Transient aerodynamic pressures and forces on ...

Original Article

Transient aerodynamic pressures and forces on trackside and overhead structures due to passing trains. Part 1: Model-scale experiments

Proc IMechE Part F: J Rail and Rapid Transit 2014, Vol 228(1) 37–70 ! IMechE 2012 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0954409712464859 pif.sagepub.com

Christopher Baker1, Sarah Jordan1, Timothy Gilbert1, Andrew Quinn1, Mark Sterling1, Terry Johnson2 and John Lane2

Abstract This is the first part of a two-part paper that describes the results of an experimental investigation to measure the aerodynamic pressure forces on structures in the vicinity of railway tracks. The investigations were carried out in order to obtain a fundamental understanding of the nature of the phenomenon and to obtain data for a variety of railway infrastructure geometries of particular relevance to the UK situation, in order to provide material for a National Annex to the relevant Eurocode. The experiments were carried out on the moving model TRAIN Rig, with models of three different sorts of trains with different nose types, and a variety of infrastructures types: vertical hoardings, overbridges, station canopies and trestle platforms. The transient loads that were measured had a characteristic form: a positive pressure peak followed by a negative pressure peak. In general the magnitudes of the two peaks were different, and varied with infrastructure type and position, as well as with train type. As would be expected, the more streamlined the train, the lower were the magnitudes of the pressure transients. A comparison of the experimental results was made with a variety of existing model- scale and full-scale data and a broad consistency was demonstrated, within the limits that the rather different experimental conditions in the various cases would allow. An analysis of the scaling of these pressure transients was carried out, and it was shown that whilst there was a reasonable coalescence around a theoretical formulation, the complexity of the flows involved meant that a general scaling formulation could not be achieved. Part 2 of this paper will consider the application of the results to the development of revised standards formulations. Keywords Train aerodynamics, aerodynamic pressures, bridges, hoardings, canopies, platforms Date received: 25 March 2012; accepted: 4 September 2012

Introduction It is well known that passing trains generate unsteady transient aerodynamic pressures. This is illustrated in Figure 1 (taken from BS EN 14067-41). Essentially as the train nose passes, the pressure rises rapidly above ambient pressure to a positive peak, then falls rapidly through ambient pressure to a negative (suction) peak, and then decays rather more slowly towards the ambient value. As the tail of the train passes, the process is reversed, with a negative peak followed by a positive peak. These pressure transients result in transient forces on trackside and overhead structures. The design requirements for these structures are contained in the Eurocode EN 1991-2:2003 ‘Traffic Loads on Bridges’.2 The data on which this code is based were originally developed by the European Railway Research Institute (ERRI) D189 committee3 and also forms the basis of a railway-specific European Committee for Standardisation standard.1 The data

are also used in the Technical Standards for Interoperability (TSI), which are being developed to allow trains to run across national boundaries within Europe. The infrastructure TSI4 directs users to this document, whereas the rolling stock TSI5 adopts rather different procedures for ensuring acceptable transient pressures from new trains, through the specification of transient pressure at specific points relative to the track. This paper reports a series of experiments that were carried out to investigate the nose pressure pulses of 1 Birmingham Centre for Railway Research and Education, University of Birmingham, UK 2 Rail Safety and Standards Board Ltd, London, UK

Corresponding author: Christopher Baker, Birmingham Centre for Railway Research and Education, University of Birmingham, Birmingham B15 2TT, UK. Email: [email protected]

Downloaded from pif.sagepub.com by guest on October 16, 2015

38

Proc IMechE Part F: J Rail and Rapid Transit 228(1)

Figure 1. Pressure transients from passing trains.

trains and their effect on trackside structures, using moving model train experiments. This investigation had two broad aims. The first was to investigate the fundamental nature of the transient pressure loading on trackside and overhead structures for a wide variety of structure shapes and different types of train. Controlled model-scale experiments allow a much greater number of tests to be carried out than were possible with the full-scale experiments and panel method calculations reported in ERRI D189/RP1 and thus enables a greater physical insight of the phenomenon to be obtained.3 The second aim was more practical, and aimed generally at improving the Eurocode by providing reliable measurement data for a number of different types of train, that would supplement the earlier measurements and replace the data from the now obsolete panel method calculations. Also BS EN 1991-2 was based on test data from operations with continental gauge rolling stock, which have significantly larger vehicle crosssections than UK rolling stock.2 Since the aerodynamic pressure loads and the loadings imposed by slipstreams are dependent on the distance from the train side or roof, they are generally overstated when applied to the UK network. Thus, there is a specific UK requirement for the development of alternative design rules to those in BS EN 1991-2, which can then be incorporated into a national annex to the code, and it is likely that such rules could result in significant economies for UK trackside and overhead structures. This work was thus funded by the Rail Safety and Standards Board with support from the UK Aerodynamics Working Group. The work described in this paper was carried out using the TRAIN Rig – a moving model aerodynamic facility in Derby, owned and operated by the University of Birmingham (http://www.birmingham.ac.uk/research/activity/railway/research/train-rig. aspx). Scale models of 1/25th size of three train types

were used, with the loads being measured on a variety of trackside structures. The technique of model testing is based on the well-established techniques of dimensional analysis, which allow the results of properly conducted model tests to be related directly to fullscale conditions. The experimental setup and data analysis techniques are set out in the section ‘Experimental methodology’. The section ‘Major experimental results’ describes the main experimental results, and these are compared as far as possible with the results of other investigations in the section ‘Validation/comparison with other sources of data’. The discussion section discusses the results in terms of the scaling of the pressure coefficient time histories, and the basic conclusions that arise from these experiments are then set out in the final section. In a companion paper (Part 2), these results will be used to investigate the adequacy of the current design criteria and to develop new, UK specific, design curves.

Experimental methodology The TRAIN Rig and experimental models The TRAIN (Transient Railway Aerodynamics INvestigation) Rig is a highly versatile moving model rig that can be used for a wide variety of aerodynamic investigations (Figure 2). In broad terms, it consists of a 150 m long track along which model vehicles can be propelled, in both directions, at speeds of up to 75 m/s. For the work described in this report, the Rig was operated at a nominal model speed of 40 m/s (which was repeatable to within 1 m/s for any one run). In the current situation it specifically allows transient static pressures to be measured on trackside structures. As will be seen later, by suitably non-dimensionalising the measured pressures


Baker et al.

39

Figure 2. The TRAIN Rig.

with the model velocity, this data can be applied directly to full-scale conditions. Experiments were carried out at a model scale of 1/25th for both the train and structure models. Three model trains were used in the experiments: . the leading car and half a trailing car of the Class 390; . a two-car Class 158 multiple unit; . a Class 66 locomotive. These three vehicles represent the leading vehicles of a streamlined passenger train, a non-streamlined passenger train and an aerodynamically rough freight train, respectively. Photographs of the 1/25th scale models are shown in Figure 3. A number of different structure models, also at 1/25th scale, were also tested. These are now described, using the coordinate system illustrated in Figure 4. This has an origin at the top of the rail level at the centre of the track with x, y and z being the along track, across track and vertical distances, respectively. On some occasions it is convenient to use the edge of the platform as the origin, and the coordinates from this point are y0 and z0 . The positions of the structures are defined by Y and Y0 laterally from the track centre and the platform edge respectively, and by h vertically from the top of rail. The structures tested were as follows.

Figure 3. Test train models: (a) Class 390; (b) Class 158; and (c) Class 66.

Horizontal structure

Y

h

1. Two-metre high hoardings, with return corners placed at the trackside and on platforms, at Y ¼ 1.45, 1.95 and 2.75 m (0.7, 1.2 and 2.0 m from the nearest rail) and Y0 ¼ 0.2, 0.7 and 1.2 m from the platform edge. Pressures were measured at the centre line of the hoarding, 0.25, 0.75, 1.25 and 1.75 m from the base. 2. Overbridges of different widths and heights mounted symmetrically above the track: specifically 10 m wide overbridges, with height above the track h ¼ 4.5, 5.0, 5.5 and 6.0 m to represent the wide structure condition implicit within BS EN

z’

z

Y’

Vertical structure

y’

y

Figure 4. Coordinate system.

1991-2, and h ¼ 4.5 m high overbridges 10.0, 6.0, 3.0 and 1.5 m wide to permit consideration of pressure variability with width. Pressures were measured 0.5, 1.5, 2.5 and 3.5 m either side of the


40


bridge centreline for the two wider bridges, and 0.38, 1.13, 1.88 and 2.63 m from the bridge centreline for the two narrower bridges. 3. Platform canopies of different heights with different back wall positions. The modelled heights above the track were h ¼ 4.0, 4.7, 5.4 and 6.0 m and the modelled back wall distances Y ¼ 3.45, 3.85, 4.25 and 4.75 m (2.7, 3.1, 3.5 and 4.0 m from the nearest rail): 16 configurations in total to permit consideration of typical UK canopy structures. Pressures were measured 0.21, 0.62, 1.04, 1.45, 1.87, 2.28, 2.70 and 3.11 m from the edge of the canopy nearest the track. 4. A trestle platform to represent newer UK platform structures, an increasing number of which are of lightweight construction. Pressures were measured 0.19, 0.56, 0.94, 1.31, 1.69, 2.06, 2.44 and 2.81 m from the edge of the platform nearest the track. Photographs of the 1/25th scale models are shown in Figure 5. Full details of both the train and the structure models can be found in Baker et al.6 The model speed of 40 m/s and scale of 1/25th resulted in a Reynolds number based on model height of 3.2 105, above the value of 2.5 105 specified in BS EN 14067-4. Preliminary tests were carried out on the overbridge structure (discussed later) at lower vehicle speeds of 20 and 30 m/s, and no noticeable changes in the results were found when they were plotted in a dimensionless form.

