Preventive Maintenance of Railway Infrastructures using GPR Ground Penetrating Radar

Transcription

1 Preventive Maintenance of Railway Infrastructures using GPR Ground Penetrating Radar Raúl Mínguez Maturana (1), Begoña Duclos Bautista (2), Álvaro Andrés Aguacil (3), Miguel Rodríguez Plaza (3) & Senén Sandoval Castaño (1) (1) Geofísica Aplicada Consultores, S.L.(Madrid, Spain), (2) Universidad Complutense de Madrid (Spain), (3) ADIF, Administrador de Infraestructuras Ferroviarias (Madrid, Spain). Abstract The increasing necessity for optimization of the maintenance resources of railway infrastructures entails the implementation of strategies for preventive maintenance in order to maximize the operational life of such infrastructures. One of the most demanding tasks in daily maintenance programs is platform repairment and particularly ballast condition assessment. Therefore a tool able to rapidly evaluate ballast and platform condition is essential. Ground penetrating radar (GPR) is a non-invasive, fast and versatile evaluation method, that allows the auscultation of long sections of railway track (hundreds of kilometers) in a very short time period. It provides high resolution images (data collected every few centimeters) and it due to its non invasive nature it can be repeatedly used over the same spot to study the evolution of a given pathology. GPR transmits an electromagnetic wave into the terrain with a propagation velocity that depends on a set of properties that can be translated into subtle variations of the platform conditions. The use of GPR has allowed to determine accurately ballast thickness, to estimate its quality (fouling degree) and to identify and limit areas of moisture accumulation. The repetition of these tests allowed to estimate the degradation speed of the platform and to evaluate the effectiveness of the maintenance works. Thus, it will be possible to plan the necessary maintenance works in the short, medium and long term, with the associated saving of costs and execution periods. Keywords: GPR, pulse velocity, auscultation, ballast thickness, contamination degree, moisture, platform condition, preventive maintenance. 1. INTRODUCTION During the last years high-performance railway lines have increased both their number and capabilities. Maintenance of these railway lines requires tools that allow for a continuous and fast assessment of its current condition. Proper analysis of the platform condition is crucial in order to reduce maintenance costs and increase operational safety levels. This article proposes GPR as a tool to support and complement standard dynamic and geometric parameters in a way that allows monitoring and diagnosis on near real-time. GPR can determine the deteriorating conditions of ballast, sub-ballast and subgrade. Databases are generated and updated along with the planned daily operations of the line in order to plan optimized maintenance works. This paper shows how GPR auscultation of the railway platform can determine continuously the geometry of the different layers identifying (among others) areas with poor drainage and areas with contaminated ballast. Joint interpretation of these new datasets and the currently existing (traditional) ones enhances our knowledge of the track status allowing to identify potential problems and to act on them in short periods of time. 2. METHODOLOGY The implementation of predictive maintenance policies requires firstly a detailed knowledge of the state of the railway platform, and secondly a comparison over time to determine performance thresholds. To achieve these goals a non-invasive diagnostic tool is needed. This tool must be able to investigate continuously large sections of track with sufficient resolution and high sampling rates. In this regard GPR has been proven in the past to be the most suitable choice

