Vibroseis evolution: May the ground force be with you

Transcription

1 Vibroseis evolution: May the ground force be with you Spencer Rowse 1* and Anthony Tinkle 1 present an examination of past and present models and concepts in those aspects of seismic data acquisition related to the modelling of the source/ earth interactions. A lthough the precision of positioning and fidelity of recording of the sensors in land seismic exploration has increased significantly over the last three decades, the same cannot be said about the accuracy in the determination of the time varying signal generated from land seismic sources. At best, the current methods of determining the source signature from vibrator sources used on land are approximate. A selective examination of representative past and present models and concepts in the modelling of the source/earth interactions suggests that current trends in land seismic acquisition (towards point source, point receiver surveying, increased emphasis on seismic attributes, and the use of distributed simultaneous and overlapping source events) are forcing a renewed focus on more accurate measurement and/or prediction of land seismic source signatures. New models are required that better describe the dynamically interactive nature of the source-earth relationship. All models are wrong, but some are useful. This aphorism by the statistician Box (Box and Draper, 1987) is meant to emphasize that any mechanistic model should have the following advantages: It contributes to our scientific understanding of the phenomenon under study. It usually provides a better basis for extrapolation (at least to conditions worthy of further experimental investigation if not through the entire range of all input variables). It tends to be parsimonious in the use of parameters and to provide better estimates of the response Paraphrasing the above, the most that can be expected from any model is that it can supply a useful approximation to reality. From this, a pragmatic geophysicist (some would say a cynical one) might ask how wrong do models have to be in order not to be useful. It is worth reminding ourselves that seismic exploration is undertaken to determine the geological structure of the earth in a desired location. This operation consists of generating a time varying impulsive or vibratory signal at different locations on, or near, the earth s surface and recording, via a multitude of sensors at different locations, each transmission of multiple generated signals over a predetermined period. In simplistic terms, an input is applied to the earth and the output response of the input signal is recorded at multiple locations and times to obtain, via a series of processing procedures, an image of the sub-surface seismic response. The fidelity of the seismic image obtained is determined by how accurately and reliably the inputs and outputs are measured, which are strongly affected by how easily the hardware permits us to undertake appropriate recordings, and the efficacy of the various processing methods employed to reduce unwanted noise and/or enhance the desired signal. Finally, through application of the art and science of seismic interpretation, a model of the local geology is deduced. Because the path from seismic image to geologic model can be tortuous, involving not just processing and interpretation, but integration of supporting information such as well logs, potential field data, related geological information, etc., it can be extremely difficult to quantify the accuracy of the geologic model. All too often, interpreting geophysicists find themselves working with a seismic image that they know is state of the art, yet of lower resolution than needed, from which they still have to interpret the best possible geological model. It is in this context that Box s aphorism may seem particularly apt. Further complicating the situation, continual evolutionary improvements in the capabilities and performance of seismic data acquisition equipment similarly tend to increase the information content of more recent surveys and to raise the standard of detail of the resulting image, frequently causing the usefulness of older interpreted geologic models to fade over time. Not surprisingly, a similar relationship may be seen to pervade many aspects of seismic prospecting practice, albeit at much lower and more technical levels. In particular, within seismic data acquisition, the computational models employed in the determination of the time varying signal generated from land seismic sources may be seen to reflect similar issues of usefulness noted above for the larger field of view. The following is an examination of past and present models and concepts in those aspects of seismic data acquisition related to the modelling of the source/earth interactions. 1 AnalySeis LLC. * Corresponding author, spencer.rowse@analyseis.com 2016 EAGE 61

