Problems, assumptions and solutions in locomotive design, traction and operational studies

The use of multibody simulations for verifying the performance of railway vehicles emerged at a commercial level in the 1980s, and has continued evolving to reliable and mature modelling techniques and software packages that are commonly used for efficient design, development and acceptance of modern railway vehicles [26]. In the case of running behaviour and vehicle dynamics approval tests for new or relocated locomotives, several standards worldwide allow substituting some of the field tests with multibody simulations (MBS) to reduce the cost, time and effort of physical tests [15, 27]. For instance, the Australian standard AS7509:2017 Rolling Stock Dynamic Behaviour [28], defines the dynamic behaviour requirements of railway vehicles, including locomotives, to reduce accelerated infrastructure degradation and reduce the risk of derailments. AS7509:2017 states that simulation tools can be used as long as key elements of the vehicle simulation are validated through physical testing, and only industry-recognised simulation software can be used, referring, for example to the Manchester Benchmarks for Rail Vehicle Simulations [29]. Using railway simulations instead for field tests is especially useful when the test requires the vehicle to transit over specific isolated or cyclic track irregularities that could be hard or expensive to reproduce in the field. Similarly, the European standard EN 14363 [30] established the tests that a railway vehicle must pass for its running behaviour to be accepted for operation. The EN 14363 Annex T.3 [30] presents two methods for validating railway vehicle models for simulation of on-track tests in vehicle homologation context. Both methods involve comparison of measured versus simulated running dynamics quantities such forces between wheel and rail, car body lateral and vertical accelerations, bogie lateral acceleration, etc. All these standards [28, 30, 31] should be considered in locomotive studies because their final application of the standards can be summarised as written in [32]: “Complying with standards can be a substantial cost for suppliers of rolling stock and equipment, and for their customers. Providing evidence of compliance, as required by the Systems Engineering model, can be very time consuming. ... Guide books and reference volumes may well be useful documents, but there is no compulsion to follow their advice. A standard, however, states how things must be done. And generally, the supplier must prove that product is in accordance with the standard.”

In addition to the use of locomotive and other railway vehicles multibody simulations (MBS) for “virtual homologations”, there are emerging applications where researchers use more complex models that include the locomotive traction system, enabling, for example predictions of wheel-rail wear and RCF [33].

In the process of formulating a locomotive dynamic model to be used in traction studies, several disciplines including electrical engineering, mechanical engineering and civil engineering are involved. In general, there are four common modelling approaches [13]: implementing a locomotive model using a general-purpose scientific commercial software like MATLAB/Simulink [34‐36]; implementing a locomotive model using in-house software [37]; implementing the locomotive model in commercial multibody software packages [13, 38, 39]; or co-simulation approaches where the multibody locomotive model and the traction mechatronic system are modelled independently in two different software packages that interchange data during the simulations [19‐22]. The latter approach allows for advanced traction and track damage studies as it allows for advanced traction models and locomotive models to be independently configured with solvers that can adapt better to the individual subsystems [21]. In this way, it is more viable to account for the wheel-rail creepages produced by acceleration and electrodynamic braking events [40]. As dynamic braking (DB) and acceleration are controlled by the traction locomotive components that form a complex mechatronic system, there have been recent co-simulation approaches to model the vehicle multibody dynamics using one solver whilst calculating the locomotive traction subsystem behaviour in parallel using an independent solver. In addition to rail damage, co-simulation of different locomotive subsystems opens new research possibilities, for instance studying innovative condition monitoring approaches to detect faults that affect several locomotive subsystems. A good example is detecting wheel polygonisation or wheel flats from the electrical signals in the locomotive traction system [41]. Additionally, it allows setting up simulations that consider not only the effects of a single locomotive, but also the influence of full train consists with several locomotives and wagons [22].

As more railway regulatory organisations continue to acknowledge the use of MBS for virtual homologations and the development of complex co-simulation studies, researchers at the Centre for Railway Engineering have proposed the Locomotive Model Acceptance Procedure (LMAP) to incorporate best practices as used in Australia in the development of locomotive MBS [15]. This section exemplifies a locomotive model acceptance procedure using an AC locomotive as an example. The locomotive is modelled as a 134-t vehicle with a Co–Co wheel arrangement, in which the car body is coupled to two bogies, each one with three independent wheelsets driven by one traction motor each. The LMAP comprises three stages that cover the locomotive model creation and debugging, dynamic tests included in railway standards, and complementary tests related to the traction system. Each stage in turn requires tests with specific procedures and acceptance criteria. The process summarised in this section is based on [13] and [15], exemplified with a recent project developed at the Centre for Railway Engineering. Figure 3 summarises the proposed LMAP.

