The experiments described in this post involved a large number of colleagues in the development and mounting of instrumentation, the sorting of the data and the analysis and interpretation of the results. Dr Andrew Quinn, Dr Dave Soper, Dr Martin Gallagher and Dr Stefanie Gillmeier of the University of Birmingham, and Mr Terry Johnson of RSSB deserve special mention. The assistance of staff at Network Rail who operated the New Measurement Train is also gratefully acknowledged.

In the previous post I discussed the pressure transients measured on the NMT due to passing trains on the WCML. In this post, I will consider the pressure transients that were measured as the NMT passed through a number of tunnels on that route. These transients can be the source of significant aural discomfort for passengers. The five tunnels that are considered are outlined in Table 1 below. They are arranged in order of length, from the shortest Preston Brook at 71m to the longest, Kilsby, at 2.2km. All are double track tunnels that cater for either fast trains only, or for mixed traffic. They were all built in the 1840s when the line was constructed and are small by today’s standards. Two of them have airshafts.

The pressures that will be presented were measured on the side of the train, although pressure varied little around the circumference in most cases. They were measured against a reference pressure in a leaky reservoir on the train itself, and a small correction was required to relate them to atmospheric pressure. The main unknown was the actual configuration of the NMT at the time the measurements were made. The NMT could run as two power cars plus anything between 3 and 5 coaches between them, giving train lengths of 115m to 161m. The pressure measurement points were 14m from the end of the power car. Also, as these were operational runs, the train speed could vary by several m/s during any one run. Thus the velocities that are given below must only be regarded indicative.

Table 1 Tunnel characteristics

We begin by considering the shortest tunnel – Preston Brook at 71m. Figure 1 shows the pressure transients measured on the NMT for the front car and the rear car, at similar speeds. Both pressure records were obtained from runs in the down (London to Glasgow) direction, as indeed is the case for all the results described below.  Consider first the case of the front car. The nose of the train enters the tunnel around 0.3s earlier than the measuring point (distance from nose / train speed) and sets up a compression (positive pressure) wave. This takes approximately 0.2s to travel along the tunnel (tunnel length / speed of sound) and then reflects as an expansion (negative pressure) wave, taking another 0.2s to reach the initial end. Thus the measurement point experiences a short period of positive pressure for about 0.1s, before the expansion wave passes over it causing a dramatic fall in pressure. The pressure then oscillates with a period of around 0.4s to 0.5s as the wave passes back and forth along the tunnel. The measurement point leaves the tunnel before the rear of the train enters. 

The situation is somewhat different when the measurements are made at the rear of the train. By the time the measurement point enters the tunnel the initial wave system caused by the train nose will have traversed the tunnel four or five times, and friction will have attenuated its magnitude significantly.  From the pressure data shown, it can be seen that the measurement point experiences a small period of positive pressure, before the expansion wave from the rear of the train passes over it causing a sharp drop in pressure. This tail wave then dominates, reflecting from the far end as a compression wave, with again an overall period of about 0.4s to 0.5s.  The data from Preston Brook thus shows the pressure variations due to single waves – either the nose wave or the tail wave. 

• 

• 

Figure 1 Preston Brook tunnel

Kensall Green tunnel is 290m long, and the train speeds are in general somewhat lower than elsewhere, as trains are accelerating away from Euston. Again, consider the pressures on the leading car first. The front of the train enters the tunnel around 0.25s before the measurement point enters. The nose compression wave will take about 1.8s to travel to the far end of the tunnel and back again. The measurement point first experiences the increase in pressure behind the initial wave, which grows steadily due to wall friction slowing the flow around the train. At around 2s, the expansion wave passes over it and the pressure falls rapidly. This wave reflects back along the tunnel and passes over the measurement point again as an expansion wave at about 3.8s. Now the tail of the train will enter the tunnel somewhere between 3 and 4s after the nose depending on the (unknown) train length. Its effect can be seen from the fact that the regular oscillations up to around 5s are disrupted by interaction between the nose and tail wave systems. For the rear measurement point, the pressure drops initially as the tail expansion wave passes over the measurement point. Thereafter the oscillation pattern is a complex interaction between nose and tail wave systems.

• 

• 

Figure 2 Kensall Green tunnel

Northchurch tunnel (figure 3) is a little longer than Kensall green and the train speeds are somewhat higher. The main difference however is that this tunnel has an airshaft at its mid point. For the front car measurements the nose enters the tunnel at a time of about 1.6s. Although the main nose wave will take around 2s to return to the start of the tunnel at a time of 3.5 to 4s, a smaller reflected wave from the airshaft can be expected to pass over the measurement point before that. This can be seen to be the case, with the pressure at the front measuring point showing a sharp fall as the expansion wave from the airshaft passes over it at a time of around 2.2s. The main expansion wave arrives at about 3.7s. The train tail enters the tunnel two or three seconds after the train nose (depending on the unknown train length) and the resulting expansion wave and its reflections at the airshaft and far end of the tunnel add to the complications. For the rear measurement point, the expansion wave due to the tail passes over it first, and then it is subject to the complex interactions between the reflecting waves. Thus the presence of an airshaft can be seen to considerably complicate the measured pressures.

