Use of Track Geometry Measurements for Maintenance Planning

Transcription

1 84 TRANSPORTATION RESEARCH RECORD 147 Use of Trak Geometry Measurements for Maintenane Planning WILLEM EBERSOHN AND ERNEST T. SELIG Trak geometry measurements are disussed as a means of evaluating the funtional ondition of trak. Examples are given of trak vertial profile geometry data from a heavy haul line obtained by a laser survey and by a trak reording ar. Bad sites with rapid geometry deterioration are ompared with good sites with stable geometry. The mid-hord ordinate measurement proedure used by the reording ar is desribed, and two methods to bakalulate the trak longitudinal profile from this measurement are presented. A definition of trak roughness based on the vertial geometry data is proposed. Examples are given of the use of roughness to identify trak setions having similar maintenane needs and to projet the amount of traffi when geometry orretions will be required. The results of the researh suggest that onsiderable potential for ost savings exists from use of trak geometry data in maintenane planning beyond that normally onsidered by the railroad where the measurements were made. Trak ondition monitoring is the primary input for maintenane planning and exeution. The ondition of a trak may be ategorized as either funtional or strutural. The funtional ondition reflets the servieability of a trak indiated by its geometry, while the strutural ondition is represented by the strength and stiffness properties of the trak whih affet the ability of the trak to perform satisfatorily. Both onditions should be monitored, but this paper will onentrate on interpreting the funtional ondition of the trak. A ompanion paper in this Reord presents the orresponding information on the strutural ondition. A trak maintenane engineer must ensure that the trak is above a ertain speified servieability level and thus should monitor the funtional ondition at regular intervals. This monitoring is done by foot, by motorized visual inspetions, or by mehanized geometry measurements. An example of the latter is the trak reording ar, usually referred to as the trak geometry ar. The reording ar is primarily being used by most railroads to identify geometry exeedanes, but a number of railroads are starting to alulate trak indies from various geometry measurements to monitor trak deterioration. However there are additional harateristis relating to vertial geometry (trak profile) whih should be onsidered for funtional ondition evaluation. This paper disusses the geometry results obtained in an extensive trak performane investigation onduted in South Afria, and shows how they ould be used to evaluate the trak funtional ondition and to determine the need for maintenane. Some of the onepts disussed in the paper may have been onsidered in the past by most railroads who use geometry ars. However, the onepts whih have been onsidered have been applied in W. Ebersohn, Department of Civil Engineering, University of Pretoria, Pretoria, 2, Republi of South Afria. E.T. Selig, Department of Civil and Environmental Engineering, Marston Hall, Box 3525, University of Massahusetts, Amherst, Mass a variety of different ways with varying degrees of reliane and suess. What the measurements represent is not always appreiated. This paper attempts to larify the onepts and ways of interpreting the geometry measurements by using examples from one operating setion of trak where a detailed investigation was onduted. SITE DESCRIPTION The Heavy Haul Coal Line in. South Afria was hosen for this investigation beause it is an ideal site for onduting applied railway researh, due to its high annual tonnage and heavy axle loads. The Coal Line of SPOORNET (South Afrian Railway Organization) links the Transvaal oal fields with the east oast of South Afria (J). The length of the line is 586 km and it was built between 1973 and The initial axle loads were limited to 185 kn (18.5 tonnes) for the wagons and 22 kn (22 tonnes) for the loomotives. The axle loads were gradually inreased to 26 kn (26 tonnes) for the wagons and kn (29 tonnes) for the loomotives in This inreased load was handled by a major upgrading program onsisting of doubling the line, strengthening the superstruture and reduing the grades for the loaded trains. Two 2-m (656-ft) setions of tangent trak on the Coal Line with reasonably uniform onditions were seleted for this study to provide both "good" and "bad" trak under the same traffi and environmental onditions. The good setion required little maintenane and the bad setion needed frequent spot maintenane. The high maintenane input in the bad setion was assoiated with substruture-related problems, resulting in rapid loss of surfae geometry. The spot maintenane onsisted of leaning the ballast and tamping to improve the geometry. Within the 2-m (656-ft) bad setion was a 1-m (328-ft) portion with most of the problems. This portion will be termed the bad site in this paper. A orresponding 1-m (328-ft) good site was seleted from within the good setion. The superstruture onsisted of ontinuously welded 6 kg/m (121 lb/yd) S6 hrome manganese steel rails fastened to the 285 kg (628 lb) onrete ties with Fist type fasteners on high density polyethylene rail seat pads. The butt (shop) welds and thermit (field) welds were evenly spaed throughout eah setion, with one additional thermit weld in the bad setion. The surfae of eah setion was visually inspeted for mud boils and surfae drainage onditions. Evidene of these problems was only observed in the first 1 m (328 ft) of the bad setion and these ourred randomly along the length of the setion. The ballast is a dolorite rushed rok with an average layer thikness of 3 mm (12 in.) below the tie in the bad setion and 415 mm (16 in.) in the good setion. More information on the strutural ondition of the various trak omponents an be obtained in the ompanion paper in this Reord.

