Original Article Rail irregularities, corrugation and acoustic roughness: characteristics, significance and effects of reprofiling Proc IMechE Part F: J Rail and Rapid Transit 226(5) 542–557 ! IMechE 2012 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0954409712443492 pif.sagepub.com Stuart L Grassie Abstract Railway noise is excited by irregularities on the running surfaces of wheels and rails, with rails being more significant for most of the wavelength and frequency range of interest. Measurements are presented that demonstrate the significance of irregularities in the 100–1000 mm wavelength range (corresponding roughly to 25–250 Hz) on ground-borne noise, and show that a reprofiling specification to address reductions in low frequency noise should address these longer wavelengths. Differences are demonstrated between irregularities in the 6.3–2000 mm wavelength range on a dedicated high speed line, and on heavy haul, mixed traffic, metro and light rail systems. Although the sample size is small for the high speed and heavy haul systems, the lowest levels of irregularity were nevertheless found on the high speed line. Heavy haul systems also have low levels of irregularity, which are below the limit specified in the acoustic standard ISO 3095. Metro systems are distinguished by corrugation, higher levels of short wavelength roughness corresponding perhaps to roughening from consistently high traction and braking, and significant differences between high and low rails in some curves. In one case shown here the high rail is amongst the smoothest in an entire sample of dozens of rails whereas the corresponding low rail is by far the most heavily corrugated and irregular. Light rail systems have relatively high levels of irregularity, particularly at short wavelengths. This may result from sanding to enhance adhesion. Reprofiling in general reduces irregularities below the level stated in ISO 3095, except at wavelengths of less than 30 mm for which irregularities are consistently greater after reprofiling than before. Reprofiling is particularly effective in reducing irregularities in the 30–500 mm wavelength range, in which irregularities are typically reduced by 10 dB but in some cases by 20–30 dB. Irregularities to the ISO 3095 limit spectrum exist immediately after some forms of grinding. The best results are obtained from reprofiling that is undertaken to an appropriate specification that is conscientiously monitored. Keywords Railways, railway noise, acoustic roughness, rail corrugation, roughness measurement, ground-borne noise Date received: 14 November 2011; accepted: 6 March 2012 Introduction Irregularities on the running surfaces of wheels and rails initiate dynamic loads and vibration of vehicle and track components. One of the most widely noticed and unfortunate consequences is noise of many types with which railways are associated and which in many circumstances constrains development of this form of transport, whose environmental credentials are otherwise exemplary. The literature in this area is vast with much of the work now being basic reference material. For example, the recent book by Thompson1 provides a topical, practical and wide-ranging view of the area by an acknowledged expert, with treatment of the different types of noise that are excited by irregularities (ground-borne, structure-borne and rolling noise) as well as squeal and aerodynamic noise, whose mechanisms of excitation are fundamentally different. Contributions to two recent ‘best practice’ manuals give a useful summary not only of mechanisms of wheel/rail noise generation but also of some of the more effective mitigation measures for both air-borne and ground-borne noise.2–4 This work is some of the RailMeasurement Ltd, Cambridge, UK Corresponding author: Stuart L Grassie, RailMeasurement Ltd, The Mount, High Street, Toft, Cambridge, CB23 2RL UK. Email: [email protected] Grassie 543 Figure 1. Simplified model of railway noise generation. most recent since a groundswell of activity in this area was initiated in the 1970s, giving rise to the first Workshop on Railway Noise in Derby in 1976. The International Workshops on Railway Noise have continued at 3-yearly intervals since, providing a rich and readily available source of reference material on wheel/rail noise. Many of the references here are from these workshops. The significance of irregularities in generation of structure-borne, ground-borne and rolling noise is commonly recognised. There are several published works showing measured spectra of roughness on rails and wheels and others where spectra have been assumed. Some of this work is discussed below. The review is intended to give not only an historical view of the area but also to explore the more significant contributions to measurement of rail roughness. The vast majority of available references are for relatively short wavelengths (typically less than about 0.3 m). This is barely adequate for representing the irregularities that contribute to structure-borne and rolling noise. It is an inadequate wavelength range for ground-borne noise, as demonstrated in the section ‘The influence of irregularities on ground-borne noise’. A few exceptions show irregularities to significantly longer wavelengths. There are some references that show differences in railhead roughness in different locations, on different types of track or the effects of grinding on corrugation. However, there is no comprehensive reference that shows the range of irregularities that can exist in different circumstances. This paper attempts to fill some of these gaps by presenting measurements from different railway systems and in different conditions. Effects of reprofiling, both grinding and milling, are shown, and spectra are presented to far longer wavelengths than the common limit of a few hundred millimetres (in the section ‘Railhead irregularities: a wide ranging survey’). An attempt is made to explain why some of the differences exist on different types of track and also to explore the limitations of reprofiling on reduction of irregularities (in the section ‘Discussion’). Although some reference is made to the literature in the complementary area of wheel roughness and irregularities, this paper concentrates on rail roughness and irregularities. The objective of work such as this is clearly to reduce railway noise. The principal area of concern here is reduction of noise from reducing irregularities on rails, with practical examples of what can be achieved. Reference is made en passant to other methods of reducing railway noise. Literature on railway noise and measurement of rail roughness ‘Railway noise’ in the present context is considered to be ground-borne, structure-borne and rolling noise. These types of noise are excited to a greater or lesser extent by