• 

• 

Figure 3 Northchurch tunnel

Shugborough tunnel passes underneath the former Shugborough estate of the Earl of Lichfield, and its portals are suitably grand (see figure 4). More prosaically, Figure 5 shows the pressure traces for the 700m long Shugborough tunnel, which has no air shafts. In some ways this is the simplest of all the traces. The front measurement point shows the initial compression wave and the gradual increase in pressure due to friction. The initial wave takes about 4s to return to the entrance, whilst the tail of the train enters the tunnel between 3 and 4s after the nose and its expansion wave will pass over the measurement point at about the same time as the reflected nose wave. The sharp drop caused by these waves at about 5s is clear. Thereafter the two wave systems can be seen to be broadly in phase and pass backwards and forwards along the tunnel producing complex peaks and troughs of pressure. The rear measurement point experiences an initial fall in pressure due to the tail expansion wave, and thereafter is subject to the pressure distributions of the complex interacting waves. 

• 

• 

Figure 5 Shugborough tunnel

Figure 6 Western portal of Shugborough tunnel

Kilsby tunnel (figure 6) is quite a fascinating construction. When it was built it was the longest tunnel in the world, at 2.225km and has numerous airshafts (table 1). The largest of these, at the 1/3 and 2/3 points are vast, and more like caverns than airshafts. A picture of one of the surface structures for these airshafts is shown in figure 7.   In addition there are 10 other open shafts distributed along the tunnel. Recent investigative work by NR has revealed that there are a number of other blind shafts or pumping shafts along the line of the tunnel, and some someway off the line, whose position cannot be precisely determined. 

In aerodynamic terms the large airshafts effectively split the tunnel into three, and this is very clear from the pressure records for both the front and rear measurement points in figure 6. Interestingly, the two measurement points on the train may well be in different sections of the tunnel at any one time, and thus subject to completely different pressure wave systems. As with Northchurch the measurements record multiple reflections from the airshafts, and the presence of three or four airshafts in each section results in highly complex flow patterns.

• 

• 

Figure 6 Kilsby tunnel

Figure 7 Kilsby tunnel airshaft

Finally consider the rear car results from Shugborough for a range of train speed (figure 8). The 40.2m/s speed data has already been shown in figure 4 and shows the expected pattern of a steep initial drop due to the passage of the train tail expansion wave and then a series of interacting wave reflections. Similarly the lowest speed 28.2m/s data show the expected pressure wave oscillations. However the mid-speed range 32.3m/s data shows no such oscillations here. It is likely in this case that the initial nose wave returns to the tunnel entrance as the train tale enters and is cancelled out by the tail expansion wave. Such an effect is of course critically dependent upon the precise values of train speed (which are not well specified for these measurements) and tunnel length, but is interesting nonetheless.

Figure 8 Shugborough tunnel pressures for several train speeds.

The experiments described in this post involved a large number of colleagues in the development and mounting of instrumentation, the sorting of the data and the analysis and interpretation of the results. Dr Andrew Quinn, Dr Dave Soper, Dr Martin Gallagher and Dr Stefanie Gillmeier of the University of Birmingham, and Mr Terry Johnson of RSSB deserve special mention. The assistance of staff at Network Rail who operated the New Measurement Train is also gratefully acknowledged.

Between 2011 and 2016, as part of a large project sponsored by the UK Engineering and Physical Sciences Research Council, colleagues at the University of Birmingham made full-scale aerodynamic measurements on the Network Rail New Measurement Train (NMT). This is a Class 43 train, with two power cars and a variable number of coaches, which is used to assess track conditions on main line railways in the UK (figure 1), on a regular two-week cycle. The aerodynamic measurements were mainly directed at measuring crosswind forces, and these results have been reported elsewhere. However during its travels around the country the NMT also detected the pressure transients caused by other passing trains, and by its own passage through tunnels. Whilst this data is of itself reliable and can be located confidently in time and space, it is not always easy to get the precise experimental conditions associated with each set of pressure transients – for example the precise NMT configuration in terms of number of coaches; the wind conditions; type and speed of passing trains etc.. Thus these results are not fully adequate for publication. Nonetheless they are of some interest if properly interpreted, and thus this blog post will present some of these results for open-air pressure transients, and the next will present some results for tunnel transients. The former are important as pressure transients caused by passing trains can cause trains to be suddenly displaced laterally causing passenger discomfort, and can also cause repeated loading on trains and trackside infrastructure, which can contribute to fatigue failure of components or structures. Tunnel pressure transients can be a source of aural discomfort to passengers, particularly in narrow tunnels – and indeed there are locations in the UK where aerodynamic speed limits have been imposed on tunnels.