2 GPR is a shallow geophysical technique capable of generating 2D- and 3D-presudoimages of the subsurface structure. The advantage over other conventional techniques such as boreholes or trenches is that it is a non-invasive method of investigation, fast and versatile, with acquisition speeds up to 80 km/h. A pulse of electromagnetic radiation (around a dominant frequency) is directed towards the surface under investigation. The frequency of the pulse (among other factors) determines the resolution and maximum depth of investigation. When the wavefront encounters a change in the electromagnetic properties of the medium due to, for example an interface between two different media (ballast subballast interface), part of the energy is reflected back to the surface and part is transmitted to greater depths. The propagation wavespeed depends on several properties of the medium that produce absorption and scattering of the signal. These variations on the pulse shape can be directly related to the various conditions of the platform. The most important parameters whose properties can be investigated with GPR in railway applications are: Thickness of the different layers (ballast, sub-ballast, subgrade). Degree of contamination of the ballast Moisture content of the ballast Possible settlement areas To determine the thickness of each layer it is necessary to identify first the reflector for each interface (e.g. ballast sub-ballast interface) and then the pulse velocity in that layer. Usually, the velocity is obtained by correlating the GPR information with data obtained from direct sampling of the platform (trenches or boreholes). The traditional approach is based on the interpolation of sparse data throughout long distances with the associated uncertainties. With the aid on GPR data this approach is no longer necessary and continuous information can be derived. It is also possible to directly measure the velocity of the GPR pulse in each of the layers using variable offset measurements like WARR (Wide Angle Refraction and Reflection) or CMP (Common Mid-Point) (Jol, 2009). Such tests were done manually in the past making data collection painfully slow. The use of modern antenna arrays significantly reduces data acquisition time making these kinds of techniques of great value. GPR pulses in ballast are affected by contamination from the fine-grained material from the subgrade. The most obvious effect is the attenuation suffered by the higher frequencies (Leng & Al-Qadi, 2009). Other visible effects are the appearance of reflectors inside the ballast, the loss of intensity of the reflector at the base of the ballast and the decrease in the propagation velocity. The pulse velocity depends on the composition of the materials and the degree of moisture and contamination (Figure 1). In the case of clean ballast, the velocity ranges between 0.12 m/ns and 0.21 m/ns. For ballast contaminated with fine-grained material ranges from 0.08 m/ns and 0.12 m/ns (Göbel et al., 1994). This property makes pulse velocity determination of outmost importance. Figure 1. Different degrees of contamination in ballast (modified after Göbel, 1994). Water content is one of the parameters that affects most the GPR pulse (Huisman et al., 2003) producing a large reduction of the pulse velocity and the appearance of reflectors of high amplitude and low frequency. Moreover, the thickening of the ballast layer indicates subsidence areas. The - 2 -

3 investigation of these zones by three-dimensional models contributes to detailed information of the morphology and extension, which allows to consider the intensity of the deformation. 3. FIELD TESTS With the aim of studying the pulse velocity variation under different ballast conditions, a series of controlled tests we carried out Contrast tests A 4,0x0,5x0,5m wooden cell was built as a ballast platform mock-up. The aim was to collect GPR data periodically over several months and compare the effects of time on the data. The following parameters were studied: (1) pulse velocity in sands and ballast, (2) ballast-sand interface identification and, (3) detection of water concentration areas. The cell was constructed over natural vegetal soil and two metallic plates were placed at the bottom of the cell. The bottom half of the wooden box was filled with medium coarse-grained sand forming an irregular layer. Over this sandy layer ballast extracted from samples taken at the ADIF Central Laboratory at Villaverde (Madrid) was deposited (Figure 2). The ballast lithologies were quartzite, amphibolite and granite. For the data acquisition a MALÅ Geoscience ground coupled GPR system with shielded antennae was used. Central frequencies of 500 MHz, 800 MHz and 2,3 GHz were selected. Data was processed using Sandmeier s REFLEX software. Figure 2. Top. Snapshots taken during the construction of the test cell. Down left. Schematic diagram showing a vertical cross-section of the test cell. Down right. 800MHz radargram and interpretation. In Figure 2 (down at right) the interface between the bottom of the ballast and the sand is depicted as a weak and discontinuous reflector. The signal in the sandy layer is clearly attenuated (loss of reflectivity). The reflector of the bottom of the sands is strong, continuous and curved (contrarily to the - 3 -