irregularities on the surface of wheels and rails. The work by Remington and his colleagues, which was undertaken for the US Department of Transportation in the 1970s, is widely recognised to be both seminal and comprehensive in its treatment of different noise sources on the railway.5–7 The model of wheel/rail noise generation proposed in Remington et al.6 is found in more elaborate form in the later review by Remington8 which followed an extremely fallow period in railway noise research in the USA. Many of the open questions posed in Remington’s review in 1988 were substantially answered by work undertaken in Europe that gave rise in particular to the TWINS model of rolling noise9 and the comprehensive treatment in Thompson’s doctoral research.10 The 1970s were also a productive period in the USA in both ground-borne11 and structure-borne noise,12,13 primarily with regard to metro systems. This work was also sponsored by the US Department of Transportation and adopted a characteristically practical approach of combining an understanding of the cause of problems with development of means of control. In the 1980s the Organisation for Research and Experiments (ORE) of the International Union of Railways undertook a wide-ranging study on ground-borne noise and vibration, which included modelling of these phenomena and practical methods for control.14 A simple and slightly more general representation of Remington and Thompson’s model that is relevant for present purposes is shown in Figure 1. 544 Proc IMechE Part F: J Rail and Rapid Transit 226(5) Table 1. Wavelength ranges (in millimetres) corresponding to frequency ranges of interest. 16 Audible ground-borne noise Structure-borne noise19 Wheel–rail rolling noise18 Range of greatest human sensitivity15 Range of human sensitivity15 25–250 Hz 100–2000 Hz 100–5000 Hz 2–5 kHz 20 Hz–20 kHz Table 2. Wavelength ranges considered in relevant European and International standards. EN ISO 3095 EN 15610 EN 13231-3 Minimum (mm) Maximum (mm) 3.15 3.15 10 630 250 1000 This distinguishes three general areas in the mechanism of railway noise generation: . excitation, from wheel and rail irregularities; . the dynamic response of the vehicle/track system, which comprises two parts: generation of dynamic forces as a result of irregularities; the response to these dynamic forces, which takes different forms depending on the specific type of noise of interest (ground- or structure-borne noise or rolling noise); . propagation of noise from the vibrating surfaces to the receiver. A basic question that arises with regard to irregularities is the wavelength range of interest. People are usually regarded as being sensitive to noise in the frequency range 20 Hz – 20 kHz, with the most sensitive range being about 2–5 kHz.15 On the other hand, the frequency range for audible ground-borne noise is about 25–250 Hz.16 The frequency range of importance for rolling noise is considered to be 200–2000 Hz by Galaitsis and Bender17 or 100–5000 Hz by Thompson.18 From the review by Wittig13 it could be assumed that the frequency range of interest for structure-borne noise is about 250–2000 Hz, whereas the more recent work by Bewes et al.19 indicates a slightly broader frequency range of about 100–2000 Hz. Frequencies of more than 2 kHz at 160 km/h are attenuated primarily by the effect of the so-called ‘contact patch filter’, which is discussed in the ORE reports14 20 m/s (72 km/h) 50 m/s (180 km/h) 80–800 10–200 4–200 2–10 1–1000 200–2000 25–500 10–500 5–25 5–5000 and elsewhere. Frequency ranges and corresponding wavelengths for a couple of train speeds, typical of metros and mainline traffic, are shown in Table 1, which illustrates the very wide range of wavelengths of irregularity that are of interest in railway noise. The wavelengths follow the well-known formula, ¼ =f, where is the train speed and f is the frequency. Three European or International standards are particularly relevant to rail irregularities and noise: . EN ISO 3095,20 which specifies the acoustic roughness of a test section to be used for acoustic typetesting of railway vehicles; . EN 15610,21 which prescribes a method of measuring railhead roughness; . EN 13231-3,22 which prescribes residual irregularities that may remain on a reprofiled rail. The 2006 version of this standard is referenced as the limits of irregularities are more relevant to noise reduction than the generous limits in the 2011 version of the standard. The wavelength ranges in which irregularities are specified in these three standards are shown in Table 2. The reprofiling standard covers the widest wavelength range of the three and is the only of the three standards that is at all adequate for groundborne noise. The wavelength range of EN 15610 is adequate for rolling noise for relatively low speed traffic but scarcely adequate for main-line passenger traffic. The vast majority of available references consider irregularities that are relevant to rolling noise or the equivalent dynamic forces. Galaitsis and Bender17 present one of the earliest sets of measurements of both wheel and rail roughness, with one-third octave spectra for wavelengths of about 1.5–150 mm, which was appropriate for rolling noise in the metro systems of interest. Combined spectra of wheel/rail roughness, for wavelengths of about 4.3–280 mm (125 Hz to 8 kHz at 125 km/h), are shown in Thompson et al.9 where these are used to validate the TWINS model of wheel/rail rolling noise. Verheijen23 presents direct and indirect measurements of rail roughness in the Grassie wavelength range 5–315 mm from the Dutch railway network. The difference in the roughness spectra between roughest and smoothest rails measured is about three orders of magnitude over a wide wavelength range. A complementary study covering a narrower wavelength range of about 1.3–80 mm is contained in Hiensch et al.24 Also, Nielsen25 has used measurements of wheel and rail roughness in the 5–160 mm wavelength range to validate a model that predicts high frequency contact forces. This work has been extended with measurements in the wavelength range 2–250 mm to suggest an acceptance criterion for rail roughness levels based on both rolling noise and development of rolling contact fatigue (RCF).26 Lower frequency contact forces are calculated by assuming that the longer wavelength rail roughness is that given by the spectrum in the standard EN IS0 3095:2005. Several sets of measurements from European railways, shown in Fodiman and Staiger27 provided the background to limits on rail roughness that have been proposed to limit noise levels from European railway operations. Power spectral densities of railhead irregularities for wavelengths of more than 20 mm are shown in Grassie28 for a field investigation in which routine measurements were made of rail corrugation