190

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 61, NO. 1, JANUARY 2012

Multisensor Railway Track Geometry Surveying System

Burak Akpinar and Engin Glal
AbstractLinear variable differential transformers (LVDTs) and inclinometers are widely used in many industrial establishments, particularly in the metrology area. These sensors are used by many engineering disciplines because of their high-precision characteristics. In addition, Global Navigation Satellite System (GNSS) receivers and total stations are widely used in geodesy. Using GNSS receivers is very popular, particularly for navigational purposes. In this paper, a new railway track geometry surveying system, which is designed by integrating the LVDT, inclinometer, GNSS receiver, and total station, is introduced. This new surveying system is an alternative to classical geodetic measurement methods that are often used for controlling the railway track geometry. Track gauge, super-elevation, gradient, and track axis coordinates, which are railway geometrical parameters, can be instantly determined while making measurements by using the new surveying system. Index TermsGeodesy, Global Positioning System, multisensor systems, rail transportation maintenance.
Fig. 1. LVDT instrument used to determine the track gauge.

I. I NTRODUCTION EPENDING on recent developments in railway systems, high-speed trains are being extensively used, and rail transportation is being increased. Reasons for this increase are high speed, economical, environment friendly, safety, and modern characteristics of railway systems. These characteristics can be continued by periodical maintenance and control measurements. Depending on different factors, deformations may occur on the superstructure of railways. Determining these deformations on time and taking precautions is very important for the safety of railway systems. Classical geodetic measurement methods and instruments are currently used to determine railway deformations. The measurement process with these methods and instruments usually takes a long time. The fast growth in railways caused the necessity to provide speed and automation in geodetic control measurements. Different studies have been done to provide speed and automation in geodetic control measurements on railways. The Rhomberg system by Bahntechnik [1] and the GnBahn system by Geo++ [2] are examples of these studies. In [3], a method for the

estimation and classication of a railroad curvature by onboard sensor data was developed. The introduced surveying system in this paper is operational on both ballast and slab tracks. Track axis coordinates, which are railway geometrical parameters, are obtained with integrated Global Navigation Satellite System (GNSS) receivers. In addition, by means of motorized total stations, the system can be operated in tunnels without interruption where it is not possible to work with GNSS receivers. A linear variable differential transformer (LVDT) by Schaevitz Sensors is used in the surveying system to determine the track gauge. The repeatability of this LVDT is 0.0006 mm. Thus, it is possible to determine the track gauge precisely. By means of an integrated biaxial inclinometer, the super-elevation and gradient values of the railway can be obtained. Thus, the standards of the International Union of Railways (UIC) on railway geometrical parameters can be achieved.

II. C OMPONENTS OF THE S URVEYING S YSTEM A. LVDT A high-precision distance-measurement system is necessary to determine the track gauge values, depending on the UIC standards. To achieve this standard, a GCD-SE-2000 LVDT by Schaevitz Sensors has been preferred. This LVDT instrument can be operated between 0 and 50.8 mm, and the repeatability of the instrument is 0.0006 mm. Fig. 1 shows the preferred LVDT instrument. The scale factor of the instrument is given in the calibration certicate as 2.5277 V dc/in. By using this scale factor, voltage values can be transformed into displacement values.

Manuscript received March 17, 2011; revised May 6, 2011; accepted May 24, 2011. Date of publication July 25, 2011; date of current version December 8, 2011. The Associate Editor coordinating the review process for this paper was Dr. Subhas Mukhopadhyay. The authors are with the Department of Geomatics Engineering, Yildiz Technical University, Istanbul 34210, Turkey (e-mail: burakpinar@gmail.com, egulal@yildiz.edu.tr). Digital Object Identier 10.1109/TIM.2011.2159417

0018-9456/$26.00 2011 IEEE

AKPINAR AND GLAL: MULTISENSOR RAILWAY TRACK GEOMETRY SURVEYING SYSTEM

191

Fig. 2. Inclinometer used to determine the super-elevation and gradient of the railway.

Fig. 4. Designed surveying system.

Fig. 3.

RTK GNSS method.

