- Email: [email protected]
Three-dimensional ground-penetrating radar methodologies for the characterization and volumetric reconstruction of underground tunneling
Construction and Building Materials 71 (2014) 551–560
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Three-dimensional ground-penetrating radar methodologies for the characterization and volumetric reconstruction of underground tunneling X. Núñez-Nieto a,d,⇑, M. Solla a,d, A. Novo b, H. Lorenzo c,d a
Defense University Center, Spanish Naval Academy, Plaza de España 2, 36920 Marín, Pontevedra, Spain GeoRadar Division, GeoSystems Business Unit Manager, IDS North America Ltd., 418 Sherbrooke Street East, Montreal, Suite 200, QC H2L 1J6, Canada Department of Natural Resources and Environmental Engineering, School of Forestry Engineering, University of Vigo, A Xunqueira s/n., 36005 Pontevedra, Spain d Applied Geotechnologies Research Group, School of Mining Engineering, University of Vigo, Campus Lagoas-Marcosende, 36310 Vigo, Spain b c
h i g h l i g h t s Documentation and characterization of underground tunneling space by GPR. Volumetric reconstruction that allows determining numerical approximation of volumes. Geometrical dimensioning of the structure from the volumetric reconstruction. Detection of moist areas from the attenuation of the GPR signal. FDTD modeling to assist in the interpretation of the field GPR data.
a r t i c l e
i n f o
Article history: Received 9 May 2014 Received in revised form 25 August 2014 Accepted 27 August 2014
Keywords: GPR Underground space FDTD modeling 3D imaging Moisture Civil engineering
a b s t r a c t This work presents the documentation and characterization of an ancient underground concrete tunnel using the ground-penetrating radar (GPR) method. Three-dimensional imaging methodologies were applied to create an accurate volumetric reconstruction of the underground tunneling space and the whole framework of galleries composing the main structure, which enabled for the dimensioning of the structure. Problems of moisture were also detected in a particular sector of the tunnel. In addition, FDTD modeling was used to improve the understanding of the GPR signal propagation and to assist the interpretation of the field GPR data. Both field and synthetic data have shown the capabilities of the method for the evaluation and characterization of this ancient construction. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Tunneling detection becomes relevant based on a civil engineering point of view, due to the momentous importance of knowing the underground space distribution. In addition to the interest with regard to their influence in the settlement and disposition of the subsoil structure [1,2], the documentation of these constructions also holds connotations of historical and cultural character, architectural interest, tourist attraction or military and defense implications [3–5]. Therefore, their exact location ⇑ Corresponding author at: Defense University Center, Spanish Naval Academy, Plaza de España 2, 36920 Marín, Pontevedra, Spain. E-mail address: [email protected] (X. Núñez-Nieto). http://dx.doi.org/10.1016/j.conbuildmat.2014.08.083 0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
and geometry provides additional information to plan and perform maintenance and repair tasks, which enable for the preservation of their structures and content in optimal conditions [6,7]. This work presents the evaluation and characterization of an existing tunnel located within the premises of the Spanish Naval Academy of Marín, Galicia, in the northwest of Spain (Fig. 1). The galleries of the tunnel develop different functions, namely: ventilation, general storage and military training site for the students belonging to the Spanish Navy Marines Corps, which is the oldest existing Naval Infantry Force in the world [8]. The ground-penetrating radar (GPR) is a geophysical commonly used technique, with a wide range of applications. There have been published numerous studies on different fields, such as military
552
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 1. Location of the prospected area in Europe and specifically in Marín (Galicia, northwest of Spain).
(mine detection [9,10], human remains and life detection [11]), civil engineering (pavement control and integrity of landing runways [12], pathologies in construction [13,14], location of buried structures [15]) and also in geology (detection of subsoil pollution [16] and freshwater bathymetries [17]). Therefore, GPR is proposed in this paper as a solution for the documentation of an underground tunnel because of its rapidity and suitability as non-destructive technique. So that, it is the ideal solution in comparison to other more invasive methods, such as excavation or underground tasting, that may deteriorate the structure of the tunnel and/or its content [18,19]. The GPR method has proven its suitability for providing high image quality results and is a well-recognized prospecting technique by the scientific community [20]. Nevertheless, there still remains some skepticism among the non-geophysical audience, promoted by the difficulty for the understanding of the GPR images. The interpretation of the radargrams (2D GPR images) is not trivial and the use of 3D processing and visualization techniques, has achieved the agreement by the non-expert community in safety underground investigations [21]. Using 3D processing and visualization produces more realistic images of the underground spaces, which allows not only for the location, but also for the 3D reconstruction of underground and buried constructions [22,23]. Therefore, an advanced 3D visualization methodology is presented in this work to provide better interpretation of the results and the volumetric reconstruction and dimensioning of the tunnel. These 3D images provide an accurate and intuitive display of the underground distribution, easily understood even by non-experts in the subject. Numerical simulation was additionally used in this work to assist in the interpretation of the field GPR data. Specifically, Finite-Difference Time-Domain (FDTD) modeling was considered, since the method has demonstrated its capabilities to improve the understanding of the electromagnetic waves propagation through the media [24].
2. Materials and methods 2.1. Tunnel description It consists of an underground tunnel with symmetrical geometry composed by a central corridor and lateral branches (Fig. 2A) that presents a final trifurcation of galleries (Fig. 2B). The main corridor presents two different levels of the tunnel, which are separated by a pronounced ascendant ramp at approximately the middle of its length (Fig. 2C). Some lateral branches present also an initial sector with ascendant slope, as illustrated in Fig. 2D. The underground construction is a concrete solid composed by several blocks seated one above the other, whose top ends in a concrete slab resulting in a sports court. Moisture and condensation problems were observed in certain areas of the ancient concrete construction (Fig. 2E), possibly caused by the poor maintenance of some water pipes existing through one of the lateral sections of the subsoil of the structure. These water pipes are supposed to lead water with chloride content from the swimming pool, near to the tunnel, to the maintenance room and vice versa. Regarding these moisture evidences, it must be also taken into account the close presence of the sea, upon which the newest military installations were built. Moreover, the geographical location of the study site could have also affected the integrity of this ancient construction. That is supported by the location of the village of Marín (Fig. 1) in the Autonomous Community of Galicia, which is characterized by adverse weather conditions with high levels of humidity. 2.2. Field data acquisition Three-dimensional (3D) GPR methodologies were performed in this work (Fig. 3A). Equidistant parallel 2D-lines in only x-direction were acquired on a regular rectangular grid (40 20 m size) with a space between profile lines of 20 cm. 0 The GPR data were collected using a RAMAC/GPR system from MALÅ A Geoscience and a central frequency of 500 MHz (Fig. 3B). The GPR survey was carried out using the common-offset mode with the antenna polarization perpendicular to the direction of data collection [25]. Data was acquired with 2 cm trace-interval or in-line spacing of traces, with reflections collected within a total time window of 75 ns and defined by 512 samples per trace. For trace-interval distance calculation, the GPR antenna was mounted on a survey cart with encoder (odometer wheel). 2.3. 3D data processing GPR-Slice software [26] was used for the three-dimensional (3D) data processing. The first step, previous to the elaboration of the 3D cube, was to filter the raw
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
553
Fig. 2. Detailed images from the interior of the tunnel: (A) main corridor and lateral branches, (B) final trifurcation of galleries at the end of the main corridor, (C) ramp of the principal corridor, (D) lateral branch presenting an ascendant slope section, and (E) evidence of condensation inside one of the lateral branches.
