Patent Application
20230228897
The present disclosure relates generally to oil field monitoring. In particular, but not by way of limitation, the present disclosure relates to systems, methods and apparatuses for spectral analysis of acoustic signals associated with drilling and completions operations.
Unconventional reservoirs include reservoirs such as tight-gas sands, gas and oil shales, coalbed methane, heavy oil and tar sands, and gas-hydrate deposits. These reservoirs have little to no porosity, thus the hydrocarbons may be trapped within fractures and pore spaces of the formation. Additionally, the hydrocarbons may be adsorbed onto organic material (e.g., of a shale formation). In some cases, these reservoirs may require special recovery operations distinct from conventional operating practices in order to mobilize and extract the oil.
The rapid development of extracting hydrocarbons from these unconventional reservoirs can be tied to the combination of horizontal drilling and induced fracturing (called “hydraulic fracturing” or simply “fracking”) of the formations. Horizontal drilling has allowed for drilling along and within hydrocarbon reservoirs of a formation to better capture the hydrocarbons trapped within the reservoirs. In some cases, increasing the number of fractures in the formation and/or increasing the size of existing fractures through fracking may serve to increase mobilization.
In some cases, modern drilling and fracturing operations may utilize a perforating gun to perforate oil and gas wells in preparation for production. Perforating guns may perforate a well's casing and surrounding rock to form tunnels via several shaped explosive charges. In some cases, these tunnels can later be expanded via high pressure fluids. Perforating guns may be controlled via a wireline (or electric line), wherein the wireline may be used to lower and raise the perforating guns (or control their horizontal position in a horizontal section of a well), as well as control firing of the charges therein.
In some circumstances, wireline sticking (or simply “sticking”) is identified as the difficulty of the wireline movement either in the upward or downward direction. Wirelines are prone to sticking for a number of reasons, including, but not limited to, a cave-in of the borehole above the drill bit; drill-cuttings settling within the borehole due to not being carried away properly; turning radius issues in deviated boreholes; adhesion of the drill string or wireline based on a lack of movement; high friction between the drill pipe and the borehole walls; or differential sticking caused by higher drilling fluid pressure than formation pressure. In a complete wireline sticking situation, neither circulation nor wireline movement may be possible. Complete sticking not only delays production as fishing operations are carried out to free the wireline and/or perforating gun, but in some cases the gun and stage of the well may have to be abandoned. Such sticking can occur as often as every 45 days and can cost an operator hundreds of thousands of dollars per jam, thus constituting a significant annual cost to fracking operations. In some areas, events related to sticking can be responsible for as much as 40% of the total well cost.
Thus, there exists a need for accurately predicting wireline sticking events to minimize downtime and optimize fracking and drilling operations.
The following presents a simplified summary relating to one or more aspects and/or embodiments disclosed herein. As such, the following summary should not be considered an extensive overview relating to all contemplated aspects and/or embodiments, nor should the following summary be regarded to identify key or critical elements relating to all contemplated aspects and/or embodiments or to delineate the scope associated with any particular aspect and/or embodiment. Accordingly, the following summary has the sole purpose to present certain concepts relating to one or more aspects and/or embodiments relating to the mechanisms disclosed herein in a simplified form to precede the detailed description presented below.
Wireline sticking is a significant cause of concern during fracking and drilling operations. This concern has been identified and dated back to at least the 1940s (see Warren, J. E. 1940. Causes, Preventions, and Recovery of Stuck Drill Pipe. API-40-030), and yet, there is a long-felt unmet need for accurately predicting wireline sticking. Besides the significant cost, wireline sticking events also hamper production due to the delays associated with fixing such sticking events. In some extreme cases, the gun and stage of a drilled well may have to be abandoned. Wirelines are prone to sticking for a number of reasons, including, but not limited to, a cave-in of the borehole above the drill bit; drill-cuttings settling within the borehole due to not being carried away properly; turning radius issues in deviated boreholes; adhesion of the drill string or wireline based on a lack of movement; high friction between the drill pipe, perf gun, and/or plug with the borehole walls; differential sticking caused by higher drilling fluid pressure than formation pressure. In some other cases, proppant sands from a previous stage of fracking may remain in the wellbore, which may increase friction on the wireline as it pushes the perforation gun and plug down for a next stage. In yet other cases, differential pressures on the wireline, or pressure difference between the hydrostatic pressure and the formation pressure, can also slow or halt wireline movement.
