|Earth and Rock Works | Earth Retaining Structures | Geotechnical Hazards | Ground Improvement | Structural Foundations | Subsurface Investigation|
|FHWA > Engineering > Geotechnical > Geotechnical Hazards > Mines > Mine Workshops|
Geophysical Technologies for Detecting Underground Coal Mine Voids:
Figure 1: Plan view of the surveyed panel. Shot point and receiver locations are indicated by black dots and ellipses, respectively. (Mason, 1981)
However, direct measurements of coal properties between sources and receivers would yield more reliable estimates of the presence of disturbances in the coal. The following example shows an experiment where channel wave tomography was used to image the distribution of CO2 in a thin low-velocity reservoir.
Tomographic Imaging of a CO2 Flood in a Low-Velocity Oil Reservoir
The advantages of CO2 injection into oil reservoirs are two-fold. It enhances the recovery of oil in place that would be difficult to produce otherwise, while it offers one method of geologic carbon sequestration. During a recent experiment in an oil field, CO2 was injected in a 10 m thick oil reservoir at a depth of 1500 m, to decrease the viscosity of the oil in place and increase production rates (Majer et al., 2001). The boreholes were drilled using directional drilling technology reaching vertically down to a depth of 1500 m before turning horizontally for a distance of 1000 m in the reservoir layer. The reservoir consisted of a limestone deposit between two anhydrite layers, thus forming, similar to the situation of a coal seam, a low-velocity channel between fast bedrock layers. To map the location of the injected CO2 it was decided to conduct a tomographic experiment, where sources and receivers were placed in opposing horizontal boreholes within the limestone reservoir. The source was piezoelectric ceramic attached at the end of coil tubing, while the receivers where hydrophones placed inside a fluid-filled coil-tubing string. Figure 5 shows a plan view of the horizontal projections of injector and producer well. The two wells were located 300 m apart, while the length of the imaged area was approximately 500 m. A total of 100 source and 72 receiver positions were used during the experiment. The amplitude spectra of a typical seismic trace and its spectral decomposition are shown in Figure 6. A total of six modes were recognized in the data. The bottom trace in the right panel of Figure 6 shows a recorded trace. The refracted P-waves arrive as small amplitudes at about 0.1 ms while the channel waves are represented by the largest amplitudes on the seismogram arriving at about 0.17 ms. In addition to the fact that the channel waves propagate directly through the low velocity reservoir (whereas the P-waves propagate along their boundary with the hostrock) their amplitudes have a much higher signal-to-noise-ratio than in the case of the P-waves, and therefore are better suited to estimate the properties of the reservoir. However, the arrival times of the channel waves are difficult to determine, because of their dispersive character and therefore it was decided to compute the energy within a time window of the 4th mode (at 850 Hz) and use this parameter to estimate the spatial distribution of the attenuation in the reservoir. It is know that gas (being more compressible than rock and water) attenuates seismic waves during propagation and thus areas of increased attenuation (decreased amplitudes) would indicate the location of the CO2 in the reservoir. The result of the attenuation imaging is given in Figure 7. The color scale indicates attenuation levels of the channel waves with 1 and 0 representing high and low attenuation, respectively. In the current case the yellow-reddish areas indicate the locations of the gaseous CO2 in the reservoir. Surface reflection and reservoir modeling studies subsequently confirmed the results of the tomography study. The linear yellow-reddish structure in the upper part of the image trending from 1000 m across the image to 800 was identified as a flow channel in the limestone reservoir. Similar results can be expected for the imaging of water or gas filled voids in coal seams.
Figure 5: Plan view of the horizontal section of the injector and producer wells. Source and receiver positions are indicated by green and blue dots respectively.
Imaging of Subsurface Voids using a Scattering Approach
In a recently developed approach, the location and size of subsurface cavities was accurately determined even in the presence of strong correlated and uncorrelated noise (Gritto and Majer, 2000). The method can be applied in cases when information about the background medium is available such as the case of a cavity in a coal seam. The knowledge of the background medium, and the information that a water- or air-filled cavity is the target, reduces the experiment to a problem with two unknowns: the location and size of the cavity. Based on this a priori information a correlation analysis in conjunction with a grid search can be performed to estimate these parameters. Figure 8 shows the results of a numerical experiment where the "field" data was generated using an analytic solution for the scattering of elastic waves by a sphere (Korneev, and Johnson, 1996). The inversion approach is based on the Mie approximation for the same scattering problem (Korneev and Johnson, 1993), which is an approximation to the exact solution in the low to intermediate frequency range. The geometry of the experiment is given in Figure 8b, where two cavities with different radii are located in the subsurface. Seismic sources and receivers are distributed in a borehole and along the surface. Figure 8a shows the scattered field for the two cavities as computed by the analytic solution. The correlation map in Figure 8b was computed using a maximum likelihood approach. Scattering functions based on the Mie approximation were computed for a representative cavity for 200 possible cavity locations in the subsurface, and correlated with the exact scattered field shown in Figure 8a. This resulted in the correct location of both cavities in the subsurface. In a second inversion step, scattering functions were computed for cavities with varying radius at the newly found locations, and subsequently correlated with the data shown in Figure 8a. The result is shown in Figure 8c, where it can be seen that the radii of the cavities are well estimated. The next case indicates what happens when the "field" data is contaminated by large-amplitude noise as seen in Figure 9a, where 300% uncorrelated white Gaussian noise was added to the data. In addition, the background velocity, which is assumed to be known, is intentionally chosen to be too high in this case (+10%). Although the low signal-to-noise ratio of the seismic data seem to render any successful data processing impossible, the correlation analysis still provides reasonable results, as the large cavity is still correctly located in space, while the location of the smaller one has migrated to the right, because of the assumption of a higher background velocity. Similarly the radii in Figure 9c are well estimated. Thus the problem of locating and estimating the size of subsurface cavities can be solved by correlation analysis even in the case of large data noise. The important fact is that enough data is collected over a certain bandwidth to be able to improve the signal-to-noise ratio by stacking the results during the correlation analysis. In a similar way, the problem of locating voids in coal seams can be approached.