Figure 1. The New Measurement Train

The results that will be presented were all obtained as the NMT passed up and down the West Coast Main Line between London and Glasgow. This is a 200km/h line, with both four-track sections (two for fast trains and two for slow trains) and two-track sections. It has branches to Birmingham, Manchester and Liverpool. In the four-track sections the NMT always travelled on the fast lines. The services that use the line are as follows. 

• 200 km/h services with limited stops between London, Birmingham, Liverpool, Manchester and Glasgow, using 9 or 11 car Class 390 Pendolino tilting trains (figure 2a)
• 200 km/h services that connect a range of towns and cities across the country using double unit Class 220 4 car Voyager trains (non-tilting) or Class 221 4 or 5 car Super Voyagers (tilting) (figure 2b), in 8 or 9 car formation. Irregular 4 or 5 coach Class 220/221 units also operate over sections of the WCML route.
• Semi-fast and commuter services operated mainly by Class 319 trains (figure 2c) south of Milton Keynes and by Class 350 trains along the whole line (figure 2d). Both of these run in single (4 car) unit or double (8 car) unit configurations. Class 350s can travel at 175 km/h and Class 319s at 160km/h. 
• A variety of freight services hauled by both electric and diesel locomotives. Perhaps the most common locomotive in use is the Class 66 (figure 2e)

• 

• 

• 

• 

• 

Figure 2 Train types on the WCML

Typical pressure transients measured on the side of the NMT for Class 390s, Class 221s, Class 350s and freight trains are shown in figure 3. It can be seen that 

• all types of train show a large positive / negative pressure transient as the nose of the passing train passes the measuring point and the passenger trains also show a negative / positive tail peak;
• no tail peak can be observed for freight trains; 
• a large transient can be observed at the gap in the centre of the double unit trains;
• between the peaks there is a small negative pressure, and the passage of individual carriages can also be discerned on the pressure traces. 

The time between the nose and tail transients can be used to determine the speed of the oncoming train, if the type of train and its length can confidently be specified. This was in general only the case for the Class 390 trains as it was usually not possible to distinguish between the different types of four and five coach trains. The results shown for the Class 350 in figure 3 are one of the few datasets where the train type could be confidently determined. 

• 

• 

• 

• 

Figure 3 Train passing pressure transient types

As noted above, it is possible to determine the speed of Class 390 Pendolinos passing the NMT. This allows the dimensionless pressure coefficient to be determined (peak to peak pressure / (0.5 x density of air x velocity of approaching train squared)), which enables the effect of velocity on the peak-to-peak pressures to be removed in a consistent fashion.  Pressure coefficient is plotted against train separation in figure 4 below. The train separation is calculated from the track spacing and the train geometry. Here again the data is less than ideal, and it was not possible to find accurate track spacings easily from NR databases. They were thus obtained from large scale digital OS maps, and are only accurate to within about 10cm. It can be seen that, as expected, the pressure coefficients fall with train separation. Perhaps the most notable point is the large variability of the data, which reflects both the uncertainties in track spacing described above, the effects of tilt and curvature and other operational variables. This level of variability is something that needs to be appreciated by both physical and computational models when assessing the engineering significance of their results. 

Two other sets of data are shown on figure 4. The first is the pressure on trackside hoardings passed by a Class 390, measured in moving model experiments. These hoardings are about half train height, so one would expect some pressure relief as the disturbed flow passes over the hoarding, and indeed the pressure coefficients lie at the bottom end of the NMT measurements.  One data point is also shown for the pressure coefficient measured in free air as a Pendolino passes. This can be seen to be about half the value of the pressure coefficients on the NMT at the same spacing.

Figure 4 Effect of train spacing on peak-to-peak pressure coefficients

Despite not being able to determine train speeds for trains other than the class 390, the pressure data measured on the train is still reliable for all train types, and the overall type of train can be identified. Thus the overall distribution of peak-to-peak pressures can be determined. This is shown in figure 5 for Class 390, 4 or 5 car multiple units and freights trains, and in figure 6 for all train types. The lower cut-off magnitude for a pressure transient to be included was 400 Pa. It can be seen that the Class 390 transients, have a distribution from 400 to 1220 Pa, with the peak being between 1000 and 1200 Pa. The freight train distribution is from 400 to 1000 Pa, with the peak in the 600 to 800 Pa interval, reflecting the fact that, although freight locomotives can be expected to be aerodynamically blunt, they move relatively slowly and the absolute transient magnitude is somewhat less than for the express passenger trains. The four- or five-coach multiple units have a very broad distribution of peak-to-peak values, and the maximum pressure transient values experienced by the NMT are caused by such trains. The maximum peak-to-peak pressures that were measured in the trials (with values of 1449 Pa and 1498 Pa) were both identified as being caused by passing blunt nosed Class 350 units travelling at 46.7 and 47.7 m/s on the smallest centre-to-centre track spacing of 3.2m. These values both exceeded the standard value of 1444 Pa. The equivalent pressure coefficients in both cases were 1.10, somewhat higher than the values shown on figure 4.