Instrumentation The model speed was measured using pairs of opposing photoelectric position finders and reflectors separated by 10 m along the TRAIN Rig track. A bespoke interface unit automatically calculated the average speed of the train through the 10 m test area based on the time taken for each of the beams to be broken. The approximate vehicle speed was predetermined from the tension in the firing cable, and a nominal value between 7.6 and 7.8 kN was used for each run. This resulted in a vehicle speed of 40 1 m/s. Generally for each of the cases listed in the previous section, two or three repeat runs were carried out. The static pressure was measured using a total of 16 Sensor Technics HCLA12X5PB differential amplified pressure transducers, connected to a Measurement Computing LGR5325 A/D converter and data logger. Transducers with a 1250 Pa range were selected to avoid pressure signal clipping. The drawback was that a lower signal-to-noise ratio resulted for the structures furthest from the train. However, the noise, position offset and other characteristics were acceptable for the required accuracy and were accounted for in post-processing and error analysis over a range of pressures. The transducers provided a linear frequency response up to a ceiling of 2 kHz, above which no higher physical frequencies were captured. This meets the requirements of BS EN 14067-4 for moving-model pressure measurements at this scale and speed. The sampling rate was

Figure 5. Photographs of modelled structures: (a) hoarding (from above); (b) overbridge (from side of track); (c) canopy; and (d) trestle platform (from side of track).


Baker et al.

41

significantly higher and linear smoothing was applied in post-processing (described in the section ‘Repeats and smoothing of data’). For the runs with the overbridge models, the hoarding models and the trestle platform model, the transducers were connected between the surface pressure tappings and a reference pressure measurement position beneath the TRAIN Rig track i.e. the pressure variation on the surface of the structure next to the track/train was measured. For the canopy models, however, the differential pressure between the top and bottom of the model was measured. For the overbridge, hoarding and trestle tests, the transducers were flush-mounted in the tappings, whereas for the canopy tests the transducers were connected remotely via 34 mm long silicone tubes with internal diameters of 1.8 mm. For all the presented data the results are plotted in conventional pressure coefficient form defined as Cp ¼

p 0:5V2

ð1Þ

where p is the measured pressure (relative to ambient), is the density of air and V is the train velocity. This format effectively removes the effect of small variations in velocity from the pressure measurements, and the theory of dimensional analysis then allows these results to be applied to the full-scale situation (assuming scale - Reynolds’ number - effects are small). The pressure coefficient variation is described in terms of a distance relative to a fixed point at the trackside, expressed in full-scale values. This procedure removes the effects of small-scale velocity changes on the time scale and makes the results immediately applicable to the full- scale situation. The figures that follow thus show the pressure coefficient distribution with reference to a fixed position on the train, the pressure coefficient effectively moving with the train. We thus have two prime experimental variables: the pressure coefficient and the distance relative to the train. A proper appreciation of the results that follow requires an indication of the uncertainty of these parameters. We consider first the pressure coefficient. From the presented definition, the three physical parameters that are used in the calculation are pressure, vehicle speed and density. Now a calibration of the pressure transducers shows that, for pressure differences of around 1000 Pa (which will be seen is in the region of the maximum pressure that was measured), the standard error was 0.6%. For the velocity, taking into account the uncertainty in the distance between the photoelectric beam detectors and their response, a standard error of around 0.2% was calculated. Taking into account the variability of atmospheric pressure and temperature, the standard error on the density was found to be 1%. Thus, the standard error on a pressure coefficient of 1.0 was (0.6% þ 2 0.2% þ 1%) ¼ 2%. This is a standard

error of 0.02 on the pressure coefficient throughout its range. For the distance the uncertainty was effectively 25 velocity (1/sampling rate), which for a velocity of 40 m/s and a sampling rate of 6250 samples/s gives an uncertainty of 0.16 m in terms of full-scale values, which is small in relation to the length of a typical vehicle (25 m).

Major experimental results Repeats and smoothing of data Figure 6(a) shows the raw data for three runs using the Class 390 running under the 10 m wide, 4.5 m high overbridge for the track centreline pressures (y ¼ 0 m). The results are plotted in pressure coefficient form and the horizontal axis is given in terms of a distance, rather than a time, and thus effectively gives the shape of the pressure distribution relative to the train position, with negative x values roughly corresponding to distances ahead of the train nose. The zero distance position in the x-direction is located at the point where the pressure transient passes through zero between the maximum and minimum peaks. This point corresponds approximately to a position on the train near where the full-body cross-sectional area is immediately adjacent to the point on the structure where pressures are being measured. The pressure coefficients show an initial rise to a high positive peak followed by a rapid fall through zero to a negative, i.e. suction, peak, before a gradual return to a zero value through a series of small oscillations. It will be seen that this pattern is repeated in most of the data sets that follow, and is consistent with that measured by other investigators: see Figure 1, for example. Two further points are apparent. First, there is very little difference between the three sets of results plotted in this way, and that second, a high-frequency oscillation can be seen on all the traces. It is thought that the latter is due to the structure model vibration caused by the passing train. Figure 6(b) shows the pressure coefficient data for hoardings but smoothed with a 10point moving average filter. The high-frequency oscillation has been eliminated. A number of other more complex filtering methods were investigated, but proved to have little advantage over the simple moving average methodology. There was very little variation between runs for almost all the data that was obtained. The exception was for some runs using the Class 66 locomotive, where there were considerable fluctuations for large positive values of distance, after the passage of the train nose, presumably due to turbulence in the wake of the train. Such distances, however, do not correspond to regions of critical (maximum) pressure coefficients, which are known to occur around the train nose. Thus, in what follows we will present data from individual runs only, although in each case multiple runs were


42


(a)

(b)

Figure 6. Effect of multiple runs and data smoothing on pressure coefficient distributions.

carried out to check basic repeatability, and a 10point moving average filter will be applied to all data.

Hoarding results Figure 7 shows summary details for all three types of train at a height of z (above top of rail) or z0 (above the platform surface) ¼ 0.25 m from the bottom of the hoarding. The results at this height are similar, if not somewhat greater than, those at other heights. The positive and negative peaks occur as would be expected, although for the trackside hoardings the negative peak is somewhat less obvious than for the platform cases, but note the two situations are not strictly comparable. The magnitudes of the coefficients decrease with distance from the nearest rail or platform edge. In general, the magnitudes of the Class 66 pressure coefficients are higher than for the Class 158 coefficients, which are themselves higher than for the Class 390 coefficients. In Figure 8, the coefficients are plotted for the hoarding positions closest to the track or the platform edge for all trains. Again these plots show the differences in the nature of the suction peaks between the trackside and platform cases, and the relative magnitudes of the peaks for the different train types can be appreciated. It is apparent from the figures that the forms of the positive pressure peaks are very smooth and consistent, but the negative pressure peak forms are less consistent, particularly for the blunt-fronted Class 158 and Class 66. In some cases a double negative peak is apparent. It is surmised that this is due to an interaction between the pressure transient and what will be a significant separation region around the nose of these vehicles.

Overbridge results Figure 9(a) shows a lateral plot of peak-to-peak pressure coefficients for the Class 390 passing beneath 10 m wide overbridges at different heights, for different distances from the centre line of the overbridge.

The results are broadly symmetric about the centre line of the bridge. The scale of the pressure transient and the flow of the nose around the vehicle are similar to the scale of the overbridge, and complex interactions might be expected to occur between the flow and the structure. Note that the pressures furthest from the centre line are effectively between the bridge legs. Figure 9(b) shows a comparison between the centre line pressure coefficients for the different overbridge heights (the average of the coefficients measured at the tappings on either side of the centre line), and it can be seen that there is a consistent drop in pressure coefficient magnitudes as the height increases. Figures 9(c) and 9(d) show similar figures for the Class 390 beneath 4.5 m high overbridges of different widths. The effect of varying width can be seen to be small in Figure 9(d), except for the smallest overbridge width of 1.5 m where the positive and negative peaks are significantly reduced. This is perhaps surprising as we might have expected an increase in load as the overall loading on the overbridge will be more coherent for a smaller structure. The possible interaction between the overbridge and the unsteady flow field may, as previously mentioned, however, also be significant in this case. Similar data is shown for the Class 158 in Figure 10 and for the Class 66 in Figure 11. In general the magnitudes of the coefficients for the Class 158 are higher than for the Class 390, and the magnitudes for the Class 66 are higher still. The same comments apply as for the Class 390, although for the smaller overbridge widths for the Class 66, the suction peak close to x ¼ 0 m, is small and dominated by a larger, unsteady peak someway downstream, which may be caused by pressure changes in the train’s boundary layer (slipstream). The differences between vehicles are illustrated in Figure 12 for centre line pressures for the 4.5 m high, 10 m wide overbridge. It should be noted that the pressure gradient between the maximum and minimum peaks is higher for the Class 66 than for the other trains (or alternatively the peaks are closer


Baker et al.