2 Methods and techniques Over the last three decades, the methods and techniques used to acquire good quality seismic data have changed dramatically as the definition of what constitutes good quality has changed. In the mid-eighties the seismic industry was transitioning from mainly recording 2D surveys with 120 channels to the newly introduced concepts and techniques of 3D acquisition. During this first decade, owing to equipment limitations, (cabled systems, number of recording channels, costs) 3D would best be described as recording a small number of widely separated orthogonal receiver and source lines using the 2D acquisition techniques developed over the previous decades to attenuate noise/increase signal. These techniques included long receiver arrays with geophone elements and source arrays with synchronized start times of 3-5 vibrators, stacking 8-16 sweeps per location and sequential recording. Due to rapid technological advances in recording equipment, distributed cable-less systems, the use of GPS for positioning and accurate timing, accompanied by decreases in cost per channel, recording equipment limitations need no longer be a concern in the design of surveys. In addition, the accuracy in positioning of sensors and fidelity in recording of the seismic data has also increased significantly over the last three decades. This increase in resolution during acquisition has been achieved by: n Reductions in bin size; n Transition from source and receiver arrays to point source point receivers; n Accurate timing and positioning due to GPS; n Increased source and receiver densities (number of shots/ sensors per km) leading to increased fold, azimuthal and offset distributions; n Increased numbers of recording channels; n 24/32 bit recording made possible by adoption of Delta Sigma recording; n Reduction in recording channel costs. Table (1) comparing the recording parameters of a state-ofthe-art acquisition crew circa 1990 to today. Arguably, the introduction of blended acquisition techniques (where two or more sources are operating with different start times) has had the largest impact on exploration over the last three decades. All the various methods of blended acquisition have step changes in productivity compared to sequential shooting techniques but introduce noise from overlapping signals that need to be removed (de-blended) before processing. Over the same period improvements to the vibrator source have also occurred. These improvements have mainly been in the mechanical performance of the source such as increased bandwidth and force output, reduction of harmonic distortion, stiffer baseplates, improved hydraulic and electronic controls and the development of non-linear and other exotic sweep designs. However, in the area of source signature estimation, the industry is still using Weighted Sum Ground Force (WSGF), introduced in the 1980s, as the output of the source/estimate of the propagating wavelet. In the marine environment accurate source signature estimates at every shot point have been common practice for decades due to an understanding of the physics involved in the interaction of the air gun bubble with the water and an environment that changes little from shotpoint to shotpoint. In the land environment this is not the case. The ground can change dramatically from location to location and is a major cause of signature change from shot to shot. However, our understanding of the physics of the interactions of the source with the ground has changed little over the last three decades. The following is an examination of the various past and present models used for the vibrator/earth system and suggestions for the future. Together with improvements in processing, these acquisition techniques have achieved greater resolution in the recorded data and improved interpretation. This is summarized in Bin size (m) 30 to 50 5 to 10 # channels ,000+ # sensor /array Source effort 4 vib 8-16 sweeps 1 vib 1 sweep Swp freq (Hz) Fold Shots per day ,000 Table 1 Comparison of state of art acquisition crew from 1990 and today. Figure 1 Representation of Rigid Plate on Half-space EAGE

