, or an investment should be made to acquire more vehicles and personnel. Maintenance also plays a role in the availability of vehicles and infrastructure. The number of vehicles that are out of order due to a defect depends on the quality and speed of the maintenance process. When the availability of the vehicles is known, they will be coupled to a service until all services are filled. The type of vehicle that is needed depends on the type of the service (inter-city, regional stopping trains, etc.). The maintenance of infrastructure can temporarily put parts of the infrastructure out of order. Depending on the duration and whether the maintenance was expected or not, it means that the maintenance task should either be included in the produced schedules and routes by the planning department or that the current operation should be rerouted and rescheduled on the go by the control center. Huisman et al. (2005) state that this can also destroy vehicle and crew schedules, which then need to be repaired if possible. This is either done in the aspect concerning the planning of services or in a more ad-hoc way by the control center in the transportation aspect, depending on the time horizon of the
maintenance task. The process of assigning and maintaining resources as described here is viewed in figure 6.
In combination with figure 2 it can be seen that the resources are assigned to a task for transportation and are released afterwards. For vehicles and infrastructure this task can also include getting ready for maintenance at the end of the trip or day. In order to make sure that disturbances do not influence the assigned tasks, the traffic controllers in the previous aspect needs to monitor and control these assignments. This requires good communication between the control center and the people that assign the resources, as well as a way to predict the consequences of actions taken by the traffic controllers.
In the section about planning services it was said that the timetable depends on the availability of infrastructure. As a consequence of the capacity analysis in the planning services aspect it might then be necessary to expand the infrastructure network in order to realize a certain service. Adjusting the rail infrastructure usually requires a capacity analysis to determine if the design, including the signaling system, is technically and financially feasible. This also goes for the geographical positioning of stops, which might have a similar influence on the capacity, but also on the number of passengers. Speed, reliability and safety are the requirements that influence the design of this process. The design of the signaling system in combination with the chosen headway (time interval between trains) determines for a big part what the maximum vehicle capacity on parts of the network is (Pachl, 2008).
21 Figure 6: Aspect of assigning and maintaining resources
Also of influence on the capacity are the length of the infrastructure and the number of tracks. The capacity over time is furthermore determined by the maximum allowed speed on each part of the infrastructure. Making the system safer might therefore in some cases go at the cost of speed. Designing the infrastructure in such a way that it has sufficient safety, capacity and speed requires a complicated analysis to make sure it fits all the requirements. In this aspect the reliability of the system is dependent on the elements that influence the punctuality. Looking at the resources, the reliability depends on their quality. For the vehicles and infrastructure this means that the chance of a malfunction or break-down should be minimised by setting high maintenance standards. For both the amount of maintenance or adjustments that need to be done depends on the usage of these resources. Predictions on maintenance can therefore be made when the total usage of these resources is known. Other factors, like the total energy usage or the nuisance that is created by the sound of the metal wheels and tracks also play a role in making a design choice. The mentioned factors are dependent on the driving profile of the vehicles. Therefore analyses are often done to
22 predict these factors when a new design choice or planning decision is made. Making such calculations based on driving behaviour requires quite some calculations and insight. For the purpose of efficiency and productivity, the assignment of vehicles should be controlled in a proper manner. From a maintenance perspective it is best to equally
distribute the number of kilometres over the vehicles that are available. This makes sure that maintenance is controllable and well balanced over time. Also the amount of deadhead kilometres (trips without passengers) to the starting point of a line should be minimised to save costs. The location from which a vehicle is assigned to a service therefore needs to be considered. This should also be taken into account in choosing the location for building a depot. A railway network often has multiple depots or shunting yards where the vehicles are parked. These have a certain vehicle storage capacity that needs to be taken into account. Only some of these depots also have maintenance capacity. Therefore it is a complex task to assign the vehicles in such a way that the mentioned constraints and requirements are met and at the same time costs are minimised. For logistical reasons, the assignment of the vehicles to specific positions on a track in the depot is also important to make sure they can leave the depot according to the sequence of the assigned services. Analysing the process of storing the vehicles in the depot is therefore of the essence for an efficient assignment of vehicles to services.