Fig. 3. (A) Prospected area and (B) GPR equipment and components: control unit, 500 MHz antenna (transmitter and receiver), laptop and odometer wheel.
GPR data files by applying the following filters: wobble removal, manual gain, band pass filter (low cutoff: 68 MHz; upper cutoff: 822 MHz), and background removal. The main objective was to amplify the received signal, as well as to remove both low and high frequency noise in the vertical and horizontal directions. Next, the 3D horizontal slices were generated by spatially averaging the squared wave amplitudes of the recorded radar reflections over the time window. The squared amplitude interpolation process creates interpolated time-slices, which are normalized to 8 bit following the color changes between different levels and not actual reflection values [27]. The number of slices depends on the length of the time window selected, the slice thickness and the overlay between slices. Thickness of horizontal slices is often set to at least one or two wavelengths of the radar pulse that is sent into the ground [27]. In any event, the slices generated should not be thicker than the size of the smallest expected target. On the other hand, gaps between the time-slice windows must be avoided because they create discontinuities which can miss targets. In this particular case, instead of generating a large set of depth slices, they are grouped together and its data combined, every 44 cm, to take into account the different depths at which the reflections appear and that could not appear with individual slices. The process involved a set of 16 horizontal slices by 60 samples thick (8.8 ns) over a time window of 482 samples (70.86 ns). Data were gridded using the inverse distance algorithm, which includes a search of all data within a fixed radius of 35 cm of the desired point to be interpolated on the grid and a smoothing factor of two. Grid cell size was set to 5 cm to produce high resolution images. For animations of pseudo-three-dimensional datasets, three interpolations were applied to
the original set of 16 slices in order to generate a smoother visualization. For time to depth conversion, a velocity of 0.1 m/ns for concrete medium was determined from hyperbolae shape analyses. 2.4. FDTD modeling Simulations were performed in this work to evaluate the hypothesis of probable moisture content in concrete, which has been achieved from the interpretation of the field GPR data. It allows for a better understanding of the observed electromagnetic signal propagation phenomena and for the characterization of the radar-wave responses under different conditions of humidity. The synthetic models (Fig. 4) were built using the GprMax software [28], which is an electromagnetic wave simulator for GPR that uses the FDTD method. The models were designed considering concrete slabs of 0.8 and 1.2 m thick ((1) and (2), respectively), since the lateral branches present different depths (Fig. 2C). The electromagnetic properties, assumed from [29] to simulate three different levels of moisture (0%, 35% and 70%), are shown in Fig. 4. Additional simulations were performed to analyze the effect of both chlorides and moisture on the GPR signal. As described in Ref. [29], two different levels of chloride contents were simulated (0.4% and 1% of cement mass) for each moisture content (35% and 70%). Modeling was created with a small spatial-step equal to 8 mm, and the excitation pulse was a Gaussian of 500 MHz center frequency. The trace-interval was 3 cm. Once generated, the synthetic data was transferred from GprMax to ReflexW software [30] using a MATLAB routine, and then filtered using a similar processing sequence to that used for the raw GPR field data.
554
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 4. Synthetic models elaborated to analyze the influence of different levels of moisture on the GPR signal: 0%, 35% and 70%. Two different levels of chloride content, 0.4% and 1% of cement mass, were also evaluated. Concrete slabs of 0.8 (1) and 1.2 (2) m thick were considered, and the values assumed for conductivity (r) and dielectric constant (K) are also shown.
Fig. 5. The sliced 3D cube created.
3. Results and discussions Different 3D visualization techniques were applied to extract information that helps in the reconstruction and characterization of the evaluated tunnel. After creating the 3D cube, for which all the processed radargrams were assembled, the resulted cube was sliced in different directions or planes (Fig. 5). The extraction of
these time-slices, or horizontal maps, provided the location, size and shape of the structural elements. However, one single time-slice did not represent the entire tunneling structure since the different galleries and branches are placed at different levels. To overcome this adversity, overlay analysis was applied (Fig. 6). Thus, the slice obtained from overlay analysis (at 70.86 ns) showed the internal structure of the tunnel,
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 6. Slice generated from overlay analysis considering the whole animation set of time-slices.
and it was possible to identify the central or main corridor, the lateral tunnel branches and the final trifurcation at the end of the main corridor, as seen in Fig. 2. The single time-slice in Fig. 6, obtained from all the strongest reflectors at different depths, can suffer a loss of subtle features when the amplitude of the reflections is not enough. The set of animations of time-slices images generated from the 3D cube can overcome this difficulty, and they provide also the extension and exact positioning in depth for each branch composing the entire structure of the tunnel. Although time-slices were obtained every 0.25 m, relevant variations were not found between two consecutive slices and, therefore, time-slices every 0.5 m were considered in Fig. 7 to simplify imaging. The points where the signal has smaller amplitude are represented in blue1 color, while red colored areas indicate areas of greater contrast in dielectric properties. Thus, the set of time-slices revealed additional information regarding targets placed at different depths such as the two levels that compose the principal corridor, displaying a first sector deeper (from 0 to 10 m in X axis), as seen in Fig. 2C. Although the tunnel is symmetric with respect to its central corridor (Fig. 2A), the lateral branches at the left side of the central axis were not easily distinguished from the radar data, and only the one corresponding to the final trifurcation was partially detected. The most probable cause for this lack of information is the attenuation of the GPR signal in this sector of the tunnel. There are different materials that produce severe GPR signal attenuation such as clay or soil cement (interfacial polarization losses), as well as moisture (water losses) [25,31]. Nevertheless, the original building materials were the same for the entire structure and there are no evidences of later reconstructions that would include possible insulating materials that may cause a loss of the electromagnetic signal. Therefore, based on a later visual inspection made and the gravitational exudations observed (Fig. 2E), the most probable hypothesis is the radar-wave attenuation due to the presence of moisture. The amount of humidity can produce a shield effect that causes the attenuation of the radar-wave signal, depending on the moisture level, that results in the subsequent loss of information [31]. Thus, FDTD simulations were built to analyze the effect of moisture on the propagation of the GPR signal and its attenuation. As seen in Fig. 8, three different levels of moisture were simulated to evaluate the behavior of this attenuation. For a rigorous comparison, all the radargrams created were processed by applying the same filtering. Observing the synthetic data produced for a concrete slab 0.8 m thick (1), it has been demonstrated that a loss