3 Figure 2 Sallas s model of (a) hydraulic vibrator actuator, (b) mechanical ground impedance. Models Miller and Pursey s (1954) classic paper predated vibroseis by a few years. In this model, a vibrating rigid plate generates a surface force on an elastic, homogeneous, isotropic half-space, as depicted in Figure 1. As the half-space is elastic, no energy is absorbed and the far-field, down-going particle velocity components are proportional to the timederivative of the ground-force. Castenet and Lavergne (1965) patented an invention that relates to a complete vibrator controlling system by means of which the force imparted to the earth by the vibrator is kept proportional to the controlling signal, whatever the nature of the ground. In this patent is the first use of the formula now commonly referred to as ground force, where: M RM x a RM and M BP x a BP = resistance of the earth where M is mass, a is acceleration, the subscript BP denotes the base plate, the subscript RM denotes the reaction mass, and the resistance of the earth is synonymous with Ground Force. Lerwill (1979; 1981) is the first to use an electrical analogy to describe both the vibroseis mechanism and the earth and suggests that The real output of the vibrator is its drive force. Therefore, in principle, the phase compensation is more correctly referred to the reaction mass where, in any case, it provides the more stable phase lock. Following this publication, a debate ensued between Lerwell and Sallas in the literature about which of the different phase control methods, reaction mass (RM), Baseplate, (BP) or weighted sum ground force (WSGF), best represents the down going wavelet. In a 1984 paper, Sallas (1987) introduces WSGF as a means of vibrator control and states that Seismic vibrator control based on ground force provides an alternative to the baseplate acceleration method used in the past. It offers the advantage of producing a more consistent downhole wavelet than the other methods studied. Furthermore, the theory upon which ground force phase control is based predicts downhole wavelet characteristics which are consistent with empirical measurements. His model and equation for ground force are reproduced in Figure 5 and below: -F G = M RM x a RM + M BP x a BP where M is mass, a is acceleration, the subscript BP denotes the base plate, the subscript RM denotes the reaction mass, and FG is the Ground Force. This equation soon became the de-facto standard in the industry to control the motions of the source with the WSGF signal phase locked to the pilot signal. Martin and Jack (1990) examined the different phase control methods (BP, RM, and WSGF) under different drive levels to conclude: The ground force model has proved to be an inadequate method of controlling vibrator phase. Nonlinear effects introduced by the vibrator, and especially its interaction with the earth, should be included in the phase control system by means of a suitable model of the vibrator and the earth. Such a system would produce consistent source signatures, irrespective of drive level and the type of material beneath the baseplate. There is a need for improved quality control on Vibroseis surveys. Continuous monitoring of all the vibrators in the source array would help to improve data quality. Too little is known about Vibroseis behaviour. In last two decades there has been much discussion on WSGF, harmonic distortion and improved models as well as improvements in vibrator performance and their control systems. Most models now represent the vibrator mechanism and earth by a combination of springs and dampers and recognize that some portion of the earth (captured mass) must be incorporated into the model where captured mass is defined as the ground mass that participates in the 2016 EAGE 63

4 motion of the vibrator baseplate as it vibrates. Ziolkowski (Ziolkowski, 2008) describes a nonlinear zone beneath the baseplate. From outside of this nonlinear zone to the sensor, the propagating wavelet exhibits linear behaviour. Other terms such as earth filter or coupling are used to describe the difference between the GF and the observed signature. Additional investigations Experiments have been conducted (Shan et al., 2009) with the use of load cells (a piezoelectric device that is employed to convert a force (weight or compression) into an electrical signal) placed beneath the BP in order to determine the true ground force (direct measure of force rather than an estimate from conventional WSGF). In the same experiment accelerometers mounted in multiple locations on the BP were also used. Compared to the WSGF estimated from the standard accelerometer mounted at the centre of the BP, the WSGF changes with the positions of the accelerometer on the BP. While load cells are impractical for field use, they show discrepancies between the fundamental force and the estimated force from WSGF at higher frequencies. Both load cells and multiple accelerometers have shown that there is apparent flexure in the BP ( like a bird flapping its wings (Wei, 2015) in the words of one author), and casting doubt on the assumptions of many models of the BP being a rigid body. Due to the need for improved resolution in modern 3D surveys, various technologies have been introduced in the last ten years either to improve the performance of the vibrator or to estimate ground parameters. These include stiffening the BP to reduce flexure, improved