Вы находитесь на странице: 1из 7

GEOPHYSICS, VOL. 78, NO. 4 (JULY-AUGUST 2013); P. P1P7, 11 FIGS. 10.1190/GEO2012-0344.

Downloaded 08/06/13 to 190.198.228.84. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Estimation of bubble time period for air-gun clusters using potential isosurfaces

Daniel Barker1 and Martin Landr1

ABSTRACT
We evaluated a method of estimating the relative bubble time period of air-gun clusters with an arbitrary number of guns. This was done by assuming incompressible flow and representing the bubbles as isosurfaces of the potential field to account for coalescence. The kinetic energy at the equilibrium radius was then compared to the equivalent energy of the single gun to estimate the relative change. The results agreed well with two-gun cluster measurements, but the lack of data does not allow us to compare with clusters containing more guns than that. We found that more compact configurations, such as a triangle instead of three guns in a line, gave a more rapid increase in the bubble time period as the gun separation decreased. This indicated that compact configurations were attractive for enhancing the low-frequency output from an air-gun cluster.

INTRODUCTION
Recently, there have been several inventive ways of configuring streamers, or using dual sensors, to remove the receiver ghost, and there have also been some innovations on source geometries with the motivation of removing the source ghost (e.g., Hill et al., 2006; Cambois et al., 2009). With these kinds of improvements to the signal, it is natural that there has also been some focus on improving the signal emitted directly from the source as well. Especially, as more complex geologies are being imaged, there has been an increased demand for more low-frequency energy (Laws et al., 2008; Hegna and Parkes, 2011; Kragh et al., 2012). This low-frequency output is related to the source bubble oscillation period in an inverse way, and the oscillation period has been shown to follow the Rayleigh-Willis formula (Willis, 1941) in which the period is propor-

tional to the volume and pressure to the power of one-third, and it is inversely proportional to the depth to the power of five-sixths. Of the possible ways of increasing the time period for an air-gun array, Hegna and Parkes (2011) conclude that increasing the total volume by either adding guns or increasing the volume of each individual gun is the only way to make a significant change. Because there is a limit to how big guns are currently manufactured, increasing the number seems most practical. One way of getting around the limited size of single guns is to emulate a bigger single gun by clustering. An air-gun cluster is identified by firing two or more guns at a very close distance to have the individual bubbles coalesce and form one bigger bubble. In addition to increasing the bubble time period, clustering can also improve the primary-to-bubble ratio, which will usually improve the source output as well. Clustering may also be considered to slow down the dynamics, as identified by the increased bubble time period, which may be a cause for increased attenuation at higher frequencies for clusters (Landr et al., 2011). The effects on the bubble time period and primary-to-bubble ratio were investigated through experiments by Strandenes and Vaage (1992) for two-gun clusters, and they were shown to be dependent on a dimensionless constant described by the separation distance of the two guns divided by the so-called equilibrium radius. The equilibrium radius is defined as the radius of the spherical bubble at its equilibrium state, where the temperature and pressure inside the bubble equal its surroundings. It is important to note that this exact state might not actually occur during the bubble oscillations, so other definitions, for instance based on adiabatic expansion, might be better. However, because we compare our results with those that Strandenes and Vaage (1992) produce, we will use their definition. Other approaches would stretch or compress the data points slightly, but they are otherwise of no significant importance to our calculation method. Barker and Landr (2012) give an analytical function for estimating the relative bubble time period increase as a function of the previously mentioned dimensionless parameter. This is limited,

Manuscript received by the Editor 23 August 2012; revised manuscript received 20 March 2013; published online 21 June 2013. 1 Norwegian University of Science and Technology, Department of Petroleum Engineering and Applied Geophysics, Trondheim, Norway. E-mail: daniel .barker@ntnu.no; martin.landro@ntnu.no. 2013 Society of Exploration Geophysicists. All rights reserved. P1

P2

Barker and Landr for assumptions based on the ideal gas law, be a function of volume and temperature, but not shape. By assuming this, we do not actually need to know exactly how the bubble expands. The reason for this is that the value will cancel because we estimate relative bubble time periods. Because the calculation is not needed, we will concentrate on estimating the relative bubble time period by considering only the kinetic energy level, assuming that the kinetic energy can be described by the bubble wall velocity U and some arbitrary function f depending only on volume, separation distance 2b, and the applicable physical constants. Although we will not assume sphericality, we will still use R as a parameter, and we let R describe the radius needed to create a spherical bubble of the same volume. This means that

