Seismic information helps predict drilling hazards, choose casing point

DRILL-BIT SEISMIC TECHNOLOGY-1

Richard Meehan
Schlumberger Cambridge Research
Cambridge

Les Nutt
Schlumberger Wireline & Testing
Houston

Neder Dutta
BP Exploration
Houston

James Menzies
Lasmo plc
London

Information derived from drill-bit seismic can help predict upcoming drilling hazards, eliminate casing strings, and ensure vital horizons are cored.

The technique requires only surface sensors and does not interfere with the drilling process.

When employed correctly, the method can lead to substantial cost savings and enhance the safety of drilling operations. This first of a two-part series provides details on the acquisition and processing of drill-bit seismic information. The second part provides examples of usage in the field.

Drill-bit seismic information can be used to:

• Locate the bit on the surface seismic section as drilling progresses. The time-to-depth values allow the position of the bit to be plotted continuously.

• Generate look-ahead images. Seismic images of the formation tops ahead of the bit can be generated, allowing upcoming horizons to be monitored in real time. The images can also be used to correlate with the surface seismic image to resolve any timing differences.

• Select casing and coring point with confidence and accuracy. In some cases, this means eliminating entire casing strings, leading to major cost savings. Reducing uncertainty in predictions of depth to formation tops can also allow more aggressive drilling with a corresponding decrease in rig time.

• Predict pore pressure at the bit. Differentiating the time-to-depth data with respect to depth provides an estimate of the formation velocity. When used in conjunction with other information, such as drilling parameters, measurement-while-drilling (MWD) data, and cuttings analysis, estimates of the pore pressure at the bit can be made. These estimates can be used to determine minimum safe mud weights

• Predict pore pressure ahead of the bit. The real-time data from drill-bit seismic, along with other measurements, can be used in conjunction with modeling techniques to predict pore pressures ahead of the bit.

• Predict depth to drilling hazards. When used with conventional high-resolution, borehole seismic-survey data, the drill-bit seismic time-to-depth information can allow continuously updated depth predictions of drilling hazards such as overpressured zones. Accurate predictions of depth to over-pressured zones and reliable estimates of current formation pressure ensure the use of safe minimum mud weights.

Basic concept

The drill-bit seismic technology uses acoustic energy radiated by a working drill bit to determine the seismic time-to-depth as the well is drilled.

The energy required for drilling is supplied to the bit by the drillstring. If a roller-cone bit is used, this rotation causes the cones to roll over the bottom of the hole. As the cones roll over, the teeth penetrate and gouge the formation, destroying the rock.

As each tooth indents the formation, it applies an axial force to the bottom of the hole and an equal and opposite force to the drillstring. The succession of axial impacts as the bit drills radiates compressional or P-waves into the formation, causes axial vibrations to travel up the drillstring.

A working roller-cone bit acts as a dipole source for P-waves,^1-3 radiating energy upwards towards the surface, and downwards ahead of the bit. At the surface, geophones, hydrophones, or a combination of both are used to detect the P-waves.

Sensors, such as accelerometers, placed near the top of the drillstring on the swivel or top drive, detect the axial vibrations traveling up the drillpipe. Although the bit-generated signal can be detected, it is continuous in nature. Because the fundamental drill-bit seismic measurement is time-to-depth, timing information must be extracted.

In general, the energy propagating through the formation travels more slowly than the axial vibrations in the drillstring. The seismic sensor signal therefore contains a time-shifted version of the drillstring sensor signal.

Correlating the drillstring sensor signal with the seismic sensor signals, a technique patented by Elf Aquitaine SA,⁴ helps to determine this difference in travel time DT_rel(Fig. 1 [76,693 bytes]). Once DT_rel is known, if the time taken for the axial vibrations to travel along the drillstring, DT_ds, can be determined, the absolute travel time from bit to surface, DT_f, can be calculated.

The time-to-depth is calculated using the direct radiation from the drill bit. The energy that propagates downwards ahead of the bit is often reflected back to the surface by impedance changes in the formation. This energy can also be detected, and processed to produce a seismic image of the formation ahead of the bit.

When used in combination with the surface seismic, such "look-ahead" images allow the approach to critical horizons to be monitored as drilling progresses.

