Advanced Noise Reduction in Acoustic Well Logging

Abstract

A method and product for reducing or eliminating noise in waveforms for the purpose of improving the subsequent acoustic log processing products. The present disclosure identifies the noise signature and removes the noise signature from the formation signal through a process of deconvolution. This product and method is particularly effective where the noise signal overlaps in the time domain and in the frequency domain with the formation signal.

Claims

1. A method for reducing or eliminating noise in near wellbore acoustic waveforms, the method comprising steps of: receiving raw acoustic data from a tool; identifying a noise spectrum from said raw acoustic data; isolating said noise spectrum by muting the said raw acoustic data outside of the time window in which the noise spectrum is dominant thereby creating a noise waveform operator; constructing an initial deconvolution operator using said noise waveform operator, repeating said noise waveform operator until said noise waveform operator and repeated noise waveform operator together is of the same parameters as said raw acoustic data; multiplying said initial deconvolution operator by a function to create a final deconvolution operator; deconvolving said final deconvolution operator from said raw acoustic data to obtain a waveform with reduced noise.

2. The method as claimed in claim 1 wherein identifying said noise waveform operator further comprises determining a wave shape by viewing frequency parameters and time parameters

3. The method as claimed in claim 2 wherein the said final deconvolution operator has the same time window as said raw acoustic data.

4. The method as claimed in claim 3 wherein said function is a mathematical function dampening amplitude of said noise waveform operator at a determined rate of decline but not changing phase of said noise waveform operator.

5. The method claimed in claim 4 further comprising: muting a white noise waveform from said noise waveform after creating said initial deconvolution operator; and adding said white noise waveform back to said initial deconvolution operator prior to deconvolving said raw waveform.

6. A Computer Program Product (CPP) for reducing noise in near wellbore acoustic waveforms, the CPP executable by a processor and tangibly embodied in a computer readable medium and containing instructions that, when executed by the processor, cause the processor to perform operations to reduce noise in near well-bore acoustic waveforms, the operations comprising: receiving raw acoustic data from a source collected near wellbore acoustic waveforms; identifying a noise spectrum from said raw acoustic data; isolating said noise spectrum by muting the said raw acoustic data outside of the time window in which the noise spectrum is dominant thereby creating a noise waveform operator; constructing an initial deconvolution operator using said noise waveform operator, repeating said noise waveform operator until said noise waveform operator and repeated noise waveform operator together is of the same parameters as said raw acoustic data; multiplying said initial deconvolution operator by a function to create a final deconvolution operator; deconvolving said final deconvolution operator from said raw acoustic data to obtain a waveform with reduced noise.

7. The CPP as claimed in claim 6 wherein identifying said noise waveform operator further comprises determining a wave shape by viewing frequency parameters and time parameters.

8. The CPP as claimed in claim 7 wherein the said final deconvolution operator has the same time window as said raw acoustic data.

9. The CPP as claimed in claim 8 wherein said function is a mathematical function dampening amplitude of said noise waveform operator at a determined rate of decline but not changing phase of said noise waveform operator.

10. The CPP claimed in claim 9 further comprising: muting a white noise waveform from said noise waveform after creating said initial deconvolution operator; and adding said white noise waveform back to said initial deconvolution operator prior to deconvolving said raw waveform.

11. A system for reducing noise in near wellbore acoustic waveforms, the noise reduction system comprising of: a plurality of sensors which are configured to sense acoustic waves corresponding to depth locations within a borehole; a processor; a plurality of program-memory locations coupled to the processor; a plurality of data-memory locations coupled to the processor; and a display device configured to be controlled by the processor, wherein the plurality of program-memory locations contains instructions that, when executed by the processor, cause the processor to perform operations comprising: receiving raw acoustic waveform data obtained from the plurality of sensors, the received raw acoustic waveform data correspond to sensed near well-bore acoustic waveforms transmitted from a source travelling through a subterranean formation; identifying a noise spectrum from said raw acoustic data; isolating said noise spectrum by muting the said raw acoustic data outside of the time window in which the noise spectrum is dominant thereby creating a noise waveform operator; constructing an initial deconvolution operator using said noise waveform operator, repeating said noise waveform operator until said noise waveform operator and repeated noise waveform operator together is of the same parameters as said raw acoustic data; multiplying said initial deconvolution operator by a function to create a final deconvolution operator; deconvolving said final deconvolution operator from said raw acoustic data to obtain a waveform with reduced noise.