The locomotive multibody model is usually built with lumped masses that interact through stiffness and damping coupling elements. Depending on the approach, the masses of bodies like traction motors are distributed into the nearby bodies to facilitate the model creation. Since it is in general difficult to get all data needed to build a good simulation model, the model developer should pay special attention in defining the input parameters for the creation of the models [42]. A good starting point is the locomotive datasheet, were information about the maximum operating speed, vehicle mass, vehicle wheelbase, distance between wheelsets and other general dimensions can be found. In this way, a general diagram that presents the bodies and the distances within them can be defined as shown in Fig. 4. When the locomotive equipment, including for example the engine, generator, fuel tank, compressor and other equipment masses and locations in the vehicle are available, it is possible to create a locomotive body representation using CAD, accounting for the equipment masses and distribution to facilitate the estimation of the car body centre of mass and rotational inertias. Similarly, using the drawings of the bogie frame, wheelsets, and bolster, it is possible to trace components general shape to obtain a three-dimensional representation that allows a better estimation of the centres of mass and rotational inertias.

Following the definition of the bodies that will form the multibody model, the rotational inertias, centres of mass and general dimensions and their degrees of freedom, the locomotive bodies and their properties are then set up in the multibody software environment. Subsequently, the primary and secondary suspension components and their characteristics are identified. Most of the MBS commercial software packages have various methods for factoring the stiffness, damping and friction nonlinearities that govern the behaviour of the suspension components, including general models for coil springs, buffers, airbags, friction blocks, bump stops etc. For instance, when using GENSYS MBS software [42] the nonlinear behaviour of a stiffness coupling between two bodies can be formulated through a piecewise function of displacement versus forces. Figure 5 shows a general view of the AC locomotive model implemented in GENSYS. Figure 6 presents a zoom in on the AC locomotive semi-steering bogie model highlighting the primary and secondary suspension components. The wheel-rail profiles that will be used in the subsequent test are also defined at this moment of the model development based on the requirements provided in AS7509:2017 Rolling Stock Dynamic Behaviour [28].

The LMAP Stage 1 process aims at guaranteeing the model was correctly formulated inside the multibody software environment and that it generates stable and reasonable results for static and dynamic scenarios with a specific solver and time step. This stage includes syntax error verification, visual model checks and quasistatic analyses to ensure the locomotive bodies and suspension components are deformed or displaced as expected. Figure 7 shows the graphic results of the vertical and lateral quasistatic analyses. For the vertical test, the car body is displaced 50 mm downwards, as shown in Fig. 7 left. The vertical displacement test is considered passed if the wheel-rail vertical forces increase in correct proportion to the total stiffness of the primary and secondary suspension systems. Similarly, for the lateral test, the car body is displaced 50 mm laterally. This test is considered passed if the bogie frame roll angle corresponds to the car body displacement direction.

For the modal test, the eigen frequencies and relative damping values are calculated for each mode found in the vehicle model [42]. The locomotive model passes the test when there are no negative damping values or high eigenvalues exceeding 5000 rad/s. Subsequently, for the time-stepping test, the locomotive model is run on an ideal tangent track at various speeds with fine and coarse time steps. The model passes the test when there are no unexpected displacements or rotations of bodies. The critical speed estimation test is conducted by running the locomotive model at a high speed on an ideal tangent track with a lateral track irregularity that triggers vehicle hunting motion. The vehicle speed is diminished at a constant rate until it stabilises. The hunting behaviour should disappear above 110% of the proposed operational speed for the model (rolling stock and track conditions for the model are defined in AS7509 [28]) to pass the test. Figure 8 shows an example of the critical speed estimation results for an AC locomotive. The simulation started at 300 km/h and diminished at a constant rate of -5 km/h per second. The top plot shows the lateral displacement of wheelsets while the bottom plot presents the vehicle speed. It is seen that, for this vehicle, the lateral movement of the wheelsets stabilised after 24.5 s corresponding to 175 km/h.