development on new and ground rails. It can be deduced from published measurements17,18 that typical wheel roughness is of a similar order of magnitude to typical rail roughness at shorter wavelengths but of very much lower amplitude at long wavelengths. This is illustrated also in Figure 2: over the majority of the frequency (or wavelength) range, irregularities on rails are more significant than those on wheels. Measurements are shown for two wheels of a passenger coach with composite brake blocks before running and after running less than 2000 km, for the first few hundred of which the brakes were dragging lightly in order to remove the periodic irregularities that resulted from reprofiling. One-third octave spectra of roughness are also shown for a test site in track that was measured in accordance with the protocol in EN IS0 3095:2005. Two instruments were used for this purpose: one based on displacement transducers that run on a stationary straight-edge29 and the so-called corrugation analysis trolley (CAT)30 (Figure 3). Another comparison of measurements from different instruments, including the two used here, is included in Jones et al.31 and was made as part of a ‘road test’ for the measuring protocol that is now contained in EN 15610:2009. It is suggested in that standard that instruments based on a chord of 1 m length produce results that are acceptable only to wavelengths of 0.100 m. Nevertheless such instruments are often used to measure longer wavelength irregularities. 545 Figure 2. Comparison of wheel and rail roughness. Figure 3. The CAT in use. The ability of an inertial instrument such as the CAT to measure long wavelength irregularities is not similarly constrained, although there are somewhat different constraints. Record lengths can be long: almost all samples are longer than 100 m and in one case a CAT record of 14 km length has been used. It is nevertheless extremely difficult to develop a traceable method of demonstrating accuracy of long wavelength measurements because instruments such as these are amongst the most accurate, if not the most accurate, available. Reference is made in the open literature to other equipment e.g. a straight-edge device to measure rail and wheel roughness,17 trolleys with displacement transducers32,33 and both a straight-edge device and an ancestor of the CAT.28 Merits and principles of different techniques are reviewed in Grassie.34 In the last few years the majority of references that include 546 measurement of railhead irregularities have used equipment that is commercially available and which is described in Diehl and Holm,29 Grassie et al.30 and elsewhere. In view of the limitations of the straight-edge equipment that is often used to measure rail roughness, the dearth of measurements of long wavelength irregularities relevant to low frequency noise and vibration is unsurprising. However, a similarly successful approach to resolving this difficulty has recently been adopted by teams in Sweden35 and in the UK.16 Measurements of irregularities has been made using a track recording car (TRC) and the CAT. In both cases, the one-third octave spectra correlate well in the wavelength range around 1–3 m which is the upper limit of wavelength measurable with the CAT and the lower limit with the TRC. The measurements presented in Bongini et al.36 demonstrate that irregularities up to at least 1 m wavelength can be measured repeatably with at least some systems that process axlebox accelerations. Chua et al.37 found reasonable correlation up to about 200 Hz of measured vibration levels on floating slab track in an underground system and those calculated assuming a spectrum of irregularities measured by British Rail. The spectrum is rather puzzling as it contains components that are a function not only of wavelength but also of velocity. Wittig13 develops a model that includes rail roughness to calculate structure-borne noise, and while he demonstrates satisfactory correlation of calculated and measured noise levels he fails to show the spectrum of roughness that has been assumed for the calculation. The influence of irregularities on ground-borne noise There is a wealth of information on the effects of shorter wavelength roughness on rolling noise that is sufficient to prescribe how particular sections of track should be ground in order to maintain noise below specified limits e.g. Asmussen et al.33 Less information is available on the effects of irregularities on groundborne noise although there is at least colloquial acceptance that irregularities exacerbate noise and vibration. For example, the ORE D151 study14 mentions that wheelflats in particular increase levels of groundborne vibration while Wilson et al.11 mention rail grinding and wheel truing amongst methods to control ground-borne noise and vibration. Neither of these studies nor more recent work demonstrates these effects. London Underground has hundreds of kilometres of track that were built before noise became a problem of public concern. On newer sections of the Underground it has been possible to use vibration-isolating trackforms e.g. the particularly successful design at the Proc IMechE Part F: J Rail and Rapid Transit 226(5) Figure 4. Correlation of reductions in ground-borne noise (dBA) and railhead roughness (dB re 1 micron), data courtesy of London Underground. Barbican development,14,38 which dates from the 1960s. This is not an option for older track, which would cost billions to rebuild. About 15–20 km of track in the most noise-sensitive locations is therefore ground annually to reduce noise. Irregularities have been monitored routinely by the grinding contractor or the Underground for more than a decade. A CAT is now used for this purpose,30 although the Underground previously used a straight-edge device similar to that described in Diehl and Holm.29 Measurements of irregularities before and after grinding have been obtained from four locations for which corresponding pre- and post-grind measurements are available of in-property noise levels. The roughness measurements have been analysed very simply to give root mean square (RMS) levels in short and long wavelength ranges (30–100 mm and 100–1000 mm) over about 300 m of track adjacent to the properties of interest. For trains at a typical maximum speed of 80 km/h, the corresponding frequency ranges are 220–740 Hz and 22–220 Hz. The reductions in noise (in dBA) are graphed in Figure 4 relative to the reductions in roughness (expressed as dB relative to 1 micron RMS). Grinding brings about a reduction in ground-borne noise of about 2–8 dB. There is a reduction in roughness of about 2–9 dB in the 100–1000 mm wavelength range and up to about 16 dB for shorter wavelengths. In one case there is a slight increase in short wavelength roughness immediately after grinding. This