43

(a)

(b)

(c)

(d)

(e)

(f)

Figure 7. Hoarding pressure coefficients at 0.25 m height at different distances from nearest rail or from platform edge: (a) Class 390, trackside; (b) Class 390 – platform; (c) Class 158, trackside; (d) Class 158 – platform; (e) Class 66, trackside and (f) Class 66 – platform.

together). This reflects the blunter nose shape of the Class 66.

Canopy results The results of Figure 13 show how the pressure coefficients vary away from the track across the canopy for the smallest back wall distance (2.7 m from the nearest rail) and the lowest canopy height (h ¼ 4 m), for all three train models. Essentially, they show the expected shape of the other results, and although the

value for the tappings nearest to the canopy edge is always higher than the other tappings, there is actually relatively little variation in coefficient across the canopy for each train type. The Class 390 results have the smallest coefficients, and the Class 66 results have the highest coefficients. One significant feature, apparent on both the Class 158 and Class 66 results are the existence of pressure coefficient oscillations of a substantial magnitude downstream of the initial maximum and minimum peaks. A simple dynamic vibration test showed that this was not due to


44


(a)

(b)

Figure 8. Comparison of pressure coefficients caused by different train models on hoardings: (a) trackside hoardings, 0.25 m from bottom of hoarding, 0.7 m from rail; and (b) platform hoardings 0.25 m above bottom of hoarding, 0.2 m from platform edge.

(a)

(b)

(c)

(d)

Figure 9. Pressure coefficients on overbridge models caused by passage of Class 390 model: (a) lateral peak-to-peak pressure variation on 10 m wide overbridges of different heights; (b) centreline pressure distributions on 10 m wide overbridges of different heights; (c) lateral peak- to-peak variations on 4.5 m high overbridges of different widths and (d) centreline pressure distributions on 4.5 m high overbridges of different widths.

structural oscillations. It could thus be surmised that this is due to a vertical flow oscillation between the platform and the canopy, but more work would be required to substantiate this conjecture. A further interesting point is that the pressure coefficients near

the edge of the canopy, although being higher than the other pressure coefficients, tend to lag the others to some extent, suggesting a distorted pressure wave within the canopy space. Figure 14 shows summary results for all three trains for the smallest and largest


Baker et al.

45

(b)

(a)

Figure 10. Pressure coefficients on overbridge models caused by passage of Class 158 model (a) lateral peak-to-peak pressure variation on 10 m wide overbridges of different heights and (b) centreline pressure distributions on 10 m wide overbridges of different heights.

(b)

(a)

(d)

(c)

Figure 11. Pressure coefficients on overbridge models caused by passage of Class 66 model: (a) lateral peak-to-peak pressure variation on 10 m wide overbridges of different heights; (b) centreline pressure distributions on 10 m wide overbridges of different heights; (c) lateral peak to peak variations on 4.5 m high overbridges of different widths and (d) centreline pressure distributions on 4.5 m high overbridges of different widths.

canopy heights, and the smallest and largest back wall distances, for the pressure tappings nearest the edge of the canopy. The results for the intermediate canopy heights and the intermediate back wall distances all fall consistently between these extremes. Again the Class 390 values are the smallest and the Class 66

values the largest. The peak magnitudes for the lowest canopies are significantly higher than for the highest as seems sensible. The oscillations in the wake can again be clearly seen. There is a suggestion that, for the Class 390 and Class 66, the frequency of the oscillation is related to canopy height, but this is not


46


Figure 12. Comparison of pressure coefficient distributions for 10 m wide, 4.5 m high overbridges for all train models.

(a)

(b)

(c)

Figure 13. Variation of pressure coefficient across 4.0 m high canopy with 2.7 m back wall: (a) Class 390; (b) Class 158 and (c) Class 66.


Baker et al.

47

(a)

(b)

(c)

Figure 14. Variation of pressure coefficient at edge of canopy for different canopy heights and back wall distances: (a) Class 390; (b) Class 158 and (c) Class 66.

the case for the Class 158, where the frequencies look the same. This suggests that the oscillation frequency is not a simple function of structure dimensions. Figure 15 shows the effect of different back wall distances on the pressures at the edge of the canopy, for the Class 66 model, with a 4.7 m high canopy. The no back wall results are also shown. It can be seen that the back wall distance, and indeed the presence or otherwise of the back wall, has only a limited effect on the measured pressures at this measurement position. This is generally true for all measurement positions. The no back wall results show a reduction in the positive pressure peak, but are otherwise very similar to those with a back wall. It is of interest to note that the pressure coefficient oscillations still exist for the open canopy, which suggests that they are due to vertical standing wave patterns in the cavity. Finally, Figure 16 shows a comparison for the pressure coefficients at the canopy edge for the lowest canopy, smallest back wall distance and for all three trains. This clearly shows the relative magnitudes of the pressure coefficient due to different train types (maximum for Class 66 and minimum for Class 390), including the downstream oscillations.

Trestle platform results In general, as might be expected, for the trestle platform the magnitudes of the coefficients decrease with distance from the edge of the platform. A comparison of the pressure coefficients for the different vehicles, for the pressure tapping nearest the edge of the platform, is shown in Figure 17. Again the high pressure and suction peaks can be seen, although the Class 66 suction peak is less distinct than for the other vehicles. As before the Class 66 produces the highest positive and negative magnitudes of pressure coefficient.

Validation/comparison with other sources of data This section discusses the validation of the current experimental results and presents a comparison with experimental data from other sources, in order to show that the results are reliable and can be used with confidence to predict full-scale pressure distributions. In this process a number of different approaches are possible:


48


Figure 15. Effect of back wall distance on pressures close to leading edge of canopy.

Figure 16. Variation of pressure coefficient at the canopy edge with different train types for a 4.0 m high canopy with a 2.7 m back wall.

. consideration of the internal consistency of results; . consideration of earlier comparisons made with experimental results in the TRANSAERO Project;7 . comparison with earlier TRAIN Rig measurements;8 . comparison with full-scale UK measurements on train sides;8 . comparison with full-scale UK measurements on platform hoardings;9

. comparison with full-scale UK measurements on a trestle platform.10

Internal consistency of results First, it will be clear from the results discussed in the section ‘Major experimental results’, that the data is internally consistent, in that it is of the same form in most cases, with the magnitudes of the coefficients being in general greatest for the Class 66 and least


Baker et al.

49

Figure 17. Comparison of pressure coefficients caused by all train models close to the edge of the trestle platform.

for the Class 390, with a consistent variation of magnitude with structure height/width, distance along and across the various structures. It is also quite repeatable on a run-to-run basis. This consistency and repeatability gives some confidence in the results.

TRANSAERO results In the 1990s, as part of the EU TRANSAERO Project, measurements were made using the TRAIN Rig with moving models of Italian ETR500 trains, with various nose shapes, and the pressures measured on the side of stationary trains.7 These results were compared with equivalent full-scale measurements of the same configurations and excellent agreement with nose and tail passing pressures was obtained, again giving some confidence in the use of the rig for the measurement of train-induced pressures. For the fullscale experiments the values of the maximum pressure peak, minimum pressure peak and time between the peaks were 0.222 kPa, 0.253 kPa and 0.122 s, while the equivalent values from the TRAIN Rig experiments were 0.220 kPa, 0.241 kPa and 0.112 s, indicating a high level of agreement.

Earlier TRAIN Rig measurements Measurements have been made in the past using the TRAIN Rig on the pressures on stationary trains as they are passed by a moving train.8 In this comparison three sets of such experimental data were used. 1. A Class 341 multiple unit passing a Mark 3 coach, with pressures measured 1.63 m above the track on the latter.

2. A Class 43 (HST) passing a freight wagon, with the pressures being measured 1.63 and 2.73 m above the track on the wagon. 3. A Class 220 passing a freight wagon, with the pressures being measured 1.63 and 2.73 m above the track on the wagon. In calculating the equivalent hoarding distances, it was assumed that the width of the moving trains and the Mark 3 coach was given by the maximum width of the W6a gauge (2.82 m) and the width of the freight container was 2.5 m. This resulted in equivalent hoarding distances from the track centre line of 1.99 m for the Class 341/Mark 3 measurements and 2.31 m for the other measurements. In view of the assumptions made in that work there could be potentially sizeable errors in these figures (of the order of 10 cm or more). The average maximum and minimum pressure coefficients from between three and six sets of measurements in each case were calculated and these were compared with the height-averaged values measured on the hoardings on the TRAIN Rig. The results of this comparison are given in Figure 18. The current TRAIN Rig data is shown as solid lines (connecting the discrete experimental points), while the various sets of earlier data are shown as experimental points. ‘Max’ refers to the maximum value of the positive pressure peak, and ‘Min’ refers to the largest magnitude of the negative pressure peak. The full-scale data is all for passenger train shapes with the leading vehicles all having relatively blunt noses. It can be seen that, in general, the present results are consistent with the earlier results, making due allowance for the variation in geometry, and the uncertainty in assigning the correct equivalent distance of


50


Figure 18. Comparison of earlier TRAIN Rig pressure coefficients measurements on the side of a stationary train with current TRAIN Rig hoarding data (TRAIN Rig measurements8 given by solid lines).

a hoarding from the track centre line for the train side results. In general, the maximum positive values correspond to the Class 390 and Class 158 data as would be expected, whereas the maximum negative values fall closest to the Class 158 data.