One of the possible mechanisms for validating the dynamics produced by the model is comparing the locomotive model wheelsets’ angle of attack (AoA) when negotiating a curve with AoA of field measurements. The wheelset AoA should have reasonable agreement for low and high-speed cases. The AoA comparison is considered as one of the field measurement validations required by dynamic behaviour standards such as AS7509 [28]. Figure 9 shows the AoA results of the wheelsets of a single three-axle bogie for simulated versus measured indicative data.

Other approaches for validation of simulation models by comparison with on-track tests can be found in [16, 17].

During this stage, the dynamic behaviour produced by the locomotive model is evaluated against the dynamics required by a particular standard. In the case of the Australian Standard AS7509 [28], for example, the verification includes hunting, twist, curve negotiation, isolated and cyclic track defects negotiation among others. For each test, the standard mentions the acceptance criteria in terms of wheel unloading, L/V forces ratio and car body vertical and lateral acceleration limits, as well as how the output signals should be postprocessed. Most of the tests of this stage are performed at the maximum locomotive design speed, with wheel and rail profiles in worn conditions and typical track irregularities superimposed over the test track features. The discrete track defect tests include flat humps that represent bridges or level crossings faults and curved dips that correspond to track subsidence caused by poor drainage. Cyclic track defect tests, on the other hand, are set at distances that trigger car body pitch, bounce and harmonic roll. Figure 10 shows the example results of the harmonic roll test for an AC locomotive. The lateral and vertical acceleration was obtained for the car body on top of the bogie centres. The acceleration signals were low-pass filtered with a fourth-order Butterworth filter with a cut-out frequency at 10 Hz.

The final stage involves the development and testing of the locomotive traction control system that represents the V-model Stage 2 shown in Fig. 2. The traction system for a diesel-electric locomotive includes a diesel engine, an AC or DC alternator, traction motors and an inverter in the case of using an AC traction motor. The system produces traction or braking forces, and it is controlled through throttle notch positions. The complete traction system mechatronic model should capture the main energy transformation processes that occur at each stage, including power delivery delays and energy losses. In the diesel engine output, for example, there are delays in response to throttle changes as a function of engine temperature, emission control requirements and other variables affecting the system power capabilities. Figure 11 shows an alternator voltage regulator, which is part of a diesel-generator plant, used for the modelling of an AC locomotive traction system. Figure 12 shows the mechatronic approach used for modelling of a whole locomotive system in Gensys and MATLAB/Simulink software products [23].

Both acceleration and DB are tested for the whole range of notch positions. The equivalent drawbar force is input to the locomotive model through a sky-hook element attached to the vehicle car body to generate load to the vehicle traction system. The main testing categories involved in the LMAP stage 3 involve a variety of acceleration and braking scenarios [13], for example gradient starting, all-weather adhesion limits, stopping distances, gradient parking, among other tests. Figure 13 shows the results of traction effort (TE) and DB tests for the full spectrum of notch positions, allowing the correct verification of the traction mechatronic system model across the whole locomotive speed range.

Discussions about wheel-rail friction include the significance of wheel-rail friction, friction characterisation and friction models.

The study of the interaction between wheel and rail is essential to explain locomotive traction studies. Such interaction is numerically explained through the wheel-rail contact model. Therefore, current locomotive traction studies have been widely focused on the development of wheel-rail contact models to use in specialised rail vehicle multibody software products. This contact model used for traction studies is the most computationally expensive part/module in railway multibody simulations. The development of an accurate and fast wheel-rail contact model is a complex and still ongoing area of railway research.

The major parts of the wheel-rail contact model are the geometrical contact module, the normal contact module, and the tangential contact module. The geometrical module deals with wheel-rail profiles and detects the location and size of the contact patches. This module requires the most computational effort (around 70%) among all the modules. The normal contact module determines the shape and dimension of contact patches and defines the normal pressure distribution on the contact patch. Finally, the tangential contact module analyses the relationship between creepage and creep forces on the contact area. These three modules interacting simultaneously produce an accurate result. In general, the most accurate methods require interaction between all three modules which requires higher computational time. On the other hand, simplified methods neglect the interaction between these modules, thus needing less computational time but leading to less exact results. For traction studies, the contact model can be classified into three categories according to the computational complexity as shown in Table 3.

