GPR ANTENNA ARRAY FOR THE INSPECTION OF RAILWAY BALLAST

Transcription

1 Proceedings of the National Seminar & Exhibition on Non-Destructive Evaluation NDE 2011, December 8-10, 2011 GPR ANTENNA ARRAY FOR THE INSPECTION OF RAILWAY BALLAST Th. Kind BAM Federal Institute for Materials Research and Testing, Division 8.2, Berlin, Germany ABSTRACT A GPR antenna array was investigated for automated measurement of wave propagation velocity as a measure of the ballast quality. This goes beyond the regular application of conventional radar systems offering only qualitative structural information. The new approach provides a fast NDT method for the classification of ballast fouling. A main advantage of the multi-offset radar is that regular ballast digging for velocity evaluation can be avoided and travel time can be directly transformed into an equivalent depth by using the automatically evaluated velocity. Fast monitoring of ballast fouling condition can reduce maintenance cost significantly by supporting the works necessary for ballast cleaning and furthermore this method reduces interruption time for a better railway service. Keywords: Railway, ballast, track bed investigation, velocity analysis PACS: Bb INTRODUCTION Railway ballast fouling reduces the safety of the wheel-track system, due to the increase of stiffness and shear strength of the degenerated rail track bed. Fast monitoring of ballast fouling condition can reduce maintenance cost significantly by supporting the works necessary for ballast cleaning [11]. The application of GPR for rail track investigation has been proven to be an useful tool for detecting layer interfaces like the ballast/subgrade interface [3-14]. It is performed by monitoring the travel time of the reflection from the ballast subgrade interface along the track. The depth of the ballast layer can be calibrated by calculating the propagation velocity of the radar signal in the ballast. Therefore depth measurements at selected points along the track have to be correlated with corresponding travel times. This kind of depth calibration is expensive because it requires digging or drilling of the ballast. The calculated depth between two calibration points depends either on a change of propagation velocity or on an actual change of the thickness of the ballast layer. This leads to an uncertainty because with local calibration it is not possible to distinguish between a change in the vertical position of the reflection interface or a change of the propagation velocity. Measurements with a multi-offset antenna arrangement can solve this problem in a similar way to the velocity analysis applied to seismic data for a common midpoint measurement [1]. BALLAST VELOCITY ANALYSIS Ballast fouling appears as a decrease of the volume of air voids inside the ballast. The reduced proportion of air within the ballast leads to an increase of the relative permittivity. Typical values for the relative permittivity of clean ballast are in the range of and for fouled ballast in the range of 4-38 [15]. The increased relative permittivity causes a decrease of the GPR propagation velocity v and can thus be measured using the common midpoint (CMP) method [1,2]. In a CMP measurement receiver and transmitter are moved in opposite directions from a center point with the same speed. The travel time t of the reflection from layered structures will increase with increased offset x between receiver and transmitter by: This formula is also valid for the movement of only one antenna, if a layer parallel to the surface causes the reflections. This antenna configuration with the movement of only one antenna is designated as common reflection surface (CRS). With this formula the difference t NMO = t- t 0 (NMO: Normal Move Out) between zero offset t0 and the travel time t can be calculated and used to compensate the increase of travel time (see Figure 1). If the velocity v is chosen properly, the corrected travel time t* is constant for each CMP trace. (1)