This distribution of peaks is of course very specific to the route under consideration and the services that operate on it. However it does suggest that, for a mixed traffic railway such as the WCML, the range of pressure transient loadings on trains themselves is very large, and if any sort of fatigue loading calculation is required, then a suitable distribution such as that shown in figure 6 needs to be determined to give the required loading information. If maximum loads are required, then these are likely to result from passings by higher speed but aerodynamically blunt trains.

Figure 5 Distribution of pressure transients experienced by the NMT by passing train type

Figure 6 Distribution of pressure transients experienced by the NMT for all train types

• 

• 

 Introduction

The book “Train Aerodynamics – Fundamentals and Applications” (hereafter referred to as TAFA) was published in early 2019, but in reality took no account of any material published after June 2018. There has however been a significant number of studies published in the second half of 2018 and 2019, and it seemed worthwhile to try to collate these in some way, and this blog post attempts to do this. Selfishly such a collation might help for any second edition of TAFA that is produced (if the sales warrant it!) but more generally it is hoped that it may prove useful to all those involved in Train Aerodynamics in one way or another in signposting ongoing work around the world.

It should be emphasized at the outset that this collation cannot properly be described as a review. A review (as I have told my graduate students for the last three decades) needs some degree of synthesis of the various reports and papers discussed. This of course requires a number of papers addressing the same issue to be available to synthesise. Looking at papers from a short time period that cover a wide range of subject matter, this is not really possible, so what follows is essentially a brief description of the work that has been carried out in 2018 and 2019, with a few interpretive comments. 

We consider the various publications roughly in the order of the Applications described in Part 2 of TAFA – train drag, pressure transient loads, slipstream loads, ballast flight, OHL issues, crosswind studies, tunnel aerodynamics and emerging issues. A brief section is included on train and tunnel ventilation that was not considered in TAFA. Some concluding reflections are included at the end of the post.

In the text, published references are linked directly to their DOI, rather than to a reference list. Those references with no DOI (recent conference publications in the main) are given in a short list at the end of the post. A full reference list can be found here. 

Train drag

One of the major issues in both experimental and computational assessments of train drag is the simulation of the ground, as the nature of this simulation can have a significant effect on the measured force. Niu et al (2018b) present some CFD analyses that usefully address this issue for a number of ground and ballast representations. Overall they show that the effect on measured drag coefficients at zero yaw angle is small (of the order of 2 to 3%) and probably not significant in comparison with the large variation in drag measured at full scale, and in different types of CFD and physical model simulations. Nonetheless the results can give some guidance for the setting up of physical model tests and CFD trials. 

There have been a small number of train drag investigations looking at the effect of modifications to different parts of the train on overall drag – nose length, inter-car gaps, bogie position, roof equipment and for freight trains, container spacing. They will be considered in turn below. 

Chen et al (2019a), using IDDES, looked at effect of nose lengths of between 5 and 10m on drag on a five-car train. The drag was shown, unsurprisingly, to decrease with nose length. Li et al (2018b) used the k-omega method to investigate the effect of inter car gap length on drag, and showed that gap lengths of less than 80mm full scale had no effect on drag. The k-epsilon CFD work of Gao et al (2019) looked at the effect of changing bogie position on the leading car of a three-car high-speed train. The results indicate that moving the front bogie back 2m from its normal position can reduce the overall drag of the front car by 10% and the overall three-car drag by 6.5%. These drag reductions will of course be a smaller proportion of the overall drag for full-length trains.

Tschepe et al (2019) investigated the drag of roof-mounted insulators through wind tunnel experiments. They observed considerable Reynolds number effects and effects of insulator position on the measured drag (which in aerodynamic terms were in the critical range), and showed that overall drag due to roof elements could be about 5% of overall train drag. Interestingly they found that soft insulators displayed a tendency to flutter, resulting in higher drag than rigid insulators.  

Maleki et al (2019) carried out a CFD investigation of the flow around containers using LES. There were major changes in the wake of containers as container spacing increased. Above a certain spacing wake closure occurs with high speed flows impinging on downstream container, resulting in increased drag. They also present potentially useful results for the optimisation of single and double stack container positions for low -drag. 

Pressure transients and loads

A small number of investigations have been reported of the pressure transients and loads caused by passing trains. Soper et al (2019) made full scale experiments on the transient pressure loads caused by high speed trains on acoustic barriers, and were able to determine the effect of variations in track distance and the nature of the overall trackside infrastructure on the measured loads. The CEN load correlations were found to represent the results well. 

Huang et al (2019) added to the small amount of data available on the loads caused by passing trains through an experimental and compressible CFD study of high speed Maglev trains passing each other, including detailed calculations of the transient pressure field around the train. 

Munoz-Paniagua and Garcia (2019) developed an optimization methodology for nose shape to optimize (i.e. minimize) the pressure transients. This involved a large number of CFD runs for different geometries that were used to train a genetic algorithm. It was shown, perhaps unsurprisingly, that nose length and bluntness were the most important parameters in the optimization.  They also considered the optimization of nose shape for cross wind performance, which will be discussed further below. 