1 For interpretation of color in Fig. 7, the reader is referred to the web version of this article.
555
of signal occurs even to relatively low levels of moisture (35%). Although the reflection generated from the bottom of the slab (at 9–13 ns) presented lower amplitude if compared with the nonmoist concrete (0%), it was perfectly distinguished even when simulating levels of 70% moisture content. On the contrary, in the case of a concrete slab 1.2 m thick (2), the GPR signal was severely attenuated when assuming moisture levels of 35% and 70%, and the reflection generated from the bottom of the slab (at 13–17 ns) was almost completely lost with levels of 35% and 70% moisture. The simulations confirmed how attenuation also depends on the thickness of the concrete slab, and they have demonstrated that, under the same moisture conditions, a loss of information occurred for concrete slabs 1.2 m thick while reflections were properly detected for concrete slabs 0.8 m thick. That could be the reason why the signal is lost from one of the lateral sectors of the tunnel, while is perfectly appreciated from the other one. Thus, the lost side would be the deeper one (Fig. 2C), where the propagation of the signal through the attenuating medium is longer. According to that, it would be probable that the whole construction presented a certain moisture level, but it was only appreciated in the deeper side. However, the lack of information observed in the GPR data might be also informing of corrosion in reinforcement and subsequent high signal attenuation. Fig. 9A supports the existence of reinforcing bars, through the inner structure over the galleries, in form of consecutive hyperbolas [32]. This hypothesis is supported by the proximity of the swimming pool maintenance room to the tunnel. The pipework lead high-chloride-content water through the tunnel walls, from the pool to the maintenance room and vice versa. It must be considered that chloride content is a more discriminating parameter than relative humidity on electromagnetic wave propagation in concrete [32]. Chloride ions expose the reinforcement to corrosion by removing, within the concrete, the protecting passivation layer on the steel surfaces. As demonstrated in Ref. [32], rebar corrosion produces partial or total loss of signal induced by high chloride content. Additional simulations were created to evaluate the influence of chloride content, in conjunction with moisture, on the propagation of the GPR signal (Fig. 10). In this case, models were built assuming a concrete slab of 0.8 m thick (Fig. 4(1)). This thickness was selected because the corresponding data obtained to analyze the effect of moisture did not show severe signal attenuation, even when considering 70% of moisture (Fig. 8(1)). The synthetic data provided for both simulated concrete conditions, either considering just the moisture effect (Fig. 8(1)) or the combined effect of moisture and chlorides (Fig. 10), were compared. This allowed confirming that chlorides have a more accentuated influence than moisture on the attenuation of the electromagnetic signal. The radar-wave response produced at the bottom of the concrete slab (at 10–12 ns) showed partial attenuation for a chloride content of 0.4% (Fig. 10(A and C)). However, this loss of signal increased with contents of 1%, showing a nearly total attenuation with contents of 70% and 1% of moisture and chlorides, respectively (Fig. 10(D)). The knowledge of the underground distribution and its dimensioning implies special interest from an engineering point of view. In Fig. 9A, the reflections produced at the top and bottom of the two galleries identified were also distinguished at 16 and 22 ns, respectively. The location of these reflections allowed for the estimation of their thicknesses, or distances between surface and top of the galleries (d1) and heights, or distances between top and bottom (d2), which resulted in 0.8 and 0.9 m, respectively (Table 1). Calculations were made from Eq. (1).
d¼v
twt 2
ð1Þ
556
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 7. Set of time-slices from 0 to 3.5 m in depth, presented with an interval approximately equal to 0.5 m, in which the internal structure of the tunnel is clearly identified.
where v is the velocity of propagation and twt is the travel-time difference (Dt) to and from the target. Fig. 9B presents a cross section extracted from the 3D cube, which coincides with the section B in Fig. 6. These images provided additional information to the common radargrams obtained when collecting data in the X axis (Fig. 3A), since they display an overall visualization of the complete length of lateral branches in the perpendicular direction. The same gallery observed in Fig. 9A from 6 to 7 m was identified in this cross section, and the reflections produced at the top and bottom of the branch were interpreted with an extension about 4 m long. The thickness (d10 ) and height (d20 ) of this branch were also estimated from Eq. (1), presenting identical values regarding those calculated from the x-line profile (Table 1). The intersection zone from 4 to 6 m at the y-line in Fig. 6 was also distinguished. However, the dimensioning from the 2D data is a difficult task that requires analyzing every single radargram and cross
section produced, which involves a considerable amount of time and dedication. To take advantage of all the available 3D techniques, the extraction of a 3D volume representing the entire tunneling space was possible by using the isosurface rendering technique (Fig. 11). This visualization technique displays surfaces of equal amplitude in the volume. A threshold value defines a maximum number of possible surfaces within the three-dimensional volume [33]. In this particular case, a threshold value of 70% provided the best results. Then, any surface with a value equal to or greater than 70% of the maximum amplitude value (threshold) was displayed. The application of this 3D imaging greatly facilitates the interpretation in the unique case of features with strong contrast with the surrounding reflections. In this work, it reproduces the shape of the hidden underground tunneling with all its galleries and lateral branches, showing an intuitive overview picture with an estimation of the dimensions of the target.
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
557
Fig. 8. Synthetic data produced to analyze the influence of moisture in concrete. Three levels of moisture content (%) were considered, and concrete slabs 0.8 (1) and 1.2 (2) m thick were simulated.
Fig. 9. (A) 500 MHz data provided along the section A (x-line) in Fig. 6, in which two lateral branches were identified and the presence of reinforcing bars over these galleries. (B) Additional cross section in the y-line generated from the 3D cube (section B marked in Fig. 6) interpreting the same gallery identified in (A) from 6 to 7 m.
Fig. 10. FDTD modeling created for a concrete slab 0.8 m thick to examine the combined effect of moisture and chloride contents on the attenuation of the GPR signal.
558
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Table 1 Dimensions (thicknesses and heights) determined from Eq. (1) for the galleries interpreted in Fig. 8. The velocities assumed for each medium considered in construction are also shown. Dimension
Material
Velocity
twt
Depth (d)
Thickness (d1–d10 ) Height (d2–d20 )
Concrete Free-space
0.1 m/ns 0.3 m/ns
16 ns 6 ns
0.8 m 0.9 m
The volumetric reconstruction (Fig. 11) also provided additional information and it has confirmed the existence of the two levels, or different depths, in the principal corridor of the tunnel. It was appreciated from the gaps of data at the location of the existent ramp that separates these two levels. As other authors have referred [34], in the case of underground structures not parallel to the surface, this loss of information is most commonly due to the scattering phenomenon, in which waves are reflected away
Fig. 11. Isosurface render generated by GPR-Slice software, showing the overall volume of the underground tunneling space.
Fig. 12. Different views of the tunnel in CAD software: (A) top view (length and width in meters), (B) front view (width in meters and height in ns), and (C) left side view (width in meters and height in ns). Additionally, the different structural elements detected are named in the top view.
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560 Table 2 The main dimensions determined for each structural element detected in Fig. 12A, which includes the main corridor of the tunnel and its different lateral branches.
559
obtained, which allowed for its geometrical dimensioning. This information becomes interesting as it provides an intuitive and easily understanding layout of the underground space distribution.
Tunnel dimensions (m) Gallery
Length
Width
Height
G0 G1 G1.1 G1.2 G2 G2.1 G2.2 G3 G3.1 G3.2 G4 G5
30.5 16.6 4.7 7.4 15.8 4.6 6.8 14.8 8.2 5.3 9.9 5.5
2.6 1.1 0.7 0.4 1.0 0.7 0.7 1.0 0.7 0.7 0.4 1.0
3.3 1.2 0.9 0.8 0.9 0.9 0.8 1.2 0.8 0.9 0.9 1.2
Acknowledgments The Military Naval Academy of Spain is acknowledged for the given facilities during this work performance. The authors also want to thank the Applied Geotechnologies Research Group of the University of Vigo for all the human and material support. Additionally, this study is a contribution to the EU funded COST Action TU-1208 ‘‘Civil Engineering Applications of Ground Penetrating Radar’’.