In the simplified contact model, the geometrical module and the normal module are ignored and a constant value for the contact patch is used. To solve the tangent task, the Polach model is implemented. The exclusion of the most computationally demanding geometrical module makes this approach relatively fast. Since the change in wheel-rail geometry is not included in the simplified model, such an approach cannot be considered accurate. To overcome some of the limitations of the simplified approach, the geometrical module is separated into online and offline segments in a hybrid approach. Those sections in a hybrid algorithm that have less influence on dynamic operation are assigned as the offline part [87] where the contact parameters are precomputed and stored in a look-up table. The parameters are retrieved by interpolation during the dynamic simulation. Therefore, a compound contact approach includes a hybrid method to solve the geometrical problem, Hertzian contact to solve the normal task in combination with the Polach theory or the Modified Fastsim to solve the tangent task. The compound approach holds better trade-offs between the computational speed and accuracy of the result, and this approach is widely accepted in traction studies. In the complex contact approach, the geometrical module uses online contact search algorithms to determine contact parameters using an iterative procedure during the dynamic simulation. Consequently, this approach requires huge computational time to solve the geometrical module. Many promising non-Hertzian algorithms [77‐80, 88] are being implemented which are gradually improving the computational speed as well as the accuracy, but further advancements in the algorithm and/or computational approach (e.g., parallel computing) are essential.

The Polach [60], Fastsim [76], lookup tables [83] and FaStrip [82] approaches are recommended for the V-model Stage 1 when considering traction studies (see Fig. 2). The Polach [60] and modified Fastsim [72] should be used for the V-model stage 2 (see Fig. 2) modelling exercises. The modified Fastsim can be used for locomotive/track damage studies in the V-model Stage 3 (see Fig. 2), but if the computational capabilities are available, then “the best option for performing wheel or track damage studies is to use the Exact theory developed by Kalker directly in the simulation procedure to solve the contact mechanics task at the wheel-rail interface” [88].

In general, wear is a process in which the surface layers of solid bodies are ruptured because of the mechanical action against each other [43]. In wheel-rail wear, the bodies in action are wheel and rail and the mechanical action appears in the form of frictional force. The quantitative definition of wheel-rail wear is the displacement of material from the wheel-rail contact surface. The wear is an unavoidable phenomenon when a train is in motion. This means the amount of wear should also be assessed in locomotive traction studies.

One of the major impacts that need to be considered due to wear in the traction studies is the change of the shape of the contacting surfaces. The change in the wheel-rail profile redistributes the contact areas on wheel and rail. However, the amount of wear is negligible in each iteration and modifying the shape of the wheel-rail profiles in every iteration involves a high computational effort. Thus, after a fixed number of iterations, the contact patch shape evolution due to the accumulated wear (i.e., wear depth) is determined and the contact surface is modified.

Many researchers around the world are working on wheel-rail wear. As a result, many wheel-rail wear models are available, including the models developed by KTH Royal Institute of Technology [89, 90], the University of Sheffield [91‐93], British Rail Research [94, 95], the University of Florence [96‐98], the Technical University of Budapest [99], and the University of Coimbra [100]. The wheel-rail wear is dominantly dependent on the friction at the wheel-rail interface. Since the interface is an open system, the friction condition is influenced by many external factors. One of the factors is a third-body interfacial layer at the wheel-rail contact interface (e.g., iron-oxide wear debris, oil spill, leaves, friction modifiers). However, most of the wear models are developed based on the limited frictional condition variation as shown in Table 4. The wear models may not be suitable to implement under different friction conditions that is not considered during development and validation. Another factor is the material used for the wheel and rail. Most of the wear models are developed for specific wheel-rail material combinations and using such models for different materials may not be applicable because of the change in wear rate. Although a wear model for each scenario would be ideal, it will be practically challenging to completely incorporate them into multibody simulations. Therefore, correction factors have been used to adjust the existing wear model according to the new scenarios [101].

The simplified concepts of parallel computing and serial computing are shown in Fig. 15. The essence of parallel computing is to use multiple computing units or computer cores to process multiple computing tasks concurrently. The primary benefit of parallel computing is faster computing speed. For a computing problem to be solved using parallel computing, the problem should be able to be partitioned into a number of independent computing tasks. The independence is often temporary, which means for certain steps of each calculation cycle, these computing tasks do not need information exchanges.