2 282 Kind : Proceedings of the National Seminar & Exhibition on Non-Destructive Evaluation Fig. 1 : CRS antenna configuration (TX: transmitter, RX: receiver) Fig. 2 : Diagram of an CMP/CRS measurement; t0 denotes the minimum travel time of the layer reflection; red dots indicates the recorded reflection of five different antenna offsets Stacking the corrected CMP traces results in an increase of the reflection amplitude at t=t 0. In the case of improper selection of velocity the reflection amplitude is spread out along the trace. A sequence of stacked traces with different applied velocities will result in a velocity spectrum along the time axis. The velocity spectrum can be calculated by using a sub-set of CMP traces, too. An example of a velocity spectrum and a comparison between a velocity spectrum calculated by different numbers of CMP traces will be given below. RAILWAY MULTI-OFFSET ANTENNA ARRAY A multi-offset array for ballast investigation was built. Array design is one of a linear type. One transmitter and up to five receivers are set up in a line (see Figure 3 and 5b). To verify the feasibility of a multi-offset array in advance, an antenna array has been simulated by using two 500MHz antennas (GSSI: SIR20) one after the other in each antenna position of the array. Tests were carried out with the linear array aligned along and perpendicular to the track. BALLAST TEST SPECIMEN For testing the concept of ballast velocity analysis, a 4m x 4m test specimen has been constructed (see Figure 4). The height of the ballast is about 80cm. Four reinforced concrete sleepers are placed in a distance of about 70cm, followed by a region without sleepers of about 140cm. MEASUREMENTS A trolley with a position wheel, which was led along the top surface of one trail, was used for the measurements. For all measurements two 500MHz antennas were used. The transmitting antenna was placed in a fixed position with an elevation of 19cm above the ballast. The receiving antenna was mounted on the trolley in the same elevation as the transmitting antenna (see Figure 5a). Ten parallel scans were recorded using the trolley. The measurement traces were triggered by the wheel every 5mm. In one complete measurement cycle an area of 1.2m x 2m was scanned. Parallel polarisation between transmitting and receiving antenna were used. The polarisations were directed along and across the track. RESULTS For the velocity analysis a scan of the receiving antenna has been used, which is in line with the transmitting antenna and along the track. Figure 6 shows the scan. Fig. 3 : CMP/CRS by an linear antenna array (a); Prototype of the linear antenna array (b)

3 NDE 2011, December 8-10, Fig. 4 : Ballast test specimen Fig. 6 : Radargram with fixed transmitter and moved receiver Fig. 5 : Multi-Offset measurement with one fixed transmitter and moved transmitter on the ballast test specimen (a); Multi-Offset measurement with a linear antenna array on a ballast railway track (b) The data were processed only with dc removal and without application of time gain. Reflections from the ballast surface an d the ballast bottom side are clearly visible. Figure 6 shows only a minor influence of the sleeper on the radargram. Only the ballast bottom side reflection is slightly affected by the sleepers. From the radargram seen in Figure 6 and a given propagation velocity the normal move out time for each trace is calculated by the offset between transmitter and receiving antenna. In a second step the time position of each trace is corrected by the corresponding normal move out time. The reflection shown in Figure 6 appears tilted. If the appropriate velocity is chosen, the reflections will be displayed as horizontal lines in the radargram. The sum of all corrected traces will increase the summed amplitude of the reflections to a maximum at the zero offset position. To find the most appropriated propagation velocity, the normal move out has been corrected and all traces in a velocity range of 5x x10 8 m/s have been summed up. Velocities greater than the speed of light are used only for visualizing the velocity spectrum. They have no practical meaning. The velocity spectrum of the summed traces is shown in Figure 7a. Two regions of high amplitudes of the summed reflections are formed around the velocities of 3x10 8 m/s in the upper right corner and around 2x10 8 m/s in the centre of the Figure 7a. The upper right region represents the ballast surface reflection and the centre region can be referred to the ballast bottom side reflection. This velocity analysis measurement shows that the multi-offset measurement of the propagation velocity inside the ballast can be used as an evaluation tool of the ballast degradation: the shifting of the areas of high amplitudes of the summed reflections within the travel time vs. propagation velocity diagram (Figure 7a) is an indication for the state of the ballast.