The above explanation is rather simplistic. In practice there are significant difficulties that must be overcome before useful information can be obtained. The most important information derived from the drill-bit seismic measurement is formation travel time.

It is necessary to ensure that all factors which may affect the accuracy, both relative and absolute, of the travel time measurement are understood. Quantifying the size of the possible timing errors gives confidence in the measurement and helps ensure the technique is correctly applied.

Particular attention must be paid to the drillstring travel-time measurement and the effects of processing on the phase of the signals. Working rigs create a great deal of noise, and sophisticated signal processing methods must be employed to extract the drill-bit signal.

Array geometry and sensor deployment

Two sets of sensors are used in the drill-bit seismic technique. The vibrations traveling in the drillstring are sensed by a multicomponent accelerometer, which is usually mounted on the swivel. The accelerometer unit is connected to the acquisition electronics by a multiconductor cable deployed along the mud hose and standpipe.

The seismic sensor array may take several forms, depending on the rig location. On land, it usually consists of individual single component geophones. Sometimes several groups of 6 or 12 geophones hardwired together are used instead.

The geophones are placed in an array pointing away from the wellhead. Normally the array is offset some distance from the wellhead. The number of sensors in the array and the sensor spacing depends upon the surface conditions.

High-quality data sets have been collected with arrays as short as 30 m, containing 12 individual geophones. On land, it is usually a simple operation to change the number and spacing of the geophones.

Coupling of the geophones to the ground is important. In some situations it may be desirable to bury the sensors in shallow holes to improve coupling. The greatest difficulties in land surveys occur when the near surface is hard and rocky.

This leads to fast surface waves that are difficult to remove during processing. Very-soft sandy surfaces also cause problems because of signal attenuation.

Offshore, either hydrophones or geophone/hydrophone pairs are used. These are deployed on the seabed, once again in a line pointing away from the wellhead, and are connected to the acquisition electronics by means of a multiconductor cable.

If the water is shallow (less than 150 m), deployment is reasonably straightforward. The cable is laid out on the back of a boat positioned close to the platform. One end of the cable is attached to the rig. The boat moves slowly away from the rig and the cable is deployed over the back of the boat.

Anchor weights attached to the cable ensure that it sinks to the bottom reasonably quickly. The direction of deployment may be constrained by the rig-anchor cables and prevailing currents. Usually the array will consist of a number of dual-sensor geophone/hydrophone units.

In deeper water, deployment is more difficult. Because the cable can drift with currents as it sinks, it is important to keep track of its whereabouts to ensure the array ends up pointing radially away from the rig.

If the array is broadside to the rig, the removal of rig-generated noise is much more difficult, with a consequent reduction in data quality. To ensure this does not happen, acoustic transponders are attached to the cable at various points along its length. The cable position can then be monitored as it sinks, and appropriate action taken if it drifts in the wrong direction.

The cable is wound on to a winch mounted at the back of the boat positioned close to the rig. The end of the cable is passed up to the rig where it is connected to the acquisition equipment.

The boat then moves slowly away from the rig, paying out cable from the winch. At the appropriate intervals, the boat and winch are stopped while anchor weights and acoustic transponders are connected to the cable. Deployment then continues.

When the sensor section of the cable is reached, the winch is stopped once again and the dual-sensor units are attached. By monitoring the output of the sensors at the rig, each unit can be checked as it is attached to the cable.

As the cable sinks, its position is monitored by means of the acoustic transponders. By continuing to move away from the rig, the boat ensures the sensor array forms a line pointing away from the rig. When the array is on the seabed, the end of the cable is dropped from the boat, and sinks to the sea floor.

This technique has been used to deploy cables successfully in almost 400 m of water. It requires careful coordination and planning, and the captain of the boat must be fully involved in this process.

Acquisition and processing

The philosophy behind the development of the drill- bit seismic technique is that it should:

• Use the minimum number of sensors
• Cause no interference to the drilling process
• Require no additional downhole hardware
• Require no rig time
• Provide real time answers.

A schematic of the acquisition hardware is shown in Fig. 2. [64,914 bytes] Two personal computers (PCs) are used to do the acquisition and processing. During drilling operations, the front-end or acquisition PC acquires a few seconds of data from the swivel accelerometer unit and each of the geophones or hydrophones.