After the participation of the computing problem, parallel computing techniques are required to communicate and coordinate the calculation process. Two different communication models are often used to achieve the communications and coordination of the partitioned computing jobs. These two models are called the distributed memory model and the shared memory model. In the former, each computing core has its own share of memory, whilst in the latter, all computing cores share the same memory field. A good example of the distributed memory model is the Message Passing Interface (MPI) technique whilst Open Multi-Processing (OpenMP) and POSIX threads (pthreads) techniques are good examples of the shared memory model. Regardless of which parallel computing model is used, the coordination and communication tasks are usually assigned to a specific computing core that is often called the master core while all other parallelised computer cores are called the slave cores.

The first case uses a 2D (longitudinal and vertical) locomotive model that considers six wheelsets, two bogie frames and a car body. The locomotive model was used with MPI to study locomotive traction and train dynamics as shown in Fig. 16. Assume that there are \(n\) locomotives in the train, then \(n+1\) computer cores are recommended to be used. In each calculation step, the train model sends out a series of parameters to individual locomotive models. Among these parameters, one of the most important is the reference value of traction force. This reference is the control command that was generated for train driving from the train model. After receiving the reference values for traction forces, each locomotive model will then proceed to calculate locomotive traction forces with the consideration of wheel-rail adhesion control and various forces in wheel-rail contact. The final traction force is the sum or traction forces generated from all wheel-rail contacts of the locomotive model. Having determined the locomotive traction force, it is then sent to the train model to proceed to the next step of simulations.

The second case uses a 3D locomotive model developed in GENSYS to study locomotive traction and train dynamics. As shown in Fig. 17, a train dynamics model and a 3D GENSYS locomotive model was connected via a co-simulation interface based on the TCP/IP technique [19]. In this case, the train dynamics model was used as the client whilst GENSYS was used as the host. The co-simulation process is similar to the one in Fig. 16 but has used the TCP/IP technique. Regarding the parallel computing, it is enabled by OpenMP instead of MPI. The utilisation of TCP/IP or OpenMP is due to various reasons: (1) GENSYS is a commercial software package which does not allow recompilation to facilitate the MPI technique, hence the TCP/IP technique was used as it does not require a recompilation of the GENSYS software. (2) Once the TCP/IP technique is used, synchronisation is not required any more from the parallel computing technique as TCP/IP has the capability of synchronising the computing processes of the train dynamics model and GENSYS. (3) When synchronisation is not required from OpenMP, the programming task is very simple to achieve parallel computing and, in this case, OpenMP’s main task is the initiate the simulations of locomotive models on individual computer cores with no changes required for the codes of GENSYS and only five lines required for the codes of the train model; and (4) MPI allows easy information exchange among multiple independent computing tasks whilst, to achieve the same function using OpenMP, is more complicated. The combination of TCP/IP and OpenMP is therefore a good choice as the information change process in this case is taken care by the TCP/IP technique.

Three simulations were conducted to study locomotive traction forces. All simulations have the same track and train configurations. The simulations are: (1) conventional longitudinal train dynamics (LTD) [22] simulations without the consideration of detailed locomotive models as shown in Figs. 16 and 17; (2) parallel computing using MPI and a 2D locomotive model as shown in Fig. 16; and (3) parallel computing using OpenMP and a 3D locomotive model as shown in Fig. 17. The wheel-rail slip limit for models that considered wheel-rail contact was set to be 7%; all simulations have the maximum adhesion limit of 0.35.

The simulated traction forces for the leading locomotive of the train are presented in Fig. 18. Two significant observations were made from the traction force results. The first observation is that the maximum traction forces simulated by 2D and 3D locomotive models on the first tangent track section (3.15–3.30 km) were evidently smaller than that by the conventional LTD model. Traction forces from the simulations at these two track sections are listed in Table 5. In these results, the adhesion force utilisation was calculated as the ratio of traction force to the maximum adhesion force available. The equivalent adhesion coefficient was calculated as the ratio of traction force to the overall gravitational force due to weight of the locomotive.

The results in Fig. 18 and Table 5 indicate that the traditional traction model (LTD model) assumes that the maximum traction force that can be achieved equals the available maximum adhesion force. However, the simulated maximum traction forces obtained by the advanced models (2D and 3D models) are about 10%–5% lower than for the traditional model (LTD).