References from the range of the receiving antennas and not collected by the GPR. The volume created was exported to a 3D CAD format. This facilitates the geometrical dimensioning of the tunnel. Fig. 12 illustrates the several target views, top (A), front (B) and side (C), obtained in CAD software from a cloud point representation. Moreover, different sections in either vertical or longitudinal directions could be provided for a complete dimensioning. The main dimensions were obtained (heights, widths and lengths), and described in Table 2. For height calculations, the conversion of time (ns) into depths (m) was made assuming the speed of the radar-wave in free-space (Table 1).
4. Conclusions The results obtained have confirmed that GPR is an effective NDT method for tunneling and underground space investigation. GPR data allowed for the identification of some of the galleries composing the structure, as well for the detection of reinforcement elements over these galleries. Otherwise, moisture problems were appreciated from GPR data and corroborated during a later visual inspection. FDTD simulations were built for a better understanding of the radar-wave propagation phenomenon under different levels of moisture. Based on the synthetic data provided, it was concluded that the signal loss could be caused by those severe moisture conditions. In addition, the lack of information would be increased because of the higher depth of the sector involved, as it implies longer propagation time of the signal through the attenuating medium. Furthermore, the attenuation could be also due to the corrosion of the pipework through the tunnel walls. It would be explained by a problem in the sealing glands of the pipe circuit, that lead high-chloride-content water from the pool, and the subsequent corrosion in reinforcement. This information might be useful for civil engineers to verify the preservation state of the tunnel and design future conservation plans. This way, maintenance tasks could be focused on the affected areas, avoiding the use of more destructive techniques, as excavation or underground tastings. The 3D GPR strategies proposed have shown its potential as a supporting tool for data interpretation. These techniques provided great detailed information from different depths and facilitated the volumetric reconstruction of the underground tunnel and its galleries framework. Moreover, using 3D methodologies resulted in a significant improvement in terms of processing time, compared with 2D ones. Finally, different views of the construction were
[1] López C. Manual de Túneles y Obras Subterráneas. Madrid (in Spanish): Entorno Gráfico; 1996. [2] Katzenbach R, Leppla S, Vogler M, Seip M, Kurze S. Soil–structure-interaction of tunnels and superstructures during construction and service time. In: 11th International conference on modern building materials, structures and techniques (MBMST 2013), Procedia Engineering, vol. 57, 2013. p. 35–44. [3] Mangol T, Penycate J. The tunnels of Cu chi: a remarkable history of war. New York: Berkley Books; 1985. [4] Juncá Ubierna JA. Tunnel heritage in Spain: roots of the underground. Tunnelling Underground Space Technol 1998;13(2):131–41. [5] Butler R, Suntikul W. Tourism and war. Abingdon, Oxon (United Kingdom): Routledge; 2013. [6] Asakura T, Kojima Y. Tunnel maintenance in Japan. Tunnelling Underground Space Technol 2003;18(2/3):161–9. [7] Yuan Y, Xiaomo J, Xian L. Predictive maintenance of shield tunnels. Tunnelling Underground Space Technol 2013;2013(38):69–86. [8] Estes KW, Heinl RD. Handbook for marine NCOs, 1995. [9] Capineri L, Ivashov S, Bechtel T, Zhuravlev A, Falorni P, Windsor C, Borgioli G, Vasiliev I, Sheyko A. Comparison of GPR sensor types for landmine detection and classification. In: 12th International conference on ground penetrating radar, Birmingham (UK), June 16–19, 2008. [10] Gonzalez-Huici MA. Strategy for landmine detection and recognition using simulated GPR responses. In: Proceedings of the 14th international conference on ground penetrating radar, Shanghai (China), June 4–8, 2012. p. 877–82. [11] Modroo JJ, Olhoeft GR. Avalanche rescue using ground penetrating radar. In: Tenth international conference on ground penetrating radar, Delft (The Netherlands), 21–24 June, 2004. [12] Szynkiewicz A, Grabowski P. GPR monitoring of pavements on airfield. In: Proc., 10th Int. Conf. on Ground Penetrating Radar, Delft (The Netherlands), June 21–24, 2004, p. 803–6. [13] Hugenschmidt J. Concrete bridge inspection with a mobile GPR system. Constr Build Mater 2002;16(3):147–54. [14] Varela-Ortiz W, Lugo Cintrón CY, Velázquez GI, Stanton TR. Load testing and GPR assessment for concrete bridges on military installations. Constr Build Mater 2013;38:1255–69. [15] Lorenzo H, Hernández MC, Cuéllar V. Selected radar images of man-made underground galleries. Archaeol Prospection 2002;9(1):1–7. [16] Chianese D, D’Emilio M, Bavusi M, Lapenna V, Macchiato M. Magnetic and ground probing radar measurements for soil pollution mapping in the industrial area of Val Basento (Basilicata Region, Southern Italy): a case study. Environ Geol 2006;49:389–404. [17] Sambuelli L, Calzoni C, Pesenti M. Waterborne GPR survey for estimating bottom-sediment variability: a survey on the Po River, Turin, Italy. Geophysics 2009;74(4):B95–B102. [18] Richards JA. Inspection, maintenance and repair of tunnels: international lessons and practice. Tunnelling Underground Space Technol 1998;13(4):369–75. [19] Maierhofer C. Nondestructive evaluation of concrete infrastructure with ground penetrating radar. J Mater Civil Eng, ASCE 2003;15(3):287–97. [20] Daniels DJ. Ground penetrating radar. London: Institution of Electrical Engineering; 2004. [21] Goodman D, Piro S. GPR remote sensing in archaeology. Geotechnologies and environment, vol. 9. Springer; 2013. [22] Neubauer W, Hinterleitner A, Doneus M, Löcker K, Trinks I, Nau E, et al. Integrated geophysical archaeological prospection resulting in the discovery of the school of gladiators in the roman town of Carnuntum in Austria. SEG Annual Meeting. Society of Exploration Geophysicists; 2012. [23] Novo A, Solla M, Montero Fenollós JL, Lorenzo H. Searching for the remains of an Early Bronze Age city at Tell Qubr Abu al-‘Atiq (Syria) through archaeological investigations and GPR imaging. J Cultural Heritage. [In press].
560
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
[24] Solla M, González-Jorge H, Varela M, Lorenzo H. Ground-penetrating radar for inspection of in-road structures and data interpretation by numerical modeling. J Constr Eng Manage 2013;139(6):749–53. [25] Annan P. GPR principles, procedures and applications. Sensors and Software, Inc.; 2003. [26] Goodman D. GPR-SLICE v7.0 Manual, from, January/2014. [27] Goodman D, Nishimura Y, Rogers JD. GPR time slices in archaeological prospection. Archaeol Prospection 1995;2:85–9. [28] Giannopoulos A. GprMax2D/3D v2.0 Manual, from March/2014. [29] Kalogeropoulos A. Non-destructive determination of chloride and water content in concrete using ground penetrating radar. Ph.D. thesis, Swiss Federal Institute of Technology in Lausanne, Suisse, France, 2012.