Full-scale UK measurements on train sides Full-scale measurement data are also available for full-scale experiments with a moving Class 390 train passing a stationary Class 390 train, with two measurement positions on the stationary train.8 This can be compared with the hoarding data obtained for the Class 390, if equivalent distances from the track can be specified. The results are shown in Figure 19. It can be seen that, allowing for the differences in geometry between the two cases, the agreement is good, with close correspondence of the positive peaks, and the full-scale negative peaks being rather greater in magnitude than the model-scale values. This is probably due to differences in the track geometry in the fullscale and model-scale cases.

Full-scale UK measurements on platform hoardings Measurements were made of the pressures on a dummy wall on the station platform at Northallerton as it was passed by a variety of train types.9 The trains were as follows: . . . .

Class 43 (HST) – one run; Class 91 service train – two runs; Class 91 test train – three runs; a freight train with tankers - one run.

The wall was 2 m away from the platform edge and measured 14 m in length and 3 m in height.

Measurements were made at heights of 0.095, 1.33 and 2.04 m above the platform. The average peakto-peak values of the pressure coefficient for each type of train were averaged over the runs, and then the height-averaged values of the peak-to-peak coefficients for each train type were calculated. These were then compared with the height-averaged values from the hoardings measured in the TRAIN Rig. The results of the comparison are shown in Figure 20. The current TRAIN Rig data is again shown as solid lines and the full-scale measurements as discrete symbols. Clearly, the full-scale tests were at a distance from the platform edge not covered by the modelscale experiments, but the results are nonetheless consistent, with the passenger train full-scale values being on plausible extrapolations of the Class 390 and Class 158 lines, and the freight train data being similarly consistent with the Class 66 data.

Full-scale UK measurements on a trestle platform Measurements were made of the pressures on the trestle platform at East Midlands Parkway Station.10 Measurements were made of the absolute pressure on the upper surface of the platform at the platform edge and 1.5 m from the edge, and differential pressure measurements were made between the upper and lower platform surfaces at the platform edge. Pressure measurements were made during the passage of Class 222 Meridian trains. There were data from eight runs in total, and when the pressure coefficients were plotted against train speed little systematic variation could be seen. The pressure coefficient results were thus averaged across all runs, and the absolute pressure values compared with the equivalent absolute values from the TRAIN Rig experiments. The results of the comparison are shown in Figure 21.


Baker et al.

51

Figure 19. Comparison of earlier full-scale pressure coefficients measurements on the side of a stationary Class 390 with current TRAIN Rig hoarding data (TRAIN Rig measurements8 given by solid lines).

Figure 20. Comparison of current TRAIN Rig hoarding data with full-scale Northallerton measurements (TRAIN Rig measurements9 given by solid lines).

The full-scale results for the streamlined Class 222 multiple unit compare well with the Class 390 data as might be expected.

Discussion In this section we discuss the fundamental nature of the pressure transients measured in this report. We first consider how these pressure transients relate to the velocity field around the nose of the train, and thus to measurements of train slipstreams (see, for example, Sterling et al.11). Second, we consider if it is possible to parameterise the results in a relatively

simply way, that might prove useful as a framework for applying these results to practical situations. First, it is clear that the pressure transients are well defined with a high level of repeatability from run to run, at least for the positive pressure part of the transient. This suggests that the transients are not affected by train-induced turbulence and can effectively be described using inviscid potential flow considerations. This is to some extent confirmed by slipstream velocity measurements (see, for example, Sanz-Andres et al.12) where the nose velocity peak is equally repeatable from run to run, in stark contrast with the flow in the train boundary layer and wake where there is very


52


Figure 21. Comparison of current TRAIN Rig data with full-scale East Midland Parkway measurements (TRAIN Rig measurements10 given by solid lines).

significant run-to-run variability caused by large-scale turbulence flow structures. Now by a simple application of Bernoulli’s equation it is possible to show that, in the inviscid flow field around the train nose, the pressure coefficient is given by Cp ¼ 1 ð1 uÞ2 v2 w2

ð2Þ

where u, v and w are the longitudinal, lateral and vertical slipstream velocities normalised with train velocity. w is always small (of the order of 0.02) and can be neglected. Typical velocity traces for u and v (from Sterling et al.11 for an ICE-1) are shown in Figure 22. Note that there are small non-zero velocity values upstream of the train, caused by low-level ambient wind flows. It can be seen that values of u and v of around 0.05 to 0.1 are measured. Such values enable us to write Cp 2u

ð3Þ

From Figure 22 it can be seen that the longitudinal velocity transient has a very similar form to the pressure transients measured here, giving some confidence in the presented analysis. This suggests that pressure transients might be inferred from full-scale velocity measurements, and vice versa. Equation (3) also implies that, as the pressure coefficient is proportional to longitudinal velocity, then if a vertical wall (or hoarding) is placed at this point then this can simply be represented by an ‘image’ source on the other side

Figure 22. Slipstream velocity traces around the nose of an ICE-1 train, 2.85 m from the track centre line, 0.5 m above top of rail (Sterling et al.11).The graph shows the longitudinal (u) and lateral (v) components of horizontal slipstream velocity normalised by the train speed; distance is measured from an arbitrary point close to the train nose.

of the wall, which would result in the following expression Cp 4u

ð4Þ

Now the data from Figure 22 corresponds to a position 0.5 m above the top of the rail, 2.1 m from the nearest rail, for a high-speed train slipstream. The maximum pressure coefficient at a point 0.25 m above the track, 2.0 m from the nearest rail shown in Figure 7 for the Class 390 (the nearest equivalent to


Baker et al.

53

Figure 23. Plot of equations (5) – load transients from the analysis of Sanz-Andres et al.12

Table 1. Scaling parameters of pressure coefficient time history Slope of regression line with Y

Class 390 hoarding

Class 158 hoarding

Class 66 hoarding

Class 390 overbridge

Class 158 overbridge

Class 66 overbridge

Y (m)

CFmax

CFmin

xmax (m)

xmin (m)

CFmax/CFmin

xmax/xmin

CFmax

CFmin

xmax

xmin

1.45 1.95 2.75 1.45 1.95 2.75 1.45 1.95 2.75 4.50 5.00 5.50 6.00 4.50 5.00 5.50 6.00 4.50 5.00 5.50 6.00

0.37 0.28 0.15 0.47 0.35 0.19 0.67 0.52 0.33 0.18 0.15 0.11 0.09 0.23 0.19 0.13 0.11 0.42 0.34 0.27 0.22

0.21 0.18 0.11 0.54 0.39 0.27 0.64 0.68 0.50 0.34 0.24 0.16 0.12 0.35 0.24 0.18 0.13 0.84 0.66 0.50 0.39

1.16 1.26 2.29 0.86 1.33 2.03 0.75 1.39 2.16 2.02 3.11 3.46 3.31 1.43 1.88 2.28 3.47 1.36 1.58 2.17 3.49

1.06 0.96 1.54 0.57 0.91 1.47 0.66 1.28 1.85 1.89 2.26 2.53 2.69 1.25 1.46 1.94 2.21 1.00 1.41 1.87 2.32

1.73 1.57 1.35 0.88 0.90 0.72 1.05 0.76 0.67 0.53 0.63 0.68 0.75 0.66 0.80 0.75 0.88 0.50 0.51 0.54 0.57

1.10 1.32 1.48 1.51 1.46 1.38 1.13 1.09 1.17 1.07 1.38 1.37 1.23 1.14 1.29 1.17 1.57 1.36 1.12 1.16 1.50

1.45

1.06

1.08

0.61

1.40

1.08

1.33

1.47

1.09

0.39

1.65

1.59

2.44

3.61

1.70

1.23

2.60

3.45

2.96

2.08

2.20

2.69

3.23

2.93

the full-scale results, although rather more streamlined than the ICE-1) is around 0.17. From Figure 22 and equation (4) the predicted value would be around 0.20, suggesting that the above argument is broadly valid.

