Kevin Smith examines the findings of the DynoTrain project, which explored using simulation to improve the process of testing and certifying rolling stock.
GIVEN the nearly 13,000 national rules and specifications that locomotives and rolling stock must meet to operate across Europe's railway network, many of which were introduced to stifle competition, it is no surprise that delays in certifying railway vehicles, particularly those intended for cross-border operations, have become all too common in recent years, much to the frustration of operators.
The European Union (EU) is attempting to address this issue through the technical pillar of its Fourth Railway Package, which empowers the European Railway Agency (ERA) to become a "one-stop-shop" for railway certification and authorisation throughout Europe. The European Commission (EC) hopes this will speed up the certification process by allowing vehicles to run across the EU using a single passport issued by the ERA, which it estimates could cut the cost and time for authorisation by 20%.
While the majority of Members of the European Parliament have leant their support to this aspect of the legislation, a research initiative aimed at pinpointing ways to improve the efficiency of the certification process itself is reaching its conclusion.
DynoTrain commenced in 2009 as one part of the TrioTrain undertaking, which included the AeroTrain study, which addressed aerodynamics, and PantoTrain, which looked at catenary interaction, with both of these projects concluding in May 2012.
TrioTrain is part of the EU's second call of the seventh framework programme which deals with key railway interoperability issues and aims to facilitate the cross-acceptance of railway certification between EU member states, close open points in Technical Specifications for Interoperability (TSIs), and reduce the time and cost of vehicle acceptance by utilising simulation for certain tests. The schemes were grouped together to share project governance through a common steering board and advisory council, and use a cross-project work package that would deal with quality assurance and regulatory acceptance of proposals for standards through CEN/CENELEC, and TSIs through ERA.
DynoTrain, which had a budget of e5.5m, involved a cross-industry consortium of participants as well as the European Rail Industry Association (Unife), which coordinated the project (see panel), and encompassed seven work packages.
The first (WP1) supported the remainder by providing an extensive database of test data and measurements from around 7500km of operations on a mixture of high-speed, main and suburban lines in Germany, France, Italy and Switzerland over four weeks in October 2010. Six test vehicles of four different types - a German Rail (DB) class 120 electric locomotive, an inter-city passenger coach, a two-axle freight wagon and a Y-series freight wagon, which were both operated under loaded and unloaded conditions - were deployed, with all utilising 10 force-measuring wheelsets.
The results from the tests are regarded as an indication of the running behaviour on different networks, with 222 parameters measured for vehicle dynamic behaviour. Data retrieved included inputs from the track and wheel-rail contact, and outputs of the vehicle reactions which were all synchronised in the same measurement train. The test train also utilised a track geometry measuring car which recorded data from the entire route as well as rail profiles, while wheel profile measurements were taken before and after the testing period, although due to the relatively small distance travelled no differences were recorded.
Data from WP1 was stored on a dedicated server hosted by German Rail (DB) in Minden and made available for analysis by all partners participating in the other work packages to investigate relationships, and validate and develop models.
This included WP2 which analysed track geometric quality and is described by Dr Yann Bezin, head of research at the University of Huddersfield's Institute of Railway Research, as "a never seen before understanding of how track quality is measured and can be compared in different countries."
Recognising the need to develop vehicles with good dynamic characteristics, which requires a detailed knowledge of expected track geometry, the project consequently looked at developing an improved assessment method for track geometry with the aim, among others, of establishing a database for the design of interoperable vehicles and data input for virtual homologation.
The study found that the current wavelength range of 3-35m is appropriate for the vehicle assessment process at least up to 200km/h and provided a method and software to create a track geometry database which is now available. Recommendations and guidance on the use of multiple regression methods broadens the range of track sections that can be included in the approval test. Bezin says this is proving particularly useful in adapting results from one network to use by another by providing estimated values for differing input conditions which reduces the need for special testing saving both time and money.
Wheel-rail contact conditions were analysed in WP3 with the aim of supplementing the database established in WP2 by offering information on the optimum rail profiles for the design of interoperable vehicles, rail profiles for virtual homologation, and a set of wheel-rail friction coefficients for virtual homologation. The study, which was led by DB, also aimed to define in-service limits of equivalent conicity for track and wheels depending on speed, to confirm the use of the radial steering index which would help to avoid additional tests on other networks and confirm the range of applications. It also sought to develop a method to measure the wheel-rail friction coefficient during track tests, which is an important factor in understanding vehicle behaviour, but is difficult to measure.