Slipstream velocities and loads

In 2018 and 2019 a significant number of papers have been published that use (in the main) CFD IDDES to calculate trains slipstream and wakes and to look at specific flow effects. These all give a great deal of information concerning the micro-nature of the flow field that it is not always easy to interpret of to put into a bigger picture. They are useful however in helping to build up a picture of the complexity of even the idealised CFD flow fields around trains. As with drag investigations described above, these studies were aimed at assessing the effects of slipstreams of changes to different parts of the train – nose and tail, gaps between double units and bogies. 

Chen et al (2019a) looked at the effect of different nose / tail lengths on drag and lift, but also looked at the effect of the slipstream behavior along and behind the train. They found that the TSI slipstream velocities decreased with tail length, with the longitudinal trailing vortices becoming weaker. In a further paper they extended this work to look at the effect of changes in nose length on the slipstreams and wakes in crosswinds (Chen et al, 2019b). Not surprisingly they found that the effect of nose length on the overall flow field was complex and quite difficult to quantify.  

Li et al (2019b) looked at the effect of the gap in double unit trains on the development of the slipstream and the wake and showed that the main effect was to increase the boundary layer and wake velocities downstream of the gap. 

Two papers from the same group in Changsha (Wang et al, 2019 and Dong et al, 2019) describe aspects of an IDDES investigation of flow around bogies. The first looks at the effect of bogie fairings on the slipstream and wake and, as might be expected, shows that bogie fairings reduce the velocities in the boundary layer and the strength of the longitudinal vorticity in the wake.  The second studies the effects of simplifying the geometry of train bogies in CFD simulations. It shows that the effects are mainly felt in the underbody flow region rather than in the wider flow around the train, and offers some suggestions for appropriate degrees of geometric simplification. 

Finally the IDDES modelling of Wang et al (2018c) should be mentioned. This is a fundamental study of the effects of bogies on the slipstreams of high-speed trains. The study shows that the generation of the strong spanwise oscillation of the wake, observed especially in the presence of bogies is due to the amplification of a natural instability of the time-mean pair of counter-rotating vortices. 

Ballast flight

The work in this area has focused on two aspects – the nature of the flow field around the train, and the accumulation of snow in the bogie area.

With regard to the first, two useful studies have been reported that address the issue of the train underbody flow, which of course controls the flight of ballast. The first of these is the thesis by Jönsson (2016), which is somewhat outside the publication time range considered here, but is nonetheless worth mentioning. The author carried out extensive measurements using PIV to measure the flow field beneath 1/50^thscale model trains, with different underbody geometries and sleeper layouts. Comparisons were made with full scale and showed that the essential aspects of the underbody flow field could be reproduced. The tests also showed how important train underbody irregularities were in increasing velocities and thus the likelihood of ballast flight. The data was also used in a simple analytical framework similar to (and earlier than, so it takes academic priority!) that included in TAFA.  

Also with regard to the underbody flow, Paz et al (2019) consider the nature of ground simulation in CFD trials, and present a method for simulating track and ballast geometries in this region, using scanned profiles of real sleeper and ballast geometries. The results show significant differences between the simulation and the normal flat ground geometries, particularly close to the ballast where higher levels of turbulence were measured. Overall the methodology as set out is potentially of great use in simulations for train authorization purposes. 

Two CFD studies, both by the same group have also been carried out to address the related problem of snow accumulation around bogies – one using a discrete phase model and IDDES (Liu et al, 2018) and one using a discrete phase model and URANS (Wang et al, 2018b). A somewhat qualitative validation of the methodology against experiments was carried out. Similar results were obtained using both methods, but the URANS work used considerably less computer resource. Unsurprisingly snow was shown to accumulate in areas of low velocity in the bogie cavity. Overall the results give useful qualitative indications of those aspects of bogie design that could be altered to reduce snow accumulation.

Overhead and pantograph systems

Two interesting studies on the dynamics of pantograph and catenary systems were reported in 2018/2019. In the first, Li et al (2018a) describes DDES calculations of the flow around and forces on pantographs on a three car high speed train at yaw angles of 0, 20 and 30 degrees, thus representing a range of cross wind velocities. As might be expected, the flow around the pantographs becomes increasingly complex and turbulent as the yaw angle increases. The aerodynamic forces on the pantograph were found to oscillate around a mean value even at zero yaw, and as the yaw angle increased a range of different dominant frequencies appeared. Whilst these results are doubtless quite specific to the case being considered, and the simulation does not fully reproduce the range of turbulent fluctuations in the atmosphere, they do show the potential for high cross winds to excite a range of pantograph oscillations.  This is an area where further work would be of significant interest.