hydraulic systems and controls, increasing the bandwidth of the vibrator at both low and high ends of the frequency sweep, and increasing the force output. Based on theories derived from civil engineering studies (Richart et al., 1970), the manufacturers of vibrator controllers can now derive estimates of ground viscosity and stiffness using the vibrator sensor signals. From these measurements, the ground velocities at each vibrator source location can be estimated (Al-Ali et al., 2003). Unfortunately, from real world tests (Ley, 2013) of different vibrator controllers, the results are inconsistent between different controllers at each location but follow the general trends in the changing surface geology. More recent attempts to improve the resolution of the seismic data have used some form of feedback from field data or predetermined model responses to either adjust the sweep parameters in the field for optimum performance, or to correct the input signal prior to processing. Zhukov (2013) analyzes a selected reflection window of the recorded field data of the n th sweep in real time and then uses a nonlinear sweep to adjust the output of the n th +1 sweep to compensate for signal attenuation or other criteria to mitigate the filter effect of near-surface heterogeneities at each source location. This methodology assumes that the radiation pattern from the vibrator is uniform over all azimuths and that the changes observed in the selected reflection window are owing entirely to changes in the surface conditions at the source. Bouchard and Ollivrin (2010) uses a digital model of the vibrator to compute ten states of the vibrator model related together by physical equations. From real time measurements of the various sensors and control mechanisms, the model parameters are constantly updated to account for changes due to variations in ground conditions and sweep frequencies. These updated measurements are then used to adjust the pilot sweep and provides a good compromise between accuracy and complexity to provide robust and fast control. The operator can select from two options for the estimated down-going signal. When operating the vibrator in raw mode, the normal WSGF is representative of the down going signal whereas in filtered mode an estimated GF derived from the model is used. Wei (2014) describes a method that uses the ratios of the accelerations of the RM and BP as the input-output frequency ratio of a complex vibrator model. From this ratio, he derives a frequency response of the system that is then compared to model data to obtain transfer functions for the vibrator mechanism, captured mass and coupling systems. Using the transfer functions and numerical analysis, values of stiffness and viscosity for the captured mass and coupling are obtained and a vibrator coupled ground filter constructed. This filter can then applied to either the Figure 3 Wei s vibrator coupled ground model EAGE

5 Figure 4 Transmissibility (ratio of input to output) of DHO for various values of damping. Figure 5 Phase difference between input and output of DHO for various values of damping. pilot or ground force signal and this filtered input signal correlated with the seismic data. While Bouchard s and Wei s papers describe different methods and models of the vibrator/earth that are impressive in their complexity, both are describing an interactive vibrator earth system where the earth resonance feeds back on to the vibrator mechanism/sensors and is incorporated into their models. However, there is no agreement on what best represents the down-going signal. Neither is there any independent testing to compare how accurate each model is at predicting the propagating wavelet. Most models now incorporate an earth component in the form of a mass, spring and damper, interacting with the vibrator mechanism. This is not a new concept. For more than four decades civil engineering studies (Richart et al., 1970) of the motion response of structures to earthquakes or strong winds have been conducted in order to mitigate damages to the structures (and people). In these models, a volume of ground beneath the structure is part of a damped harmonic oscillator (DHO) consisting of a mass, spring and damper attached to a support. The properties of this DHO, its natural frequency and damping, are determined by the mass of the structure, its area in contact with the ground ( footing ), and the ground properties of density, Shear modulus and Poisson s ratio. From these properties the response of the support to a known time varying force applied to the structure can be accurately determined. Figures 4 and 5 display the transmissibility and phase response of a DHO for various values of damping plotted against frequency ratio of input to natural frequency. Note that the response of any DHO system