12. The system as claimed in claim 11 wherein identifying said noise waveform operator further comprises determining a wave shape by viewing frequency parameters and time parameters.

13. The system as claimed in claim 12 wherein the said final deconvolution operator has the same time window as said raw acoustic data.

14. The system as claimed in claim 13 wherein said function is a mathematical function dampening amplitude of said noise waveform operator at a determined rate of decline but not changing phase of said noise waveform operator.

15. The system claimed in claim 14 further comprising: muting a white noise waveform from said noise waveform after creating said initial deconvolution operator; and adding said white noise waveform back to said initial deconvolution operator prior to deconvolving said raw waveform.

Description

BRIEF DESCRIPTION OF DRAWINGS

[0012] The following detailed description of embodiments of the invention will be better understood when read in conjunction with the accompanying drawings. For the purpose of illustrating the invention, there is shown in the drawings, embodiments which are presently preferred. In the drawings, the left-most digit(s) of a reference number indicates the drawing in which the reference number first appears. The same reference numbers have been used throughout the drawings to indicate similar elements of the drawings.

[0013] FIG. 1 illustrates a raw waveform data from a receiver 1, in accordance with an embodiment of the present invention;

[0014] FIG. 2 illustrates an example of the raw waveform data and its corresponding frequency spectrum from a receiver 1, in accordance with the embodiment of the present invention;

[0015] FIG. 3 illustrates the isolated noise spectrum from the receiver data shown in FIG. 1, in accordance with the embodiment of the present invention;

[0016] FIG. 4 illustrates a noise waveform minus the static noise from the receiver data should in FIG. 1, in accordance with the embodiment of the present invention;

[0017] FIG. 5 illustrates the deconvolved raw waveform, in accordance with the embodiment of the present invention;

[0018] FIG. 6 illustrates the flowchart of the method outlined in the present invention, in accordance with the embodiment of the present invention.

DETAILED DESCRIPTION

[0019] For a thorough understanding of the present disclosure, reference is to be made to the following detailed description in connection with the above-mentioned drawings. Although the present disclosure is described in connection with exemplary embodiment, the present disclosure is not intended to be limited to the specific forms set forth herein. It is understood that various omissions and substitutions of equivalents are contemplated as circumstances may suggest or render expedient, but these are intended to cover the application or implementation without departing from the spirit or scope of the present disclosure. Further, it will nevertheless be understood that no limitation in the scope of the disclosure is thereby intended, such alterations and further modifications in the figures and such further applications of the principles of the disclosure as illustrated therein being contemplated as would normally occur to one skilled in the art to which the disclosure relates. Also, it is to be understood that the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. Further, reference herein to one embodiment or an embodiment means that a particular feature, characteristic, or function described in connection with the embodiment is included in at least one embodiment of the disclosure. Furthermore, the appearances of such phrase at various places herein are not necessarily all referring to the same embodiment. The terms a and an herein do not denote a limitation of quantity, but rather denote the presence of at least one of the referenced item.

[0020] The present disclosure provides a product and method for reducing a noise signal or noise signals in near-wellbore acoustic datasets. The present disclosure is particularly used in connection with hydrocarbon exploration and extraction. Acoustic logs can be used at a number of stages in the exploration, extraction, and development process. For example, prior to fully developing a field, seismic and acoustic logging data is collected on the field. This data is then used to create reservoir models, determine the cost-effectiveness of exploiting the field, and other data-driven models used in connection with planning and developing fields. Finding and producing hydrocarbons efficiently and effectively requires an understanding of the formation characteristics, such as the hardness or softness of the formation, and an accurate idea of the depths and locations of geological properties and layers. The acoustic logging waveform data provides information about the stresses within the subterranean formation. Also, importantly, sonic and seismic data is collected and shot in the time domain with no initial, direct correlation to depth. After processing the acoustic logs, the tie from time to depth is computed, which is necessary to map out the subterranean formation and the reservoir accurately. These previous two mentioned uses are intermediate products derived from raw acoustic waveform data. Accurate maps and reservoir information are essential in placing wells, creating a field development plan, and in determining the size and scope of the reservoir. The present disclosure relates to improving the accuracy of any processing done with near-wellbore acoustic, or sonic, wave measurements and datasets by reducing or removing noise in the raw dataset. Where raw dataset, is the waveform data prior to processing, filtering, or noise reduction, designated as .sub.raw.