Secondly, Xie and Zhi (2019) report wind tunnel results for the dynamic behavior of catenary systems, including the effect of wire tension on the natural frequency and displacements of the contact wire in a range of cross wind conditions.  A large scale if somewhat crude simulation of the near ground atmospheric boundary layer was used. The authors discuss the possibility of resonant oscillations occurring between the train pantograph system and the overhead wire. Deflections of 6cm at mean wind speeds of 17m/s were measured. This is again a topic where further work would be useful – in particular a study of the interaction between pantographs and the overhead wires in high crosswinds would be very interesting and potentially very significant. 

Crosswind

The number of new studies on trains in crosswinds has increased significantly in recent years and shows no sign of slowing down. In 2018/19 results of studies have been published on wind conditions near the track, train aerodynamic forces in cyclonic winds, train aerodynamic forces in tornado winds, the effects of wind shelter and full-scale measurement of wheel unloading risk. These will be considered in turn below.  It is perhaps worth noting at this point however that many of the CFD calculations described below, even though they give a detailed description of unsteady flow fields, nonetheless do not fully simulate the turbulence structure of the oncoming atmospheric boundary layer. These results thus not fully represent reality, where the flow structures around the train can be expected to be disrupted by the oncoming turbulence. They are nonetheless useful in giving an overall impression of the flow field and the effects of different geometry changes for example. 

Zhang at al (2019b) used IDDES, calibrated against wind tunnel tests, to predict the flow speed up over embankments of varying geometry, to enable a rational siting  of warning anemometers. This work usefully adds to the data available for the wind speed up over railway embankments.  Hu et al (2019) also consider wind characteristics in terms of the nature of the wind relative to a moving vehicle. The analytical framework they have produced is more extensive and rigorous than those developed by earlier authors and has the potential to be used to generate realistic time series of velocity that could be used in  overturning calculations. 

A number of studies have been reported that enlarge the database of aerodynamic coefficients of trains in cross winds. Noguchia et al (2019) provide experimental and LES data for the crosswind forces and moments on a range of conventional train types on embankments.  Guo et al (2019) present the results of IDDES calculations that investigated the difference in cross wind pressures and forces for both a 6 car single unit and a 6 car double unit high-speed train. They also provide extensive discussions of the nature of the wake and the unsteady flow.  The effect of the gap between the two units was shown to have a significant effect on the crosswind forces on coaches in the centre of the formation, and also affected the primary frequencies of oscillation.  Lin et al (2019) report the results of a benchmark test to measure crosswind forces and moments with two different trains in three different wind tunnels, all of which were carried out to confirm to the CEN guidelines.  Systematic differences in results between nominally similar tests are observed, and seem to be associated with blockage and boundary layer effects in the wind tunnels. These results were presented in brief at a conference, and a full write up of the results should prove extremely interesting and is eagerly awaited by the author. 

The opimisation work of Munoz-Paniagua and  García (2019) has already been mentioned above. As with the pressure transients they found that nose length was the major factor determining the crosswind forces and moments.

The above studies were concerned with trains in cyclonic winds. A couple of studies have been reported where pressures on trains were measured using Tornado Vortex Generators – that of Bourriez et al (2019) with a moving model, and that of Cao et al (2019) with a stationary model. Both these studies have major scaling issues, where the train scale and tornado vortex scale do not match. Nonetheless they give interesting indicative results. Further developments can be expected in this field in the future. 

In view of the importance of reducing crosswind forces and moments, the literature describing the effect of wind barriers on trains in cross winds is surprisingly sparse. Recent studies have gone some way to remedy this – Mohebbi et al (2019), Niu et al (2018a), Hashmi et al (2019) used a variety of CFD techniques to investigate the effect of wind fences on train forces; Misu et al (2019) used equivalent wind tunnel tests; He et al (2018) and Flamand et al (2019) both considered train / bridge systems, where the cost to providing shelter on the train in increasing the loads on the bridge were considered. Wu et al (2019) looked at a case where wind shielding effects were undesirable, when a train runs in the wake of a bridge tower. Through the use of a simple low speed moving model rig to measure transient train forces, and a dynamic model of the wind / bridge / vehicle system they concluded that the shielding effect could have an adverse impact on both the running safety and riding comfort of the train.

Finally in terms of assessing safety and risk, two interesting full-scale experimental techniques have been developed. Wei et al (2018) report a method for measuring wheel unloading by making continuous measurement of the accelerations and displacements of the wheel set using relatively simple equipment, rather than the more conventional measurements made by instrumenting the wheels themselves (although such measurements are themselves quite innovative and difficult). The results from full-scale experiments on trains for the derailment and loading coefficients as they move in and out of the shelter provided by wind breaks are impressive and indicate that the methodology may be of some use in the future. Similarly, Lu et al (2019) use measurements from primary suspension. Good agreement with instrumented wheelset data was demonstrated provided a suitable calibration was carried out. 

Tunnels

In 2018 and 2019 a number of papers have been published on tunnel aerodynamics, addressing the issues of pressure transients and micro-pressure waves, tunnel velocities and structural loading. We will consider each of these issues in turn.