2 Ebersohn and Selig 85 FUNCTIONAL CONDITION MEASUREMENTS The funtional ondition of the test setions was monitored by surveying the trak in both the loaded and the unloaded onditions, and measuring the trak geometry with a reording ar. This allowed a omparison of the absolute funtional ondition, in a loaded and unloaded state, with the geometry ar results. Unloaded Trak Longitudinal Profiles The unloaded trak profiles were measured using a digital level taking elevation readings on the top of both rails at every seond tie along the trak. The profiles were measured before tamping, diretly after tamping and during subsequent traffi. The top setion of Table 1 gives the umulative traffi up to eah set of measurements. The values in parentheses indiate the traffi sine the previous tamp. The profiles were tied into an elevation referene system to determine the absolute settlement. As a typial example, the unloaded absolute longitudinal profile of both sites for both rails over the worst portions of both setions are indiated in Figure 1. The wavelengths and amplitudes of the trak indiated in Figure 1 represent the atual wave form in the longitudinal profile of the two rails. Loaded Trak Longitudinal Profiles Loaded profiles were measured using the same digital level taking elevation readings at every seond tie on bar-oded sales fixed to both sides of the rear axle of a two-truk oal wagon with a 26-kN (26-tonne) axle load. Only two sets of loaded profile measurements ould be obtained as indiated in the shedule given in the middle setion of Table 1. Figure 2 shows a typial example of the loaded absolute longitudinal profile of both rails. These lines represent the atual defleted wave forms for both rails, as followed by a loaded axle. TABLE 1 Funtional Condition Measuring Shedule a Unloaded Profile Shedule in Cumulative Traffi Bad Site MGT GN Good Site MGT GN Before Tamp (11.72).. Before Tamp (11.72).. After Tamp.. After Tamp Loaded Profile Shedule in Cumulative Traffi GT Bad Site GN Good Site MGT GN Before Tamp (13.162).. Before Tamp (13.162).. After Tamp. O.O After Tamp Car Shedule in Cumulative Traffi GT GN Good Site MGT.. Before Tamp (8.47) After Tamp GN Trak Reording Car Measurements The trak geometry was measured using a Plasser EMV8 reording ar. The mid-hord ordinate readings are taken over a 7 m (23 ft) hord length at every 25 mm (9.8 in) along the rail. Midhord ordinate measurements will be referred to as the "mid-hord profiles" of the trak. Versine (alignment) and ant (ross-level) measurements were also obtained, but only the profile results will be used in this paper. The ar is equipped with an adjustable axle load onfiguration, allowing different axle loads to be applied to the trak depending on its line lassifiation. The bottom setion of Table 1 gives the umulative traffi of eah measuring run. The runs were measured at both 26 kn (26 tonne) and 16 kn (16 tonne) axle loads. Only the 26 kn (26 tonne) axle results will be disussed in this paper. Figure 3 shows the mid-hord profile measurements for both sites before.and after tamping and after 6.5 MGT (65 GN) traffi. Mid-Chord Pro.file Measuring Priniple The frame of the EMV8 is used as an inertial referene to measure the mid-hord ordinate. The mid-hord profile is a funtion of the atual wave form of the trak from whih the absolute position of the top of rail is eliminated. In effet the profile obtained with the geometry ar mid-hord ordinate reading represents the atual geometry passed through a distorted filter of trak wave-lengths. The quantities measured are designated by z in Figure 4, whih shows the effet of the loaded wheels relative to the unloaded measuring trolley. The mid-hord ordinate hn is alulated using the following equation: h n = Zn-I + Zn+I + 2 Zn (1) The symmetrially spaed measuring positions have the advantage that any longitudinal rotation of the vehile frame anels out in the alulation of hn. The mid-hord ordinate measuring priniple does not represent the atual profile of the trak as mentioned above. Figure 5 shows the mid-hord ordinate response to an irregularity that is shorter than the hord being used to measure it. Note that the shape of the profile does not look like the atual shape of the irregularity. The response of a three-point mid-hord ordinate measuring system to sinusoidal inputs is as follows: where A. = wavelength of a sinusoidal omponent of a longitudinal rail profile, l = half the hord length of the measuring system, hn(a.) = hord system response to wavelength, and Zn(A.) = longitudinal profile amplitude for a wave length. (2) Note: Values in ( ) indiate traffi in MGT sine previous tamp. The response funtion H(A.) is given by the following:

3 e 2 Bad Site Rail A. 1 RailB en :; C1> e 2 Good Site Rail A g 1 RailB t/) :.Q -ro ::; Q) Cl -1-2 :-._ "> Distane (m) FIGURE 1 Unloaded absolute longitudinal profiles for both rails. 2 Bad Site Rail A e RailB. 1 en s: -1 C1> -2 e en ;s a s: -1 C1> Good Site Rail A RailB Distane (m) FIGURE2 Loaded absolute longitudinal profiles for both rails.

4 FIGURE 3 Mid-hord profiles for inner rail. \(Rotation 35mm + 35mm () Zn- l Zn Zn+ 1 FIGURE 4 Defletion measured by geometry ar. Measuring Point. / 1 tto!l_ ' """' ih35/ - t: Rail Elevation Plot Mid Chord Ordinate Plot FIGURE 5 Mid-hord ordinate response to trak irregularity.

5 88 TRANSPORTATION RESEARCH RECORD 147 hn(a) ( 27rl ) H('A.) = Zn(A) = 1 - COS T (3) This is demonstrated in Figure 6. This funtion is also known as the transfer funtion. Beause the mid-hord measurement is repeatable, the hange in mid-hord profile measurements an be used to quantify the hange in the trak funtional ondition. The hange in the loaded and unloaded longitudinal profile measurements an also be used to quantify the hange in the trak ondition. TRACK ROUGHNESS Trak roughness is a measure of the vertial (or horizontal) deviation from a referene alignment over a length of trak. Roughness may be expressed, for example, as the number of exeedanes, standard deviations or the sum of squares of profile deviation measurements. Undulations in the trak ause vibrations in a moving train. The frequenies of the vibrations depend on the speed, weight and suspension harateristis of the train as well as the damping harateristis of the trak. Longer wavelengths and lower speeds ause lower frequenies and have a less deleterious effet on the trak omponents. A ertain wavelength exists above whih the indued frequenies an be ignored, and so the wavelengths longer than this an be filtered out of the funtional ondition measurements. Figure 7(top) shows an example of the before-tamp unloaded longitudinal profile measurement, together with the profiles obtained by using a 7.8 m (26 ft) and 23.4 m (77 ft) filter length. The filtering was aomplished by subtrating the mean of the elevation readings over the filter length from the elevation reading at the point under onsideration. This has the effet of removing wavelengths longer than the filter length. The sum of squares alulation for roughness R 2 for a surveyed profile is given by where d; is the differene between the elevation of the point being measured and filtered mean elevation measurements and n is the number of measurements in the length of trak under onsideration. It should be reognized that the value of roughness for a given geometry depends on three parameters: (a) the length of trak being inluded in the alulation, (b) the smoothing or filtering funtion hosen, and () the spaing of the measurements within the length of trak, and thus the number of measurements. The mid-hord ordinate obtained with the geometry ar is already a filtered measurement, although a distorted one as shown in Figure 6. Hene for alulating roughness with Equation 4, d; is taken as the mid-hord ordinate diretly. The mean of the ordinate readings over a length of trak should not be subtrated from eah d; as in alulating standard deviation, although this mean is approximately zero if a long enough length of trak is used. Figure?(bottom) ompares the mid-hord profile and the unloaded longitudinal profile, with the ordinates alulated from the unloaded survey using a 7.8 m (26 ft) hord length. The wavelengths and the peaks and valleys in these two profiles orrespond, but the amplitudes vary. BACKCALCULATION OF LONGITUDINAL PROFILES FROM MID-CHORD ORDINATE MEASUREMENTS If the funtional ondition of the trak is measured only to determine whether the ondition of the trak is deteriorating, it is not neessary to measure the absolute longitudinal profile of the trak. Instead measures suh as roughness and exeedanes obtained with the geometry an be used. However, if the geometry measurements are needed to identify the different trak defets and how they (4) CD 1.8 "O Q. 1.6 E <C 1.4 "ii a:: 1.2 a; 1.. =.8 a:: 'E L-----L----'...Jlll.... i_l...l.----'--....,...._..._, Wavelength (m) FIGURE 6 Response funtion for mid-hord ordinate measurements.