Wireline sticking rarely occurs instantaneously; rather, the stuck wireline event is in most cases preceded by changes in one or more wireline parameters that foreshadow a stuck wireline event, with changes occurring sometimes minutes or even hours prior to the stuck wireline event. For example, increasing torque, decreasing drill string revolutions per minute (RPM), and/or decreasing drilling fluid flow may be indicative of an upcoming stuck wireline event.
While some attempts have been made to identify characteristics of static well pressure, which is used as one indicator of wireline sticking, these attempts are not only difficult to read, but also lacking in their ability to predict wireline friction. For instance, WEATHERFORD uses hydraulics and torque-and-drag software to determine deviation of real-time data from a real-time model as well as trend analysis of real-time data, with parameters such as pump pressure, flow rate, torque, rotary speed, hookload and drag, and weight on bit (see Salminen, et al. STUCK-PIPE PREDICTION BY USE OF AUTOMATED REAL-TIME MODELING AND DATA ANALYSIS. September 2017. Society of Petroleum Engineers, SPE Drilling & Completion).
In some other cases, statistical analytics of previous sticking events have been used to predict wireline sticking (see U.S. Pat. No. 9,970,266; see also Weakley, Use of Stuck Pipe Statistics To Reduce the Occurrence of Stuck Pipe, Sep. 23-26, 1990, Society of Petroleum Engineers, SPE Annual Technical Conference and Exhibition), as well as artificial neural network analytics of existing data have been used (Jahanbakhshi et al, Intelligent Prediction of Differential Pipe Sticking by Support Vector Machine Compared With Conventional Artificial Neural Networks: An Example of Iranian Offshore Oil Fields, December 2012, Society of Petroleum Engineers, SPE Drilling & Completion) but these are far from real-time. At least one group has used linear regression based on “training wells” to build models that can be used to predict wireline sticking based on a current hook load value for a given bit depth (U.S. Patent Application No. 2017/0306726). Landmark Graphics Corp. has used real-time measurements of wireline hookload moving averages to short interval hookload moving average to identify sticking events (U.S. Pat. No. 10,436,010).
In yet other cases, the prior art has focused on avoiding, rather than predicting, sticking. For instance, Wheater describes mechanical means of avoiding sticking through the use of standoffs which help to relieve wireline pressure against sidewalls of the borehole, and thereby reduces cable key-setting, high wireline cable drag, and friction that could lead to sticking (see U.S. Pat. No. 10,066,449). However, mechanical means of avoiding sticking are not guaranteed, and may merely decrease the likelihood of sticking, rather than eliminate it completely. Additionally, in some circumstances, mechanical means may also increase the cost and complexity of the system due to the use of non-standard components/parts and/or additional parts.
The novel embodiments described herein are directed to assisting drilling operators in avoiding stuck wireline events by providing an indication of a likelihood of a future stuck wireline event sufficiently far in advance that one or more corrective measures may be taken. Embodiments are also directed to predicting and preventing stuck wireline events by identifying increased wireline friction associated with imminent wireline sticking.
In one aspect of the disclosure, acoustic or vibration data can be monitored at the wellhead, a circulating fluid line, or a standpipe of the well, and when vibrations in the time domain exceed a threshold, an indicator can be returned to the operator or a controller can adjust fracking operation parameters. The threshold can be an average amplitude or a rate of increase in acoustic or vibration signals. In a second example, spectral analysis of acoustic or vibration signals in a well may help operators in identifying component failures. Additionally or alternatively, spectral analysis may allow operators to preemptively remove or replace a component from operation before failure. In a third example, analyzing acoustic or vibration data in the frequency domain may provide operators with insight on casing wear, which may allow them to identify excessive casing wear.