The second and more interesting observation that was made from the results in Fig. 18 is that the traction force simulated by the 3D locomotive model slightly increased during curve negotiation. This seems to negate the normal impression that traction forces reduce during curving [102]. Further investigation into the wheel-rail contact has indicated one explanation which shows that, during curve negotiation, two contact points have been detected on both right and left wheels of the first wheelset at different times. Furthermore, evident longitudinal creep forces were generated from the second contact point on the right wheel. The total creep forces generated by the first contact point were decreasing during curve negotiation. However, at the same time, additional creep forces were generated from the second contact point. The force reductions at the first contact points were smaller than the forces generated from the second contact points. This is the reason why the traction forces increased during curve negotiation in Fig. 18.

A hybrid train operation is considered to provide greater energy efficiency when some of the energy produced by the application of dynamic braking can be reutilised which would otherwise be wasted as heat energy. The idea of using dual-mode locomotives has been considered since the 1970s, with one feasibility study being performed by the US Federal Railroad Administration (FRA) in 1979 [106]. Considering heavy haul train operational scenarios in Australia, heavy haul trains can accommodate a large energy storage system (ESS) which has the potential to allow the use of hybrid technology efficiently. One of the challenges is to operate the ESS at a suitable operational state that allows rapid charging which ensures reliable performances as well as longevity of the equipment. Nelson [107] reported that the battery system should be operated at a nominal 50% state of charge (SoC) level with the hybrid operations limited between 20% and 80% of the full SoC level (Fig. 19).

Hybrid locomotives are assumed as having similar weights to diesel locomotives considering the fact that the battery system mass including inverter should be equal to 50–60 t [108, 109]. An example of design parameters of a hybrid locomotive based on mechanical parameters of a diesel-electric locomotive is shown in Table 6.

Advancements in the modelling techniques allow an investigation of the progressive SoC level of a hybrid locomotive at the design stage using techniques developed in [20, 23, 110]. Three simulation environments would need to be used to represent full train operation embedded in the control algorithm of a locomotive. Co-simulation approaches between train and Multibody Simulation (MBS) of the locomotive, and co-simulation between MBS and another program containing a code generating algorithm of the electrical part has been proposed to evaluate the SoC of hybrid locomotives in [104, 105]. Modelling of traction system requires several blocks of models including a model of traction motors, inverters and control algorithms, inverter DC bus and alternator including excitation and rectifier subsystems, and details are available in [111]. The ESS sub-block for the hybrid locomotive uses a lookup table to generate voltage input based on speed and notch positions as discussed in [111]. The voltage parameter is then converted to current and then to power. The power parameter is converted to energy which is integrated by a numerical solver to evaluate the SoC of the hybrid system at every timestep of simulation.

The technique of assessing SoC based on train and locomotive operation leads to the development of a method to further post-process the results of simulations to allow assessing locomotive design outcomes such as the ESS control algorithm and the volume and mass of a battery system. An example method is presented in Fig. 20 to demonstrate the applicability of the advanced simulation technique in assessing a locomotive design. As the method suggests, the first step involves obtaining the data of the railway network including track geometry, train configuration, and operational modes. Step 2 involves modelling the train configuration on the specific track route to obtain the speed and notch information necessary to implement in the MBS model. Step 3 includes developing the MBS model representing the mechanical part of the locomotive that takes inputs from both train simulation and the code generated program. The results from the MBS can be analysed along with the inputs from the electrical model to provide locomotive design outcomes. The electrical part can be modified to suit the proposed battery system which would allow better approximation of the mass and space requirement of the battery system. The information on driving strategy would also be useful to assess the ESS control algorithm.

Wayside charging stations have been considered as an option for hybrid locomotives where continuous electrification on the rail network is not available. The digital twin technique can be applied to assess the location of wayside stations, and example plots (Figs. 21, 22) are added from reference [105]. One case of simulation showed that SoC can be reduced from 100% to 60% on the first leg of a journey in the loaded condition which represents a mine to port heavy haul freight transport condition. As 60% SoC was found sufficient for the return leg of the journey between port and mine in the empty condition, one proposal was to setup a charging station at the mine (case 2, Fig. 22). As an alternative, two further simulation cases were performed (Fig. 22) and it was observed that it would also be possible to install the charging station at the port. Thus, it follows that the simulation tool has good prospects to analyse the different requirements of train and hybrid locomotive operations.