The accelerometer signal is correlated with itself and each of the seismic sensor signals, and the resulting waveforms are stored. Another data segment is then acquired, the signals cross correlated, and the waveforms added to the previous results.

This correlation and stacking process continues for a fixed number of data segments. Once this number of segments has been reached, the resulting waveforms, called the intermediate stacks, are written to disk. The number of data segments is chosen to ensure there is at least one intermediate stack for every meter of hole drilled.

The second PC is used to perform the rest of the processing and analysis. The raw and processed data can be stored on DAT tape in a variety of formats.

Because the drill-bit seismic technique uses a small number of sensors and operates in a very noisy environment, sophisticated signal processing techniques must be invoked to ensure accurate and reliable answers. Fig. 3 [63,953 bytes] shows a flow diagram of the major processing steps.

In order to obtain absolute time-to-depth information, it is essential to determine the time taken for the bit signal to travel along the drillstring. If the drillstring were a uniform pipe, the drillstring travel-time determination would be easily calculated.

The acoustic velocity of a rod wave in steel is given by:

sq. rt.E/ p (1)

where: E is Young's modulus and p is density, generally on the order of 5,100 m/sec.

Unfortunately, the drillstring is not a uniform pipe. Most of it is made up of 9.5 m sections of drillpipe that are joined together by threaded connections at each end called tooljoints.

The tooljoints are short, on the order of 0.5 m, but are much thicker than the body of the drillpipe. The tooljoints have a large effect on the transmission of vibrations along the drillstring.⁵

At the frequencies of interest here, below 200 hz, these add mass to the system without a corresponding increase in stiffness. This causes a significant reduction in the effective acoustic velocity.

The actual acoustic velocity for any particular drillstring depends upon the detailed geometry of the drillstring components, and as such, is sensitive to the grade and wear state of the drillpipe.

A robust way of determining drillstring travel time is to calculate it from the accelerometer measurement. The drillstring can be considered as a one-dimensional, equal travel-time layered system (Fig. 4 [87,918 bytes]). The acoustic impedance of each layer depends upon the layer material properties and cross-sectional area.

Such a system can be represented by the reflection coefficients that characterize the impedance contrasts at layer boundaries. For an impulse applied at the bit, the output of a sensor on the swivel can be written as a ratio of polynomials in z, where z is the delay operator.⁶

A unique relationship exists between the coefficients of the z polynomials and the reflection coefficients.^6-8 The sensor signal can thus be processed to produce a reflection coefficient series, or drillstring image.⁹

Identification of any particular interface in this image, for example that between the drill collars and the heavyweight pipe, allows the travel time from that point to the surface sensor to be determined. Identification of the reflection coefficient caused by the impedance contrast between the bit and the formation gives the travel time from bit to swivel along the drillstring.

An example of a drillstring image calculated from real data is shown in Fig. 5. [144,541 bytes] This data set was obtained when drilling on land through predominantly Tertiary and Cretaceous clastics and limestones, using a 121/4-in. roller-cone bit.

The moveout of the reflections influenced by the BHA components can be seen as depth increases. The impedance contrast caused by the bit/formation interface can be clearly identified in each trace, providing the drillstring travel time.

Drillstring multiple removal

The seismic-sensor signals are cross correlated with the drillstring sensor signal. As the bit-generated signal travels along the drillstring, it undergoes multiple reflections caused by changes in the cross section of the drillstring components.

This introduces multiples into the drillstring sensor signal that cause multiple peaks in the cross correlations. Before any time picking can be performed, these multiples must be removed. This is achieved by designing a filter based on the drillstring imaging processing. The multiple removal operation is then carried out after correlation.

It should be noted that the signals detected by the seismic sensors also contain multiples caused by the drillstring structure. This occurs because some of the energy that travels up the drillstring is reflected downwards by impedance changes.

When this reflected energy reaches the bit, some of it is transmitted into the formation and is then detected by the seismic sensors. These reradiated multiples are usually quite low in amplitude and do not interfere with the time picking process, thus, no attempt is made to remove them.

Drilling rigs are noisy. The most troublesome source of noise is heavy machinery such as mud pumps and diesel engines. On land, these machines generate large surface waves that are detected by the geophones.