Downloaded 08/06/13 to 190.198.228.84. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

however, to the case in which the two bubbles are still separated at the time they reach their equilibrium radius. When considering clusters with more than two guns, as Hopperstad et al. (2012) suggest, the magnitude of the clustering effects is not so well documented, although one would expect the main trends to be the same. Of the existing ways to model air-gun clusters, modeling full coalescence is usually avoided (as by Laws et al., 1990; Li et al., 2011), or it is handled by boundary integral methods (Cox, 2003), which, although giving reasonable results, tend to be too slow and unstable for practical use. With the increased focus on clustering at ranges where this is to be expected, this is as an area in which improvements are needed. We improve the method that Barker and Landr (2012) describe to give a relative bubble time period estimate when the bubbles are coalescing and also to be usable for more than two guns. We will do this by replacing the spherical bubble assumption with an assumption that the bubble is an isosurface of the velocity potential. This implicitly requires the fluid flow to be incompressible. Although the isosurface used does not necessarily equal the physical shape of the bubble, such as boundary integral methods may model (e.g., Cox et al., 2004), it is practical because it allows us to have a smooth transition between separate bubbles and one big merged bubble as well as not having to calculate the expansion until the equilibrium volume is obtained. It should be noted that we do not include standard interaction effects by superpositioning the pressure signal from the other gun, as Ziolkowski et al. (1982) describe, which is commonly used when modeling air-gun arrays and in this case could have been approximated by using the same calculations as for compressible models, mainly because it is hard to define in a meaningful way when the bubbles coalesce and it requires actually modeling the dynamics, which we want to avoid. The results that Barker and Landr (2012) present also seem to indicate that it is not necessary, because they match measured data reasonably well. Although the overall goal is to incorporate the knowledge into a full model of air-gun clusters with full coalescence, here we will limit the investigation to improving on the estimate that Barker and Landr (2012) present, and we will extend the method to more than two guns.

 R

3V 4

1
3

(2)

regardless of the shape of the volume (V ). We then choose to represent the kinetic energy as

Ek f R; bE0 ;

(3)

where E0 is the single-bubble kinetic energy that Rayleigh (1917) gives:

E0 2 R3 U 2 ;

(4)

with being the density of water. Considering the results shown by Strandenes and Vaage (1992), it should then be possible to further simplify this expression to rely only on the relationship between the radius and separation distance, which we will denote by

2b ; R

(5)

so that the expression can be written as

ESTIMATING THE BUBBLE TIME PERIOD FROM KINETIC ENERGY


Barker and Landr (2012) estimate the bubble time period by comparing the kinetic energy at the equilibrium radius with the associated potential energy at t 0 and obtaining an estimate of the velocity at that state described as U EQ . Although the potential energy is explicitly specified by Barker and Landr (2012), given the assumption of adiabatic expansion, this is not actually necessary as long as one assumes that the work having been done to the fluid by the bubbles expansion will only be dependent on the volume at the given state. Other dependencies could also be allowed, as long as they do not change when considering a cluster. This is the case when the work (W ) is given as

Ek f E0 :

(6)

As we assume the kinetic energy is the same at the equilibrium radius for the single gun and one of the guns in the cluster, because the difference in potential energy should be the same, we have
2 3 2 f EQ 2 R3 EQ;cluster U EQ;cluster 2 REQ;single U EQ;single ;

(7)
and because the equilibrium volume is not dependent on whether the gun is in a cluster or not, we obtain
2 f EQ U2 EQ;cluster U EQ;single ;

(8)

V EQ

P P dV;

(1)

and, by assuming that the time used for a full oscillation is proportional to the difference between the maximum and minimum radius, divided by the bubble wall velocity at equilibrium radius,

where P P is the difference between the internal bubble pressure and the hydrostatic pressure. The reason for this is that P will,

T estimate

Rmax Rmin ; UEQ

(9)

Bubble period of clustered guns where the underlying assumption is that the average bubble wall speed scales in the same way as the velocity at equilibrium radius. The relative bubble time period can then be estimated by

P3

1 Ek 2

dS Z dS

(17)

Downloaded 08/06/13 to 190.198.228.84. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

T relative

UEQ;single q f EQ : UEQ;cluster

(10)

1 ^ C 2

(18)