[0021] In seismic or seismology data processing, a variety of techniques have been developed and are routinely used to reduce noise in datasets or to extract relevant signals from noise due to the many environmental factors affecting the data collected. Acoustic logging, as an industry, however, relies primarily on frequency filtering to reduce noise and amplify relevant signals. In more noisy waveforms collected, acoustic log processing experts also rely on the addition of filtering in the time domain, also known as muting. More advanced methods have been seen as unnecessary in the past due to the apparent stability of the wellbore, the confined environment, and the small-scale measurement taking place. While the reliance on frequency filtering and time domain filtering provides sometimes passable results, the unique challenges brought by unconventionals, the new tools that take measurements while drilling, the often more rugged geologic formations, and the prevalence of directional drilling in unconventional hydrocarbon exploration have created a need for more effective filtering and noise reduction techniques. Additionally, the economic margins for hydrocarbon extraction have become smaller, making it more imperative to make use of as much of the data as possible and tool advancements have also created more sensitive instrumentation that also picks up more noise as well as more relevant signal. Therefore, there is a need in the industry to adopt more effective filtering techniques leading to more effective noise reduction.

[0022] As briefly mentioned above, the current industry standard is to do one of the three following things with raw, sonic log data, where sonic and acoustic are often used interchangeably in the industry, after it has been collected: 1) raw data which has no filtering; 2) data which has undergone frequency domain filtering; and 3) the most advanced current method, data which has undergone time domain and frequency domain filtering. These techniques, however, fall short when encountering a noise signal that overwhelms the formation signal or when encountering noise signals that overlap in time and frequency with the formation signal. The noise signal often arrives preceding or dominates the waveform data after (anteceding) the formation signal. These windows in which the noise signal is dominant or the only present signal, provides an opportunity to isolate the noise signal and study its characteristics in isolation from the formation signal. The noise signal reverberates causing interference and often partially or completely masks the formation signal, overlapping with the formation signal in both time domain and in frequency domain. This overlap makes it more difficult to distinguish and separate the formation and noise signals. The present disclosure provides a method and product for reducing or eliminating noise in these cases and presents an alternative to the other methods used.

[0023] For the noise reduction system, product and method, before any processing occurs, the noise must be defined and identified to distinguish it from the signal. We defined the noise as relatively unresponsive to the formation (monochromatic waveform), meaning that although the formation signal would change throughout the dataset and interval in response to the formation, the noise signal would be relatively unresponsive because it was being caused by static factors in the wellbore, such as the tool, or was not affected by the formation, or was not traveling through the formation. The frequency spectrum of the full, raw waveform was studied to help identify the noise, as seen in FIG. 2. This was supported by studying a broad variety of datasets and in the literature previously published in the field. Since the noise is relatively unaffected by the formation, the noise should appear as a strong forerunner when viewing the waveforms in the time domain (as seen in FIG. 1), as it reaches the tool receivers before the formation signal. This noise, since relatively unaffected by the more dynamic formation, retains the same waveform shape or character throughout the time domain and throughout the measured interval. This is in contrast to the signal of interest, which is much more responsive to the formation and change throughout the logged interval. In some cases, the noise waveforms reverberate and dampen with time as they are reflected and refracted off of the wellbore walls. In these cases, although the waveforms retain their shape or wavelength, they lessen in amplitude with time. Regardless of the rate of decline of the amplitude of these noise waveforms, any attempt to reduce the effect of noise waveforms must, not only remove or reduce the effect of the initial arrival of the noise signal, but also the subsequent, dampened noise overlapping with the formation signal. The noise waveform can be identified by choosing from a library of noise signatures based on, for example, the tool or formation; by identifying the unique noise signature for that dataset; or by constructing a set of parameters to mathematically identify the noise signature.

[0024] Once the noise waveform (.sub.noise) has been defined and identified, the identified noise signal and the reverberating dampened signal must be isolated. Enough of the noise waveform must be isolated to sub construct a deconvolution operator composed of the noise and the dampened signal, to then remove the noise signal or reduce its effect on the signal of interest while retaining as much of the signal of interest's (.sub.formation) character as much as possible. Of note, the aim of known predictive deconvolution routines is to extract the signal shape from the noise surrounding it. In acoustic logging, by contrast, we do not have a predetermined shape for our formation and so cannot use these known predictive deconvolution routines. While we could assume a predetermined shape, this would not be ideal and it is preferable to leave the formation signal as unaltered as possible to achieve the most accurate processed results. That said, from our research and experimentation in identifying the noise signal, we determined that the noise signal is of predetermined shape, where predetermined means that it can be determined from a particular dataset or for a category of datasets. Therefore, to begin the process, we isolated .sub.noise. .sub.noise was isolated in time as seen in FIG. 3 since, as stated above, .sub.noise arrives earlier in time than the formation signal.