to a known time varying input signal can be fully described by a knowledge of the natural frequency and damping of the system. Transmissibility of the system is defined as either the ratio of the output force measured at the support to an input force applied to the mass or the ratio of the displacement of the mass to a displacement of the support. In Figure 4 the horizontal scale is shown as the ratio of input frequency to the natural frequency, f n (no damping present) of the system. For frequencies up to times the natural frequency, the transmissibility of the system is >1 for all values of damping and this region of amplified response is commonly called resonance. Note that the maximum resonance value varies with damping, being coincident with the natural frequency only at very low values of damping. For input frequencies >1.414fn for all values of damping, the transmissibility of the system is less than 1. In the civil engineering model, the spring and damper elements are attached to the support and mass of the structure. When a time varying force is applied to the mass, the forces that are induced on the support are the vector sum of the spring and damper forces. From Figure 5 the phase difference between the input force and the transmitted force depends on the ratio of input to natural frequency and damping, being ~0 for very low frequencies but ~180 degrees at high frequency ratios. Motions induced in the spring element cause compressions and rarefactions of the earth volume that act as a form of feedback on the footing and structure. In various areas of the world, the combination of earth properties and building design can result in a 2016 EAGE 65

6 Figure 6 Vibrator mechanism and interactive volume represented by spring, (k), and damper, (d), connected to support. system with very low damping that, unless external dampers are installed, will result in excessive motion or structural failure when subjected to forces such as earthquake or high winds. Conceptually, a vibrator can be considered as a structure in firm contact with a volume of earth as represented in Figure 6. The civil engineering models would suggest that as such, the vibrator and the volume of earth will interactively behave as a DHO system. Within the interactive volume of earth beneath the BP, motions of the earth particles are determined by the mass of the vibrator mechanism and earth properties as well as the static and time varying forces generated by the vibrator. Outside of this interacting volume, the motions of the earth can be described by normal wave equations and are dependent on the ground properties only. The boundary between the interactive volume and the surrounding undisturbed media, represented by the dashed line in Figure 6, is the support component of the DHO and, being remote from the BP, is a better representation of the propagating wavelet than surface measurements such as WSGF. Initial results from efforts to specifically model the output of the vibrator-earth interactive system from the perspective of a classical DHO system have been reported by Tinkle and Rowse (2010) and Heath and Rowse (2015), who noted good agreement between predicted (modelled) output and shallow downhole measurements at lower frequencies. Summary and conclusion The models of source signature prediction have progressed from the purely theoretical with ideal properties (rigid, elastic, homogeneous, isotropic) that are unlikely to be encountered in the real world, to complex systems com- prised of the vibrator mechanism interacting with a volume of earth in the vicinity of the BP. In these models, both the vibrator and earth mechanisms are represented by multiple masses, springs and dampers. Current thinking is that the time varying forces generated by the vibrator act on a nonrigid BP and pass through and are modified by an interacting volume of earth beneath the BP. Models have been used in attempts to derive this transfer function, filter or true ground force using the various sensor outputs that are located on the vibrator mechanism. These different terminologies are an acknowledgement that model estimates are still not proven to be accurate representations of the output force being transferred to the ground (far field) due to inadequate knowledge of the dynamic behaviour of the vibrator mechanism with the interacting volume of ground beneath the BP. In processing, to compensate for our lack of knowledge concerning the input signal, assumptions are made or obtained from other sources to estimate the input signal. These assumptions can be derived from well logs, statistical estimation from seismic data, assuming minimum phase wavelets or relying on the experience of the processor. Some processors have expressed the view that lack of an accurate input signal is unimportant as variations