Now let us consider the parameterisation of the experimental results. The basic questions are how and why the pressure transients vary with train type, distance from the train, etc. We base our discussions around the two perhaps most fundamental cases of


54


(a)

(b)

Figure 24. Normalised pressure transients (a) trackside hoardings and (b) overbridges.

those studied: the trackside hoardings and the 10 m wide overbridge, that is the vertical surfaces next to the track and horizontal surfaces above the track. We also use as a framework for the discussion the work of Sanz-Andres et al.,12 which sets out a potential flow calculation of the forces on a vertical pedestrian barrier parallel to the track as a train passes by it. While the situation considered here is considerably more complex that that analysed by Sanz-Andres et al.,12 it is felt that that analysis could act as a framework for the consideration of the parameterisation of the current results. Sanz-Andres et al.12 gave the following expression for the transient pressure force coefficient on a vertical structure next to the track

CF ¼

3HA ðx=YÞ 4Y3 1 þ ðx=YÞ2 2:5

ð5Þ

where H is the height of the barrier and A is the area of the train behind the nose. In terms of the current results, this suggests that the distance scales on the distance from the train centre Y, and that the pressure/ force coefficients should scale on the cross-sectional

size of the train (presumably allowing for any largescale flow separation around the nose) and with Y3. This curve, plotted as 4CF Y2 =3HA gainst x=Y is shown in Figure 23. It can be seen to be of the expected form, and is anti-symmetric around the origin. Now clearly the experimental results shown in the section ‘Major experimental results’ lack this degree of symmetry with the magnitudes of the positive and negative x-direction pressure coefficient time histories being significantly different. We thus explore in what follows whether the traces for the negative and positive x-direction portions of these curves scale in the way suggested by equation (5). To enable this comparison to be made the hoarding and overbridge data has been analysed to give overall forces in the x-direction through either finding the height-averaged pressure coefficient (for the vertical hoardings) or the pressure coefficient averaged over a 2 m length at the centre of the overbridge span. These have then been aligned so that the zero-crossing between the positive and negative peaks is at the origin in the x-direction. Table 1 shows the maximum and minimum values for each case considered together with the distances of these peaks from the


Baker et al.

55

origin. The ratio of the magnitudes of the maximum and minimum peaks and the ratio of the distances to the peaks are given, as an indication of the lack of symmetry. In addition, for each set of cases for any one train, the power-law exponents for the variation of the maximum and minimum coefficients with respect to Y (for the hoarding) or bridge height h, are given. Similarly, the power law exponents for the variation of the position of the peaks with Y are also given. It can be seen from Table 1 that there is a considerable degree of asymmetry, both in the maximum and minimum values and the positions of these values. While some trends are apparent in the data, it is difficult to make generalisations. The slopes of the best-fit lines with Y show that for the vertical hoardings, the peak values fall off with distance from the train with a value of the exponent somewhere between one and two whereas for the overbridge case the exponent takes on values of two and a half to four the latter being closer to the value of three expected from equation (5). Both the positions of the maximum and minimum peaks increase with distance from the train (i.e. the peaks become more spread out and diffuse), although the variation with Y or h has an exponent of rather greater than the value of unity that would be expected by the scaling with distance from the track centre line suggested previously. These points being made, the experimental data was then normalised with the magnitudes of the positive and negative peaks, and the magnitudes of the peak positions, thereby effectively assuming different scaling for positive and negative values of x. The results are shown for the three classes of configuration in Figure 24 and compared with the form of equation (5). It can be seen that the data fits this form well between the positive and negative peaks, but deviates significantly for regions outside the peaks. Inspection of the results shows that in general the greatest deviation is for small values of Y or h, i.e. with the structures close to the train. Thus, it can be concluded that the scaling of the pressure coefficient time histories is complex and not amenable to a general parameterisation, although the theoretical curves of Sanz-Andres et al.12 do give a useful framework for analysis and discussion.

The experimental programme was undertaken in connection with the RSSB-funded research project T750 ‘Review of Euronorm design requirements for trackside and overhead structures subjected to transient aerodynamic loads’, which was sponsored by the railway industry ‘Aerodynamics GB Working Group’. Permission of BSI for the use of figure 1 is also acknowledged.

Conclusions

References

From the data presented in the preceding sections the following main conclusions can be drawn.

1. BS EN 14067-4:2005þA1:2009. BSI railway applications — aerodynamics — Part 4: Requirements and test procedures for aerodynamics on open track. 2. BS EN 1991-2; 2003. BSI Eurocode 1 Actions on structures – Part 2: Traffic loads on bridges. 3. ERRI D189/RP1, 1994. Loading due to dynamic pressure and suction from railway traffic. Effect of the slipstreams of passing trains on structures adjacent to the track. 4. TSI Directive 96/48/EC 2008/217/E, 2008. Interoperability of the trans-European high speed rail system. Infrastructure sub-system.

1. The use of the TRAIN Rig methodology has been shown to be a robust way of obtaining aerodynamic loading on a wide variety of trackside structures in an efficient manner, with the results showing good run-to-run repeatability. 2. The nose pressure coefficient distribution caused by passing trains is of the expected type, with a positive pressure peak followed by a

3.

4. 5.

6.

7.

8.

9.

negative pressure peak. In general, the peaks are not symmetric, that is they do not have the same magnitudes. In general the surface pressure coefficients generated by the Class 66 freight locomotive are greater than those generated by the Class 158 multiple units, which are themselves greater than those generated by the Class 390 Pendolino. For the hoarding structures, the trackside negative peak is very indistinct. The pressure coefficients across the overbridges, show a roughly parabolic fall off from the centre line. The coefficients fall as overbridge height increases, but are insensitive to the width of the bridge in the along track direction, except for the smallest bridge widths. The canopy pressure coefficients show little variation across the canopy, except very close to the canopy edge. The effect of back wall distance on the canopy pressures is also small. For the blunter trains, a vertical standing pressure wave appears to be generated in the canopy/platform space. A comparison of the current results with a range of earlier measurements and calculations at both model-scale and full-scale show a reasonable agreement, although the nature of many of the earlier results makes a precise comparison difficult. There are indications that the surface pressure transients on vertical surfaces such as hoarding are closely correlated with slipstream velocities. The scaling of the pressure transient time histories is complex and not amenable to easy generalisation. That being said, the theoretical approach of Sanz-Andres et al.12 offers a potentially useful framework for the consideration of these results.

Funding


56


5. TSI Directive 96/48/EC 2008/232/E, 2008. Interoperability of the trans-European high speed rail system. Rolling stock sub-system. 6. Baker CJ, Gilbert T, Jordan S, et al. Review of Euronorm design requirements for trackside And overhead structures subjected to transient aerodynamic loads. Report on RSSB Project T750, 2011. London, UK: Rail Safety and Standards Board. 7. Johnson T and Dalley S. 1/25 scale moving model tests for the TRANSAERO Project. In: Schulte-Werning B, Gregoire R and Malfatti A (eds) TRANSAERO- A European initiative on transient aerodynamics for railway system optimisation. Berlin, Germany: Springer-Verlag, 2002, pp.123–135. 8. Rail Safety and Standards Board. AeroTRAIN project database of UK pressure loading data. Report, 2010. London, UK: Rail Safety and Standards Board. 9. Figura GI. Trackside safety tests at Northallerton. British Rail research report RR AER 014, 1993. Derby, UK: British Rail. 10. Johnson T. Aerodynamic forces on East Midlands Parkway station platforms. DeltaRail Report ES-2009025 Issue 1, 2009. Derby, UK: DeltaRail. 11. Sterling M, Baker CJ, Jordan SC and Johnson T. A study of the slipstreams of high speed passenger trains and freight trains. Proc IMechE, Part F: J Rail Rapid Transit 2008; 222: 177–193. 12. Sanz-Andres A, Laveron CA and Baker C. Vehicleinduced loads on pedestrian barriers. J Wind Engng Ind Aerodyn 2004; 92: 403–426.

CFmax CFmin Cp h H p u v w V x xMAX xMIN y y0 Y Y0 z z0

maximum integrated force coefficient minimum integrated force coefficient pressure coefficient p/0.5v2 distance from top of rail to overbridge/ canopy (m) height of vertical structure in analysis of Sanz-Andres et al.12 (m) pressure relative to ambient (Pa) longitudinal slipstream velocity (m/s) lateral slipstream velocity (m/s) vertical slipstream velocity (m/s) train/model velocity (m/s) distance along the track (m) distance of maximum pressure peak from pressure zero crossing point (m) distance of minimum pressure peak from pressure zero-crossing point (m) lateral distance from centre of track (m) lateral distance from edge of platform (m) lateral distance of vertical structures from centre of track (m) lateral distance of vertical structures from platform edge (m) vertical distance from the track (m) vertical distance from top of platform (m) density of air (kg/m3)

Appendix 1 Notation A CF

area of train in analysis of Sanz-Andres et al.12 (m2) integrated force coefficient


Baker et al.