In a fourth example, acoustic or vibration signals can be monitored at the wellhead, a circulating fluid line, or a standpipe of the well, and when vibrations in the time domain exceed a threshold and specific spectral signals are identified in the frequency domain of the vibration data, an indicator can be returned to the operator or a controller can adjust fracking operation parameters. The threshold can be an average amplitude or a rate of increase in acoustic or vibration signals.
In a fifth example, acoustic or vibration signals can be monitored at the wellhead, a circulating fluid line, or a standpipe of the well, and when specific spectral signals are identified in the frequency domain, an indicator can be returned to the operator, or a controller can adjust fracking operation parameters. The threshold can be an average amplitude or a rate of increase in acoustic or vibration signals.
Some embodiments of the disclosure may relate to a system for preventing wireline sticking during hydraulic fracturing operations, the system comprising: a sensor coupled to a fracking wellhead, circulating fluid line, or standpipe of a well and configured to convert acoustic vibrations measured in fracking fluid in the fracking wellhead, circulating fluid line, or standpipe into an electrical signal in a time domain; a memory configured to store the electrical signal; a converter configured to access the electrical signal from the memory and convert the electrical signal in a window of time into a current frequency domain spectrum; a machine-learning system configured to classify the current frequency domain spectrum as associated with increasing wireline friction, the machine-learning system trained on previous frequency domain spectra measured during previous wireline operations and previously classified by the machine-learning system; and a user interface configured to return an indication of the increasing wireline friction to an operator of the hydraulic fracturing operations.
Some other embodiments of the disclosure may relate to a method of preventing wireline sticking during hydraulic fracturing operations, the method comprising: providing a sensor coupled to a wellhead, circulating fluid line, or standpipe of a well and configured to convert acoustic vibrations in fracking fluid in the wellhead, circulating fluid line, or standpipe into an electrical signal in a time domain; recording the electrical signal to a memory; converting the electrical signal in the memory for a window of time to a current frequency domain spectrum comprising an amplitude spike at one or more frequencies; analyzing the current frequency domain spectrum via a machine-learning system trained on previous frequency domain spectra measured during previous wireline operations and previously classified by the machine-learning system; classifying the current frequency domain spectrum as associated with increased wireline friction; and returning an indication of the increasing wireline friction to a well operator.
In yet other embodiments, the disclosure may relate to a method of preventing wireline sticking, the method comprising: starting a wireline operation on a fracking stage of a well; measuring acoustic vibrations in fracking fluid in a wellhead, circulating fluid line, or standpipe of the well; converting the acoustic vibrations into an electrical signal in a time domain; recording the electrical signal to a memory; converting the electrical signal in the memory for a window of time to a current frequency domain spectrum comprising an amplitude spike at one or more frequencies; analyzing the current frequency domain spectrum via a machine-learning system trained on previous frequency domain spectra measured during previous wireline stages and previously classified by the machine-learning system; classifying the current frequency domain spectrum as associated with increased wireline friction; and adjusting a parameter of the wireline operation based on the increased wireline friction.
These and other features, and characteristics of the present technology, as well as the methods of operation and functions of the related elements of structure and the combination of parts and economies of manufacture, will become more apparent upon consideration of the following description and the appended claims with reference to the accompanying drawings, all of which form a part of this specification, wherein like reference numerals designate corresponding parts in the various figures. It is to be expressly understood, however, that the drawings are for the purpose of illustration and description only and are not intended as a definition of the limits of the disclosure. As used in the specification and in the claims, the singular form of ‘a’, ‘an’, and ‘the’ include plural referents unless the context clearly dictates otherwise.
Various objects and advantages and a more complete understanding of the present disclosure are apparent and more readily appreciated by referring to the following detailed description and to the appended claims when taken in conjunction with the accompanying drawings:
The word “exemplary” is used herein to mean “serving as an example, instance, or illustration.” Any embodiment described herein as “exemplary” is not necessarily to be construed as preferred or advantageous over other embodiments.