23
3.
Railway simulation
Chapter 2 shows that to successfully operate a railway system a great number of analyses need to be made in order to support the design, operation and evaluation of the system. These analyses are required on different levels of decision-making and for both internal and external communication. They include many variables that can change over time, which makes analysis by hand very complex. Analytical computer tools are more and more being used by railway companies for design, control and evaluation purposes (Hansen, 2008). Computer simulation is a method that replicates real processes in an abstract way in order to be able to study the behaviour of these processes under varying conditions. According to Siefer (2008) simulation is already an important day by day tool for railway companies. Before looking further into how existing railway simulation applications support decision-making, the advantages and disadvantages of using railway simulation will be discussed here, followed by a brief introduction to the methods that are used in railway simulation models. This will help to understand how these models work.
3.1
Advantages and disadvantages of simulation
To determine if simulation is the right tool to use for specific analyses it is good to know what the advantages and disadvantages of simulation are.
Advantages
Siefer calls railway simulation ‘a virtual laboratory in which infrastructure and timetables can be changed and modified in different ways’. It allows to reproduce the processes in a railway system, like the ones described in chapter 2, and to test a great number of scenarios.
According to Vromans (2005) an advantage of simulation over analytical models that use algorithms is that it is possible to add a higher level of detail. The algorithms in these analytical models are often not suited to be equipped with a high amount of details. He also states that analytical models are often focussed on smaller parts of a network, or have restrictive assumptions about the model. Simulation is able to recreate whole networks in high detail.
24 Railway operations are very interactive and have a strict and dense schedule on which they operate; there is practically no room to experiment with different configurations to see what the resulting interactions are. With simulation it is possible to define different scenarios to test any combination of trains, infrastructure and timetables without the risk of disturbing the operation. These combinations require the variation of a great number of parameters, which is near to impossible to do by hand. Computer simulation offers the possibility to analyse large networks in a relatively short amount of time.
When a new part of infrastructure has to be designed it is hard to estimate what the impact of the design on the daily operation will be. Building a prototype is not possible due to a long construction time and high costs and impact on the operation. Simulation gives a good alternative to test these designs beforehand. Almost the same goes for timetables; the impact of a new timetable is not entirely known until it is actually implemented. Through simulation it is possible to test the timetable while making the risk of unforeseen
disturbances significantly smaller.
Simulation also allows the use of different kind of distributions, which makes analyses on variations in, for example, disturbances possible. By performing multiple runs the outcomes can be tested on statistical reliability. This also enables to study extreme cases as well as averages.
Although computer simulation programs for railway systems are often not as transparent as analysing the process by hand, they do often offer a way to get an understanding of the processes and relationships at the higher aggregation levels in a railway system. When the model is built in a structured way, for example comparable to the structure of the Delft Systems Approach, it is better suited for understanding and explaining the processes that occur. A visualisation of the processes is also often possible and helps to explain and understand what is happening in the model.
Disadvantages
Apart from advantages, the use of simulation also has some disadvantages. With every increase in the level of detail there is also a bigger demand for data and knowledge about the system (Thompson, 2001). Donald Knuth, seen as the father of informatics, once said the following: “we often fail to realise how little we know about a thing until we attempt to simulate it on a computer”. This very well illustrates the effort and knowledge that needs to
25 be put into building an accurate simulation model. The data and knowledge need to be available for use before the simulation model can be developed.
Although the use of simulation might later on save a railway company a lot of money by being able to test designs and evaluate the system, according to Thomspon there still can be high initial costs associated with developing such a model. For example when the amount of available data is not sufficient and a system that collects these data needs to be installed, or when a lack of knowledge needs to be compensated by buying it. Sometimes additional computer infrastructure needs to be installed. These costs need to be taken into account when considering to develop a simulation model. Also the time taken for development can be substantial, since there are multiple disciplines that need to work together on a complex model.