Tomographic imaging has been successfully applied in the past to estimate properties of the medium under investigation. The better the source and receiver coverage, the more reliable are the results of the imaging. Inherent problems associated with tomographic methods are limited access and long propagation distances between sources and receivers. The longer the wavelength of the seismic wave the farther is its maximum propagation distance, but the lower is the resolution of the wave. In contrast, short wavelengths provide high resolution but their propagation is limited to short distances. Thus, a good balance is required between resolving power and sufficient propagation distance to image the medium under consideration.
Seismic detection of voids in coal seams can be based on two principles: direct and indirect measurements. If seismic P-waves are used for detection, indirect parameters will be measured such as induced stress patterns in the roof or floor of the coal seam, because the P-waves propagate along the coal/bedrock interface and thus are not directly affected by the presence of a void. Experience is needed to interpret the resulting velocity patterns to estimate the location and size of a cavity. The migration of stress through rock is a diffuse process and anisotropic stress pattern, present even in apparently homogeneous rock, are to be expected in mining areas. The general layout of coal mines, including multiple levels of tunnels to mine vertical sequences of coal seams, gives rise to this complicated stress pattern. In some instances a calibration of the method in areas with known voids is preferable to determine the characteristics of the resulting velocity pattern.
If channel waves are used, the obstruction by voids located in their path becomes apparent in terms of increased travel times and attenuated amplitudes, relative to channel waves that had a clear path through the coal. Because the obstruction of the travel path can be observed in most tomographic surveys, the estimation of location and size of voids becomes more reliable. Channel waves exhibit more bandwidth than P-waves. The higher the transmitted seam waves frequencies, the higher the resolution of the survey. The phases of high frequencies are concentrated within the area of the coal seam whereas lower frequencies carry energy outside as well as within the seam (Gritto and Dresen, 1992). If phases with lower frequencies encounter an object during propagation, part of the channel wave is reflected back, while the part propagating in the bedrock adjacent to the coal seam continues unaffected. These phases will interfere constructively again, and create a new channel wave several wavelengths behind the disturbance. The only noticeable difference is a reduction in high frequency energy, requiring a careful analysis of the transmission seismograms for these effects. In a best-case scenario, high frequency sources (e.g. small explosive charges) should be used such that the frequency content is high enough and energy is confined to the coal seam. However, these high frequencies will be attenuated quickly and thus a compromise between resolution and distance of propagation must be found individually for each case.
Void detection in coal mines is a difficult task, because of the complicated geology encountered in previously mined areas. Collapsing roof structures produce strong heterogeneities that scatter rather than reflect energy in a coherent from. This results in the recording of seismic waves that resemble more typical noise patterns than coherent data containing the necessary information about the voids. Recent developments in scattering theory, however, resulted in promising techniques to utilize the scattered energy to determine the location and size of voids in the subsurface.
A combination of the above mentioned techniques will likely yield the highest success rate, as they are complimentary to each other. The data can be collected in a single seismic experiment, as long as the specifics of each technique are taken into consideration (i.e., use of sources with sufficient bandwidth, use of 3-component geophones, etc.). Therefore, the costs of data collection will not be higher than for any other seismic experiment that uses only one of the above mentioned techniques.
This work was supported by the Director, Office of Science, Office of Basic Energy Sciences, Division of Engineering and Geosciences, of the US Department of Energy under Contract No. DE-AC03-76SF00098. Data processing was performed at the Center for Computational Seismology, LBNL, which is supported under the same contract.
Buchanan, D.J., Davis, R., Jackson, P.J. and Taylor, P.M. (1981): Fault location by channel waves seismology in United Kingdom coal seams, geophysics, Vol. 46, pp. 994-1002.
Dresen L. and Rueter, H. (1994): Seismic coal exploration, Part B: In-seam seismics, Vol. 16B, eds. Helbig, K. and Treitel, S., Pergamon, pp. 433.
Greenhalgh, S.A., Burns, D. and Mason, I. (1986): A cross-hole and face-to-borehole in-seam seismic experiment at Invincible Colliery, Australia, Geophysical Prospecting, Vol. 34, pp. 30-55.
Gritto, R. and Dresen, L. (1992): Seismic modelling of seam waves excited by energy transmission into a seam, Geophysical Prospecting, Vol. 40, pp. 671-699.
Gritto, R. and Majer E.L., (2000): Seismic Mapping of Subsurface Cavities, Conference Proceedings of the Annual Symposium on the Application of Geophysics to Engineering and Environmental Problems (SAGEEP), Arlington, VA, pp. 1215-1244.
Korneev V.A. and L.R. Johnson, (1993): Scattering of elastic waves by a spherical inclusion - I. Theory and numerical results, Geophys. J. Int., 115, pp. 230-250.
Korneev V.A. and L.R. Johnson, (1996): Scattering of P and S waves by a spherically symmetric inclusion, PAGEOPH., 147, pp. 675-718.
Majer, E.L., Daley, T.M., Gritto, R., Korneev, V.A., and Li, G., (2001): Application of cross well and single well seismic methods for mapping CO2 movement, Proceedings, GSA, 97th Annual Meeting.
Mason, I., (1981): Algebraic reconstruction of a two-dimensional velocity inhomogeneity in the High Hazle seam of Thoresby colliery, Geophysics, Vol. 46, pp. 298-308.
Nolet, G. (ed.), (1987): Seismic Tomography, D. Reidel Publishing Company.