Preliminary note: the flowcharts and block diagrams in the following figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, some blocks in these flowcharts or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The present disclosure relates generally to oil field monitoring. In particular, but not by way of limitation, the present disclosure relates to systems, methods and apparatuses for spectral analysis of acoustic or vibration signals received at a well head.
Sources of Acoustic and/or Vibration Pressure and Acoustic/Vibration Sensors
Existing pressure sensing techniques for oil field monitoring involve recording pressure changes (e.g., absolute changes over long periods of time) with reference to an absolute pressure of fluid in the well. However, currently used static pressure sensors usually have a slow sample rate (e.g., slower than 1 Hz) and provide very little signal amplitude when indications of wireline friction arise. Given the high cost of delays and stage abandonment that can result from wireline sticking, there is a need for systems and methods that can more accurately predict wireline sticking and do so with greater lead time to sticking events.
In some cases, analyzing fluctuations or vibrations in the fluid in a well (e.g., fracking fluid) in a frequency domain, rather than a time domain, may serve to provide a more accurate understanding of wireline sticking. In some cases, this analysis may involve acquiring dynamic acoustic or vibration pressure data from the well's fluid (e.g., fracking fluid) and converting it into a frequency spectrum or frequency domain. In some circumstances, the analysis can focus on repeating patterns, which may have a better correlation to underground events, and may travel through and be more easily discerned through thousands of feet of rock and sand formations, than one-off changes in absolute pressure (events that often take hours to register). In some cases, this spectrum may also be referred to as an acoustic or vibration spectral frequency signature (or frequency signature). In some embodiments, the analysis may comprise generating machine learning (ML) models, or other artificial intelligence (AI) models, and training the models to recognize the acoustic or vibration signatures of different events. One non-limiting example of an event may comprise identifying increasing wireline friction or wireline sticking. Once models have been trained to recognize the acoustic or vibration signatures of different wireline events, real-time acquired data may be compared to the model or analyzed by the model for real-time assessment of wireline friction or sticking. Current techniques involve operators making wireline decisions based on slow feedback parameters such as speed of descent and tension on the wireline, parameters that lag wireline events by significant amounts over indications available when acoustic or vibration signals in the fracking fluid are monitored. The systems and methods disclosed herein alleviate some of the deficiencies of current wireline monitoring techniques by utilizing real-time quantitative and qualitative analysis of acoustic and vibration signals in the fracking fluid, in either the time domain, frequency domain, or both, to more accurately assess wireline friction and sticking, and to optionally provide warnings to operators and/or automated and optimized control of wireline operations.
In some cases, the techniques described in this disclosure may utilize a high frequency (e.g., greater than 1 kHz) acoustic or vibration sensor directly coupled to a well, for instance at the well head, circulating fluid line, or standpipe. This acoustic or vibration sensor may be in direct physical contact with fluid in the well, the well casing, the well head pipe, the circulating fluid line, the standpipe, or the well pad (e.g., vibration sensors on the well pad can obviate the need to have direct contact with fluid in the well). In some embodiments, a vibration sensor need not be directly coupled to a component of the well, but instead can indirectly measure vibrations in the fluid. For instance, a laser reflecting off a surface of the well, such as a viewing window, could measure vibrations of the viewing window using optical methods. For the sake of brevity, an acoustic or vibration sensor may be used wherever the term acoustic sensor is seen in this disclosure. In some examples, the high frequency acoustic or vibration sensor may provide a digital or analog signal indicative of high frequency pressure fluctuations. Additionally or alternatively, the signal may be indicative of vibrations in the fluid. In some examples, this signal may be passed to a conversion and analysis component, or a converter (e.g., spectrum analyzer), configured to identify frequency components of the signal (e.g., via an algorithm that transforms pressure or vibration data in the time domain to the frequency domain such as a Fast Fourier Transform (FFT) and compares the frequency domain signal to previously measured frequency domain signals or signatures).