DESCRIBING THE KINETIC ENERGY


To describe the kinetic energy of the system, we need some model of the water movement, and we will use the regular superpositioned velocity potential that Barker and Landr (2012) use. This assumes that the water is incompressible and the potential for the two-gun case is given by

1 ^ _ V; C 2

(19)

_ _ V V p ; 2 2 4 r 4 r 4b 4br cos

(11)

where we have used the fact that the integral of the normal velocity over the whole bubble surface is equal to the rate of change _. of volume V Again, we choose to describe our constant relative to the singlebubble case, in which the isosurface is given by
2 ^ R U C RU: r rR

(20)

where we have chosen to represent it in spherical coordinates, as described in Figure 1. The center of one of the bubbles is placed at the origin, and the other is placed at 2b. A dot denotes differentiation with respect to time, and with our definition of R, we have

_ 4 R2 U; V

_ U R:

(12)

^ C RU, where C is introduced to account Now, we choose C for deviations from the single-bubble case, and it is influenced by the works of by Strandenes and Vaage (1992) and Barker and Landr (2012); we assume that C is a function of only. This means that we can express the energy as

Because is a velocity potential, we have that, for any point in the fluid,

Ek

u ; Z Ek

3 CVU2 2

(21)

(13) C 2 R3 U2 ; (22)

and the total kinetic energy in the fluid domain, denoted by , is

1 2 u dV 2 Z

(14)

1 2

dV:

(15)

By applying Greens identity (McOwen, 2003), this volume integral can be transformed to a surface integral over the bubble wall (),

1 Ek 2

dS;

(16)

where denotes the normal derivative with respect to the bubble wall, pointing inward.

ESTIMATING THE BUBBLE TIME PERIOD BY POTENTIAL ISOSURFACES


Because the integral required to calculate the kinetic energy uses the value of , it seems natural to let the shape of the bubble(s) be described by an isosurface of the potential, meaning a surface where ^ , which can be defined by . In that case, the con is a constant C stant can be moved out of the integral and we are left only with the surface integral of the normal derivative of , which, by harmonic theory, should be proportional to the number of sources inside _ ; e.g., the surface and their respective strength (equivalent to V Kellogg, 1929), so

which is equal to the single-bubble kinetic energy when C 1. A quick run through the bubble time period estimate routine, described by equation 10, shows that p the estimated relative bubble time period then becomes C , and the only problem remaining is to determine the value of C. The assumption that C is a function of only can now be verified numerically, so we can calculate C by using any set of parameters matching a given . For simplicity, we choose R 1 and b 2. By equation 2, we note that the volume inside the isosurface must be 4 3 (in the case of coalescing bubbles, we only consider the volume on one side of the plane of symmetry), which can only be the case for a specific C-value. For a given C, we then need the corresponding volume V C. Sadly, this cannot be given analytically, and it will need to be calculated numerically by locating the points on the surface (where RU C) and then numerically integrating the volume inside.

Figure 1. Model used for calculating kinetic energy of a twobubble system, with two spheres representing the bubbles.

P4

Barker and Landr Let us now examine the extreme values of . First, we notice that the limit as goes to zero is equivalent to the relative bubble time period of a gun with twice the size, which makes sense physically because this is the logical representation of two guns firing exactly on top of each other. Second, as expected, the relative time period tends toward 1 when becomes big because the dynamics will become more and more similar to a single gun. The isosurface method can also be applied to clusters containing more than two guns, as long as they retain the symmetry of all being at the same state at the same time. This means that they will have to

With such a method in place, we can then find C numerically by solving V C 4 3. This means that we 1) Pick a start value for C. 2) Find the isosurface by using the potential field (equation 11 for the two-gun case), and integrate to find the volume inside. Any standard integration method can be used for this, but the exact method will be dependent on the case because there will be different limits depending on the symmetries. This can be done either for one of the bubbles or for the whole volume followed by dividing by the number of guns to find the volume corresponding to one bubble. 3) Compare to the desired volume, 4 3. If the volume is within the desired precision, we have found C. If not, adjust C (decrease if the volume is too small; increase if it is too big) and jump to step 1. These steps can, of course, be solved more efficiently by the use of any standard optimization routine.