The work on pressure transients in tunnels has been carried out using both experimental and computational methods.  The experimental work of Heine et al (2019) using a moving model rig investigates the effect of wall cavities (for cross passages) on tunnel pressures. These cavities were installed to reduce the pressure load on interconnecting doors, but interestingly the results show that, by creating extra surfaces for pressure waves to reflect from, the pressure loading on the doors can actually increase under certain circumstances.  

Iliadis et al (2018) also used a moving model rig to look at pressure transients as blunt freight trains entered a tunnel, with measurements both on the tunnel wall and the train. The major point to emerge is that for certain freight train loading situations, gaps in the train formation can result in significant tunnel entry pressure transients, and the maximum pressure in the tunnel might not always be associated with the entry of the train nose as for passenger trains. 

Li et al (2019c) report a k-epsilon CFD investigation of the pressure waves in tunnels with variable cross section, but with sudden transitions between the sections. Unsurprisingly they show that a complex series of pressures results from these transitions, but on the whole the magnitudes of the pressures are reduced from the single area case.  Wang et al (2018) describe a similar investigation using k-omega CFD, but with gradually varying area rather than abrupt transitions. The pressure magnitudes are again reduced as would be expected. Both of these also showed that there was the potential for reducing the gradient of the initial pressure wave, which is the main parameter of importance in the generation of micro-pressure waves, using such approaches.

Micro-pressure waves were considered in more detail in the work of Saito (2019) who used the results of moving model rig experiments and a simple analytical formulation of tunnel entry pressures. He investigated the optimum area and length of unvented entrance hoods, and derived some useful design guidelines. 

Another method of reducing the strength of micro-pressure waves is to use ballast rather than slab track in tunnels. Fukuda et al (2019) report on the results of full-scale tests where experiments were measured before and after slab track in a tunnel was replaced by ballasted track. Significant reductions in pressure gradient were observed. A simple methodology for predicting these pressure gradient reductions has been derived. 

The propagation of micro-pressure waves themselves has been considered by Zhang et al (2018) and Zhang et al (2019a).  The former developed straightforward analytical models of pressure magnitudes around the tunnel exit portal, which were calibrated against experimental and CFD data.  The latter made measurements of the relatively small pressure amplitudes at the exit of a tunnel simulation using a moving model rig.

In some situations it is necessary to determine the velocity transients in tunnels as well as the pressure transients, particularly when loads on structures or people are required. These issues have been addressed by Jiang et al (2019) who carried out a URANS study of the slipstreams generated by different train types in a double track tunnel, providing a great deal of detailed information concerning the nature of the slipstream variation with height above the ground and distance along the tunnel. The work of Iliadis et al (2019) mentioned above for freight trains, also made such measurements. Kikuchi et al (2019) took a different approach and developed a simple calculation method based on unsteady incompressible flow for the flow over the roof of a train. This method takes into account the unsteady boundary layer development along the train roof, and allows the velocities to be determined that can be used to assess pantograph performance in confined tunnel situations. 

Finally a couple of papers have appeared that have addressed the issues of loads on trains and tunnel infrastructure directly. The first, by Lu et al (2018) looked at the fatigue loading caused by the pressure waves due to two high speed trains passing in a tunnel, and involved the use of unsteady k-epsilon CFD and a finite element model of the car body to determine the loads at specific points on the vehicle. Courtine at al (2019) carried out full scale and model scale experiments to analyse designs of “deflectors” to be placed around lorries in the Channel Tunnel, particularly with regard to the effect that they might have on the loads on the soft-sided trailers. 

Emerging Issues

The final chapter of TAFA briefly summarises a number of emerging issues. Of the ones discussed, there are two that have seen further work published in 2018/19 – evacuated tube transport, and snow drifting.

With regard to the former, two investigations have investigated the shock wave formation around such tube trains for very high speed vehicles using different methodologies. Zhou et al (2019) use the compressible Navier-Stokes equation for flow around an axisymmetric body in a tube, whilst Niu et al (2019) use a variety of different CFD methods, and also consider heating of the flow. Both reveal complex shock wave patterns, particularly behind the vehicle, and investigate the choked flow region in particular. Both sets of results serve to emphasise the complexity of the flows around such vehicles and indicate the formidable challenges that still remain before this type of transport can be implemented. 

The issue of wind blown sand around railways has been addressed by a group at Torino in Italy. The paper by Raffaele and Bruno (2018) presents an outline of a probabilistic method for assessing the accumulation of sand around railways, while the second by Bruno et al (2018) presents a thorough review of current literature and methodologies. Clearly it is not possible to easily summarise a review paper – suffice it to say that this provides and excellent starting point for those who wish to delve into this subject further. 

Ventilation

One area of study that was not included in TAFA, and in retrospect probably should have been, is that of ventilation of tunnels and stations, and the internal ventilation of the trains themselves. The need for such studies is becoming increasingly urgent as the poor air quality in underground stations and on trains becomes more and more apparent. Whilst no attempt will be made to present a full review of this subject here, it is worthwhile to set out the work that has been done in the 2018/19 period at least.