is a very small sample of measurements that were taken with the independent objectives of monitoring London Underground’s grinding specification and quantifying the benefits of grinding on noise reduction. It is nevertheless demonstrated that there is moderately good correlation of the reduction in ground-borne Grassie 547 Figure 5. Irregularities on rails after reprofiling (a) one-third octave spectra of roughness and (b) relative to ISO 3095 limit. noise with reductions in railhead irregularities in the 100–1000 mm wavelength range, corresponding roughly to the 25–250 Hz frequency range given in Table 1. On the other hand, there is no correlation between ground-borne noise and shorter wavelength irregularities. If such a correlation were demonstrated more widely, it would suggest that a reprofiling specification for reduction of ground-borne noise should not only address the wavelength range corresponding to frequencies of about 25–250 Hz but should also ensure that reprofiling is undertaken to that specification. Railhead irregularities: A wide ranging survey It had become clear to the author from his work on rail reprofiling and corrugation in particular that there may be interesting differences and similarities between 548 Proc IMechE Part F: J Rail and Rapid Transit 226(5) Figure 6. Change in irregularities on rails as a result of reprofiling. irregularities on different types of railway track. However, it would be impractical for anyone to measure sufficient track to discover whether trends observed from a relatively small sample were indeed significant. Fortunately there are many users of the CAT instrument worldwide, and in 2010 these users were asked if they could provide measurements from different railway systems and in particular of track before and after reprofiling. An excellent variety of data was provided by many users of this equipment, including research workers, consultants, suppliers, railway systems and reprofiling contractors. There are measurements from many railway systems and of rail in different condition. Measurements were provided from rail that had been reprofiled using equipment from four different suppliers of grinding trains and one milling train. Those who contributed data are acknowledged here, as promised, but there is no indication of which measurements come from which railway or from which type of reprofiling train. It was clear from the measurements of reprofiled rail that a similarly high and similarly low standard of reprofiled rail can be supplied by every type of reprofiling equipment: it is not the reprofiling equipment itself that is significant in providing an excellent or poor finish, but how this equipment is used. Measurements of the effects of reprofiling and from different ‘types’ of railway system are presented in the following two subsections. Effects of reprofiling As the CAT is widely used for quality assurance of reprofiled rail, many sets of measurements were obtained for this condition and in all but a few cases the record length is significantly more than 100 m. The basic presentation of data is of the one-third octave spectra (Figure 5(a)). Post-reprofiling measurements are shown for one site on a high speed line, a mixed passenger and freight railway and a light rail system, and for eight sites on metros. One site on a metro was milled, all other sites were ground with conventional modules that rotate about an axis normal to the rail. Data are shown only if both rails had been measured. The solid bold line in Figure 5(a) corresponds to the rail roughness limit spectrum in EN IS0 3095:2005, which can be expressed as follows LISO 3095:2005 ¼ 9:7 þ 18:5 log 10 ¼ 9:7 for 510 for 10 4 53:15 ð1Þ where L is the roughness limit in decibels and is the wavelength in millimetres. This equation allows the ISO 3095 line to be extended to wavelengths beyond the 630 mm in the standard.20 Grassie 549 Figure 7. Irregularities on a dedicated high speed railway (note scale is 30 dB to þ15 dB). Figure 8. Irregularities on heavy haul lines. The broken bold line in Figure 5(a) is an equivalent spectrum for the RMS limits on allowable longitudinal irregularities after reprofiling that are specified in EN 13231-3:2006. Although the limits are for broad wavelength ranges of e.g. 30–100 mm and 100–300 mm, these can be recast to give the following spectrum, which is of similar form to that in EN IS0 3095:2005. 550 Proc IMechE Part F: J Rail and Rapid Transit 226(5) Figure 9. Irregularities on mixed passenger and freight lines. Figure 10. Irregularities on metro lines. LEN 132313:2006 ¼ 0:1695 þ 20 log 31:5 ¼ 5:0515 for 255510 for 531:5 ð2Þ The standards EN IS0 3095:2005 and EN 132313:2006 were developed with different objectives but there is surprisingly close correlation of the two limit spectra. The greatest difference is in the 10–30 mm wavelength range, the reasons for this behaviour are analysed in the ‘Discussion’ section. An alternative and more compact presentation of the data, which is essentially identical to that of Berggren et al.35 is made in Figure 5(b). The limit spectrum from Grassie 551 Figure 11. Irregularities on light rail and tram systems. ISO 3095 (equation (1)) has been subtracted from the data, thereby showing trends far more readily. In particular if the post-reprofiling roughness spectrum lies below the 0 dB line, the site has been reprofiled to the limit specified in ISO 3095:2005. Similarly, if the spectrum lies below the bold dashed line, the site has been reprofiled to the standard specified in EN 13231-3:2006. No attempt has been made in Figures 5(a) and (b) to differentiate between measurements from reprofiled rails at a total of 11 sites on four different types of railway system. This has been done deliberately to show the typical range of results that is obtained. Although a range of finish is obtained, residual irregularities exceed the EN 13231-3:2006 limit in only a few cases, regardless of the type of track or type of reprofiling equipment used. However, in almost all cases the level of short wavelength irregularities ( < 30 mm) on reprofiled rails exceeds the ISO 3095:2005 limit. Measurements were also available both before and after reprofiling for the single sites on the dedicated high speed and light rail systems, for five sites on metros and for a single rail on a mixed traffic system. In this case the measurements are presented as the difference in spectra of irregularities before and after reprofiling (Figure 6). Where measurements lie above the 0 dB line, reprofiling has brought about an increase in irregularities; the converse is the case if measurements lie below the 0 dB line. Irregularities on different types of railway system The measurements presented in Figures 7 to 11 respectively are for roughness spectra on the following types of railway systems: . Figure 7: dedicated high speed (four measurements from one line, pre- and post-grind); . Figure 8: heavy haul (four measurements from both rails on two lines); . Figure 9: mixed passenger and freight (12 measurements from both rails on six lines); . Figure 10: metros (eight measurements from both rails on four lines); . Figure 11: light rail or tram systems (eight measurements from six lines). Differences between the roughness spectra and the ISO 3095 limit spectrum are shown, as in Figure 5(b), for the wavelength range 6.3–2000 mm. A few measurements are available to shorter wavelengths (Figure 9). For the dedicated high speed line the measurements are the same as those used in the section ‘Effects of reprofiling’. Otherwise the measurements were not taken directly before or after reprofiling. A ‘line’ may be a railway system or a particular line on a railway system e.g. the two heavy haul lines were in Australia and the USA, whereas three of the tram ‘lines’ were on the same tram system. For all cases except the tram/light rail systems, spectra are shown for both 552 Proc IMechE Part F: J Rail and Rapid Transit 226(5) Figure 12. Typical irregularities on different railway systems. left and right rails at a site. The scale on the y-axis of all graphs is (20 dB/þ25 dB) except for Figure 7, where it is (30 dB/þ15 dB). In Figure 12 a selection of measurements has been made for each type of railway system, with measurements shown for both high and low rails in a metro curve. Discussion Irregularities on the dedicated high speed line before grinding are not only lower than the ISO 3095 limit throughout the wavelength range 6.3–2000 mm (Figure 7) but also by far the lowest of the measurements shown (Figure 12). Both pre- and post-grind measurements show evidence of an irregularity at fastener spacing at 630 mm and corrugation at about 125 mm. The irregularity at about 60 mm in pre-grind measurements may be light corrugation. Irregularities after grinding of this site are worse almost throughout the wavelength range, except insofar as the light irregularity at about 60 mm is removed. Irregularities at less than 60 mm wavelength have been particularly severely affected by grinding. Heavy haul lines are usually ground routinely to reduce RCF damage. Track geometry is also maintained to a high standard to enable high loads to be carried without damaging the track. Data are available from only two lines, albeit from different parts of the world on railways carrying different traffic (coal trains and mixed freight). It nevertheless appears (Figure 8) that routine grinding contributes to a relatively low roughness spectrum for most of the wavelength range, with irregularities in one case being consistently below the ISO 3095 limit, while a combination of grinding and maintenance of track geometry contributes to long wavelength irregularities that are barely 5 dB above the ISO 3095 limit: a factor of 1.8 on the RMS amplitude. For both lines there is a prominent effect of sleeper spacing at about 630 mm. Dawn39 has commented on the fact that there is a prominent peak in groundborne vibration (which would be reflected in groundborne noise) when the sleeper-passing frequency coincides with what is now known as the ‘P2 resonance’ of the unsprung mass on the track stiffness. This resonance is typically at 50–100 Hz, although it can be significantly lower (as in Dawn’s tests) for vehicles with a leaf-spring suspension that is shorted out by stiction. For one railway shown here there are prominent residual grinding marks at 25 mm, corresponding to a grinding speed of about 5 km/h. This is relatively slow grinding on a heavy haul railway. Spectral peaks from grinding have been noted elsewhere.36,40 The generally low level of irregularities except at longer wavelengths is doubtless a consequence of low traffic speeds, so that so-called ‘pinned–pinned resonance’ corrugation,41 for example, would be eliminated by contact patch filtering. There is a great variety of irregularities on mixed traffic railways (Figure 9). In almost all cases irregularities are lower than the ISO 3095 limit for wavelengths of less than about 30 mm, rising steadily with wavelength to 5–20 dB above this limit (corresponding to an increase in mean square or RMS irregularities of Grassie about 3.2–100 or 1.8–10, respectively) for wavelengths of more than a metre. This is a very wide range that probably reflects differing maintenance practices. In most cases the roughness for wavelengths of less than 10 mm is similar to that on the high speed line (Figure 7). At one site the roughness is relatively high for wavelengths of less than 50 mm. These measurements are from a railway at high latitude where sand, which is used for much of the year to obtain adequate adhesion, has probably caused this high broadband level of roughness. There is significant corrugation on many records in the 40–100 mm wavelength range. This probably reflects pinned–pinned resonance corrugation, with the broad wavelength range resulting from the broad speed range for these mixed traffic railways. In some cases there are irregularities at about 10 mm wavelength that may be residual grinding marks from relatively low speed grinding (about 2 km/h). Although not directly apparent from these measurements, it can be deduced (as pointed out by Hölzl and Werner40) that corrugation could be initiated if grinding took place at a speed that introduced a peak in the roughness spectrum close to the corrugation wavelength. This is a hazard that occurs quite frequently, particularly on metro lines, but is relatively easily avoided. Irregularities on metro lines are generally higher than the ISO 3095 limit at all wavelengths, with very prominent peaks in some examples that correspond to corrugation (Figure 10). Corrugation wavelengths are often well defined, reflecting the tighter speed control that is typical of metros. The worst corrugation and the smoothest rail in this sample of measurements are actually from the same site (see also Figure 12). The level of about 25 dB above the ISO 3095 limit in the 100 mm one-third octave wavelength range corresponds to an RMS corrugation level of almost 50 microns, which is a severe corrugation indeed at such wavelengths. The level of irregularities on the high rail is about 30 dB lower for much of the wavelength range. Even at wavelengths of less than 10 mm, roughness is below the ISO 3095 limit only on this single sample of metro rails. One possibility is that sustained high traction and braking on metro systems roughen the rails. Apart from a couple of outliers, six of the eight sets of data lie within about 5 dB of one another for wavelengths of more than 100 mm. This is remarkable consistency given that the four metro lines are in different countries. For almost all samples from tram