B. Biaxial Inclinometer A biaxial inclinometer by HLPlanar Technik has been preferred to determine the super-elevation and gradient of the railway. Fig. 2 shows the preferred inclinometer. Depending on the UIC standards, the maximum allowed super-elevation value is 300 mm in the curves where the curve radius is bigger than 700 m [4]. When the curve radius is decreased, the maximum super-elevation value is decreased. Super-elevation values up to 384 mm can be determined by the preferred inclinometer on railways where the track gauge value is 1435 mm. The precision of the super-elevation values on these railways is between 0.7 and 2 mm. The UIC standards for super-elevation can be achieved with this precision. The designed surveying system operates at walking speed (0.51 m/s). To determine the super-elevation, centripetal acceleration on the inclinometer can be neglected. Its contribution is not signicant for walking speed. C. GNSS Receiver Dual-frequency real-time kinematic (RTK) GNSS receivers have been used on the surveying system to determine the track axis coordinates. In the RTK GNSS measurement method, reference and rover receivers are used. The reference receiver has a radio transmitter that transmits the carrier-phase correction data. The rover receiver has a radio receiver that receives the corrections transmitted from the reference receiver [5][7]. Fig. 3 shows the position determination with the RTK GNSS method. D. Total Station The surveying system is also equipped with a motorized total station. Thus, the surveying system can be operated in tunnels

Fig. 5. Parts of the surveying system.

Fig. 6. LVDT-mounted underside of the surveying system.

without interruption where it is not possible to work with GNSS receivers. The integrated total station has an automatic target recognition (ATR) ability. A reector on the surveying system can be tracked by the ATR total station when moving on the railway. With the ATR, the operator only needs to roughly point with the optical sight and trigger a measurement. The infrared beam transmitted from the total station telescope is reected back by the prism and instantly analyzed. The total station moves the telescope, ne points to the center of the prism and measures. When the lock mode is enabled on the total station, the moving reector can be tracked. After the initial ATR measurement, the total station remains locked onto the reector and follows it as it moves. Measurements can be taken at any time. As intelligent software predicts the reector movement, the total station continues to track even if obstructions cause short interruptions to the beam. If long interruptions should

192

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 61, NO. 1, JANUARY 2012

Fig. 7. Data processing procedure.

Fig. 8. Determining the track axis coordinates where the super-elevation and gradient are zero.

cause a complete loss of lock, the operator must just quickly point again with the optical sight. The designed surveying system is shown in Fig. 4. The surveying system is formed as three parts (see Fig. 5). A box containing sensor connectors, a radio receiver that receives the data transmitted from the total station, and cables for LVDT and inclinometer connections are placed in the rst part. The biaxial inclinometer is mounted in the second part of the system. The Y -axis of the inclinometer is oriented toward the railway track axis direction. The X-axis of the inclinometer is perpendicular to the Y -axis and oriented on the lateral direction of the track axis to determine the super-elevation. The third part of the system is used to store the apparatus. The LVDT is mounted to the underside of the system to determine the track gauge (see Fig. 6). The axis of the LVDT is superposed with the X-axis of the inclinometer. All these axis conditions were metrologically tested in the laboratory and in the surveying area with high-precision measurement equipment. III. DATA P ROCESSING Outlier detection is the rst phase of data processing. The median absolute deviation (MAD) ltering technique is used for outlier detection [8]. When the possible outliers on LVDT,

inclinometer, and GNSS/total station data are removed, the super-elevation, gradient, track gauge, and track axis coordinates are determined in the second phase. This phase is called as the determination of railway geometrical parameters. These determined parameters are stored in a database by combining with Coordinated Universal Time (UTC). Thus, time synchronization of data is possible. If the prior railway project exists, synchronized data can be compared with prior project data of the railway by using the Kalman ltering technique [8]. Fig. 7 shows the owchart of the data processing procedure. A. Determining the Track Axis Coordinates Track axis coordinates are determined with the GNSS receivers and total station. A survey pole that carries the GNSS receiver antenna and reector is mounted on the surveying system superposed with the track axis. Thus, if the super-elevation and gradient of the railway is zero, track axis coordinates can be directly determined. This condition is shown in Fig. 8. In curves or inclined railways, track axis coordinates cannot be determined from the GNSS and total station measurement directly. Because of the lateral and longitudinal inclination, measured values do not reect the projected track axis coordinates. If the lateral inclination angle is and the longitudinal

AKPINAR AND GLAL: MULTISENSOR RAILWAY TRACK GEOMETRY SURVEYING SYSTEM

193

Fig. 9.