Again data from WP1 was utilised, with the variations in rail inclination in Germany and Switzerland (1/40) and Italy and France (1/20), particularly useful. After establishing a reference wheel profile to assess the track contribution, measured wheel and rail profiles were used to develop probable conicity maps which were used to identify the separate contribution of the wheel and rail profiles to the combined conicity values. This was an essential step in developing in-service limits or a target system due to the ability to conduct evaluations of the various scenarios' economic impact. The measured lateral and vertical wheel-rail forces were also compared with simulations to assess the practicality of developing an algorithm to understand wheel-rail friction values.
Most significantly the study found that conditions for in-service wheel-rail contact were similar on all networks considered. In addition the equivalent conicity of the majority of tracks is between 0.1 and 0.25 and is not dependent on speed, while, depending on the wear condition of the rails, very high and very low values of equivalent conicity can be found in all networks. As a result the distinction between different rail inclinations outlined in the vehicle approval process is unnecessary and it is sufficient to carry out dedicated tests of running stability on track sections with adequate layers of equivalent conicity. This analysis was taken as a common basis to support recommendations to close the open point for in-service limits on equivalent conicity in the draft locomotives and passenger, and infrastructure TSIs.
Track loading limit requirements is one of the primary hindrances to simplifying cross-acceptance of rolling stock which, while operating successfully on one or more national networks, currently requires retesting before approval is granted to operate on another network. As a result WP4 studied this phenomenon to see whether additional tests are required with the aim of improving the process of cross-acceptance by reducing the number of tests, developing limit values for infrastructure construction and maintenance, clearly identifying local requirements, and offering proposals for cross-acceptance processes, including through the use of simulations and operating limits depending on infrastructure conditions.
The project compiled data from a detailed review of current requirements as outlined in TSIs, EN and UIC standards as well as other national rules, while any relationship between these requirements and limit values relating to track construction and maintenance standards were also recorded.
The study showed that all parameters used in national standards are useful for assessing a vehicle's influence on track forces and the subsequent deterioration of train and track components. In addition, for line speeds greater than 160km/h the current track construction standards are similar, meaning that a one-size-fits-all approach is viable. In some countries, weaker track conditions on lower speed lines were often apparent. However, the project took this into account by introducing a lower limit on one of the vehicle assessment parameters. The use of multiple regression analysis allows an evaluation of the estimated maximum value of different parameters for different target conditions. As a result it is possible to compare these with the appropriate limit value, or with values for existing, or comparable, vehicles, with this method applicable to test data or results from a validated dynamic model, which will enable a reduction in the amount of testing required.
These work packages have hinted at the use of simulations to replace line testing in certain situations. WP5 took this a step further by utilising existing track dynamics simulation tools to provide an insight into the interaction between the vehicle and track at a range of frequencies not covered elsewhere. The study concentrated on locomotives and freight wagons and was intended to understand the relationship between track loading limits currently in force in the EU and the actual loading on various track components.
The partner rolling stock manufacturers involved in the project filled in detailed questionnaires which provided a comprehensive report on the building and validation of vehicle models, with a particular emphasis on the different suspension elements and modelling methods in use. Bench tests were also carried out to compare the results against the models, while the different vehicles used during WP1 were modelled by several partners using the static test data, where available, and used to simulate sections from the WP1 runs.
Comparisons were made between the simulations and the test results using a range of modelling levels and complexities, which helped to inform a review of the significance of different modelling options and the need to utilise data showing track geometry and wheel-rail profiles.
The partners produced a wide range of comparisons between the simulations and the test results and the project was successful at quickly identifying quantities with large deviations, which is useful for pinpointing specific track sections which offered the large deviation. The system can also provide a valuable summary of the range of options available for modelling standard suspension components, together with the advantages and limitations of each.
Utilising simulation was the critical component of WP6 which investigated how a virtual test track tool (VTT) could supplement or replace track testing to assess dynamic behaviour. The project looked at how these methods might extend the validity of the test results beyond their original conditions and offer significant cost and time savings.
To achieve this, a detailed study using a range of mathematical techniques was conducted to evaluate the importance of different vehicle suspension and operating parameters to the results of simulations. This could then inform the level of modification that was considered reasonable for approval to take place solely through simulation.
WP6 focused exclusively on extending existing certifications using virtual tools meaning that the proposed procedure can only be used on vehicles and track that are similar to systems for which on-track measurements are available. The project's primary objective was to provide a representation of the dynamic response of the system using simulation that is at least as precise as that given by the measurement.
To successfully introduce simulation into the certification process it must be highly representative of the system's physical behaviour. The model must also be fully validated and the simulations must be based on a representative set of excitations. The VTT was thus developed and adapted to the requirements outlined in EN 14363, which relates to vehicle-track interaction, to enable the use of a representative set of track data developed in WP2 and WP1 measurements, while the tool can also use any suitable data from track recording vehicles to replicate the requirements of most commercial railway vehicle dynamics software and virtual test tracks representing different networks. The tool takes into account the vehicle's area of application and selects track sections that meet EN requirements for a test route. These are then combined into a track data file applicable for vehicle dynamic simulations, reducing the time it takes to assemble the simulations.