The amplitude of this surface noise can be several orders of magnitude greater than the drill bit signal. Offshore, the vibrations caused by the pumps couple from the rig structure to the water and travel down to the seabed sensors.

Jack up rigs can also excite surface waves that travel along the sea floor. Usually, rig-generated noise is more troublesome on land than offshore

Traditional seismic-acquisition methods reduce surface waves by using large arrays of sensors. Summing the output from a large array reinforces those signals that arrive at each sensor simultaneously (the drill-bit signal), while attenuating those signals that travel across the array (surface waves).

Such techniques provide limited attenuation of surface noise; however, they are not particularly effective for drill-bit seismic acquisition. Because of the particularly large amplitude of the rig-generated noise, and the need to keep the number of sensors to a minimum, especially for offshore surveys, alternative techniques must be used.

Powerful adaptive digital filters are used to attenuate the rig-generated noise. Each individual geophone or hydrophone output is digitized. These digital signals are then processed by a multiple input/multiple output unconstrained digital beam former based upon adaptive interference canceling techniques.^{10 11}

By careful design of the filter structure, signals with a specified range of moveouts across the array are protected, while noise wavefields with different moveouts are suppressed.

Fig. 6 [127,174 bytes] shows the effectiveness of the technique when applied to hydrophone data. These data were recorded during a drill-bit seismic survey offshore Viet Nam. The water depth was about 140 m, and the array consisted of individual hydrophones arranged in a line pointing away from the rig.

The nearest hydrophone was about 210 m from the wellhead. The plot on the left shows the cross correlations of each of the hydrophone signals with the drillstring sensor. The noise generated by the rig machinery can be seen propagating across the array.

The plot on the right shows the result of applying the digital beamformer. The direct arrival lags at about 0.39 sec. The event at 0.57 sec is the reflection of the direct arrival from the sea surface.

Because each individual channel is preserved, the moveout of the direct arrival across the array can be determined, and any necessary alignment correction made before summing the traces together.

Source, sensor compensation

The drill-bit source consists of impulses in force, the rig sensor measures acceleration, and the seismic sensors measure pressure (hydrophones) or particle velocity (geophones). The accelerometer phase response is flat over the bandwith of interest; however, the seismic sensors exhibit a lot of phase distortion, particularly at low frequencies. Corrections must be applied to deal with all the above effects.

In the offshore environment, the seismic sensors are laid on the seabed. Normally, dual hydrophone/geophone sensor units are used. Hydrophones are sensitive to pressure variations in the water while the geophones are sensitive to particle velocity.

As the bit-generated signal travels upwards, it reaches the seabed sensors, then continues on towards the sea surface where it is reflected back downwards. The sea surface is an almost perfect reflector, and hence has a reflection coefficient of magnitude 1.

The reflection coefficient is negative for pressure, and positive for particle velocity. When this reflected signal reaches the sea floor, some of it is reflected back up towards the surface, and some is transmitted down into the formation.

Because the impedance of the sea floor material is usually greater than the impedance of the water, the sea-floor reflection coefficient is positive for pressure and negative for particle velocity. Because the sensors are above this interface, they measure the sum of the signal reflected from the sea surface, and the signal reflected from the sea floor.

If the direct signal from the bit has an amplitude of 1, and if the sea floor has a reflection coefficient of magnitude k, then the reflected signal measured at the sea floor by the hydrophone has an amplitude of -(1 + k). For the geophone this reflected signal has amplitude (1 + k).

By combining the hydrophone and geophone signals together with the appropriate scaling, the sea surface multiple can be canceled out as shown in Fig. 7 [135,456 bytes]. The top plot shows the hydrophone data. The first and second sea surface multiples are visible. The middle plot shows the geophone data recorded over the same interval. The combination of the two data sets (bottom plot) cancels the sea surface multiples.

Offset, water depth corrections

The seismic sensors are usually offset some distance from the wellhead. In order to refer the measured time-to-depth to vertical, zero-offset time-to-depth, a correction must be applied to the data.

This correction can be estimated from the measured time-to-depth if the offset of each sensor from the wellhead is known. On land, it is a trivial task to measure the actual distance from the wellhead to the sensors.