Simulation of processes is a method that solely reproduces the processes as they have been described, it is not a tool that makes suggestions of how to improve certain processes or outcomes. This is not necessarily a disadvantage of simulation, but needs to be taken into account when deciding to use it. It can however be outfitted with algorithms that are able to do this, although this is a different discipline.
3.2
Methods of railway simulation
Simulation has evolved over the years as a result of different scientific approaches. This has lead to the different classifications of railway simulation models that are used today (Siefer, 2008). The scale of the model is split up in macro-, meso- and microscopic simulation. This often holds a strong relationship with the planning time horizon (stragetic, tactical and operational planning) that the software is used for. Also the analytical approach
(deterministic or stochastic) and the processing technique (synchronous or asynchronous) are mentioned. Also the distinction between continuous and discrete models is generally found in literature.
Macro-, meso- and microscopic simulation
Simulation models in the railway industry are made for different reasons and these influence the level of detail that is needed in the model. Often railway simulation is used for planning related purposes, which are classified in literature according to the time horizon; strategic,
26 tactical and operational level planning (Siefer, 2008; Huisman et al., 2005). According to Huisman et al., strategic level planning in railway companies generally includes a time horizon between 2 to 20 years. It involves the purchasing of rolling stock, the hiring and training of crew and the planning of lines and routes, including expansion of infrastructure. Tactical planning is usually done once or a few times per year and includes the creation and analysis of timetables and crew duties (Marinov et al., 2012). For operational planning, some tasks can be done a few times per year, but others are done on a daily basis. Detailing the timetable with the routing through stations for example is often done after the basic timetable has been created. Also the crew schedules are created from the available duties and the vehicles are coupled to the created trips. When deviations occur due to disturbances in the network, the daily schedules are adjusted to bring back the operations to normal. The levels of detail that are required to perform these different stages of planning vary for every task. In general the planning process from strategic to operational requires an increasing level of detail. For example, for determining routes and connections in the network only the stops and the tracks connecting them are required without much further detail. Often average running times and capacity per track are used here. In infrastructure modelling, this is called a macroscopic level (Radtke, 2008; Siefer, 2008; Marinov et al., 2012; Gille et al., 2010). For the creation of timetables and the determining of exact running times a high level of detail is required. All the elements that influence the speed, like
signalling systems, speed limit differences, lengths and maybe even gradients need to be taken into account. Also the layout of every track needs to be detailed, also within stations. In the mentioned literature this is referred to as a microscopic level. Depending of the size of the network, this high level of detail can slow the speed of the model. However, the capacity of computers is quickly rising and simulation speeds rise with them. An infrastructure model with a level of detail that is in between macro- and microscopic is called a mesoscopic model. According to Radtke (2008) an ideal situation is that a macroscopic model is generated from microscopic data, this will increase the accuracy of the averages used in macroscopic models and prevent the user from having to maintain multiple infrastructure databases. Also the model can then be scaled as needed for performing multiple kinds of analyses.
27 Deterministic vs. stochastic
Deterministic models determine the progress in the model according to given pre-defined quantities without randomness, whereas stochastic models base their estimations on distributions for all kinds of processes. In railway applications stochastic distributions are often used to model distributed arrival and departure times, fluctuating passenger numbers and disturbances.
Synchronous vs. asynchronous
Synchronous simulation models run all processes in the model at the same time. Every event is processed as time advances. This represents the processes in the same way as they occur in reality. This is required for modelling the interactions between trains, so it is suited for evaluation purposes.
Asynchronous railway simulations are based on priority rules for different processes. For example, high priority trains are calculated first, the lowest priority trains last. This makes sure that the high priority trains have as few disturbances as possible. The lowest priority trains are fitted into the time slots that remain unassigned in the end. This kind of simulation is suited for timetable design as it reproduces the train prioritisation process of timetable design, but for this reason this method is not suited for analysis purposes.