Deviations caused by the (a) super-elevation and (b) gradient.

Fig. 10. Inclinometer axes.

inclination angle is , which are obtained from the inclinometer, deviations from the projected track axis are dened by E = hG sin B = hG sin . (1) (2)
Fig. 11. Track axis coordinates in the + and + condition.

These conditions are shown in Fig. 9. and angles are obtained from inclinometer measurements. Angle values on X (X, +X) and Y (Y, +Y ) directions are determined by the biaxial inclinometer. The Y -axis of the inclinometer is oriented toward the railway track axis direction so that the angle values on the Y -axis of the inclinometer reects the gradient values of the railway. The X-axis of the inclinometer is perpendicular to the Y -axis and oriented toward the lateral direction of the track axis to determine the superelevation. While operating with the surveying system on the railway, four different conditions occurred. These conditions are the leftright super-elevation and the plusminus gradient. The inclinometer axes and four conditions are shown in Fig. 10. Determining the rail track axis coordinates is shown in Figs. 1114 considering these four conditions. According to Fig. 9(a), the track axis elevation is dened by HOE = HA hG cos where HOE HA (3)

Fig. 12. Track axis coordinates in the and + condition.

track axis elevation; elevation of the upper side of the survey pole obtained by the GNSS or total station;

elevation of the GNSS antenna or reector from the railway plane. By subtracting the hG cos value caused by the superelevation from the HA elevation value, the track axis elevation can be determined. In [9], it is stated that the maximum allowed gradient value is 0.60% on railways. Taking this value into account, the hG sin value dened in Fig. 9(b) is calculated smaller than 1 mm, so that the deviation caused by the gradient

hG

194

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 61, NO. 1, JANUARY 2012

Fig. 15.

Time-tagged GNSS, LVDT, and inclinometer data.

and by using
Fig. 13. Track axis coordinates in the + and condition.

X1 = X1 + S11 cos t1 1 Y1 = Y1 + S11 sin t1 1 the azimuth in the and + condition is dened in t1 1 = t1 2 200 1

(8) (9)

(10)

and the track axis coordinates in and + condition is determined, respectively (see Fig. 12). The azimuth in + and (Fig. 13), and (Fig. 14) conditions is dened in (12) and (13) respectively t2 1 = arctg X1 X2 Y1 Y2 (11) (12) (13)

t1 1 = t2 1 + 200 1 t1 1 = t2 1 + 200 + 1 .

Fig. 14. Track axis coordinates in the and condition.

B. Time Synchronization of Data Collected by Different Sensors The designed surveying system can operate while moving on a railway. During the operation, data collected by different sensors are stored in a database in a computer. Time synchronization of data has to be done to determine the super-elevation, gradient, and track axis coordinates on a certain point. Kinematic multisensor systems such as the surveying system designed in this paper need careful time treatment. Depending on the speed of the surveying system, the accuracy of time synchronization between data collected by different sensors affects the calculated position accuracy [10]. To store all the sensor data in the database with a time tag, a time source is needed. If GNSS receivers are used for positioning, UTC derived from the National Marine Electronics Association sentences in GNSS data can be used for synchronization [11], [12]. By using this UTC, data collected by different sensors can be stored in a database with a time tag. Fig. 15 illustrates the time-tagged GNSS, LVDT, and inclinometer data. For the shown example, GNSS data are sampled with 1 Hz, whereas LVDT and inclinometer data are sampled

of the railway can be eliminated to determine the track axis elevation. Depending on the inclinometer axes in Fig. 10, when the right super-elevation (+) and plus gradient (+) condition occurred, the actual track axis on point 1 is measured on point 1 because of the deviation caused by the super-elevation and gradient. This condition is demonstrated in Fig. 11. The actual track axis coordinates on point 1 are determined by considering the offset values dened in (1) and (2). By considering 1 = arctg t1 2 = arctg S11 = h sin E = B h sin X2 X1 Y2 Y1 (4) (5) (6) (7)