4 flat bottom of the cell). This apparent discrepancy in the shape of the cell bottom is due to differences in the pulse velocity associated with the different layer thicknesses. The pulse velocity is approximately 0.18 m/ns in the ballast layer and 0.10 m/ns in the sand layer. In those areas of the cell where the thickness of the ballast layer is larger the bottom reflector appears shallower (higher average velocity resulting in smaller traveltimes) whereas in those areas where the sandy layer is thicker the bottom reflector appears deeper. Figure 3 shows four radargrams acquired with the 800 MHz antenna on different months under diverse meteorological conditions. For example, data shown at Figure 3C was acquired under wet conditions and the ballast-sand interface is readily visible. This effect is due the large velocity contrast between the ballast and the wet sand. The undulations of the cell-bottom reflector are more pronounced. This agrees with the fact that there is a larger velocity contrast between the ballast and the sand. Figure 3. Radargrams of the test cell acquired with a 800 MHz MALA antenna. A) Right after building the cell. B) A month later (January 2010). C) Five months later (April 2010) after heavy rains. D) A year later (January 2011) after several weeks with no rain. Results in radargram D (Figure 3D) correspond to data acquired in dry conditions. The bottom cell reflector appears almost flat. This indicates that the velocity contrast between sand and ballast is smaller. Finally, if we look closely to the shape of the bottom reflector in all cases one can see that in the three first cases there is a reverberation whereas in the last case the reflector is sharper, indicating a drier condition of the underlying ground Calibration tests Once proven the importance of the velocity in the correct determination of any given geometry, the influence of fine-grained contamination in ballast is analyzed. For this purpose a test to quantify the influence of both water content and sand content in ballast pulse velocity was designed. These tests were carried out at the ADIF Maintenance Base of Brihuega (Guadalajara) belonging to the High-Speed Track between Madrid and Barcelona. A 400 liter PVC box was initially filled with ballast to the top (Figure 4). Then different amounts of water and sand were added. A metallic plat was placed at the bottom to create a strong reflector

5 A 500 and 800 MHz set of MALÅ Geoscience shielded antennae were used. Measurements were: 1) clean ballast (without water and sand), 2) ballast contaminated with 10, 20, 25 and 30% in weight of sand in dry conditions, and 3) ballast contaminated with the same amount (in weight percentage as before) of sand varying the amount of water. Water was added at 5 liter intervals to wet the ballast until saturation was reached. After a first analysis of the recorded data, the 800 MHz antenna was discarded because the penetration limit of this antenna falls short for the selected box (87 cm). Figure 4. Pictures taken during the ballast calibration tests at the ADIF Maintenance Base at Brihuega. Pulse velocities range between 0.17 m/ns for clean ballast and 0.10 m/ns for highly contaminated ballast. The presence of water (without sand) does not influence significantly on the pulse velocity as compared with the presence of sand. This result is explained because in normal conditions ballast suffers fouling and degradation resulting in fine-grained material filling the gaps between the ballast. This fine-grained material absorbs water decreasing substantially the draining properties of clean ballast. Figure 5. Pulse velocities measured in ballast using a 500 MHz GPR antenna with different amounts of contamination - 5 -

6 Pulse velocities follow an uniform and negative linear trend when plotted against water and sand contamination (Figure 5). Velocities close to 0.10 m/ns are measured when the critical limit of 25% (in weight) of fine-grained material and 20 liters of water is reached (5% en volume) Track tests. Comparison of different antennae. To determine the most suitable GPR system from the available ones in the market a series of tests were carried out. A railway line with heterogeneous conditions in ballast thickness and degree of contamination was selected. The selected area is 100 km south of Madrid in the conventional railway line that connects Madrid and Valencia de Alcántara. A total amount of 50km of GPR data was gathered in one single night. The railway line is not electrified and the steel rails seat atop of concrete bi-block sleepers. Among several tests to find out the performance of each GPR system used, a resolution test was done in a 1km section close to the town of Rielves. This railway section has in the center a 400m embankment that has suffered many geotechnical problems in the past. Equipment from three different manufacturers were tested. The first test was carried out in June 2009 with a Ground Coupled system (250 and 500 MHz) from MALA Geoscience. Data was acquired every 2cm referencing the data with a submetric GPS. The second system was the Norwegian built 3D-Radar composed by an array of 21 antennae spaced 7.5cm apart (across measuring direction) with step-frequency technology allowing to acquire data between 100 and 3000 MHz. The GPR array was installed in the front of an ADIF maintenance train (Figure 6). Tests were performed at low speed (4 km/h for maximum resolution) and high speed (80km/h for low resolution). Positioning was achieved using a submetric GPS. The last system employed in this work was from the Italian manufacturer IDS composed by three aircoupled 400 MHZ antennae. The system was installed as well in the front of an ADIF maintenance train. Measurements were done every 6 and 12cm at acquisition speeds between 60 and 80 km/h. Figure 6. 3D-Radar GPR system tested mounted in the front side of an ADIF maintenance train. Figure 7 shows radargrams along the center of the track for the same section of track. The broad spectrum of the dataset acquired with the 3D-Radar system produces well-defined reflectors with weaker amplitudes than the two other systems. The IDS data produce high quality - 6 -