in input signal all get processed out during normal processing operations. However, if accurate estimates of the source signature are unknown and assumptions are made in order to perform seismic deconvolution, how accurately will the de-convolved traces estimate the Earth s geology? With an accurate estimate of the input signature, improvements in the accuracy and resolution of processed data should be obtained. Other applications for an improved wavelet estimate include source QC (quality control), deconvolution, multiples attenuation, tying reflection data to wells, modelling and inversion, AVO analysis, reservoir monitoring (4D, for example, for CO 2 sequestration monitoring) and the analysis of multi-component recordings. For the last 50 years, the assumption has been that the far field down-going particle velocity components are proportional to the time-derivative of the ground force. Most recent models now incorporate some form of interacting volume or filter between the base plate and the propagating wavelet in the far-field. Clearly, this marks a significant departure from the original concept of ground force. Vibroseis appears to have advanced mainly by a series of small evolutionary steps but the process has not been inclusive of all aspects of this complex source. Recent work suggests that the classic models of the force output of the source are being overtaken by that evolution, and in terms of Box s aphorism, may be in danger of being rendered less than useful. The time appears ripe for more inclusive research into an updated understanding of the underlying [geo]physics EAGE

7 References Al-Ali, M., Hasting-James, R., Makkami, M., and Korvin, G. [2003] Vibrator attribute leading velocity estimation. The Leading Edge, 22 (5), Bouchard, D. and Ollivrin, G. [2010] Developments in vibrator control, Geophysical Prospecting, 58, Box, G.E.P. and Draper, N.R. [1987] Empirical Model-Building and Response Surfaces. Wiley, New York, USA. Castenet, A. and Lavergne, M. [1965] Vibrator controlling system. US Patent 3, 208, 550. Heath, R. and Rowse, S. [2015] Enhanced Vibroseis: The next step in improving land 2D-4D. ASEG-PESA Annual Meeting, Poster. Lerwill, W.E. [1979] Seismic Sources on Land, Developments in Geophysical Exploration Methods. Applied Science Publishers, London. Lerwill, W.E. [1981] The Amplitude and Phase Response of a Seismic Vibrator. Geophysical Prospecting, 29, Ley, R., Adolfs, W., Bridle, R., Al-Homaili, M., Vesnaver, A. and Ras, P. [2006] Ground viscosity and stiffness measurements for near surface seismic velocity. Geophysical Prospecting, 54 (6), Martin, J.E. and Jack, L.G. [1990] The behaviour of a seismic vibrator using different phase control methods and drive levels. First Break, 8 (11), Miller, G.F. and Pursey, H. [1954] The field and radiation impedance of mechanical radiators on the free surface of a semi-infinite isotropic solid. Proceedings of the Royal Society A, 223, Richart, F.E., Woods, R.D. and Hall, J.R. [1970] Vibrations of Soils and Foundations. Prentice Hall, New Jersey, USA. Sallas, J.J. [1984] Seismic vibrator control and the downgoing P-wave. Geophysics, 49 (6), Shan, S., Eick, P.M., Brewer, J.D., Zhu, X. and Shaw, S.A. [2009] Load Cell System Test Experience: Measuring the Vibrator Ground Force on Acquisition, SEG Annual Meeting, Expanded Abstracts. Tinkle, A and Rowse, S. [2010] Toward a simplified model of vibrator seismic source performance: preliminary results. SEG Annual Meeting, Expanded Abstracts. Wei, Z, [2014] Seismic Data Filtering Based on Vibrator-Coupled Ground Model, US 8, 909, 480. Wei, Z. [2015] Improving Vibroseis Data Quality with the Vibratorground Model. 77 th EAGE Conference & Exhibition, Extended Abstracts. Zhukov, A. [2013] The adaptive vibroseis technology: hardware, software and outcomes. SEG Annual Meeting, Expanded Abstracts. Ziolkowski. A [2008] Vibroseis Processing for Truer Amplitudes: Do we dare to replace cross-correlation by deconvolution? Vibroseis Workshop, Abstracts. EAGE Workshop on Velocities: Reducing Uncertainties in Depths April 2016 Kuala Lumpur, Malaysia Programme highlight A strong technical programme is being put together; currently featuring Keynote speakers Dr Mac Al Chalabi (PetroTech), Etienne Robein (ex-total), Dr Eric Verschuur (Delft University of Technology) and Dr. Tariq Alkhalifah (Kaust University). Abstracts are expected from King Abdullah University of Science and Technology (KAUST), Kyoto University, National University of Singapore (NUS), University of Western Australia (UWA), Shell, Woodside, PETRONAS, INPEX, Schlumberger, CGG, Petroleum Geo-Services (PGS), Paradigm, DownUnder GeoSolutions (DUG), Ikon Science and other industry players. Short course April 2016 Seismic Velocities and Depth Conversion Instructor: Mac Al Chalabi Register now! 2016 EAGE 67