57

Transient aerodynamic pressures and forces on trackside and overhead structures due to passing trains. Part 2: Standards applications Christopher Baker1, Sarah Jordan1, Timothy Gilbert1, Andrew Quinn1, Mark Sterling1, Terry Johnson2 and John Lane2

Abstract This paper is the second part of a two-part paper that describes the results of an experimental investigation to measure the load transients on railway structures due to passing trains. Part 1 described the model-scale experiments that were carried out and the results that were obtained. In this paper the results are further analysed in terms of the development of new formulations for standards that would be relevant to trackside structure geometries commonly found in the UK. Specifically, it compares the experimental results with the formulations of the existing standards, and shows that they are very conservative for the UK situation. This is largely due to the fact that the standard formulations were derived from experiments and calculations based on European train sizes and track geometries, whereas the trains in the UK are somewhat smaller. Two methods of correcting for different train size were evaluated and both were shown to bring the loads predicted from the standards closer to the experimental results. Formulae were then derived from the experimental data for the aerodynamic loading on a variety of typical trackside structures that may be useful in future standards revisions. Keywords Train aerodynamics, aerodynamic pressures, bridges, hoardings, canopies, platforms Date received: 25 March 2012; accepted: 4 September 2012

Introduction Part 1 of this paper described a series of aerodynamic experiments carried out using the moving model TRAIN Rig to measure the transient aerodynamic pressure loads on trackside and overhead structures. The rationale for this tests, fully described in Part 1, was to measure the transient loading on a wider range of trackside and overhead structure geometries than was possible in earlier full-scale experiments, for a number of different train types, so that an in- depth understanding of the flow phenomena could be achieved, and also, more practically, to provide information on the loading of structures of relevance to UK operating conditions, to supplement the current code which was developed for continental loading gauges (BS EN 1991-2 and BS EN 4067-4 and to provide material for a UK National Annex to these standards.1,2 The tests were carried out using models of three different types of train (a streamlined Class 390 Pendolino leading car, a two- car blunt-fronted Class 158 multiple unit and a Class 66 freight locomotive) for a number of different trackside and overhead structures (hoardings, overbridges, canopies and trestle platforms). Part 1 described the nature of these experiments, and outlined the results that were

obtained. The results were compared, as far as possible, with the results from other full-scale and modelscale experiments, and a broad level of consistency and reliability was demonstrated. From these results the following main conclusions could be drawn concerning the general nature of these pressure transients. 1. The use of the TRAINR Rig methodology has been shown to be a robust way of obtaining aerodynamic loading on a wide variety of trackside structures in an efficient manner, with the results showing good run-to-run repeatability. 2. The nose pressure coefficient distribution caused by passing trains is of the expected type, with a positive pressure peak followed by a negative pressure peak. In general, the peaks are not symmetric,

1 Birmingham Centre for Railway Research and Education, University of Birmingham, UK 2 Rail Safety and Standards Board Ltd, London, UK

Corresponding author: Christopher Baker, Birmingham Centre for Railway Research and Education, University of Birmingham, Birmingham B15 2TT, UK. Email: [email protected]


58


Figure 1. Vertical structure next to the track in BS EN 4067-4 (hoarding).

that is the positive and negative peaks do not have the same magnitudes. 3. In general the surface pressure coefficients generated by the Class 66 freight locomotive are greater than those generated by the Class 158 multiple units, which are themselves greater than those generated by the Class 390 Pendolino. 4. A comparison of the current results with a range of earlier measurements and calculations at both model-scale and full-scale show a reasonable agreement, although the nature of many of the earlier results makes a precise comparison difficult. This paper uses the experimental results to consider a number of issues relevant to the codification of aerodynamic pressure transient loading on structures. The section ‘The current codification procedure’ sets out the current codification procedure used in BS EN 1991-2 and BS EN 4067-4. The section ‘Direct comparison of results with code values’ then presents a comparison of the experimental loading data with the uncorrected code values, and the code values corrected for the difference in continental and UK loading gauges through two different methods. The section ‘Analysis of experimental data’ then casts the experimental data into a format that is directly comparable with the codification method and this is then used in the section ‘Development of new codification format’ to develop a series of relationships for structural loading in UK conditions that will form the basis of a proposal for a National Annex to the code. Finally, conclusions are drawn.

The current procedure outlined in BS EN 4067-4 for describing the train-induced aerodynamic loads on structures makes the basic (design) assumption that the pressure wave for the forces on vertical structures next to the track is of the form shown in Figure 1, for horizontal structures above the track it has the form shown in Figure 2) and for canopy structures it has the form shown in Figure 3. In each case, the pressure pulse is assumed to consist of a constant positive pressure that extends for a distance of 5 m ahead of the train for vertical structures (Figure 1) or from the front of the train nose for horizontal structures (Figure 2), with a constant negative pressure of 5 m length immediately following. There would thus appear to be some discrepancy between the assumed position of the pressure transient relative to the train for vertical and horizontal structures, although in reality this will not be of importance, since the load on the structure is required, without reference to train position. The magnitudes of the positive and negative pressures are assumed to be the same. From the results presented in Part 1, this can be seen to be only a very rough approximation of reality, with the measured pressure distributions showing rather sharper positive and negative peaks (i.e. less than 5 m in length), and while for some of the results the magnitudes of the positive and negative peaks are similar, this is far from the general situation. This point having been made, in each case, the loadings on the structures are calculated from the following expressions. For vertical structures next to the track (hoardings) p1k ¼ 0:5V2 k1 Cp1 , Cp1 ¼

The current codification procedure One aspect of the current work was to use the experimental data to suggest revisions to current codification procedures to cover UK-specific conditions.

2:5 þ 0:02 ðY þ 0:25Þ2

ð1Þ

where is the density of air and V is the train speed, Cp1 is a pressure coefficient and Y is defined in Figure 1. k1 takes on a value of 1.0 for freight


Baker et al.

59

Figure 2. Horizontal structure above the track in BS EN 4067-4 (overbridges).

Figure 3. Horizontal structure adjacent to the track in BS EN 4067-4 (canopies).

trains, 0.85 for conventional passenger trains and 0.6 for streamlined high-speed trains. For horizontal structures above the track (overbridges) p2k ¼ 0:5V2 k2 Cp2 , Cp2 ¼

2:0 þ 0:015 ðh 3:10Þ2

ð2Þ

and h is defined in Figure 2. k2 takes the same values as k1. For canopy-like structures p3k ¼ 0:5V2 k3 Cp3 , Cp3 ¼ k3 ¼

1:5 þ 0:015, ðY þ 0:25Þ2

7:5 h 3:7

where h is again defined in Figure 3. k3 is dependent on the value of h and there is thus assumed to be no dependence upon different types of train in this case.

The dependence upon h is only applicable within the range of h used in the measurements i.e. 4.0 < h < 6.0 m In attempting to understand the genesis of the above method, the experiments of ERRI D189/RP1 were analysed.3 The nature of the experimental data that was obtained was not altogether clear. It would seem that different methods were used to find the spatially averaged loads on horizontal and vertical structures. For the former, a simple averaging of the lateral pressure distribution was carried out. For the vertical stuctures, however, a ‘moment averaging’ was carried out about the base of the structure i.e. the moments of the measured pressures around the base were calculated and then divided by the height of the structure to give a pseudo-force value. From the report it is, however, not clear how the 5 m long peak values were obtained. It appears likely that for vertical structures next to the track, the experimental pressure data was averaged over a 5 m length across


60


the maximum and minimum peaks, to take account of the effective spacing between fence supports, but for horizontal structures, the maximum peak value was taken to apply over the entire 5 m length, and that these led to the values in the code. Also, it should be noted that for over track structures such as overbridges, the peak values of pressure measured at the centre of the track were assumed to act over up to a 10 m length either side of the track centre. The results of Part 1 for 10 m span overbridges show a very considerable falling off in pressure magnitude away from the centreline and this is also likely to be the case for longer span overbridges, potentially making this a very conservative assumption. While the discussed procedure can be criticised on a number of grounds, in what follows the experimental results are analysed in a manner that is consistent with this methodology. For all the results, the positive and negative pressure peaks are calculated as the average values over 5 m lengths before and after the main zero-crossing point. For horizontal structures, either the average value of such peaks in the lateral direction are calculated (canopies), or the distribution in this direction is considered (trestle platform and overbridges). For vertical structures (hoardings) the ‘moment averaged’ value is calculated over the structure height (although in reality this differs little from the arithmentically averaged value in almost all cases).

Direct comparison of results with code values In this section, we compare the experimental results outlined earlier with the provisions of the code for the three types of structure for which the code makes allowance: vertical structures parallel to the track (hoardings), horizontal structures above the track (overbridges) and horizontal structures to the side of the track (canopies). Before making this comparison, it must be noted that the code values were determined for the larger continental loading gauge, and thus for a structure at a fixed distance from the side of the track or above the rail, the effective separation between train and structure is thus higher for UK trains than for continental trains, and some allowance needs to be made for this point. Two such methods were used in this study. In the first, an increment has been added to the distance from the track to the structure when using the equations set out in the section ‘The current codification procedure’: effectively the difference between the continental G1 gauge and the GB W6A gauge, giving a lateral increment of 0.24 m (the difference between the maximum half-width of the considered gauges) and a vertical increment of 0.36 m (the difference between the maximum heights), allowing corrections to be made for structures at both the side of and above the track. This has been applied to the hoarding, bridge and canopy results. The second approach