6 Ebersohn and Selig 89 3 BAD SITE -- Longitudinal Profile using 7.8 m Filter --- Longitudinal Profile using 23.4 m Filter Surveyed Profile e :; G> 15 Ci CJ :e 1 5 e :; G> 15 Ci CJ :e 1 G> > R 2 (7.s)=3.4 mm 2 R 2 < 23.4i=7.8 mm Mid-Chord Profile Calulated over 7.8 m Surveyed Profile 1.. s 5 a 'E. -5 "O i 5-1 R2(Mid-Chord)=6. mm Tie No Distane (m) FIGURE 7 Atual longitudinal profile at bad site ompared with filtered profile and mid-hord profile. e =e develop, the wavelengths and the assoiated amplitudes in the trak should be determined. A typial lassifiation of wavelengths as used by the Netherlands Railways is given in Table 2. Two approahes an be followed to determine the longitudinal profile by onverting the mid-hord ordinate measurements obtained from the geometry ar. The first is geometri analysis, and the seond is a Fourier transformation analysis using the transfer funtion that exists between hn and Zn. TABLE2 Wavelength Classifiation Ourring in: Wavelength Phenomenon Monitoring Tehnique m Rail.1to.3 Conugations Axle box.1to.1 Poor weld & wheels and rail 2 to 3 Rolling defets aeleration Ballast 3 to25 Versine top gauge twist Chord or inertial measurements 25 to 7 Alignment & profile Date levels 7 tooo Aliment & Erofile Laser levels Geometri Approah To reonstrut the longitudinal profile from mid-hord measurements using the geometri approah, the relationship as indiated in Figure 8(top) will be used. If the positions of Points 1 and 2 are known, the position of Point 3 an be determined by extending the line joining the first two points and finding that Point 3 is twie the mid-hord offset from this line. The alulation of the position Z 3 on the profile with the variables defined in Figure 8 is given by the following: Figure 8(bottom) shows the proedure followed to determine the position of eah point by suessively alulating values of h at intervals of l. In its general form the absolute position of a single (5) point on the line relative to the first two points is given by the following: Zn = Zi + (n - 2) (Zi - Z1) - n-1 2 L (n - r) h, (6) r =2 Equation 6 depends on the position of the first two points as well as the differene between their positions. A small error in the measurement of these two positions will result in a umulative error in the bakalulated Z value. This onsideration, as well as the fat that the longitudinal profile indued trak fores are less sensitive to longer wavelengths, suggests that the longitudinal profile should be alulated by filtering out the longer wavelengths.