Continuous vs. discrete
The way time progresses in a model can be either continuous or discrete. Continuous models run processes continuously over time, which is according to reality, while discrete models progress either through fixed time steps or by an event-driven progression of time. The first lets time progress in constant intervals, during which changes are made. Event-based simulation lets time progress unequally depending on the occurrence of an event.
28
4.
Existing applications
From chapter 2 it can be concluded that managing a railway system in general consists of planning the desired operation according to the requirements and afterwards trying to
maintain this planned scenario during operation or quickly finding the best alternative in case of disturbances. Many interrelated facets influence these processes, which makes it difficult to determine the most efficient and effective way to perform them. It is therefore easily explained why railway companies look for supporting tools to improve the quality of these processes and the output.
For the last two decades computer tools for timetable design are being used by railway planners and operators to design and analyse these timetables (Hansen, 2008). Analytical tools like software packages based on specific algorithms and the application of operations research are now widely used for the design and optimisation of timetables and crew schedules. For specific theoretic purposes, like the initial creation of duties or trips, these systems work very well. However, according to Hansen (2008) these methods are based on simplified or isolated parts of a railway system and often fail to include all dependencies, which may result in not achieving the desired operational results. Wood & Robertson (2002) state that also methods of assessment for different purposes like safety and cost or revenue exist, but that these are also restricted to a subset of these options.
As stated earlier, the simulation of processes replicates the behaviour of a system, making it possible to better understand the resulting behaviour and output and use these insights for making adjustments. Siefer says that today various railway companies already use simulation very successfully in their planning process. Some commercial simulation packages have been developed for the use in railway systems, together with some scientific models that are applied at companies. Many of these models are actually a combination of simulation and optimisation algorithms, in order to use the simulation output for further adjustments. According to Siefer (2008) and Hansen (2010) railway simulation tools can be used for, or support the following purposes:
- Timetable construction and testing - Operations analysis
29 This corresponds to a large part of the processes described in chapter 2. The PROPER model in figure 3 shows three aspects of a railway system; planning services, transporting
passengers and assigning resources. Timetable construction and testing is a part of the planning services aspect and includes the processes in figure 4. The analysis of operations corresponds to the second aspect, that of transporting passengers. This is used to simulate the actual process of transportation. Rescheduling is the interaction between these two aspects, when disturbances or conflicts occur in reality and the current operation needs to be rescheduled. This mostly happens both in the control centre and corresponds to the control function and the processes in the transportation aspect of figure 5, but also requires
communication between this control centre and the other two aspects of planning and assigning in order to tune the control actions to the demand and the available resources. This is indicated by the control function above the three aspects in figure 3.
Timetable construction and testing
Timetable construction and testing is what most software applications were used for when they were first introduced in railway systems and is still often the main purpose for using simulation and optimisation packages. This means most of the simulation software that is available is mainly focused on this process.
Simulation and optimisation tools assume that the demand for transportation is calculated in advance through the counting of passengers or use of questionnaires (Marinov et al, 2012). Also demand models can be used to predict the growth in demand on the basis of historical data. Planning and optimisation tools are then used to determine the routes and create a timetable. The creation of a timetable is an iterative process; after the initial creation it is often tested with the use of simulation to find the delays that result from the pure timetable itself (Siefer, 2008), from where it can again be further improved. Also artificial delays can be inserted to check the timetable stability and robustness. Demitz et al. (2004) used the commercial simulation and optimisation software RailSys to analyse and improve the timetable of 2003 in North Rhine-Westphalia, by adjusting the waiting time at stations. Another research using RailSys for evaluation purposes was done by Rudolf & Radtke (2006), in which they simulated two large networks in Germany to test strategies for optimal
allocation of buffer times (allowances). This allowed them to test the punctuality and speed as a result of these strategies. It is further used at a number of railway companies
30 length, switches, speed, radius, etc.), signalling system characteristics, vehicle characteristics and initial routes and timetable to detect and solve initial conflicts in the timetable through optimisation algorithms, after which a macro- or microscopic, synchronous simulation can be run to accurately determine further propagation of delays and resulting conflicts (rmcon.de, 2014). This simulation determines arrival, departure and passing times from the behaviour of all elements, from which further statistics can be derived. Also pure running times can be calculated from inserting single trains. In addition it includes some algorithmic optimisers to optimise the timetable using the results of the simulation. Simulation here helps to determine the resulting punctuality of trains as a consequences of changes made to the input. This process is still completely deterministic, since the behaviour is completely determined by the implemented logic. The model has a simple animation included to see where trains are located during the simulation.