[30] Sandmeier KJ. ReflexW v7.5 Manual, from March/2014. [31] Laurens S, Balayssac JP, Rhazi J, Klysz G, Arliguie G. Non-destructive evaluation of concrete moisture by GPR: experimental study and direct modeling. Mater Struct 2005;38:827–32. [32] Saarenketo T, Maijala P, Leppälä A. Recommendations for guidelines for the use of GPR in bridge deck surveys. Finland: Roadscanners Oy; 2011. [33] Leckebusch J. Ground-penetrating radar: a modern three-dimensional prospection method. Archaeol Prospection 2003;10:213–40. [34] Conyers LB. In: Introduction to GPR. Altamira Press; 2004.
Contents lists available at ScienceDirect
Construction and Building Materials journal homepage: www.elsevier.com/locate/conbuildmat
Three-dimensional ground-penetrating radar methodologies for the characterization and volumetric reconstruction of underground tunneling X. Núñez-Nieto a,d,⇑, M. Solla a,d, A. Novo b, H. Lorenzo c,d a
Defense University Center, Spanish Naval Academy, Plaza de España 2, 36920 Marín, Pontevedra, Spain GeoRadar Division, GeoSystems Business Unit Manager, IDS North America Ltd., 418 Sherbrooke Street East, Montreal, Suite 200, QC H2L 1J6, Canada Department of Natural Resources and Environmental Engineering, School of Forestry Engineering, University of Vigo, A Xunqueira s/n., 36005 Pontevedra, Spain d Applied Geotechnologies Research Group, School of Mining Engineering, University of Vigo, Campus Lagoas-Marcosende, 36310 Vigo, Spain b c
h i g h l i g h t s Documentation and characterization of underground tunneling space by GPR. Volumetric reconstruction that allows determining numerical approximation of volumes. Geometrical dimensioning of the structure from the volumetric reconstruction. Detection of moist areas from the attenuation of the GPR signal. FDTD modeling to assist in the interpretation of the field GPR data.
a r t i c l e
i n f o
Article history: Received 9 May 2014 Received in revised form 25 August 2014 Accepted 27 August 2014
Keywords: GPR Underground space FDTD modeling 3D imaging Moisture Civil engineering
a b s t r a c t This work presents the documentation and characterization of an ancient underground concrete tunnel using the ground-penetrating radar (GPR) method. Three-dimensional imaging methodologies were applied to create an accurate volumetric reconstruction of the underground tunneling space and the whole framework of galleries composing the main structure, which enabled for the dimensioning of the structure. Problems of moisture were also detected in a particular sector of the tunnel. In addition, FDTD modeling was used to improve the understanding of the GPR signal propagation and to assist the interpretation of the field GPR data. Both field and synthetic data have shown the capabilities of the method for the evaluation and characterization of this ancient construction. Ó 2014 Elsevier Ltd. All rights reserved.
1. Introduction Tunneling detection becomes relevant based on a civil engineering point of view, due to the momentous importance of knowing the underground space distribution. In addition to the interest with regard to their influence in the settlement and disposition of the subsoil structure [1,2], the documentation of these constructions also holds connotations of historical and cultural character, architectural interest, tourist attraction or military and defense implications [3–5]. Therefore, their exact location ⇑ Corresponding author at: Defense University Center, Spanish Naval Academy, Plaza de España 2, 36920 Marín, Pontevedra, Spain. E-mail address: [email protected] (X. Núñez-Nieto). http://dx.doi.org/10.1016/j.conbuildmat.2014.08.083 0950-0618/Ó 2014 Elsevier Ltd. All rights reserved.
and geometry provides additional information to plan and perform maintenance and repair tasks, which enable for the preservation of their structures and content in optimal conditions [6,7]. This work presents the evaluation and characterization of an existing tunnel located within the premises of the Spanish Naval Academy of Marín, Galicia, in the northwest of Spain (Fig. 1). The galleries of the tunnel develop different functions, namely: ventilation, general storage and military training site for the students belonging to the Spanish Navy Marines Corps, which is the oldest existing Naval Infantry Force in the world [8]. The ground-penetrating radar (GPR) is a geophysical commonly used technique, with a wide range of applications. There have been published numerous studies on different fields, such as military
552
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 1. Location of the prospected area in Europe and specifically in Marín (Galicia, northwest of Spain).
(mine detection [9,10], human remains and life detection [11]), civil engineering (pavement control and integrity of landing runways [12], pathologies in construction [13,14], location of buried structures [15]) and also in geology (detection of subsoil pollution [16] and freshwater bathymetries [17]). Therefore, GPR is proposed in this paper as a solution for the documentation of an underground tunnel because of its rapidity and suitability as non-destructive technique. So that, it is the ideal solution in comparison to other more invasive methods, such as excavation or underground tasting, that may deteriorate the structure of the tunnel and/or its content [18,19]. The GPR method has proven its suitability for providing high image quality results and is a well-recognized prospecting technique by the scientific community [20]. Nevertheless, there still remains some skepticism among the non-geophysical audience, promoted by the difficulty for the understanding of the GPR images. The interpretation of the radargrams (2D GPR images) is not trivial and the use of 3D processing and visualization techniques, has achieved the agreement by the non-expert community in safety underground investigations [21]. Using 3D processing and visualization produces more realistic images of the underground spaces, which allows not only for the location, but also for the 3D reconstruction of underground and buried constructions [22,23]. Therefore, an advanced 3D visualization methodology is presented in this work to provide better interpretation of the results and the volumetric reconstruction and dimensioning of the tunnel. These 3D images provide an accurate and intuitive display of the underground distribution, easily understood even by non-experts in the subject. Numerical simulation was additionally used in this work to assist in the interpretation of the field GPR data. Specifically, Finite-Difference Time-Domain (FDTD) modeling was considered, since the method has demonstrated its capabilities to improve the understanding of the electromagnetic waves propagation through the media [24].
2. Materials and methods 2.1. Tunnel description It consists of an underground tunnel with symmetrical geometry composed by a central corridor and lateral branches (Fig. 2A) that presents a final trifurcation of galleries (Fig. 2B). The main corridor presents two different levels of the tunnel, which are separated by a pronounced ascendant ramp at approximately the middle of its length (Fig. 2C). Some lateral branches present also an initial sector with ascendant slope, as illustrated in Fig. 2D. The underground construction is a concrete solid composed by several blocks seated one above the other, whose top ends in a concrete slab resulting in a sports court. Moisture and condensation problems were observed in certain areas of the ancient concrete construction (Fig. 2E), possibly caused by the poor maintenance of some water pipes existing through one of the lateral sections of the subsoil of the structure. These water pipes are supposed to lead water with chloride content from the swimming pool, near to the tunnel, to the maintenance room and vice versa. Regarding these moisture evidences, it must be also taken into account the close presence of the sea, upon which the newest military installations were built. Moreover, the geographical location of the study site could have also affected the integrity of this ancient construction. That is supported by the location of the village of Marín (Fig. 1) in the Autonomous Community of Galicia, which is characterized by adverse weather conditions with high levels of humidity. 2.2. Field data acquisition Three-dimensional (3D) GPR methodologies were performed in this work (Fig. 3A). Equidistant parallel 2D-lines in only x-direction were acquired on a regular rectangular grid (40 20 m size) with a space between profile lines of 20 cm. 0 The GPR data were collected using a RAMAC/GPR system from MALÅ A Geoscience and a central frequency of 500 MHz (Fig. 3B). The GPR survey was carried out using the common-offset mode with the antenna polarization perpendicular to the direction of data collection [25]. Data was acquired with 2 cm trace-interval or in-line spacing of traces, with reflections collected within a total time window of 75 ns and defined by 512 samples per trace. For trace-interval distance calculation, the GPR antenna was mounted on a survey cart with encoder (odometer wheel). 2.3. 3D data processing GPR-Slice software [26] was used for the three-dimensional (3D) data processing. The first step, previous to the elaboration of the 3D cube, was to filter the raw
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
553
Fig. 2. Detailed images from the interior of the tunnel: (A) main corridor and lateral branches, (B) final trifurcation of galleries at the end of the main corridor, (C) ramp of the principal corridor, (D) lateral branch presenting an ascendant slope section, and (E) evidence of condensation inside one of the lateral branches.