Recommendations
The study subsequently makes several recommendations for the successful application of virtual certification. Among them is the standardisation of VTT, and the development of a standard track database, like that collated during DynoTrain. This would be representative of different countries and line speeds and would require maintenance over time as new regions and traffic types are added. The plan also recommends the use of a variable friction coefficient in simulation as the current system only takes account of dry conditions and to use pairs of measured rail profiles. A sensitivity analysis of the vehicle/track model is encouraged to verify its design and robustness, and to determine the elements that have to be carefully modelled.
Bringing together the findings and practises identified in each of the work packages was conducted in WP7, which was intended to add value to the DynoTrain project by assuring acceptance by each of the participants.
As a result it scientifically assessed the uncertainties raised in each work package, particularly when simulation was proposed to replace testing, to maintain or improve existing safety levels, and to arrive at convincing proposals for standards. The participants also communicated with ERA, CEN and National Safety Authorities to promote the uptake of the findings and methods in the certification process.
Success
It appears these efforts have brought some success. In parallel with DynoTrain, ERA is in the process of revising TSIs for locomotives and passenger rolling stock, and infrastructure. It has also set up a dynamics working party to cover some of the interface issues found between these two documents. In addition CEN TC256 is working to amend various standards encountered during the project, with DynoTrain partners also involved in the process to revise EN 14363 from the 2005 revision used during the project.
"The revisions include a lot more references to vehicle design and the use of simulation, how to use simulation to support testing, and to fast track the certification of a new vehicle," Bezin says. "We envisage the revisions to be approved fairly soon."
Bezin says another major success of the programme is the "terabytes of data" compiled during WP1, which has the potential to be a beneficial resource to research projects for many years to come.
"We only scratched the surface of what is available during the DynoTrain project," he explains. "We feel it will prove to be a significant source of data during subsequent research projects and in follow up work to what we discovered during this project."
He adds that there is scope to extend this study to accommodate additional networks and this could be a future role for ERA as it acquires a more significant role in rolling stock certification in Europe.
"This database provides a good basis and during the project we also received input from Trafikverket in Sweden and Banedanmark about track geometry and conditions on their individual networks," Bezin says. "The VTT encourages the use of different data from different places into one test track to provide a mixture of geometrics so it is possible to observe the behaviour of different track conditions on vehicles on a specific route to inform the certification process of the potential results before actual track testing takes place.
"Collating a database from across Europe would have great value in providing more data to show how a vehicle may perform in certain conditions which will aid both design work and the certification process. ERA could build and maintain such a database, which would be an extremely useful tool during design and certification."
Indeed Bezin says that in his view a major success of the project was to improve cooperation between different regions by establishing a cross-industry collective that shared knowledge and understandings of specific concepts from their own unique perspectives. The project then might not only ease the process of approving rolling stock for operation across borders, but open up the boundaries between various facets of the railway industry to the benefit of the industry as a whole as the movement towards greater interoperability gathers pace.
Dynotrain members
• Rolling stock manufacturers: Alstom, AnsaldoBreda, Bombardier, CAF, Siemens
• Operators/infrastructure managers: German Rail (DB), Network Rail (NR), French Rail Network (RFF), French National Railways (SNCF), Trenitalia
• Universities/research centres: Huddersfield University, Politecnico di Milano, KTH Stockholm, Berlin Technical University, University of Rome
• Others: Ceit, Ineco, RSSB, UIC, Unife, Alma, Inrets, IFSTTAR
Aerotrain and Pantotrain
CONCLUDING in May 2012, AeroTrain and PantoTrain have offered significant respective improvements to the understanding of aerodynamic performance of railway vehicles, and the interaction between the pantograph and overhead catenary.
Findings from AeroTrain have enabled several open points in TSIs relating to aerodynamic performance to be closed with some of the results included in the new TSI Loc and Pas. The findings have also benefited knowledge in the fields of open air pressure pulse, ballast projection, crosswinds, trains in tunnels and slipstream effects. For example the project demonstrated that computational fluid dynamics simulation can reliably measure the pressure pulse at the head of the train, while a measurement technique was developed to assess the aerodynamic load in relation to ballast disturbance, and proposals for the simplification of TSI methodology to assess slipstream effects were offered.
PantoTrain, which proposed to transfer as much pantograph/catenary certification away from on-track testing to laboratory and simulation testing as possible, has similarly offered some significant steps forward. These include providing a basis for validating any software tool that may be used for virtual certification, the compilation of a comprehensive database of European pantographs and catenaries, as well as the consider