Offshore it is more difficult because there is no direct access to the seabed sensors. When acoustic transponders are used during deployment, they can give the precise location of the cable with respect to the wellhead. However, deployment in shallow water (less than 150 m) does not usually require acoustic transponders.

In this case, the distance from wellhead to sensors can be calculated by firing an air gun in two or more known positions and recording the airgun signal as detected by the sensors. Triangulation techniques can then be used to locate the sensors.

In the offshore case, a correction must also be made for the water depth. Traditional surface seismic data are referred to mean sea level, but the drill bit seismic sensors are usually located on the seabed.

In order to relate the measured time-to-depth to the surface-seismic two-way time, the time taken for the bit signal to travel from the seabed to the sea surface must be known. This depends upon the water depth and the acoustic velocity of the water.

The depth is usually available from the site survey data, and apart from tidal variations, does not change significantly over time. The acoustic velocity, however, can change significantly with time.

For seawater, acoustic velocity is primarily a function of temperature, density, and salinity,¹² all of which show seasonal variations. The size of the variations depends upon location and climatic conditions.

To ensure that the time-to-depth measurement is not biased by seawater properties, the water-depth correction should be calculated with reference to the seawater acoustic velocity, obtained at the time of the original surface seismic survey.

Once the drillstring travel time and offset and water depth corrections have been established, the cross correlations can be shifted to "one-way time."

The data can also be further processed to produce a look-ahead image. Standard borehole seismic processing techniques may be used to do this; however, improved performance can be obtained by using semblance-weighted deconvolution.¹³

Acknowledgments

The authors wish to thank BP Exploration, Statoil, Norsk Hydro, Lasmo, Union Texas Petroleum, and Itochi Oil Exploration Co. Ltd. for permission to use their data.

They also wish to thank Bill Borland, Bill Underhill, Scott Leaney, Julian Drew, Shoichi Nakanishi, Masahiro Kamata, Kevin Dodds, Lisa Stewart, Aung Than Oo, Nguyen Huu Ngu, Frode Ronning, and Jamie Paterson of Schlumberger; Eamonn Doyle of Norsk Hydro; Chris Einchcomb of BP; and Alastair Sharp of Lasmo for their helpful comments.

The development of the drill-bit seismic service was partially funded by the Thermie program of the Commision of European Communities under contract OG 046/93.

References

1. Hardage, B.A., Crosswell Seismology and Reverse VSP, Geophysical Press Ltd., London, 1992, p. 41.
2. White, J.E., Underground Sound, Elsevier Science Publishers, 1983, p. 194.
3. Rector, J. W., and Hardage, B.A., "Radiation pattern and seismic waves generated by a working roller-cone drill bit," Geophysics, 57, No. 10, October 1992, pp. 1319-33.
4. Staron, P., Arens, G., and Gros, P., "Method for instantaneous acoustic logging within a well bore," International Patent Application No. WO 85/05695, May 20, 1985.
5. Drumheller, D.S., "Acoustical properties of drillstrings," Journal of the Acoustical Society of America, Vol. 85, No. 3, March 1989.
6. Claerbout, J.F., Fundamentals of Geophysical Data Processing, Blackwell Scientific Publications, California, 1985, p. 1.
7. Marple, S.L. Jr., Digital Spectral Analysis with Applications, Prentice Hall, New Jersey, 1987, p. 181.
8. Rabiner, L.R., and Schafer, R.W., Digital Processing of Speech Signals, Prentice Hall, New Jersey, 1978, p. 441.
9. Booer, A.K., and Meehan, R.J., "Drillstring Imaging: An Interpretation of Surface Drilling Vibrations," SPE paper 23889, presented at the IADC/SPE Drilling Conference, New Orleans, Feb. 18-21, 1992.
10. Widrow, B., and Stearns, S.D, Adaptive Signal Processing, Prentice Hall, New Jersey, 1985.
11. Haykin, S., Adaptive Filter Theory, Prentice Hall, New Jersey, 1986.
12. Chen and Millero, "The Sound Speed in Seawater," Journal of the Acoustical Society of America, Vol. 62, 1977, pp. 1129-35.
13. Haldorsen, J., Miller, D., and Walsh, J., "Multichannel Weiner deconvolution of vertical seismic profiles," Geophysics, Vol. 59, No. 10, October 1994.

The Authors