used the methodology developed in Johnson and Dalley.4 They demonstrated that the pressure coefficients on a stationary train passed by a moving train scaled with the inverse square of the separation distance between the vehicles. This methodology was applied to the hoarding results, with the predicted pressure coefficients from equation (1) being multiplied by ((Y-WG1)/(Y-WW6A))2, where Y is the distance from the track centreline to the hoarding, WG1 is the semi-width of the continental G1 gauge and WW6A is the semi-width of the UK W6A gauge. This correction should be regarded as only a first approximation, since the actual vehicles may well be smaller than the relevant gauge. Figures 4(a) to (c) show such a comparison between the hoarding experimental pressure coefficient data and the code provisions, with no corrections and the distance and pressure coefficient corrections, respectively. Note that the code values, shown as continuous lines for positive values and broken lines for negative values, do not apply for a distance from the track centreline of less than 2.3 m. The symmetric nature of the code values for the positive and negative pressure peaks can be seen. The conservative nature of the uncorrected code provisions is clear. In general, the experimental results are in broad agreement with the code values for both types of correction, and it is not altogether obvious which is to be preferred. Figure 5 shows the comparison with the experimental pressure coefficient results for 10 m wide overbridges and the code values, with no correction and using the distance increment correction to the code values. Note that in accordance with the methodology used in deriving the code values, the actual maximum experimental peak values are used, rather than the 5 m average values. The experimental values are the centre line values (y ¼ 0), and make no allowance for the fall off in pressure away from the centreline described in Part 1 of this paper. The uncorrected code values are again clearly conservative. For the Class 390 and Class 158 the corrected code values are still conservative, although for the Class 66, the experimental negative peaks are significantly higher than the corrected code values. The reasons for this rather poor agreement may well lie in the type of analysis used in the evaluation of the experimental data from ERRI D189/RP1 for the European Committee for Standardisation code, that effectively assumed a symmetry between positive and negative peaks. However, the precise nature of this analysis is not clear from ERRI D189/RP1 in terms of how this concept of symmetry is applied. Figure 6 shows a comparison between the experimental results for the 4.7 m high (above the track) open canopy pressure coefficient distribution and the equivalent code values, again with no correction and with the distance correction applied to the code values. In this case, both the horizontal and vertical


Baker et al.

61

(a)

(b)

(c)

Figure 4. Comparison of experimental pressure coefficient results with code values for 2 m high trackside hoardings: (a) no correction; (b) distance increment correction; (c) pressure coefficient correction, after Johnson and Dalley4 - applied to the pressure coefficient values from BS EN 4067-4 and given in equation (1)). Note BS EN 4067-4 only valid for distances from track centre >2.3 m).

(a)

(b)

Figure 5. Comparison of experimental pressure coefficient results with corrected code values for 10 m wide overbridges (a) no correction and (b) distance increment correction.

gauge increments have been applied. Experimental results were only available for the freight train case, and the code makes no provision for trains of different types. There is again reasonable agreement for the corrected values, although they are again non-conservative for the negative peak values. Thus, it can be seen that there is a broad equivalence between the current results and the

corrected code values, although the lack of transparency in the calculating of code values from experimental results, and the necessity to impose corrections for loading gauge differences, makes it difficult to say any more than this. Nonetheless it does appear that the corrected Class 66 experimental values can at times exceed those indicated in the code.


62


(a)

(b)

Figure 6. Comparison of experimental pressure coefficient results with corrected code values for 4.7 m high canopy (a) no correction and (b) distance increment correction.

Analysis of experimental data In this section an analysis of the experimental data of Part 1 of this paper is presented in a form that is consistent with that used in BS EN 14067-4 i.e. with the maximum values calculated over a 5 m distance before and after the major zero crossing between the positive and negative peaks. For the sake of consistency this procedure is carried out for all structures, although, as pointed out above, it seems that for overbridges the code is based on peak rather than average values.

Hoardings experimental data Figure 7 shows the experimental pressure coefficient data for trackside-mounted and platform-mounted hoardings. These values were obtained by the ‘moment average’ method as outlined previously. As before the Class 66 peaks have the largest magnitudes and the Class 390 the smallest. The positive pressure peak drops off with hoarding distance from the track/ platform edge in a consistent way. However, the behaviour of the negative peak is somewhat different, particularly for the trackside hoardings, where the variation with distance from the track is complex. There is significant positive/negative peak asymmetry here, reflecting the asymmetry in the pressure distributions.

Overbridge experimental data Figure 8 shows the maximum and minimum peak pressure coefficients for overbridges. For the 10 m wide overbridges of different heights, there can be seen to be a reduction in coefficient values as the height is increased, with the Class 66 pressure magnitudes being the largest and the Class 390 the smallest. For the 4.5 m high overbridge of different widths, there can be seen to be little change to the coefficients for widths greater than 3 m, but the 1.5 m width values

are less. However, we must question whether the use of a 5 m averaging length is appropriate for overbridges with a width in the train direction of less than this amount. For bridges less than 10 m in width, an averaging length of half the bridge width might be most appropriate, which would result in rather greater loads. The results are broadly symmetric, with similar values for the positive and negative peaks, although there is a tendency for the latter to fall off less as the height increases.

Canopy experimental data Figures 9(a) to (c) show the average canopy forces for the different trains, plotted in terms of both canopy height and back wall distance. Only the smallest and largest back wall distances and heights are shown for clarity, the data for the other tests carried out falling between these extremes. An examination of the figures shows clearly there is a decrease in the maximum peaks with height as seems reasonable, but such an effect is less apparent for the negative peaks. In general, the closer the back wall, the higher the canopy loadings.

Trestle platform results Figure 10 shows the maximum and minimum peak pressure coefficients analysed in the same way as above for the trestle platform results. The results are broadly as expected, with a decrease in coefficient from the platform edge, and the Class 66 coefficients being the highest and the Class 390 the smallest. The results again show an asymmetry, with the negative peaks being larger than the positive peaks.

Development of new codification format Outline In this section we will consider how the experimental results can be put into a suitable format for


Baker et al.

63

(a)

(b)

Figure 7. Maximum pressure coefficients for trackside and platform mounted hoardings (a) trackside-mounted hoardings and (b) platform-mounted hoardings.

(b)

(a)

Figure 8. Maximum pressure coefficients for overbridges (a) 10 m wide overbridges of different heights and (b) 4.5 m high overbridges of different widths.

supplementing the existing information in BS EN 14067. In particular we look to obtain curves of the following form used in the code. For vertical structures next to the track (hoardings) p1k ¼ 0:5V2 kCp1 , Cp1 ¼ f1 ðY or Y0 Þ

ð4Þ

where Y is the distance of the hoarding from the track centreline and Y’ is the distance from the platform edge. For horizontal structures above the track (overbridges) p2k ¼ 0:5V2 kCp2 , Cp2 ¼ f2 ðh, WÞ

ð5Þ

where h is the overbridge height above the track and W is the width of the overbridge in the along track (x) direction. For canopies with back walls p3k ¼ 0:5V2 kCp3 , Cp3 ¼ f3 ðh, YÞ

ð6Þ

where h is the height of the canopy above the track and Y is the distance of the back wall from the track centreline.

For trestle platforms, we consider curves for local loading of the form p4k ¼ 0:5v2 kCp4 , Cp4 ¼ f4 ðy0 Þ

ð7Þ

where y’ is the distance from the edge of the platform. Note that the factor that describes the effect of train type k will be taken to be the same in each case in order to eliminate the effect of train type. In addition for the overbridge case, it might be useful to parameterise the variation of load with distance across the track, although this is not taken into account for the existing codification process in BS EN 1991-2 and BS EN 14067. In the next section, we first consider the determination of the factor k that describes the effect of different train types. In the following four sections we consider the case of hoardings, overbridges, canopies and trestle platforms, respectively, and develop the appropriate form of the functions Cp1 to Cp4.


64


(a)

(b)

(c)

Figure 9. Maximum pressure coefficients for canopies (numbers in legend indicate back wall distance from nearest rail).

Figure 10. Maximum pressure coefficient distribution for trestle platform.

The effect of train type To determine the effect of train type, for each of the train/structure configurations described in Part 1, four ratios were found: the ratios of the peak positive and negative pressure coefficients for the Class 390 and Class 158 trains, to the equivalent values for the

Class 66 train. These ratios were then used, giving each of the configurations equal weight, to find the average ratios for the Class 390 peaks to those for the Class 66 peaks, and the same ratio for the Class 158 peaks to the Class 66 peaks. The tabulated results are shown in Table 1. Note that this weighting of the different configurations is purely arbitrary, but the


Baker et al.

65

Table 1. Ratios of 5 m average peaks. Bridge variable height

Bridge variable width

Trestle

Hoarding trackside

Hoarding platform

Canopy 1

Canopy 2

Canopy 3

Canopy 4

4

4

1

3

3

4

4

4

4

Class 390 positive peak

0.421

0.409

0.473

0.494

0.547

0.417

0.45

0.412

0.462

0.447

0.426

Class 390 negative peak Class 158 positive peak

0.535 0.521

0.707

0.256 0.442

0.357 0.474

0.402 0.6

0.368 0.58

0.323 0.603

0.291 0.574

0.277 0.605

0.404 0.562

0.526

Class 158 negative peak

0.569

0.442

0.474

0.6

0.496

0.448

0.461

0.415

0.489

Number of configurations/ weighting in calculating average values

(a)

Average values

(b)

Figure 11. Maximum pressure coefficients for hoardings scaled with k factors to allow for the effect of different train types. (Vertical axis shows pressure coefficient/k; (a) trackside hoardings and (b) platform hoardings, solid and dotted lines show envelopes to data.)

results are not particularly sensitive to the weighting used. These ratios, (i.e. k values), have values of 0.43 for the Class 390 to Class 66 loads, and 0.53 for the Class 158 to Class 66 loads (and 1.0 of course for the Class 66). Note that these are significantly less than would be expected from BS EN 14067 (0.65 for streamlined trains and 0.80 for ordinary passenger trains), but reflect the experimental results that were obtained. It may well be that this difference arises because the locomotives that were used in the freight train experiments of ERRI D189/RP1, on which the results of BS EN 1991-2 and BS EN 14067 were based, SNCF BB7200 and BB9200, have frontal shapes that, while blunt, are a little more rounded than those of the Class 66. Also, note the large spread of the results from the different configurations shown in Table 1.