or light rail systems, irregularities are not only significantly higher than the ISO 3095 limit throughout the wavelength range (Figure 11) but also greater than irregularities on other types of railway system except for the heavily corrugated low rail on a metro (Figure 12). There is also a wide range of irregularities at wavelengths of longer than about 150 mm, which is particularly 553 surprising given that three of the samples are from the same tram system. There is evidence of corrugation in some samples, with a quite well defined wavelength that reflects fairly tight speed control on such systems. Irregularities are significantly higher for almost all samples at short wavelengths, which may also reflect the common use of sanding and its effect on roughening rails. In almost all the cases examined here the level of irregularities on a rail after reprofiling are significantly below the EN 13231-3:2006 limit, as shown in Figure 5(b). In many cases reprofiling reduces irregularities up to 15 dB below this limit in the 30–500 mm wavelength range. In the vast majority of cases, reprofiling also reduces irregularities below the ISO 3095 limit except for wavelengths of less than about 30 mm in which the ‘grinding signature’ and general roughness from the grinding stones (and, in one case, milling cutters) makes it difficult to achieve an extremely smooth surface. It is clear from Figures 5(a) and (b) that the higher level of irregularities permitted by EN 132313:2006 in the 10–30 mm wavelength range compared to EN ISO 3095 is not only necessary for conventional reprofiling work but also sufficient. In two cases grinding has left a level of irregularities significantly higher than the limit of EN 13231-3:2006 for all wavelengths less than 100 mm. This is a poor standard of work. Oscillating grinding is often used to reduce irregularities at short wavelengths33,42 although a technique known as ‘offset grinding’ is also quite effective. The effects of conventional and offset grinding are illustrated in Figure 13 from measurements made over a 400 m test site that was ground specifically to see whether offset grinding could reduce irregularities to the ISO 3095 limit. Measurements were made immediately after grinding and before the passage of traffic. Irregularities were low before grinding throughout most of the wavelength range, although greater at short wavelengths (<30 mm) than would occur from normal traffic e.g. Figure 12. Conventional grinding significantly reduced irregularities in the 40–500 mm wavelength range with a slight reduction at longer wavelength. Irregularities at short wavelength (<30 mm) were greater after conventional grinding, with a prominent ‘grinding signature’ at 25 mm. Two passes of offset grinding reduce irregularities over the full wavelength range 6–300 mm, with a reduction of up to 8 dB for wavelengths of less than 30 mm, which are critical to rolling noise. The roughness spectrum after offset grinding exceeds the ISO 3095 limit for three adjacent one-third octave bands (10, 12.5 and 16 mm), which is permissible according to the standard;20 the maximum exceedence is 2.5 dB. The extent to which reprofiling reduces initial irregularities is informative. In almost all cases reprofiling 554 Proc IMechE Part F: J Rail and Rapid Transit 226(5) Figure 13. Effects of conventional and ‘offset’ grinding. Figure 14. Development of irregularities after reprofiling. reduces irregularities in the 30–1000 mm wavelength range by up to 30 dB, and typically by 10 dB. For wavelengths of less than 30 mm there is often an increase in roughness of more than 10 dB. For irregularities of more than 1 m wavelength reprofiling is also relatively ineffective: there is a typical reduction in irregularities of less than 5 dB (less than 50% in RMS amplitude). The effects of milling are very similar to that of typical grinding, with a reduction of about 10 dB in the 30–300 mm wavelength range and irregularities on the finished rail that are within those of EN 13231-3:2006. In one exceptional case (the high speed line of Figure 7) grinding increased irregularities at all wavelengths on rails that were initially extremely smooth: it is questionable why this track was ground. It follows from the above observation that if reprofiling increases the level of short wavelength irregularities ( < 30 mm, say), these irregularities typically decrease under traffic. At longer wavelengths irregularities sometimes increase and sometimes decrease depending on the type of track and traffic. This effect has been shown in measurements presented by Hölzl and Werner40 for a site immediately after grinding and a year later. It is also illustrated here in a more comprehensive set of measurements for a 100 m length of metro line which was known to corrugate both quickly and severely. Four sets of measurements are shown in Figure 14, from the day following grinding through to 92 days afterwards. Although the track was ground extremely well, there is a very small residual corrugation at 50 mm wavelength. Other prominent peaks in the one-third octave spectrum after grinding arise from the periodic grinding marks at 16 mm, with a harmonic at 8 mm. Traffic gives rise to a significant and continuous increase in the level of irregularities at wavelengths of longer than about 30 mm (particularly at the 50 mm corrugation wavelength) and a very pronounced decrease in irregularities at shorter wavelengths. The level of irregularities at the principal wavelength of the grinding signature (16 mm) is reduced by more than 10 dB in 3 months. This difference in behaviour is explained, for example, by Hempelmann and Knothe43 in their discussion of Grassie corrugation development. Longer wavelength irregularities or corrugation result from structural dynamics of the vehicle/track system. Shorter wavelength irregularities, for which the corresponding frequencies do not influence the structural dynamic behaviour, are reduced simply as a result of wear. In essence, wear that results from traffic reduces the level of roughness at all wavelengths unless structural dynamic behaviour exists to increase the level. If such behaviour does exist corrugation typically develops. The spectrum of corrugation is narrower the narrower the range of vehicle speeds over the site. With such a diversity of data from reprofiled track it is difficult to determine all of the reasons for the differences. Nevertheless one finding was clear. A consistently good standard of reprofiled rail was obtained where there existed a specification that was not only clearly stated but also monitored using equipment that was adequate for the purpose. This was the case with the equipment used here, but is surprisingly uncommon even on the majority of European mainline railway systems. In fact the relaxation between the 2006 