Downloaded 08/06/13 to 190.198.228.84. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

ESTIMATED PERIODS FOR SYMMETRIC CLUSTERS


We will start by considering the two-gun cluster, because the results can be compared with the experimental measurements that Strandenes and Vaage (1992) present. An example isosurface in this case is shown in Figure 2 compared to a single spherical bubble, and Figure 3 shows a selection of isosurfaces using the same separation distance, but with increasing volume. We would like to point out that the surface should be treated as an abstraction because a real cluster firing will not create a bubble looking like that, at least not during its first expansion, although it will be more similar at the contraction after the bubbles have reached their first maximum volume. However, because we will be using the shape to approximate the bubble over an entire period as a kind of mean shape, it should be a reasonable approximation. In any case, we assume that it will probably be better than taking the bubbles to be spherical and disjoint. Figure 4 shows the estimated bubble time period along with the experimental data that Strandenes and Vaage (1992) obtain. It is interesting to note that we do not need to estimate any bubble size nor type of expansion (adiabatic is most common) to create the estimate because it is entirely based on . The reason for all other assumptions being neglected is that we calculate the relative change.

Figure 3. A selection of isosurfaces using the same separation distance, but with increasing volume.

Figure 2. Example of the isosurface used for the two-bubble model, which in the spherical model would have with a radius equal to 1/2 of the separation distance ( 2). Note that because the surface has merged two bubbles, its volume is twice that of the reference sphere.

Figure 4. Estimated time period along with data points extracted from Strandenes and Vaage (1992).

Bubble period of clustered guns be placed in a circle, or at the vertices of a regular polyhedron. This induces an ambiguity in how to define the separation distance because there are several ways of doing it. One way could be to use the distance to the nearest gun, but we prefer to use the distance to the center of the sphere all guns will be placed on as b, which is consistent with the two-gun case. Apart from possible ambiguity in the definition of b, the real distinction from the two-gun case is that the rotational symmetry along the line between the two guns is lost and the surface will need to be calculated in three dimensions. This can be seen in Figure 5, which shows an example of a three-source isosurface. In general, this increases the computation time needed to calculate C. The corresponding estimated relative bubble time period is shown, along with four- and five-gun clusters, in Figure 6. Again we see that the limit as tends to zero is the relative bubble time period of a single gun n times the size of the original. In addition, the plateau, which we denote the flat section at the start of every curve where the relative time period is approximately equal to what we would expect of a gun n times the size, becomes bigger. The data that Strandenes and Vaage (1992) present show the maximum primary-to-bubble ratio for two-gun clusters being attained for a -value slightly bigger than the end of the plateau. If this is the case for n-gun clusters as well, it might be necessary to increase the separation distance as more guns are added if it is desired to keep this ratio optimal. It is worth noting that the shape of the n-gun bubble time period estimates is somewhat similar to the two-gun case, but it is not possible to transform one to the other by scaling and stretching, as shown in Figure 7. There, the two-gun curve has been modified

P5

Downloaded 08/06/13 to 190.198.228.84. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

by scaling to match the relative time period of the three-gun case at 0 and , and has been stretched by a factor of 4/3. Although the curves look similar, they are definitely not identical. This means that although it would be most correct to calculate the curves, it is possible to estimate one from the other, given some control points. With four guns, we have a choice for configuration. Either the guns can be placed on a circle (resulting in the cluster being a square), or they can be placed in the vertices of a tetrahedron. Although the latter might be somewhat unfeasible for use in a real air-gun array, it is still interesting to consider the relative difference between the two configurations, which is displayed in Figure 8, whereas Figure 9 shows the respective isosurfaces of the two configurations for a -value slightly smaller than 4. It should be noted that, for the tetrahedral case, the guns are not placed on the same horizontal plane, so there is a difference in hydrostatic pressure for the different guns. This has been assumed to be negligible because when the bubbles quickly coalesce, they will merge into one bubble anyway, but this assumption could introduce errors for big . The two configurations behave in a similar way for small , but the relative time period of the tetrahedron configuration decays slightly

Figure 5. Example of the isosurface used for a three-gun cluster.

Figure 7. An attempt to replicate the bubble time period curve for three guns, by stretching and amplifying the two-gun curve. The two-gun curve is modified to match at 0 and , and stretched by a factor of 4/3.

Figure 6. Estimated relative bubble time period for clusters with different number of guns in a circle configuration. The value is plotted logarithmically to display convergence to 1.

Figure 8. Estimated relative bubble time period for the two possible configurations of a four-gun cluster.