Fig. 3. (A) Prospected area and (B) GPR equipment and components: control unit, 500 MHz antenna (transmitter and receiver), laptop and odometer wheel.
GPR data files by applying the following filters: wobble removal, manual gain, band pass filter (low cutoff: 68 MHz; upper cutoff: 822 MHz), and background removal. The main objective was to amplify the received signal, as well as to remove both low and high frequency noise in the vertical and horizontal directions. Next, the 3D horizontal slices were generated by spatially averaging the squared wave amplitudes of the recorded radar reflections over the time window. The squared amplitude interpolation process creates interpolated time-slices, which are normalized to 8 bit following the color changes between different levels and not actual reflection values [27]. The number of slices depends on the length of the time window selected, the slice thickness and the overlay between slices. Thickness of horizontal slices is often set to at least one or two wavelengths of the radar pulse that is sent into the ground [27]. In any event, the slices generated should not be thicker than the size of the smallest expected target. On the other hand, gaps between the time-slice windows must be avoided because they create discontinuities which can miss targets. In this particular case, instead of generating a large set of depth slices, they are grouped together and its data combined, every 44 cm, to take into account the different depths at which the reflections appear and that could not appear with individual slices. The process involved a set of 16 horizontal slices by 60 samples thick (8.8 ns) over a time window of 482 samples (70.86 ns). Data were gridded using the inverse distance algorithm, which includes a search of all data within a fixed radius of 35 cm of the desired point to be interpolated on the grid and a smoothing factor of two. Grid cell size was set to 5 cm to produce high resolution images. For animations of pseudo-three-dimensional datasets, three interpolations were applied to
the original set of 16 slices in order to generate a smoother visualization. For time to depth conversion, a velocity of 0.1 m/ns for concrete medium was determined from hyperbolae shape analyses. 2.4. FDTD modeling Simulations were performed in this work to evaluate the hypothesis of probable moisture content in concrete, which has been achieved from the interpretation of the field GPR data. It allows for a better understanding of the observed electromagnetic signal propagation phenomena and for the characterization of the radar-wave responses under different conditions of humidity. The synthetic models (Fig. 4) were built using the GprMax software [28], which is an electromagnetic wave simulator for GPR that uses the FDTD method. The models were designed considering concrete slabs of 0.8 and 1.2 m thick ((1) and (2), respectively), since the lateral branches present different depths (Fig. 2C). The electromagnetic properties, assumed from [29] to simulate three different levels of moisture (0%, 35% and 70%), are shown in Fig. 4. Additional simulations were performed to analyze the effect of both chlorides and moisture on the GPR signal. As described in Ref. [29], two different levels of chloride contents were simulated (0.4% and 1% of cement mass) for each moisture content (35% and 70%). Modeling was created with a small spatial-step equal to 8 mm, and the excitation pulse was a Gaussian of 500 MHz center frequency. The trace-interval was 3 cm. Once generated, the synthetic data was transferred from GprMax to ReflexW software [30] using a MATLAB routine, and then filtered using a similar processing sequence to that used for the raw GPR field data.
554
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 4. Synthetic models elaborated to analyze the influence of different levels of moisture on the GPR signal: 0%, 35% and 70%. Two different levels of chloride content, 0.4% and 1% of cement mass, were also evaluated. Concrete slabs of 0.8 (1) and 1.2 (2) m thick were considered, and the values assumed for conductivity (r) and dielectric constant (K) are also shown.
Fig. 5. The sliced 3D cube created.
3. Results and discussions Different 3D visualization techniques were applied to extract information that helps in the reconstruction and characterization of the evaluated tunnel. After creating the 3D cube, for which all the processed radargrams were assembled, the resulted cube was sliced in different directions or planes (Fig. 5). The extraction of
these time-slices, or horizontal maps, provided the location, size and shape of the structural elements. However, one single time-slice did not represent the entire tunneling structure since the different galleries and branches are placed at different levels. To overcome this adversity, overlay analysis was applied (Fig. 6). Thus, the slice obtained from overlay analysis (at 70.86 ns) showed the internal structure of the tunnel,
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 6. Slice generated from overlay analysis considering the whole animation set of time-slices.
and it was possible to identify the central or main corridor, the lateral tunnel branches and the final trifurcation at the end of the main corridor, as seen in Fig. 2. The single time-slice in Fig. 6, obtained from all the strongest reflectors at different depths, can suffer a loss of subtle features when the amplitude of the reflections is not enough. The set of animations of time-slices images generated from the 3D cube can overcome this difficulty, and they provide also the extension and exact positioning in depth for each branch composing the entire structure of the tunnel. Although time-slices were obtained every 0.25 m, relevant variations were not found between two consecutive slices and, therefore, time-slices every 0.5 m were considered in Fig. 7 to simplify imaging. The points where the signal has smaller amplitude are represented in blue1 color, while red colored areas indicate areas of greater contrast in dielectric properties. Thus, the set of time-slices revealed additional information regarding targets placed at different depths such as the two levels that compose the principal corridor, displaying a first sector deeper (from 0 to 10 m in X axis), as seen in Fig. 2C. Although the tunnel is symmetric with respect to its central corridor (Fig. 2A), the lateral branches at the left side of the central axis were not easily distinguished from the radar data, and only the one corresponding to the final trifurcation was partially detected. The most probable cause for this lack of information is the attenuation of the GPR signal in this sector of the tunnel. There are different materials that produce severe GPR signal attenuation such as clay or soil cement (interfacial polarization losses), as well as moisture (water losses) [25,31]. Nevertheless, the original building materials were the same for the entire structure and there are no evidences of later reconstructions that would include possible insulating materials that may cause a loss of the electromagnetic signal. Therefore, based on a later visual inspection made and the gravitational exudations observed (Fig. 2E), the most probable hypothesis is the radar-wave attenuation due to the presence of moisture. The amount of humidity can produce a shield effect that causes the attenuation of the radar-wave signal, depending on the moisture level, that results in the subsequent loss of information [31]. Thus, FDTD simulations were built to analyze the effect of moisture on the propagation of the GPR signal and its attenuation. As seen in Fig. 8, three different levels of moisture were simulated to evaluate the behavior of this attenuation. For a rigorous comparison, all the radargrams created were processed by applying the same filtering. Observing the synthetic data produced for a concrete slab 0.8 m thick (1), it has been demonstrated that a loss