7 data although they lack the high frequency content component clearly visible in the 3D-Radar example. Lastly, the MALA dataset has less resolution and the data present strong reverberations. The goal of any GPR system used in railway platform investigations is to provide the maximum number of parameters about the current status of the different layers that compose the platform. In this regard, the 3D-Radar system produced the sharpest image and offers the possibility to create three-dimensional images of the platform and measure directly pulse velocities in a CMP configuration. Figure 7. Top. Radargram corresponding to the center cannel of the 3D-Radar system with frequencies between 100 and 3000 MHz. Center. Radargram acquired with the 500 MHz ground-coupled MALA system. Down. Radargram corresponding to the 400 MHz air-coupled IDS system Track tests. Testbed The next step was to select a section of railway to be renovated completely and use it as a testbed to measure different properties before and after renovation woks were completed. The renovation consisted in: Programmed substitution of rails, ties and power lines. Restoration of the ballast. Detection and repairment of subgrade anomalies. The restoration works aimed to lengthen the operational live of the railway track, minimizing the frequency of future conservation efforts reducing at the same time the associated economical costs to operate the line. The testbed was located at the section Tembleque-Huerta de Valdecarábanos belonging to the conventional railway line between Madrid and Alcazar de San Juán. The ballast layer was completely renovated using 70% of the original material. After the restoration works the ballast thickness should be 30 cm. The investigated section is 2km long between the and distance markers. The data acquisition was achieved using a V2429 GPR antenna from 3D-Radar composed by 29 channels spaced 7.5 cm across the measuring direction. In one single profile a total swath of 2.1m was - 7 -

8 analyzed. CMP measurements were also acquired obtaining a continuous profile measure of the pulse velocity before and after the restoration works. The reduced dimensions of the testbed (2 km) and the presence of maintenance machinery during the tests suggested the measurements to be acquired in walking mode (Figure 8) allowing maximum resolution data. The antennae were suspended on a light aluminum frame. Data was references again using a submetric GPS. Acquisition parameters were kept constant before and after the restoration works for comparison of the resulting datasets: Zero-offset measurements using the 29 channels with a spatial sampling of 7.5cm along (and across) the track. The frequencies acquired range between 100 and 1000 MHz. With this configuration, a high-resolution 3D model of the ballast layer was constructed. CMP measurements. With a maximum offset of 2.1m velocities were measured every 10cm along the track. Figure 8. 3D-RADAR system pushed manually along the testbed. To compare the ballast layer before and after the restoration works several parameters were measured and different indexes were subsequently calculated. These indexes were calculated for every single data point to quantify objectively properties in the datasets that otherwise could lead to subjective results. Combining different indexes areas with clean and fouled ballast as well as areas with high water content could be identified. Three different indexes were defined based on the amplitude and frequency content of the data. The amplitude indexes are: 1. Data inflection points 2. Numerical integration of the radargram (in absolute value) 3. Maximum amplitude (around the ballast bottom reflector) The frequency indexes are: 1. Maximum frequency of the whole radargram 2. Maximum frequency of the radargram above the ballast bottom reflector 3. Maximum frequency of the radargram below the ballast bottom reflector The interpretation of each index is now described. Number of inflection points. This parameter counts the number of inflection points (both maxima and minima) along a single data measurement in the corresponding time window of the ballast layer. The number of inflection points is associated with the scattering of the pulse in the ballast layer - 8 -