and 2011 versions of the reprofiling standard EN 13231-3 is in part a consequence of the fact that equipment on most European rail grinding trains is unable to demonstrate that the requirements of EN 13231-3:2006 have been satisfied. The converse is also the case i.e. where a standard does not exist, is irrelevant or is inadequately monitored, poor results such as the worst of those shown in Figure 5 can result. An example that illustrates this is that some of the worst post-grind measurements in Figure 5 and also some of the best come from the same metro system, before and after a grinding specification was adopted and monitored based on EN 13231-3:2006. An example of an irrelevant grinding specification is one that specifies a requirement for arithmetic mean roughness RA in the belief that this affects noise. ‘Roughness’ of this form is reduced quickly by plastic deformation as demonstrated, for example, in field tests44 that provided input to EN 13231-3:2006. In these tests RA was reduced by an order of magnitude within the running band after the passage of less than 50,000 tonnes of traffic. Conclusions Irregularities on wheels and rails excite dynamic loads and give rise to noise. For most cases and for most of the wavelength range of interest, rails contribute significantly more to the overall roughness spectrum than wheels. Although it is recognised that irregularities contribute to ground-borne noise, measurements that demonstrate this are extremely scarce. A contribution to this area is made here using measurements taken by London Underground staff of railhead irregularities 555 and in-property noise levels before and after grinding of the adjacent underground track. It is demonstrated that there is satisfactory correlation of reductions in ground-borne noise levels with reductions in railhead roughness in the 100–1000 mm wavelength range, corresponding roughly to the 25–250 Hz frequency range. Conversely there is no correlation between reductions in noise and short wavelength irregularities. A wide variety of measurements is presented of railhead irregularities in the 6.3–2000 mm wavelength range. These come from different types of railway system (high speed, heavy haul, mixed passenger and freight, metro and light rail) worldwide. A comparison is also made of the limiting spectra of railhead irregularities specified in European standards for reprofiling of rails and acoustic type testing of vehicles. These were extrapolated to longer wavelengths for the purposes of this paper. There are interesting differences in the roughness spectra from different types of railway system. The level of irregularities on the two heavy haul lines for which measurements were available are compliant with the ISO 3095 limit throughout the wavelength range specified in the standard. Irregularities at sleeper spacing and residual ‘grinding signature’ are more prominent than on other types of line. Consistently the lowest irregularities were found on a high speed line and on the high rail of a metro curve. Even at wavelengths of more than a metre, irregularities on the high speed line were about 10 dB below the limit extrapolated from the ISO 3095 limit spectrum. Conversely, the highest level of irregularities was found, as a reflection of corrugation, on the low rail of the same metro curve in which the high rail had such a low level of irregularities. Mixed traffic railways have a relatively high level of long wavelength irregularities, occasional signs of irregularities at sleeper spacing, broad peaks in the roughness spectra reflecting the presence of corrugation, and low levels of short wavelength roughness. Light rail or tram systems have a high level of irregularities throughout the wavelength range, particularly at short wavelengths. The short wavelength roughness on tram systems, and also on one of the mixed traffic lines examined, probably results from sanding to enhance adhesion. There is evidence that where traffic speeds are well controlled, corrugation occurs over a narrower wavelength range, whereas on mixed traffic lines, for example, the wavelengths are less well defined. The changes in irregularities brought about by reprofiling are shown for work undertaken by several grinding contractors with different equipment in different parts of the world. In the majority of cases the finished rail is within the requirements of EN 132313:2006. This standard is substantially identical to the acoustic standard ISO 3095 except that it allows a 556 higher limit of irregularities in the 10–30 mm wavelength range. The higher limit for 10–30 mm is not only necessary in order to accommodate the irregularities that result from conventional reprofiling (both grinding and milling) but also sufficient for the vast majority of cases. Reprofiling in general is particularly effective at reducing irregularities in the 30–500 mm wavelength range, typically by 10 dB but in some cases by 20–30 dB. It has a small influence on reducing irregularities at more than 1 m wavelength, and brings about a significant increase in irregularities for wavelengths of less than about 30 mm. Irregularities in this wavelength range are subsequently reduced by wear. Some reprofiling is done to an extremely poor standard and in one exceptional case left the rail rougher at all wavelengths than it was initially. Results are shown from one technique that can achieve a roughness within the requirements of ISO 3095 immediately after reprofiling. The effect of traffic following reprofiling is typically to increase irregularities at wavelengths longer than about 30 mm and reduce irregularities rapidly at shorter wavelengths. Although reprofiling usually reduces irregularities to satisfactory limits, some of the worst results here were obtained where there existed an inappropriate standard that was inadequately monitored. Conversely some of the best results were obtained where there was an appropriate standard that addressed the irregularities of concern and which was conscientiously monitored. Most reprofiling work is done to a satisfactory standard, but in some cases this is more by accident than design. Funding This work received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Acknowledgements It would have been impossible to produce this paper without generous contributions from many colleagues. Those who contributed data include ARTC, Citirail, Corus/Tata Rail Technologies, IRT (Monash), London Underground, Loram Maintenance of Way, Metro Medellin, Queensland Rail, Schweerbau GmbH, SNCF, Sumitomo Metal Technologies and Wiener Linien. The author is particularly grateful to John Edwards and James Shepherd of London Underground for providing the measurements of both noise and irregularities before and after grinding, which are the basis of the section ‘The influence of irregularities on ground-borne noise’. The original catalyst for the paper was an invitation from Prof Simon Iwnicki of Manchester Metropolitan University to give the opening presentation at a seminar on rail corrugation in April 2010. Proc IMechE Part F: J Rail and Rapid Transit 226(5) References 1. Thompson DJ. Railway noise and vibration: mechanisms, modelling and means of control. Oxford, UK: Elsevier, 2009. 