1 For interpretation of color in Fig. 7, the reader is referred to the web version of this article.
555
of signal occurs even to relatively low levels of moisture (35%). Although the reflection generated from the bottom of the slab (at 9–13 ns) presented lower amplitude if compared with the nonmoist concrete (0%), it was perfectly distinguished even when simulating levels of 70% moisture content. On the contrary, in the case of a concrete slab 1.2 m thick (2), the GPR signal was severely attenuated when assuming moisture levels of 35% and 70%, and the reflection generated from the bottom of the slab (at 13–17 ns) was almost completely lost with levels of 35% and 70% moisture. The simulations confirmed how attenuation also depends on the thickness of the concrete slab, and they have demonstrated that, under the same moisture conditions, a loss of information occurred for concrete slabs 1.2 m thick while reflections were properly detected for concrete slabs 0.8 m thick. That could be the reason why the signal is lost from one of the lateral sectors of the tunnel, while is perfectly appreciated from the other one. Thus, the lost side would be the deeper one (Fig. 2C), where the propagation of the signal through the attenuating medium is longer. According to that, it would be probable that the whole construction presented a certain moisture level, but it was only appreciated in the deeper side. However, the lack of information observed in the GPR data might be also informing of corrosion in reinforcement and subsequent high signal attenuation. Fig. 9A supports the existence of reinforcing bars, through the inner structure over the galleries, in form of consecutive hyperbolas [32]. This hypothesis is supported by the proximity of the swimming pool maintenance room to the tunnel. The pipework lead high-chloride-content water through the tunnel walls, from the pool to the maintenance room and vice versa. It must be considered that chloride content is a more discriminating parameter than relative humidity on electromagnetic wave propagation in concrete [32]. Chloride ions expose the reinforcement to corrosion by removing, within the concrete, the protecting passivation layer on the steel surfaces. As demonstrated in Ref. [32], rebar corrosion produces partial or total loss of signal induced by high chloride content. Additional simulations were created to evaluate the influence of chloride content, in conjunction with moisture, on the propagation of the GPR signal (Fig. 10). In this case, models were built assuming a concrete slab of 0.8 m thick (Fig. 4(1)). This thickness was selected because the corresponding data obtained to analyze the effect of moisture did not show severe signal attenuation, even when considering 70% of moisture (Fig. 8(1)). The synthetic data provided for both simulated concrete conditions, either considering just the moisture effect (Fig. 8(1)) or the combined effect of moisture and chlorides (Fig. 10), were compared. This allowed confirming that chlorides have a more accentuated influence than moisture on the attenuation of the electromagnetic signal. The radar-wave response produced at the bottom of the concrete slab (at 10–12 ns) showed partial attenuation for a chloride content of 0.4% (Fig. 10(A and C)). However, this loss of signal increased with contents of 1%, showing a nearly total attenuation with contents of 70% and 1% of moisture and chlorides, respectively (Fig. 10(D)). The knowledge of the underground distribution and its dimensioning implies special interest from an engineering point of view. In Fig. 9A, the reflections produced at the top and bottom of the two galleries identified were also distinguished at 16 and 22 ns, respectively. The location of these reflections allowed for the estimation of their thicknesses, or distances between surface and top of the galleries (d1) and heights, or distances between top and bottom (d2), which resulted in 0.8 and 0.9 m, respectively (Table 1). Calculations were made from Eq. (1).
d¼v
twt 2
ð1Þ
556
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Fig. 7. Set of time-slices from 0 to 3.5 m in depth, presented with an interval approximately equal to 0.5 m, in which the internal structure of the tunnel is clearly identified.
where v is the velocity of propagation and twt is the travel-time difference (Dt) to and from the target. Fig. 9B presents a cross section extracted from the 3D cube, which coincides with the section B in Fig. 6. These images provided additional information to the common radargrams obtained when collecting data in the X axis (Fig. 3A), since they display an overall visualization of the complete length of lateral branches in the perpendicular direction. The same gallery observed in Fig. 9A from 6 to 7 m was identified in this cross section, and the reflections produced at the top and bottom of the branch were interpreted with an extension about 4 m long. The thickness (d10 ) and height (d20 ) of this branch were also estimated from Eq. (1), presenting identical values regarding those calculated from the x-line profile (Table 1). The intersection zone from 4 to 6 m at the y-line in Fig. 6 was also distinguished. However, the dimensioning from the 2D data is a difficult task that requires analyzing every single radargram and cross
section produced, which involves a considerable amount of time and dedication. To take advantage of all the available 3D techniques, the extraction of a 3D volume representing the entire tunneling space was possible by using the isosurface rendering technique (Fig. 11). This visualization technique displays surfaces of equal amplitude in the volume. A threshold value defines a maximum number of possible surfaces within the three-dimensional volume [33]. In this particular case, a threshold value of 70% provided the best results. Then, any surface with a value equal to or greater than 70% of the maximum amplitude value (threshold) was displayed. The application of this 3D imaging greatly facilitates the interpretation in the unique case of features with strong contrast with the surrounding reflections. In this work, it reproduces the shape of the hidden underground tunneling with all its galleries and lateral branches, showing an intuitive overview picture with an estimation of the dimensions of the target.
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
557
Fig. 8. Synthetic data produced to analyze the influence of moisture in concrete. Three levels of moisture content (%) were considered, and concrete slabs 0.8 (1) and 1.2 (2) m thick were simulated.
Fig. 9. (A) 500 MHz data provided along the section A (x-line) in Fig. 6, in which two lateral branches were identified and the presence of reinforcing bars over these galleries. (B) Additional cross section in the y-line generated from the 3D cube (section B marked in Fig. 6) interpreting the same gallery identified in (A) from 6 to 7 m.
Fig. 10. FDTD modeling created for a concrete slab 0.8 m thick to examine the combined effect of moisture and chloride contents on the attenuation of the GPR signal.
558
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
Table 1 Dimensions (thicknesses and heights) determined from Eq. (1) for the galleries interpreted in Fig. 8. The velocities assumed for each medium considered in construction are also shown. Dimension
Material
Velocity
twt
Depth (d)
Thickness (d1–d10 ) Height (d2–d20 )
Concrete Free-space
0.1 m/ns 0.3 m/ns
16 ns 6 ns
0.8 m 0.9 m
The volumetric reconstruction (Fig. 11) also provided additional information and it has confirmed the existence of the two levels, or different depths, in the principal corridor of the tunnel. It was appreciated from the gaps of data at the location of the existent ramp that separates these two levels. As other authors have referred [34], in the case of underground structures not parallel to the surface, this loss of information is most commonly due to the scattering phenomenon, in which waves are reflected away
Fig. 11. Isosurface render generated by GPR-Slice software, showing the overall volume of the underground tunneling space.
Fig. 12. Different views of the tunnel in CAD software: (A) top view (length and width in meters), (B) front view (width in meters and height in ns), and (C) left side view (width in meters and height in ns). Additionally, the different structural elements detected are named in the top view.
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560 Table 2 The main dimensions determined for each structural element detected in Fig. 12A, which includes the main corridor of the tunnel and its different lateral branches.
559
obtained, which allowed for its geometrical dimensioning. This information becomes interesting as it provides an intuitive and easily understanding layout of the underground space distribution.