Hoardings Figure 11 shows the hoarding experimental peak coefficients (Cpe) divided by the average k values for each train type. An envelope curve is also shown for the trackside and platform hoardings that encompass the data in a conservative way. Note that the curve

envelopes are symmetrical i.e. the lower value is simply the negative upper value. Such an approach preserves the format of BS EN 14067 but clearly involves significant conservatism. These curves are given by Cp1 ¼

6:0 ðY þ 1:75Þ2

ð8Þ

for the trackside-mounted hoardings and Cp1 ¼

ðY0

3:8 þ 2:5Þ2

ð9Þ

for the platform-mounted hoardings.

Overbridges We note firstly from Figure 8, that changes in overbridge width do not cause significant changes in loading, and thus the pressure coefficient in equation (5) will be taken to be a function of overbridge height only. Again the experimental pressure coefficient data was scaled with the appropriate k factors for each train type and upper and lower envelope curves derived. The results are shown in Figure 12.


66


Figure 12. Maximum pressure coefficients for overbridges scaled with k factors to allow for the effect of different train types (solid and dotted lines show envelopes to data).

The upper and lower envelope curves are given by the following form Cp2 ¼

6 ðh 1Þ2

ð10Þ

Again the conservatism forced by the symmetry assumption can be seen. Although no specifically allowed form for lateral pressure variation was included in the formulations for the codification values, it is also possible to fit envelope curves to the lateral variation of pressure coefficient with loading for the overbridges that were tested. Such a method results (Figure 13) in the following envelope curve Cp2 ðyÞ ¼ Cp2 ð0Þð1 0:03y2 Þ

ð11Þ

It should, however, be noted at this point that these results apply to a bridge across a single track only. For two-track running it will not be appropriate to allow for the lateral variation in the same way, as it is possible that trains may exist on both tracks at the same time, although, at the cost of a little complication, the lateral variation could be applied to the structure over both tracks on the outer side of the tracks, with a constant load value in the central portion.

Canopies The loads on canopies with back walls are clearly a function of the height of the canopy and the back wall distance. To find the variation with back wall distance, the ratios were formed of the loads at each

back wall distance from the nearest track to that at a distance of 2.7 m from the nearest track. The results are shown in Figure 14. Each point on this figure is an average of the ratio for four points: one for each canopy height for a particular train type and back wall distance. The envelope curve is given by r ¼ 1 0:1ðY 3:45Þ2

ð12Þ

where Y is the distance from the track’s centreline. The experimental pressure coefficients were then divided by the appropriate values of k and r and the results are shown in Figure 15. An upper and a lower envelope line is also shown and is given by Cp3 6 ¼ r ðh 0:1Þ2

ð13Þ

which leads to Cp3 ¼

6 1 0:1ðY 3:45Þ2 2 ðh 0:1Þ

ð14Þ

Trestle platform Figure 16 shows the trestle platform experimental results divided by the appropriate value of k to remove the effect of train type, together with upper and lower envelope curves. These are given by the following expression Cp4 ¼

3 ðy0 þ 2:2Þ2


ð15Þ

Baker et al.

67

Figure 13. Lateral variation pressure coefficient ratios for the overbridge results, scaled with centreline values.

Concluding remarks There are a number of basic assumptions made in the codification process used in BS EN 14067 that are worthy of comment at this point. 1. The positive and negative pressure peaks can appropriately be averaged over a 5 m distance. This is potentially a non-conservative assumption in situations where it may be necessary to consider a shorter distance (e.g. narrow overbridges), as the absolute peak values can be much higher than the 5 m average values. Also, such a procedure is certainly not appropriate for a structure of less than 5 m width, (such as many pedestrian overbridges). It might therefore be appropriate to consider

whether or not a shorter averaging length might be more appropriate for some structures and some trains. 2. The codification method contains an implicit assumption of symmetry between the positive and negative peaks, although the precise nature of this assumption is unclear i.e. whether the absolute maximum or absolute minimum value is chosen as the defining case. The results presented here show that this is rarely the case, with the positive peak being generally larger in magnitude than the negative peak, and this assumption leads to major conservatism in the derivation of design curves; a pressure coefficient envelope that comes close to, say, the positive peaks, may significantly overestimate the negative peaks, and vice versa.


68


Figure 14. Ratio of canopy loads to load with a back wall distance 2.7 m from nearest track (3.45 m from track centreline) (solid and dotted lines show envelopes to data).

Figure 15. Canopy experimental pressures divided by k and r to correct for train type and back wall distance (solid and dotted lines show envelopes to data).

It may be worth considering whether the benefits of relaxing this assumption outweigh the increased code complexity that would result. 3. For horizontal structures above the track, such as overbridges, the assumption made in the code that the peak track centreline pressure can be applied over a 20 m span has been shown to be very conservative, both from the current experiments which show a significant decrease in pressures away from the track centreline, and from a comparison with the full- scale measurements made on an overbridge (Part 1). For a bridge over a single track, it would seem appropriate to make some

allowance for this through the use of a length reduction factor in the code, although for two tracks, with the possibility of a train on both tracks beneath the bridge at the same time, such an allowance is probably not appropriate. 4. The code implicitly assumes static loading throughout. Now while this is entirely appropriate in most cases, for some lightweight or flexible structures other information is needed to enable, say, fatigue calculations to be carried out. Such calculations would not only require relative short-term loadings, but would also require some information on the frequencies associated with the


Baker et al.

69

Figure 16. Experimental pressure coefficients for trestle platform divided by k value (solid and dotted lines show envelopes to data).

pressure transient events. Extensive information of this type could be extracted from the current experimental data. Also, it is clear from reading the report of ERRI D189/RP1 on which BS EN 14067 is based, that much of the codified data was based on the use of theoretical panel method computational fluid dynamics (CFD) calculations from the early 1990s. These were essentially unverified at the time, and in the light of current developments in CFD their adequacy cannot necessarily be assumed. These results either need to be independently verified or replaced by more reliable data. There is a case that could be made for an extensive series of tests using the TRAIN Rig methodology to measure the loads on structures caused by continental gauge trains, in order to replace some of the less reliable data in the current code. Thus, in conclusion it can be seen that the outputs from the project have highlighted a number of areas where further work could offer benefits to the railway industry. These are as follows. 1. There is a need to investigate whether the degree of conservatism in the various assumptions, and in differences between the existing pressure curves in BS EN 1991-2 and BS EN 14067 and the experimental results, make any practical or economic difference in the design of typical railway structures. 2. A robust ‘gauge’ correction method can be developed and verified based on the experimental results and some checks for the actual distance between real trains and line- side structures rather than the distance from the train ‘gauge’. This method could be included within the UK National Annex to BS EN 1991-2 and BS EN 14067.

3. There is a need to identify appropriate fatigue load spectra (magnitude and frequency) for fatigue-sensitive structures. Funding The experimental programme was undertaken in connection with the RSSB-funded research project T750 ‘Review of Euronorm design requirements for trackside and overhead structures subjected to transient aerodynamic loads’, which was sponsored by the railway industry ‘Aerodynamics GB Working Group’. Permission of BSI for the use of figures 1, 2 and 3 is also acknowledged.

References 1. BS EN 1991-2; 2003. BSI Eurocode 1 Actions on structures – Part 2: Traffic loads on bridges. 2. BS EN 14067-4:2005þA1:2009. BSI railway applications — aerodynamics — Part 4: Requirements and test procedures for aerodynamics on open track. 3. ERRI D189/RP1, 1994. Loading due to dynamic pressure and suction from railway traffic. Effect of the slipstreams of passing trains on structures adjacent to the track. 4. Johnson T and Dalley S. 1/25 scale moving model tests for the TRANSAERO Project. In: B Schulte-Werning, R Gregoire and A Malfatti (eds) In: TRANSAERO- A European initiative on transient aerodynamics for railway system optimisation. Berlin, Germany: Springer-Verlag, 2002, pp.123–135.

Appendix 1 Notation Cpi h


code pressure coefficient (i ¼ 1 to 4) distance from top of rail to overbridge/ canopy

70 k k1 k2 k3 pik r V W WG1 WW6A

Proc IMechE Part F: J Rail and Rapid Transit 228(1) parameter that specifies the effect of train type k parameter in equation (1) k parameter in equation (2) k parameter in equation (3) peak pressures used in code ratio defined in equation (12) train/model velocity width of overbridge in x-direction semi-width of G1 loading gauge semi-width of W6A loading gauge

z z0

distance along the track lateral distance from centre of track lateral distance from edge of platform lateral distance of vertical structures from centre of track lateral distance of vertical structures from platform edge vertical distance from the track vertical distance from top of platform

density of air

x y y0 Y Y0