(h sin )2 + (h sin )2

t1 1 = t1 2 + 200 + 1

AKPINAR AND GLAL: MULTISENSOR RAILWAY TRACK GEOMETRY SURVEYING SYSTEM

195

TABLE I DIFFERENCES BETWEEN FILTERED FORWARD AND BACKWARD MEASUREMENTS WHEN THE GNSS RECEIVER WAS USED FOR POSITIONING

TABLE II DIFFERENCES BETWEEN REAL PARAMETERS AND FILTERED AVERAGES OF F ORWARD AND B ACKWARD M EASUREMENTS W HEN THE GNSS R ECEIVER W AS U SED FOR P OSITIONING

with 10 Hz. In Fig. 15, the rst sentence contains UTC, LVDT data, inclinometer data, horizontal coordinates, elevation, and quality information of GNSS data, respectively. If only the total station is used for positioning, UTC cannot be derived. In this case, a data acquisition card connected to the LVDT and inclinometer can be triggered by computer time. In this case, UTC in Fig. 15 is replaced with personal computer time. Thus, it is possible to store all sensor data in a database with time tags. IV. E XPERIMENTAL T EST To test the accuracy of the designed surveying system, forward and backward measurements had been done on a railway that contains curves and straight lines and is 420 m long. In the rst stage of the data processing section, the MAD ltering technique was used for outlier detection. Outliers were detected in the LVDT and inclinometer data on 142 m, and these outliers were eliminated. Forward and backward measurements were ltered by using the customized kinematic model of the Kalman ltering technique [8], and ltered values were compared with known railway geometrical parameters, which are called real parameters in this paper. The maximum, minimum, and mean values and standard deviations related to the differences between ltered forward and backward measurements are given in Table I to determine the inner accuracy of the surveying system when the GNSS receiver was used for positioning. The maximum, minimum, and mean values and standard deviations related to the differences between the real parameters and the average of forward and backward measurements are given in Table II to determine the outer accuracy of the surveying system when the GNSS receiver was used for positioning. The same tests were done when the total station was used for positioning. The results are given in Tables III and IV. The differences between the real parameters and the Kalman ltered averages of forward and backward measurements were compared with the 3 criteria, which has 99% probabilities in a normal distribution. The results of this comparison are given in Figs. 1621. values used in this comparison were obtained from the standard deviations of GNSS, total station, LVDT, and inclinometer measurements. All of the difference values related to

TABLE III DIFFERENCES BETWEEN FILTERED FORWARD AND BACKWARD MEASUREMENTS WHEN THE TOTAL STATION WAS USED FOR POSITIONING

TABLE IV DIFFERENCES BETWEEN REAL PARAMETERS AND FILTERED AVERAGES OF F ORWARD AND B ACKWARD M EASUREMENTS W HEN THE TOTAL STATION WAS USED FOR POSITIONING

Fig. 16. Differences between real parameters and ltered averages of forward and backward measurements for the track gauge.

the horizontal position, elevation, super-elevation, and track gauge are smaller than 3 criteria. Thus, it can be said that the differences between real parameters and ltered averages of forward and backward measurements are insignicant. In addition, the outer accuracy of the surveying system had been achieved. V. C ONCLUSION Depending on the UIC standards, the tolerance values for the horizontal track axis geometry on ballast and slab tracks are 25 and 15 mm, respectively. The tolerance value for the vertical track axis geometry is 1015 mm for both types

196

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT, VOL. 61, NO. 1, JANUARY 2012

Fig. 17. Differences between real parameters and ltered averages of forward and backward measurements for the super-elevation.

Fig. 21. Differences between real parameters and ltered averages of forward and backward measurements for the vertical position obtained by the total station.

Fig. 18. Differences between real parameters and ltered averages of forward and backward measurements for the horizontal position obtained by the GNSS receiver.