Tunnel dimensions (m) Gallery
Length
Width
Height
G0 G1 G1.1 G1.2 G2 G2.1 G2.2 G3 G3.1 G3.2 G4 G5
30.5 16.6 4.7 7.4 15.8 4.6 6.8 14.8 8.2 5.3 9.9 5.5
2.6 1.1 0.7 0.4 1.0 0.7 0.7 1.0 0.7 0.7 0.4 1.0
3.3 1.2 0.9 0.8 0.9 0.9 0.8 1.2 0.8 0.9 0.9 1.2
Acknowledgments The Military Naval Academy of Spain is acknowledged for the given facilities during this work performance. The authors also want to thank the Applied Geotechnologies Research Group of the University of Vigo for all the human and material support. Additionally, this study is a contribution to the EU funded COST Action TU-1208 ‘‘Civil Engineering Applications of Ground Penetrating Radar’’.
References from the range of the receiving antennas and not collected by the GPR. The volume created was exported to a 3D CAD format. This facilitates the geometrical dimensioning of the tunnel. Fig. 12 illustrates the several target views, top (A), front (B) and side (C), obtained in CAD software from a cloud point representation. Moreover, different sections in either vertical or longitudinal directions could be provided for a complete dimensioning. The main dimensions were obtained (heights, widths and lengths), and described in Table 2. For height calculations, the conversion of time (ns) into depths (m) was made assuming the speed of the radar-wave in free-space (Table 1).
4. Conclusions The results obtained have confirmed that GPR is an effective NDT method for tunneling and underground space investigation. GPR data allowed for the identification of some of the galleries composing the structure, as well for the detection of reinforcement elements over these galleries. Otherwise, moisture problems were appreciated from GPR data and corroborated during a later visual inspection. FDTD simulations were built for a better understanding of the radar-wave propagation phenomenon under different levels of moisture. Based on the synthetic data provided, it was concluded that the signal loss could be caused by those severe moisture conditions. In addition, the lack of information would be increased because of the higher depth of the sector involved, as it implies longer propagation time of the signal through the attenuating medium. Furthermore, the attenuation could be also due to the corrosion of the pipework through the tunnel walls. It would be explained by a problem in the sealing glands of the pipe circuit, that lead high-chloride-content water from the pool, and the subsequent corrosion in reinforcement. This information might be useful for civil engineers to verify the preservation state of the tunnel and design future conservation plans. This way, maintenance tasks could be focused on the affected areas, avoiding the use of more destructive techniques, as excavation or underground tastings. The 3D GPR strategies proposed have shown its potential as a supporting tool for data interpretation. These techniques provided great detailed information from different depths and facilitated the volumetric reconstruction of the underground tunnel and its galleries framework. Moreover, using 3D methodologies resulted in a significant improvement in terms of processing time, compared with 2D ones. Finally, different views of the construction were
[1] López C. Manual de Túneles y Obras Subterráneas. Madrid (in Spanish): Entorno Gráfico; 1996. [2] Katzenbach R, Leppla S, Vogler M, Seip M, Kurze S. Soil–structure-interaction of tunnels and superstructures during construction and service time. In: 11th International conference on modern building materials, structures and techniques (MBMST 2013), Procedia Engineering, vol. 57, 2013. p. 35–44. [3] Mangol T, Penycate J. The tunnels of Cu chi: a remarkable history of war. New York: Berkley Books; 1985. [4] Juncá Ubierna JA. Tunnel heritage in Spain: roots of the underground. Tunnelling Underground Space Technol 1998;13(2):131–41. [5] Butler R, Suntikul W. Tourism and war. Abingdon, Oxon (United Kingdom): Routledge; 2013. [6] Asakura T, Kojima Y. Tunnel maintenance in Japan. Tunnelling Underground Space Technol 2003;18(2/3):161–9. [7] Yuan Y, Xiaomo J, Xian L. Predictive maintenance of shield tunnels. Tunnelling Underground Space Technol 2013;2013(38):69–86. [8] Estes KW, Heinl RD. Handbook for marine NCOs, 1995. [9] Capineri L, Ivashov S, Bechtel T, Zhuravlev A, Falorni P, Windsor C, Borgioli G, Vasiliev I, Sheyko A. Comparison of GPR sensor types for landmine detection and classification. In: 12th International conference on ground penetrating radar, Birmingham (UK), June 16–19, 2008. [10] Gonzalez-Huici MA. Strategy for landmine detection and recognition using simulated GPR responses. In: Proceedings of the 14th international conference on ground penetrating radar, Shanghai (China), June 4–8, 2012. p. 877–82. [11] Modroo JJ, Olhoeft GR. Avalanche rescue using ground penetrating radar. In: Tenth international conference on ground penetrating radar, Delft (The Netherlands), 21–24 June, 2004. [12] Szynkiewicz A, Grabowski P. GPR monitoring of pavements on airfield. In: Proc., 10th Int. Conf. on Ground Penetrating Radar, Delft (The Netherlands), June 21–24, 2004, p. 803–6. [13] Hugenschmidt J. Concrete bridge inspection with a mobile GPR system. Constr Build Mater 2002;16(3):147–54. [14] Varela-Ortiz W, Lugo Cintrón CY, Velázquez GI, Stanton TR. Load testing and GPR assessment for concrete bridges on military installations. Constr Build Mater 2013;38:1255–69. [15] Lorenzo H, Hernández MC, Cuéllar V. Selected radar images of man-made underground galleries. Archaeol Prospection 2002;9(1):1–7. [16] Chianese D, D’Emilio M, Bavusi M, Lapenna V, Macchiato M. Magnetic and ground probing radar measurements for soil pollution mapping in the industrial area of Val Basento (Basilicata Region, Southern Italy): a case study. Environ Geol 2006;49:389–404. [17] Sambuelli L, Calzoni C, Pesenti M. Waterborne GPR survey for estimating bottom-sediment variability: a survey on the Po River, Turin, Italy. Geophysics 2009;74(4):B95–B102. [18] Richards JA. Inspection, maintenance and repair of tunnels: international lessons and practice. Tunnelling Underground Space Technol 1998;13(4):369–75. [19] Maierhofer C. Nondestructive evaluation of concrete infrastructure with ground penetrating radar. J Mater Civil Eng, ASCE 2003;15(3):287–97. [20] Daniels DJ. Ground penetrating radar. London: Institution of Electrical Engineering; 2004. [21] Goodman D, Piro S. GPR remote sensing in archaeology. Geotechnologies and environment, vol. 9. Springer; 2013. [22] Neubauer W, Hinterleitner A, Doneus M, Löcker K, Trinks I, Nau E, et al. Integrated geophysical archaeological prospection resulting in the discovery of the school of gladiators in the roman town of Carnuntum in Austria. SEG Annual Meeting. Society of Exploration Geophysicists; 2012. [23] Novo A, Solla M, Montero Fenollós JL, Lorenzo H. Searching for the remains of an Early Bronze Age city at Tell Qubr Abu al-‘Atiq (Syria) through archaeological investigations and GPR imaging. J Cultural Heritage. [In press]
560
X. Núñez-Nieto et al. / Construction and Building Materials 71 (2014) 551–560
[24] Solla M, González-Jorge H, Varela M, Lorenzo H. Ground-penetrating radar for inspection of in-road structures and data interpretation by numerical modeling. J Constr Eng Manage 2013;139(6):749–53. [25] Annan P. GPR principles, procedures and applications. Sensors and Software, Inc.; 2003. [26] Goodman D. GPR-SLICE v7.0 Manual, from
[30] Sandmeier KJ. ReflexW v7.5 Manual, from