OpenTrack is developed by the ETH Zurich in Switzerland and is a microscopic, synchronous model that works in both a discrete and continuous way (Nash & Huerlimann, 2004). The train motion is modelled in a continuous way, whereas the signal states and delay
distributions are discrete elements. It also uses infrastructure, signalling system (including tram systems), vehicle, and timetable data as an input to simulate operations, as is shown in figure 7. Different signal types are available and can be changed in the model. Just as RailSys, it outputs different kind of graphs based on the arrival, departure and passing times of all trains. In addition it also has an animation functionality, which is slightly more detailed than the animation of RailSys as it also shows the block sections of the signalling system that are occupied and the state of the signals. It also has the possibility to view the influence of different weather conditions on the traction of the vehicles.
Next to RailSys and OpenTrack there are more simulation models that have a comparable or more limited structure and functionality with regard to timetable evaluation. Examples are RailPlan (trapezegroup.co.uk, 2014), VISION (McGuire et al., 1994), FALKO (siemens.nl, 2014), RAILSIM (railsim.com, 2014) and SIMONE (Middelkoop, D. & Bouwman, M., 2001). The first four were developed for general use, the last one was developed for analysing the Dutch rail network and timetables.
Also asynchronous simulation models are available. According to Siefer (2008) this simulation method is suitable for the generation of timetables on the basis of train priorities. It first generates high-priority train runs, after these are done it continues with lower priority trains. The only well-known models are STRESI and BABSI. The first generates timetables by letting
31 trains run in a prioritised order and enables operational testing. BABSI allows for timetable stability testing, conflict detection and route suggestion (via.rwth-aachen.de, 2014). Siefer also states that asynchronous railway simulation is only used in research institutes, since it is not suited for simulating realistic behaviour and creating timetables with minimal overall delay.
Figure 7: OpenTrack lay-out (Nash & Huerlimann, 2004)
Operations analysis
Usually after the initial timetable has been created, the simulation of operations is used for testing scenarios with different kinds of lay-out or parameters. It can be used for testing infrastructure options, single lines, signalling systems, vehicle characteristics or any
combination of options by adjusting these elements before running the simulation. Moreover it is possible to introduce (stochastic) variables that represent, for example, disturbances. It is then possible to predict how the system will react in everyday operation. The propagation of delays through the system as a consequence of these initial delays can be evaluated. According to Siefer “the operational simulation results can be used for the evaluation of the performance of a whole network”. Most simulation systems that were mentioned in the previous section also include these functionalities. Demitz et al. also used the simulation of
32 disturbances in their research to further analyse the influence of the size of buffer times on the propagation of delays during daily operation. Lindfeldt (2010) used RailSys to “evaluate the possible total effects of reduced primary delays” on the Western Main Line in Sweden. Also Hansen (2010) mentions the possibility to use STRESI, RailSys and OpenTrack to determine waiting times as a result of random delays. OpenTrack even allows to insert delays, or vehicle break-downs, at specific points per individual vehicle (Nash & Huerlimann, 2004).
In addition, some simulation software also supports the blocking of infrastructure, e.g. for maintenance purposes or accident simulation. OpenTrack and RailSys are amongst them. Others also have the possibility to track the circulation of vehicles, to be able to see how these are distributed over end-points in the network at the end of the day.
Many simulation models also use the simulated operational data to calculate energy usage and some even cost functions. Through formulas that use the simulated speed profiles and travel times, the different energy consumptions and peak loads can be determined.
Capacities of energy supply points along the infrastructure are often included in the analysis. Among others, simulation packages that include energy consumption are OpenTrack,
VISION, FALKO and RAILSIM.