[0025] As we have determined our basic method, this is the overarching equation:

.sub.formation=.sub.raw+.sub.decon

where, .sub.formation is the final, deconvolved waveform that reflects the formation signal, where .sub.raw is the initial, unfiltered waveform,

.sub.decon=.sub.static(.sub.noise.sub.static)+(fcnX)*(.sub.noise.sub.static)

where fcnX is a function to account for the change in amplitude of .sub.noise over time, and where .sub.static is the static prior to the noise waveform.

[0026] In the raw waveform (.sub.raw), there is also pure static that occurs prior to the onset of the recorded waveform. This must be removed, or muted, to construct the deconvolution operator (.sub.decon), as pictured in FIG. 4. The pure static is also called fore-noise. This may be muted prior to construction of the operator so that any computer products or algorithms do not factor the static into the computations below; however, alternative embodiments may not need this static to be muted while constructing the deconvolution operator.

[0027] Now that the .sub.noise and .sub.static have been identified and isolated, we must construct the remainder of the deconvolution operator, which consists of the dampening noise waveform. This dampening noise waveform must be added to the initial, isolated noise waveform. To compute the dampening noise waveform, the initial noise waveform will be repeated until the desired matrix size or waveform parameters are achieved and the initial noise waveform will be multiplied by a function that dampens the initial noise waveform over time window desired, beginning with the function not dampening the initial noise waveform at all and ending with the total dampening of the initial noise waveform.

[0028] As noted above, in some cases, the noise reverberates in time; therefore, the dampened noise waveform will retain the shape of .sub.noise but lessen in amplitude through the time domain. This dampening can be defined as a function, either a statistical function based on something like a Gaussian distribution or can be a step function, by which .sub.noise is multiplied by. As more time elapses in the recording window, the function multiplied by .sub.noise will approach zero. The rate of decline may be algorithmic or may be linear, and the preference depends on the particular dataset or the type of noise in question. For example, where the aim is to remove noise signal associated with a casing in a well, that noise signal will most likely exhibit a more linear, steep rate of decline whereas noise associated with an irregular wellbore will most likely exhibit a logarithmic, less steep rate of decline.

[0029] The initial noise waveform with the dampened noise waveforms must be added to the initial static, if muted earlier, until the resulting operator (.sub.decon) is of equal parameters as of the original, raw waveform. The main parameter is the time window. The waveforms may be viewed as a matrix, so the raw waveform matrix and the constructed deconvolution operator matrix are of the same size. Now, that we have our matrix of predetermined shape consisting of purely the noise signal we can perform a deconvolution routine to remove the noise signal from the raw waveform.

[0030] The resulting waveform, which is our final, deconvolved waveform, has noise eliminated or reduced and should be pure formation signal, or nearly pure formation signal, as seen in FIG. 5. The validity of the deconvolved waveform can be verified using well-known processing techniques such as semblance or cross-correlation. The strength of the correlation and the comparison of the results with similar computations from other datasets or measured parameters allow for determination of accuracy.

[0031] FIG. 6 describes the method for reducing noise in collected near well-bore acoustic waveforms:

At S1 the process starts;
At S2 raw acoustic waveform data received from a tool traveling through a subterranean formation is received;
At S3 the noise spectrum is identified within the received raw acoustic log data;
At S4 the noise signal is isolated in time, where the noise signal may be specific to the formation, to the tool, or to the particular dataset;
At S5 the static or forenoise is muted or discounted from the isolated noise signal, where the static is simply white noise, for the purpose of creating the noise waveform operator;
At S6 the noise waveform is repeated until the static and the repeated noise waveforms together are of the same parameters as the raw acoustic waveform;
At S7 the repeated noise waveforms are multiplied by a function for the purpose of creating a rate of decline, where the rate of decline starting with having no dampening effect on the noise waveforms to having a complete dampening effect on the noise waveform;
At S8 the raw waveform is deconvolved using the constructed deconvolution operator to achieve a waveform with reduced or eliminated noise; and
At S9 the process ends.

[0032] Preferred embodiments are described herein, including the best mode known to the inventor. Of course, variations of those preferred embodiments will become apparent to those of ordinary skill in the art upon reading the foregoing description. The inventor expects skilled artisans to employ such variations as appropriate. Moreover, any combination of the above-described elements in all possible variations thereof is encompassed unless otherwise indicated herein or otherwise clearly contradicted by context.