First of all let us consider the ventilation of tunnels and underground stations. Izadi et al (2019) present the results of a unsteady RANS calculation of velocities and pressures in a small number of stations connected by a single tunnel. Pressures and velocities were predicted for a variety of train operating scenarios using both a simple axisymmetric geometry and a more complex three-dimensional geometry. Various fan operating strategies were also discussed. Koc et al (2018) use a simpler one-dimensional approach, but applied to a more complex network of tunnels and stations. A methodology for using ANNs to predict pressures and velocities in complex situations is presented. Zarnaghsh et al (2019) use a finite volume technique to investigate the behavior of tunnel ventilation fans and in particular the interaction between the velocity field produced by trains and those produced by the fans themselves. It was shown that the passage of trains could result in the operating characteristics of the fans moving significantly away from their nominal operating point. 

Work on the ventilation of trains has also been reported. Abadi et al (2019) report a CFD k-omega study to improve the performance of a roof mounted ventilation system under crosswind conditions, by varying the inlet geometry. Li et al (2019a) report a DES similar study to investigate the performance of an ACU with roof inlets on a high speed train at different speeds.

Concluding reflections

Firstly some paper statistics are of interest, but note the figures that follow are somewhat arbitrary in many ways and reflect both the chosen publication dates and the sources consulted. There were 55 items accessed in total, all but one from between mid-2018 to the end of 2019.  Of these, 28 were from Chinese institutions, reflecting the fact that research there is driven by the rapid development of the Chinese high-speed rail network, with the rest being spread fairly evenly around Europe, the Middle East and the Pacific Rim. Of the Chinese papers, 18 were from the Central South University at Changsha, 7 from South West Jiaotong University at Chengdu and 3 from elsewhere (although there was some overlap of authorship). In terms of technique used, 30 of the papers were mainly based on CFD studies, although some included experimental verification of one sort or another.  Of the Chinese papers 21 were CFD studies, including 16 from Central South University. There was thus a significant imbalance between the use of CFD in China and elsewhere. 

The growing use of CFD techniques is thus notable, and in particular the IDDES technique seems to be becoming the most favoured. This growth is wholly understandable and will no doubt continue. Arising from this, it seems to the author that there is a growing need for a small set of freely available well documented validation cases, ideally from full scale experiments for a range of train types, that investigators can use routinely to prove their techniques. At the moment the validations used are somewhat ad hoc, and perhaps a more systematic approach would give greater confidence in the results, and also allow research papers to be reduced somewhat in length, as the details of the validation cases would not be required. It would be interesting to hear from CFD practitioners on such a possibility.

Finally it must be remembered that CFD simulations, in the same way as physical models, can only offer a simplified representation of the flow around full-scale trains, and need to be interpreted in this light. There is a tendency amongst some authors (and I name no names!) to quote numerical results to higher levels of accuracy than is either sensible or useful when the uncertainty of the full-scale situation is considered.  Just as with physical model tests, the role of the engineer in interpreting CFD results in terms of the reality of the operating railway is crucial. 

References without DOIs

Bourriez F, Soper D, Baker C and Sterling M (2019) Physical model measurement of tornado induced forces on trains, Proceedings of the 15^thInternational Conference on Wind Engineering, Beijing

Cao J, Cao S, Ge Y (2019) Experimental investigations of wind load distributions on a high-speed train under tornado-like vortices, Proceedings of the 15^thInternational Conference on Wind Engineering, Beijing

Courtine S, Aguinaga S, , Bouchet J-P, Brunel C (2019) Aerodynamic improvement of superstructure interacting with trucks carried in the Channel Tunnel, 15^thInternational Conference on Wind Engineering, Beijing

Flamand O, Gattas M, van Stuijvenberg J, te Morsche N, Jongstr B, Knapp G, Benthem J (2019) Improving the wind protection of High Speed trains whilst decreasing wind loads on the supporting bridge, 15^thInternational Conference on Wind Engineering, Beijing

Jönsson M (2016) Particle image velocimetry of the undercarriage flow of downscaled train models in a water-towing tank for the assessment of ballast flight, PhD Thesis, Technical University of Hamburg

Lin P, Brambilla E, Rocchi D, Tomasini G (2019) Measurement of the aerodynamic coefficients of high-speed railway vehicles: benchmark between different wind tunnels, 15^thInternational Conference on Wind Engineering, Beijing

Misu Y, Takeda S, Nagumo Y, Doi K  (2019) Estimation Effectiveness of Windbreak Fences at Rolling Moment on Leeward Rails, 15^thInternational Conference on Wind Engineering, Beijing

Raffaele L, Bruno L (2018) Probabilistic assessment of windblown sand accumulation around railways”, Proceedings Italian Conference on wind Engineering 

Wu J, Tang Q, Li X (2019) Shielding effect of bridge tower on aerodynamic characteristics and running safety of high-speed train with wind tunnel tests, 15^thInternational Conference on Wind Engineering, Beijing