Rescheduling
After the timetable and operational parameters have been tested and tuned to fit all requirements, the resulting schedules are applied during daily operation. Real disturbances then occur which can cause deviations in these schedules that cannot be automatically repaired in time. This requires the rescheduling of individual routes and departure times until a feasible solution within the initial timetable is found. It is a difficult process to quickly find a good solution due to the many dependencies between services. According to Jacobs (2008) “the automation of the traffic regulation process improves the productivity, quality and flexibility of traffic management”. This indicates that simulation can play a role here.
Jacobs names asynchronous and synchronous simulation as two possible means for conflict resolution. He says that traditionally manual asynchronous methods are used to solve quick rescheduling problems, since these follow a natural course of action for scheduling by using train priorities to insert trains in the schedule without conflicts. However, these do not take connectivity and circulation of trains into account and cause large delays for low-priority
33 trains. Synchronous simulation shows all conflicts in chronological order, but only allows to identify very obvious conflicts in advance (like two trains using the same track). He states that synchronous simulation is more and more being combined with asynchronous methods to combine the advantages of both.
He names the example of the DB Netz AG control centres in Germany, that use a simulation tool that is fed with real-time data to determine the positions of vehicles and is then able to highlight future conflicts through simulation. However, these still have to be solved through other means. This is mainly done manually or with algorithmic supporting tools. Simulation can then be used again to test the different alternatives based on the real-time operational state. As of today, some existing simulation tools are outfitted with (asynchronous)
automatic rescheduling rules to support this process, but it still requires evaluating different options by simulating them and selecting the best alternative. This costs some time, which might not be available when the disruption is unexpected.
The measures that are taken can have an effect on the circulation of rolling stock and crews. For rescheduling purposes it is also important to know what happens to these ‘resources’ as rescheduling actions are made. In most simulation packages it is possible to see where a train ends up after being rescheduled, although output formats are not yet really designed for easy analysis of this feature, so they have to be analysed individually. RailSys does include an additional planning tool to assign vehicles to a simulated timetable in a circulation pattern. Consequences for crew circulation are however still not included in existing
simulation software. Simulation methods for rescheduling therefore still lack some required functionality.
34
5.
Current developments and possible improvements in railway
simulation
The previous chapter discussed the current practice of using simulation for design and evaluation purposes in railway systems. Some functionalities that are not yet available were also mentioned. This chapter describes possible improvements on current possibilities of simulation models, supported by developments found in literature.
Rescheduling
The use of simulation in railway environments started with supporting the initial long- and medium-term planning processes by focussing on the timetables and infrastructure
evaluation. Siefer says that there is still great potential for development of railway
simulation and “in the future might be used regularly in traffic control centres”. This would require that simulation tools are both able to start with real-time operational data and to use this data to analyse different strategies within a short time-frame. It should present the consequences in a clear way and easy to interpret. Rescheduling algorithms do already exist in research, but purely focus on the optimisation of the combination of infrastructure and trains. For example, Wegele et al. (2010) compares two laboratory real-time support tools based on operations research and branch and bound methods, which only minimise train delays based on the input of infrastructure, signalling systems and train data.
In addition, the consequences that the rescheduling action has for the circulation of rolling stock, but also for personnel should be taken into account. Rolling stock is often needed at a specific location for maintenance, or to make sure that sufficient vehicles of a certain type are available at each location. Rescheduling can disturb this balance and this should be taken into account. Available simulation packages also do not offer the functionality to include crew schedules and to simulate how they move from train to train. This makes it hard to predict the consequences of a rescheduling action for the personnel. They could end-up at a required location late, or even at the wrong location. Also the question remains if the duties that result from these actions still comply to the collective labour agreements. Therefore both rolling stock and personnel need to be included in the simulation results in a clear overview. Incorporating these functionalities in a simulation model would contribute to the effectiveness of the usage of vehicles and personnel. Also flexibility will be greatly improved, by being able to respond to disturbances of these schedules quicker. By testing the different