4 284 Kind : Proceedings of the National Seminar & Exhibition on Non-Destructive Evaluation Fig. 7 : Velocities spectrum (a); (b) (d) calculated from a limited number of 8,6 and 4 traces For a multi-offset antenna array the number of traces and the lateral position of the traces are limited by the scale of a single antenna. Choosing an antenna in the frequency range of about MHz will limit the number of antennas to 4 10 elements per meter inside the array. For fast acquisition a multioffset antenna array should collect all traces for one velocity analysis at one array position at the same time. The velocity analysis has been carried out with the data shown in figure 4 but with a sub set of traces in order to test the impact of reduced trace numbers on the quality of the velocity analysis. 4, 6 and 8 traces with equal distance between the traces have been selected. The velocity analysis was done as described above for the results presented in Figure 7a. Figure 7b-7d shows the radargrams together with spectra of the propagation velocity. CONCLUSION Conventional propagation velocity analysis can be applied for the investigation of ballast track beds. For a fast acquisition on a railway inspection train, a multi-offset array can be used, which is acquiring the necessary traces for velocity analysis in one step and with minor impact to the quality of the velocity spectrum. In a ongoing investigation it will be investigated, if a decrease of propagation velocity caused by fouling of the ballast will lead to a significant shift of the high amplitudes in the velocity spectrum towards lower velocities on the left side of the spectrum.

5 NDE 2011, December 8-10, ACKNOWLEDGEMENTS Part of this work was part of the project SAFERAIL, which was founded by the European Union in the 7 th Framework program. REFERENCES 1 Yilmaz, Öz, Seismic Data Analysis, Vol.1, SEG, (2001) 2 Daniels, D.J., Ground Penetrating Radar, 2 nd Edition, IEE, London, UK, (2004) 3 Olhoeft, G.R., GPR Evaluation of Railway truck Substructure Conditions, Ninth Intrnational Conference on Ground Penetrating Radar, Santa Barbara, US, (2002) 4 Olhoeft, G.R., Smith, S., Hyslip, J. P., Selig Jr., E.T., GPR in Railroad Investigations, Tenth Intrnational Conference on Ground Penetrating Radar, Delft, NL, (2004) 5 Al-Nuaimy, W., Eriksen, A., Gagoyne, J., Train-Mounted GPR for High-Speed Rail Trackbed Inspection, Tenth Intrnational Conference on Ground Penetrating Radar, Delft, NL, (2004) 6 Jack, R., Jackson, P., Imaging attributes of railway track formation and ballast using ground probing radar, NDT&E International, Elsevier, Vol. 32, pp , (1999) 7 Clark, M., Gordon, M., Mike, F., Issues over high-speed non-invasive monitoring of railway trackbed, NDT&E International, Elsevier, Vol. 37, pp , (2004) 8 Sussmann, T.R., Selig, E. T., Hyslip, J. P., Railway track condition indicators from ground penetrating radar, NDT&E International, Elsevier, Vol. 36, pp , (2003) 9 Gallagher, G.P., Leiper, Q., Williamson, R., Clark, M.R., Forde, M.C., The application of time domain ground penetrating radar to evaluate railway track ballast, NDT&E International, Elsevier, Vol. 32, pp , (1999) 10 Hugenschmidt, J., Ballast Inspection Using GPR, Proc 2 nd Int l Railway Conference, London, UK (1999) 11 Selig, E.T. and Waters, J. M., Track Geotechnology and Substructure Management, Telford, England (1994) 12 Kathage, A., Nissen, J., White, G., Bell N., Fast Inspection of Railway Ballast by Means of Impulse GPR Equipped with Horn Antennas, Railway Engineering-2005, London, UK (2005) 13 Smekal, A., Berggren, E., Hrubec, K., Track-Substructure Inestigations Using GPR and Track Loading Vehicle, Railway Engineering-2003, London, UK (2003) 14 Saarenketo, T., Silvast, M.,Noukka, J., Using GPR on Railways to Identify Frost Susceptible Areas, Railway Engineering-2003, London, UK (2003) 15 Clark, M.R., McCann, D.M., Forde, M.C., GPR as a Tool for the Characterisation of Ballast, Railway Engineering- 2003, London, UK (2003)