Considering the results in Tables III and IV, it can be said that the total station can be also used to determine the horizontal and vertical track axis geometry. Moreover, using the total station instead of the RTK GNSS method enhances the accuracy of the track axis geometry. Depending on the UIC standards, the uncertainty values for the super-elevation and track gauge cannot exceed 5 and 1 mm, respectively. Considering the results in Tables I and II, it can be said that the designed surveying system can be used to determine the super-elevation, track gauge, and gradient values. The surveying system can be further improved by integrating different sensors. By integrating the inertial navigation systems and odometers into the surveying system, it will be possible to dead reckoning when the GNSS and total station signals interrupt. R EFERENCES
[1] M. Dnisch, H. Kuhlmann, and W. Mhlenbrink, Baubegleitendes Festpunktfeld bei der Einrichtung und Kontrolle der Festen Fahrbahn, Wichmann, Heidelberg, 2000, AVN 10/200. [2] I. Milev and L. Gruendig, Rail track data base of German railThe future automated maintenance, in Proc. INGEO FIG Regional Central Eastern Eur. Conf. Eng. Surv., Bratislava, Slovakia, Nov. 1113, 2004, pp. 18. [3] J. Trehag, P. Handel, and M. gren, Onboard estimation and classication of a railroad curvature, IEEE Trans. Instrum. Meas., vol. 59, no. 3, pp. 653660, Mar. 2010. [4] L. Beales, Track system requirements, Railway Group Standards, GC/RT5021, Railway Safety, London, Oct. 2003. [5] R. T. Lemmon, The inuence of the number of satellites on the accuracy of RTK GPS positions, Australian Surv., vol. 44, no. 1, pp. 6470, 1999. [6] B. Hofmann-Wellenhof, H. Lichtenegger, and J. Collins, GPS-Theory and Practice, 5th ed. Wien, NY: Springer-Verlag, 2001. [7] E. Glal and B. Akpinar, Applications of GPS based machine guidance systems in open pit mining operations, in Proc. Int. Conf. Modern Manage. Mine Producing, Geology Environ. Protect., Varna, Bulgaria, Jun. 913, 2003. [8] B. Akpinar, A new measurement system design for determining the geometrical changes on railways, Ph.D. dissertation, Yildiz Technical Univ., Istanbul, Turkey, 2009. [9] DLH General Directorate, Light Rail System Criteria, 1995, Republic of Turkey, Ministry of Transport and Communications, Ankara. [10] J. Stephen and G. Lachapelle, Development and testing of GPSaugmented multi-sensor vehicle navigation system, J. Navig., vol. 54, no. 2, pp. 297319, May 2001. [11] R. Glaus, G. Peels, U. Mler, and A. Geiger, Precise rail track surveying, GPS World, pp. 12182004, May 1. [12] B. Li, C. Rizos, H. K. Lee, and K. H. Lee, A GPS-slaved time synchronization system for hybrid navigation, GPS Solutions, vol. 10, no. 3, pp. 207217, Jul. 2006.

Fig. 19. Differences between real parameters and ltered averages of forward and backward measurements for the vertical position obtained by the GNSS receiver.

Fig. 20. Differences between real parameters and ltered averages of forward and backward measurements for the horizontal position obtained by the total station.

of tracks. Considering the results in Tables I and II, it can be said that the RTK GNSS method can be used to determine the horizontal and vertical track axis geometry.

AKPINAR AND GLAL: MULTISENSOR RAILWAY TRACK GEOMETRY SURVEYING SYSTEM

197

Burak Akpinar received the Ph.D. degree in geomatics engineering from Yildiz Technical University (YTU), Istanbul, Turkey, in 2009. Since 2002, he has been a Research Assistant with the Department of Geomatics Engineering; Division of Surveying Techniques, YTU. His research interest includes engineering surveying, surveying instruments, and deformation measurements on railways.

Engin Glal received the Ph.D. degree in geodesy from Leibniz Universitt Hannover, Hannover, Germany, in 1997. From 1998 to 2009, he was an Assistant Professor with the Department of Geomatics Engineering; Division of Surveying Techniques, Yildiz Technical University, Istanbul, Turkey. Since 2009, he has been an Associate Professor with the same department. His research interest includes deformation measurements and the analysis of engineering structures, system analysis, and engineering surveying.