P6

Barker and Landr

Downloaded 08/06/13 to 190.198.228.84. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

which created the restriction of geometries of either a circle or a regular polyhedron. However, when creating a cluster with a lot of guns, a linear cluster might be desirable. In this case, the symmetry is lost, as can be seen when comparing an interior gun to one of the guns at the edges. Also, depending on the separation distanESTIMATING TIME PERIODS FOR INLINE ces between the guns, there might not be one single defined period CLUSTERS for all the guns. However, if we concern ourselves with the mean The presented method was designed to estimate the relative bubbehavior of the cluster, we can still estimate the relative bubble time ble time periods in clusters in which all bubbles behave identically, period change by selecting one C such that the sum of volumes of all the bubbles equals the desired volume and letting U represent the mean bubble wall velocity of all guns. These assumptions are most likely more valid when is small enough to have the bubbles coalesce into one big bubble. An estimate for a three-gun cluster is shown in Figure 10, where we have defined b to be half the separation distance between each of the guns. We note that the limit for 0 is equal for the two configurations, as it should be, and that the relative bubble time period of the inline cluster drops off a lot faster than the for the triangle configuration. This Figure 9. Visualization of the two allowed configurations of a four-gun cluster. (a) The guns are arranged in a square. (b) A tetrahedron configuration is displayed. is reasonable because the inline cluster spans a greater length, such that two of the guns are further apart from each other in that configuration. Figure 11 shows how adding more guns influences the bubble time period of inline clusters. The most interesting aspect, when compared to the symmetric clusters, is that the bubble time period plateau becomes smaller as the number of guns increases. If the relationship between bubble time period and primary-to-bubble is valid here as well, keeping that ratio will be harder because the guns will need to be moved closer to each other as more guns are added. Of course, this is offset by the fact that it is practically a lot easier to build an inline cluster as opposed to circular or polyhedra-shaped clusters, the only possible exception being a cube cluster consisting of eight guns. faster. This can be explained by the guns being in a less tight configuration for the tetrahedron. Again, one of the curves cannot be easily transformed into the other.

DISCUSSION
Figure 10. Comparison of an inline configuration and a triangle configuration for a three-gun cluster. In our estimates, we use isosurfaces of the potential field as the representation for the coalescing bubbles created by several air guns fired together in a cluster. Although this representation will give the simplest mathematical result, it should be noted that it is not a necessary assumption for the method because the kinetic energy can be calculated from an arbitrary shape (and potential). Because the actual physical shape of the bubbles will be different from the isosurface shapes, using other representations may improve the results. As observed from Figure 10, there is a significant difference between the inline configuration and the triangle configuration of a three-gun cluster. This observation could serve as an excellent way to test the validity of the assumptions in our method. Unfortunately, we do not have access to such data; however, a dedicated experiment comparing bubble time periods for these two configurations of a three-gun cluster should be a good experimental test. If experimental data support this observation, it furthermore indicates that the triangular configuration is more efficient as a low-frequency source, because longer bubble time periods are achieved. Furthermore, the proposed method can then be extended to predict how the bubble time period might vary as a function of various configurations also for n-gun clusters.

Figure 11. Comparison of inline configurations with two to five guns.

Bubble period of clustered guns Because the basis of the estimation is based on conservation of energy, it is not possible to estimate the primary-to-bubble ratio directly. It should be possible to create a full air-gun model using the isosurface scheme by differentiating the energy balance, with the single-gun case being equivalent to the Rayleigh equation, but additional terms will be needed in that case to explain dampening. In addition, to stop two coalescing bubbles from separating again when they contract, including movements of the bubble will probably be necessary. This movement may also be part of the dynamics causing clustered air guns to generally have better primary-tobubble ratios than single guns because it is a likely cause of energy loss. Because the model assumes an incompressible fluid and other current air-gun models usually do not, this might also have to be corrected.

P7

sortium at NTNU. We would also like to thank the reviewers and the assistant editor, J. Carcione, for their feedback and consequent improvement of the paper.