7 9 TRANSPORTATION RESEARCH RECORD 147 where y = roughness (mm 2 ), x = traffi (MGT), a = sale fator, b = shape fator, and r = initial roughness (mm 2 ). FIGURE 8 Geometrial approah for bakalulation of longitudinal profile. Thus moving average over (2s + 1) points about Zn is used to alulate the longitudinal profile as follows: 1 +s Z.(n) =Zn - ( 2 s + 1) m -s Zn+m (7) where (2s + 1) is the number of readings in the averaging distane or filter length and Z.(n) is the true longitudinal profile at the nth point whih is a funtion of s. Thus there are S readings inluded on eah side of the nth point. Fourier Approah To reonstrut the longitudinal profile from mid-hord measurements using the Fourier approah, hn is transformed to. a Fourier equation after whih the transfer funtion given in Equation 4 is used to obtain the Zn values. The two approahes were found by the writers to give omparable values for wave amplitudes, and hene show promise for overoming the amplitude distortions aused by the mid-hord measuring system. However, approahes (geometri and Fourier) will only give atual wave forms with the long wavelengths filtered out. From Figures 1 and 2 it an be seen that the predominant lengths of the long waves are between 3 m and 1 m (1 ft and 33 ft). FUNCTIONAL PERFORMANCE OF TEST SITES The loaded and unloaded roughness using a 13-m (43-ft) filter length for the surveyed measurements and the mid-hord profile roughness were determined for eah test site. Figure 9 shows these results for the unloaded and the mid-hord profile measurements. The trends are the same but the values differ due to the differene in measuring priniple as explained above. To desribe the roughness growth rate of eah site mathematially, a power funtion was fitted to the roughness values using statistial urve-fitting methods. The power funtion starts at the initial roughness after tamping at MGT, and rises at a rate desribed by a and bas follows: y=r+axb (8) This funtion fit the bad site data in Figure 9 very well, but not the good site. Not enough loaded roughness measurements are available to desribe a trend, but the roughness values for the bad and good sites before tamping are 19.5 mm 2 (.3 in. 2 ) and 3. mm 2 (.5 in. 2 ) respetively, and after.6 MGT of traffi they are 1.1 mm 2 (.2 in. 2 ) and 7.1 mm 2 (.1 in. 2 ). After an initial set of profile measurements, smoothing tamping was performed in both setions to provide an opportunity for monitoring the trak funtional deterioration starting with a newly tamped trak. As depited in Figure 9, for the bad site the roughness was greatly redued by tamping, but the roughness quikly returned to an unaeptable level. In ontrast, for the good site the tamping atually made the roughness worse. With further traffi the roughness inreased and then quikly stabilized, even showing a tendeny to derease. Smoothing tamping of the bad setion did not solve the ause of the bad site geometry problems and it also did not gain any meaningful length of tamping yle. Tamping disturbed the stable onditions in the good site, indiating that tamping should not be done on a setion of trak if it is in a stable ondition with an aeptable geometry. The primary appliation of the geometry ar results is to monitor the deterioration of the funtional ondition of a trak, either on an overall system basis or on a short length for a maintenane projet. In both instanes the geometry results should be used to identify setions with the same inherent geometry harateristis. This an be done by using either the mid-hord profiles or the bakalulated longitudinal profiles to study the rate at whih roughness develops. Care should be taken to identify singular features (suh as bridges, signals, and level rossings) during suh an investigation. Grouping setions with the same inherent geometry harateristis is benefiial in maintenane planning as it may represent areas with similar maintenane needs. The setions identified should then be investigated to determine the orret maintenane input. A.high roughness does not mean that the trak should just be tamped. Some other maintenane ation may also be required to make the improved geometry last. As an example, one 7-m (1.1-mi) length of trak, whih inluded the two test setions, was investigated to identify setions with similar harateristis using the mid-hord profile measurements. Figure 1 shows the mid-hord ordinates with.dotted lines. The trak features along the length of the trak are indiated with the solid vertial lines. A running roughness profile is alulated to identify setions of trak whih show similar roughness levels. Eah roughness point on the profile represents 2 measurements spread over a 5-m (164-ft) length of trak. The point is plotted at a distane oiniding with the enter of the 5-m (164-ft) length. In pratie, the length of trak used in a roughness alulation should probably be the minimum length of trak onsidered for maintenane work. The roughness plot in Figure 1 learly identifies different setions, but it gives no indiation of what the ause of the problems are in the bad setions.