2. Thompson DJ and Jones CJC. Noise and vibration from the wheel-rail interface. In: R Lewis and O Olofsson (eds) Wheel/rail interface handbook. Oxford, UK: Woodhead Publishing Ltd, 2009, pp.477–509. 3. Jones CJC. Wheel-rail interface noise and vibration. In: F Schmid, M Burstow, S Clark, B Eickhoff, M Hiensch, SS Hsu and S Kent (eds) Wheel-rail best practice handbook, First edn. Birmingham, UK: University of Birmingham Press, 2010, section 4.8, pp.4–58 to 4–70. 4. Jones CJC and Jones RRK. Management of noise and vibration. In: F Schmid, M Burstow, S Clark, B Eickhoff, M Hiensch, SS Hsu and S Kent (eds) Wheel-rail best practice handbook, First edn. Birmingham, UK: University of Birmingham Press, 2010, section 4.9, pp.4–7 to 4–80. 5. Remington PJ, Rudd MJ, Ver IL, et al. Wheel/rail noise, parts I-V. J Sound Vib 1976; 46: 359–451. 6. Remington PJ, Rudd MJ and Ver IL. Wheel/rail noise and vibration, volume 1: mechanics of wheel/rail noise generation. Report no. UMTA-MA-06-0025-75-10, National Technical Information Service, Springfield, VA, 1975. 7. Remington PJ, Rudd MJ and Ver IL. Wheel/rail noise and vibration, volume 2: applications to control of wheel/ rail noise. Report no. UMTA-MA-06-0025-75-11, National Technical Information Service, Springfield, VA, 1975. 8. Remington PJ. Wheel/rail rolling noise What do we know? What don’t we know? Where do we go from here? J Sound Vib 1988; 120: 203–226. 9. Thompson DJ, Hemsworth B, Vincent N, et al. Experimental validation of the TWINS prediction program for rolling noise, parts 1 and 2. J Sound Vib 1996; 193: 123–147. 10. Thompson DJ. Wheel-rail noise generation, parts I-V. J Sound Vib 1993; 161: 387–482. 11. Wilson GP, Saurenman HJ and Nelson JT. Control of ground-borne noise and vibration. J Sound Vib 1983; 87: 339–350. 12. Kurzweil LG. Prediction and control of noise from railway bridges and tracked transit elevated structures. J Sound Vib 1977; 51: 419–439. 13. Wittig LE. Railway elevated structure noise – a review. J Sound Vib 1983; 87: 249–271. 14. Organisation for Research and Experiments of the International Union of Railways. Vibrations transmitted through the ground, reports 1–9, 1981–1984. 15. Kinsler LE and Frey AR. Fundamentals of acoustics, 2nd edition. New York: John Wiley and Sons, 1962, pp.392–398. 16. Triepaischajonsak N, Thompson DJ, Jones CJC, et al. Ground vibration from trains: experimental parameters and validation of a numerical model. Proc IMechE, Part F: J Rail Rapid Transit 2011; 225: 140–153. Grassie 17. Galaitsis AG and Bender EK. Wheel/rail noise – part V: measurement of wheel and rail roughness. J Sound Vib 1976; 46: 437–451. 18. Thompson DJ. On the relationship between wheel and rail surface roughness and rolling noise. J Sound Vib 1996; 193: 149–160. 19. Bewes OG, Thompson DJ, Jones CJC and Wang A. Calculation of noise from railway bridges and viaducts: experimental validation of a rapid calculation model. J Sound Vib 2006; 293: 933–943. 20. EN ISO 3095:2005. Railway applications – acoustics – measurement of noise emitted by railbound vehicles. 21. EN 15610:2009. Railway applications – noise emission – rail roughness measurement related to rolling noise generation. 22. EN 13231-3:2006. Railway applications – track – acceptance of works – part 3: acceptance of rail grinding, milling and planing work in track. 23. Verheijen E. A survey on roughness measurements. J Sound Vib 2006; 293: 784–794. 24. Hiensch M, Nielsen JCO and Verheijen E. Rail corrugation in The Netherlands – measurements and simulations. Wear 2002; 253: 140–149. 25. Nielsen JCO. High-frequency vertical wheel-rail contact forces – validation of a prediction model by field testing. Wear 2008; 265: 1465–1471. 26. Nielsen JCO and Ekberg A. Acceptance criterion for rail roughness level spectrum based on assessment of rolling contact fatigue and rolling noise. Wear 2011; 271: 319–327. 27. Fodiman P and Staiger M. Improvement of the noise technical specifications for interoperability: the input of the NOEMIE project. J Sound Vib 2006; 293: 475–484. 28. Grassie SL. Short wavelength rail corrugation: field trials and measuring technology. Wear 1996; 191: 149–160. 29. Diehl RJ and Holm P. Roughness measurements – have the necessities changed? J Sound Vib 2006; 293: 777–783. 30. Grassie SL, Saxon MJ and Smith JD. Measurement of longitudinal rail irregularities and criteria for acceptable grinding. J Sound Vib 1999; 227: 949–964. 31. Jones CJC, Létourneaux F and Fodiman P. Testing a new rail roughness measurement standard, http://intellagence.eu.com/acoustics2008/acoustics2008/cd1/data/articles/001997.pdf. (accessed 14 September 2011). 557 32. Van Ruiten CJM. A new method for the measurement of wheel/rail roughness. J Sound Vib 1988; 120: 287–295. 33. Asmussen B, Onnich H, Strube R, et al. Status and perspectives of the ‘‘specially monitored track’’. J Sound Vib 2006; 293: 1070–1077. 34. Grassie SL. Measurement of railhead irregularities: a comparison of different techniques. Wear 1996; 191: 245–251. 35. Berggren EG, Li MXD and Spännar J. A new approach to the analysis and presentation of vertical track geometry quality and rail roughness. Wear 2008; 265: 1488–1496. 36. Bongini E, Grassie SL and Saxon MJ. ‘‘Noise mapping’’ of a railway network: validation and use of a system based on measurement of axlebox vibration. Notes Numer Fluid Mech Multidiscip Des 2011; 118: 511–517. 37. Chua KH, Balendra T and Lo KW. Groundborne vibrations due to trains in tunnels. Earthq Engng Struct Dyn 1992; 21: 45–460. 38. Grootenhuis P. The attenuation of noise and ground vibrations from railways. J Environ Sci 1967; 14–19. 39. Dawn TM. Ground vibrations from heavy freight trains. J Sound Vib 1983; 87: 351–356. 40. Hölzl G and Werner K. Die leise schienenoberfläche durch schleifen? ETR 1993; 42: 659–664. 41. Grassie SL. Rail corrugation: characteristics, causes and treatments. Proc IMechE, Part F: J Rail Rapid Transit 2009; 223: 581–596. 42. Hartleben D. Lärmminderung durch oszillierendes schleifen. ZEV Rail Glasers Annalen 2005; 129(8): 324–329. 43. Hempelmann K and Knothe K. An extended linear model for the prediction of short-pitch corrugation. Wear 1996; 191: 161–169. 44. Grassie SL. Requirements for transverse railhead profile and railhead roughness following grinding. In: The sixth international heavy haul railway conference, strategies beyond 2000, Cape Town, South Africa, 6–10 April 1997, pp.549–564.
© Copyright 2026 Paperzz