REFERENCES
Barker, D., and M. Landr, 2012, Simple expression for the bubble-time period of two clustered air guns: Geophysics, 77, no. 1, A1A3, doi: 10.1190/geo2011-0169.1. Cambois, G., A. Long, G. Parkes, T. Lundsten, A. Mattsson, and E. Fromyr, 2009, Multi-level airgun array: A simple and effective way to enhance the low frequency content of marine seismic data: 79th Annual International Meeting, SEG, Expanded Abstracts, 152156. Cox, E., A. Pearson, J. Blake, and S. Otto, 2004, Comparison of methods for modelling the behaviour of bubbles produced by marine seismic airguns: Geophysical Prospecting, 52, 451477, doi: 10.1111/j.1365-2478.2004 .00425.x. Cox, E. J., 2003, The source signature due to the close interaction of marine seismic airguns: Ph.D. thesis, University of Birmingham, School of Mathematics and Statistics. Hegna, S., and G. Parkes, 2011, The low frequency output of marine air-gun arrays: 81st Annual International Meeting, SEG, Expanded Abstracts, 7781. Hill, D., L. Combee, and J. Bacon, 2006, Over/under acquisition and data processing: The next quantum leap in seismic technology?: First Break, 24, 8196. Hopperstad, J., R. Laws, and E. Kragh, 2012, Hypercluster of airguns More low frequencies for the same quantity of air: 74th Annual International Conference and Exhibition, EAGE, Extended Abstracts, Z011. Kellogg, O., 1929, Foundations of potential theory: Springer. Kragh, E., R. Laws, J. Hopperstad, and A. Kireev, 2012, Reducing the size of the seismic source with a 4C towed-marine streamer: 74th Annual International Conference and Exhibition, EAGE, Extended Abstracts, Z014. Landr, M., L. Amundsen, and D. Barker, 2011, High-frequency signals from air-gun arrays: Geophysics, 76, no. 4, Q19Q27, doi: 10.1190/1 .3590215. Laws, R. M., L. Hatton, and M. Haartsen, 1990, Computer modelling of clustered airguns: First Break, 8, 331338, doi: 10.3997/1365-2397 .1990017. Laws, R. M., E. Kragh, and G. Morgan, 2008, Are seismic sources too loud?: 70th Annual International Conference and Exhibition, EAGE, Extended Abstracts, B026. Li, G., M. Cao, H. Chen, and C. Ni, 2011, Modelling the signature of clustered airguns and analysis on the directivity of an airgun array: Journal of Geophysics and Engineering, 8, 9298, doi: 10.1088/1742-2132/8/1/011. McOwen, R. C., 2003, Partial differential equations: Methods and applications, 2nd ed.: Prentice-Hall. Rayleigh, O., 1917, On the pressure developed in a liquid during the collapse of a spherical cavity: Philosophical Magazine, 34, 9498, doi: 10.1080/ 14786440808635681. Strandenes, S., and S. Vaage, 1992, Signatures from clustered airguns: First Break, 10, 305312, doi: 10.3997/1365-2397.1992015. Willis, H., 1941, Underwater explosions, time interval between successive explosions: Technical report, British Report WA-47-21. Ziolkowski, A., G. Parkes, L. Hatton, and T. Haugland, 1982, The signature of an air gun array: Computation from near-field measurements including interactions: Geophysics, 47, 14131421, doi: 10.1190/1.1441289.

Downloaded 08/06/13 to 190.198.228.84. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

CONCLUSION
We suggested a simple way of estimating the bubble time period of air-gun clusters, and we verified that the results agree with experimental data for the two-gun case. This was done by introducing nonspherical isosurfaces as an approximation for the shape of coalescing bubbles throughout an oscillation. The isosurface method can be extended for clusters with more than two guns. Estimated bubble time periods for n-gun clusters show reasonable behavior as the cluster separation distance changes. However, due to the lack of experimental data for clusters with more than two guns, these estimates are not verified. If we choose to trust the extension to bigger clusters, we see that our model gives some interesting results regarding the cluster configuration. For instance, we see that a three-gun triangle configuration is more effective than the corresponding inline configuration with respect to increasing the bubble time period. This can be explained by the triangle being more compact, causing the interactions to be stronger. This general principle seems to hold for clusters with an even larger number of guns as well, according to the modeling. However, it is important to note that the biggest increase in the bubble time period might not yield the best primary-to-bubble ratio, a parameter which is not predicted or estimated by the proposed method.

ACKNOWLEDGMENTS
We acknowledge Petroleum Geo-Services (PGS) for financing the Ph.D. study of D. Barker. M. Landr acknowledges financial support from the Norwegian Research Council to the ROSE Con-

Вам также может понравиться