9 which is directly related with the degree of contamination of the ballast layer. Numerical integration of the radargram (in absolute value). It involves calculating the area occupied by the radargram curve (in absolute value) at a single point. High values of the area correspond to larger amplitudes of the radargram. This is associated with fine-grained contamination of the ballast layer. Maximum amplitude (around the ballast bottom reflector). It is the maximum signal amplitude (in a time window of 2ns around the ballast bottom reflector). When the ballast is contaminated the reflectivity at the bottom interface is low. It also provides information about the humidity below the reflector because high amplitude reverberations below the reflector often indicate high concentrations of water. Maximum frequency of the whole radargram. The dominant frequency is calculated for the whole radargram. Maximum frequency of the whole radargram above the ballast bottom reflector. Indicates the dominant frequency in the ballast at each point. High values correspond with clean ballast whereas low values correspond with contaminated areas. Maximum frequency of the whole radargram below the ballast bottom reflector. Indicates the dominant frequency below the ballast at each point. From the CMP measurements the pulse velocity variation has been calculated inside the ballast layer before and after the restoration works. Figure 9 shows the radargrams before and after the restoration works. It is worth noting that the radargram corresponding to the renovated section was acquired after several days of intense rain which can be seen with the presence of the large reverberations below the ballast reflector. Figure 9. Radargrams corresponding to the central channel of the 3D-Radar antenna before (top) and after (down) of the restoration works. Figure 10 show the calculated indexes for both tests before and after the restoration works. The following conclusions can be derived: 1. The ballast bottom reflector after the restoration woks is more homogeneous and flatter than before. No significant deformations are seen in the ballast layer after being restored. 2. An increase in the pulse velocity (around 15-30%) is measured after the restoration works. 3. There is a general reduction trend in the indexes that are based on the amplitudes of the - 9 -

10 radargrams (right column in Figure 10) associated with the reduction in contamination inside the ballast layer. 4. Indexes based on the frequency content of the radargrams do not show apparent changes between both surveys. However, a clear reduction of the frequencies can be observed (Figure 10, bottom left) under the ballast layer associated with the increase of humidity existing during the survey after the restoration works. Figure 10. Six different indexes for the ballast layer calculated before and after restoration works. 4. CONCLUSIONS GPR ability as an efficient tool for ballast and subgrade condition assessment has been tested. CMP measurements allow to determine the pulse velocity inside the ballast layer. Using this parameter the total thickness of the ballast layer and its degree of contamination can be determined. Knowledge about ballast thickness variations with time will allow to quickly identify those areas under subsidence and correction measurements will be applied before further pathologies are developed. The final goal is to program efficiently maintenance works and increase the safety of day-to-day operations. Pulse velocities are very sensitive to humidity and contamination of the ballast layer. This property makes pulse velocity determination a tool to pin down areas prone to suffer pathologies in the near future. Finally, a better understanding of the ballast and subgrade current conditions is crucial to understand its future evolution and to plan ahead of time (even before the pathology shows up) maintenance works reducing economical costs and increasing operational safety levels

11 Acknowledgements This research was supported by the Ministerio de Fomento (Ministry of Public Works) and Ministerio de Ciencia e Innovación (Ministry of Science and Innovation). Our gratitude to ICYFSA, the maintenance staff of ADIF and the Universidad Complutense of Madrid. References Göbel, C., Hellmann, R., Petzhold, H. Georadar model and in-situ investigations for inspection of railways tracks. Proceedings 5th International Conference on Ground Penetrating Radar. Kitchener, Canada. June (1994). Huisman, J.A., Hubbard, S.S., Redman, J.D. and Annan A.P. Measuring Soil water Content with Ground Penetrating Radar: A Review. Vadose Zone Journal, 2: (2003). Soil Science Society of America. Jol. H. M., Ground Penetrating Radar: Theory and Applications. Elsevier Science. First Edition, pp (2009). Leng Z. and Al-Qadi I. Dielectric Constant Measurement of Railroad Ballast and Application of STFT for GPR Data Analysis. NDT&E International. Nantes, France. June 30th July 3rd (2009)