8 Ebersohn and Selig MGT = Million Gross Tonnes 1MGT = 1 GN e "&:,,; 5.. ::n ::J a: 2.5 Before Tamp Bad Site Good a=.74 b = Before Tamp Good Site 22.5 Before Tamp N' 2. Bad Site E N a: 15.,,; en Q) Unloaded Longitudinal Profile Roughness ::n ::J a= 1.97 D 7.5 b =.26 a: Good D 5. After Tamp Good Site 2.5 After Tamp Bad Site Before Tamfeo Good Si e Traffi (MGT) Mid-Chord Profile Roughness FIGURE9 Roughness hange with maintenane and traffi. Comparison of suessive reording ar measurements, ombined with the investigation of problem setions, an provide a means to estimate future tamping shedules for eah setion. Thus more effetive sheduling of tamping operations an be ahieved. This an be done by determining the inrease in roughness based on the mid-hord profile or using a bakalulated longitudinal profile. The advantage of the latter is that it represents the atual amplitudes and wavelengths, whih an be used to investigate the inrease in wavelength amplitudes that should relate to the ause of geometry deterioration. CONCLUSIONS The trak reording ar is typially used to detet geometry speifiation exeedanes. However, it has muh more potential. This paper investigated the uses of vertial profile geometry. The mid-hord ordinate gives a measure of the geometry with long wavelengths removed, although the amplitudes of the shorter waves are altered from their atual values. These mid-hord measurements an be used to alulate the atual vertial profile geometry with long wavelengths removed. The lengths and amplitudes of the individual waves then an be determined from these profiles to provide a basis for diagnosing the soure of the geometry problems. Trak roughness an also be alulated from the mid-hord ordinate measurements or from the bakalulated profiles using the definition proposed in the paper. Observing hanges in roughness with traffi will provide a basis for projeting future maintenane requirements. A moving roughness profile of the trak an easily be generated from the mid-hord ordinate measurements or, with more effort, from the bakalulated profile. The roughness profile has great potential for distinguishing setions of trak with the same inherent geometry harateristis. The individual setions an then be examined to determine the orret maintenane ations. Loaded and unloaded longitudinal profiles obtained by surveying means give a more aurate representation of the trak geometry than is available from the reording ar. Loaded profiles are preferred beause they represent the trak as seen by the trains. However, surveying tehniques or their equivalent are presently limited to loal setion investigations beause they annot ollet the data over long distanes as quikly and effiiently as the reording ar.

9 92 TRANSPORTATION RESEARCH RECORD 147 Travel Diretion of Loaded Trains..,, e Cl) e ii: 5. a. 2.5 'E i f s 5. D:: ui 2.5 Ill Cl). :. C> :::J -2.5 a: Embankment Bridge Embankment Cut Shallow Cut Vaiying Fill Good Site R=2.s R2=3.6 =7.9 li=4.2 fi=s.1 R=1.s R2= Distane* 1 (km) FIGURE 1 Maintenane setion identifiation using moving roughness. The moving roughness analysis approah was extended to many kilometers of oal line trak. The preliminary onlusions were as follows: For 2 perent of the setion the improved surfae geometry following tamping would not last very long; to redue maintenane ost, some other remedial ation would be needed in addition to tamping. For 5 perent of the setion tamping was not needed beause the surfae was satisfatory. Tamping would provide a long enough lasting surfae to be effetive by itself for only 3 perent of the setion. Without this analysis the entire setion may have been sheduled for tamping on a regular basis as ditated by loalized problem areas. Thus, a signifiant potential for better use of maintenane resoures is possible. ACKNOWLEDGMENT The authors wish to express their appreiation to SPOORNET, whose support made this paper possible, and to all the personnel from the Trak Testing Center and the University of Pretoria Civil Engineering Department who took part in the investigation. The analysis tehniques were developed as part of researh at the University of Massahusetts, supported by the Assoiation of Amerian Railroads. REFERENCE 1. Kuys, W. C. Trak Struture Failure Under 26-Ton Axle Loads on South Afrian Transport Servies Coal Line. Pro., 4th International Heavy Haul Railway Conferene, Brisbane, Australia, Sept. 1989, pp Publiation of the paper sponsored by Committee on Railway Maintenane.