In some examples, the acoustic sensors described throughout this disclosure may or may not have a reference pressure. Furthermore, the acoustic sensors may be configured to measure at least changes in pressure. Thus, in some cases, the acoustic sensors may be configured to measure absolute pressure in addition to pressure changes (i.e., if a reference pressure is being used). Additionally or alternatively, for instance, if no reference pressure is being used, the acoustic sensors may be used in parallel with an absolute pressure sensor. In some cases, the absolute pressure sensor may be configured to measure static or absolute pressures, where the absolute pressure may be used as a baseline (or reference) for the higher sensitivity data from the acoustic sensor.
It should be noted that throughout this disclosure a vibration sensor may be used in addition (or as an alternative) to a high frequency acoustic pressure sensor.
As previously noted, the term acoustic sensor may be broadly used to refer to a high frequency acoustic pressure sensor and/or a vibration sensor (e.g., sampling at ˜1 kHz or greater). One non-limiting example of a vibration sensor may comprise a piezoelectric vibration sensor. In some cases, piezoelectric vibration sensors may be configured to generate a current or voltage proportional to an amount of piezoelectric material movement. The piezoelectric material can be in direct physical contact with the fluid in the well or may be physically coupled to a protective membrane that is in direct physical contact with the fluid. Either way, vibrations in the fluid may be transmitted to the piezoelectric material, which may cause movement or vibrations of the piezoelectric material. Movement of the piezoelectric material may generate a current or voltage, where the current or voltage may be proportional to the amount of vibration or movement of the piezoelectric material. The ICP Pressure Sensor, Model Number 113B23, is one non-limiting example of an acoustic or vibration sensor.
In some cases, the generated current or voltage may be recorded and stored, and there may be a 1 to 1 mapping of vibration data to current or voltage data. The measured current or voltage readings may be used to determine vibration data, for instance, by mapping the current or voltage readings to corresponding vibration values in a look-up table. In some cases, raw data may comprise one or more of the mapped vibration data, and the measured current and/or voltage readings. This raw data can be passed through a transform operation such as a Fourier Transform, and further analyzed in the frequency domain (e.g., via a spectrum analyzer), further described below.
In some examples, a transform component and/or a conversion and analysis component (e.g., converter or spectrum analyzer) may be implemented as a software program, firmware module, hardware comprising analog circuits, or a combination thereof. In some embodiments, a conversion function (e.g., Fourier Transform) may comprise the use of wavelet analysis techniques. Further, wavelet analysis may refer to the use of a custom function that is stretched and scaled. Further, wavelet analysis may facilitate in optimizing analysis of detailed timing of events in a signal.
For the purposes of this disclosure, a conversion and analysis component (e.g., spectrum analyzer) may be configured to measure the magnitude of an input signal at different frequencies. Said another way, the conversion and analysis component may analyze signals in the frequency domain, rather than the time domain. Typically, the conversion and analysis component may receive electrical signals as an input. In some other cases, the conversion and analysis component may receive acoustic or vibration signals via an appropriate transducer. In some embodiments, the conversion and analysis component may utilize a Fourier Transform or another applicable transform algorithm to convert raw acoustic or vibration data from the time domain to the frequency domain.
Fracking pads may include one or more acoustic sensors (e.g., one acoustic sensor for each well head) or one or more static pressure sensors and one or more acoustic sensors (e.g., one static and one acoustic sensor for each well head). The acoustic sensors may be high frequency pressure sensors (e.g., sampling at ˜1 kHz or greater). Each fracking pad may include a transceiver for transmitting raw data from its sensor(s) to a local or cloud-based conversion and analysis component. Additionally or alternatively, the raw data may be transmitted to a processing resource that receives and analyses outputs from various conversion and analysis components. In one embodiment, a set of pads may comprise a master transceiver configured to receive data from one or more other pads on a local network. Each pad can transmit raw data or converted data (i.e., frequency domain data) to the master transceiver, and the master transceiver may transmit (i.e., relay) the data received on the local network to a cloud-based resource, such as a server farm where more complex analysis takes place (e.g., comparison to a model; training a model).
Further, the acoustic sensors may be coupled to one or more conversion and analysis components. In some cases, the number of conversion and analysis components may vary (e.g., one for each pad, one for each well head, or one for a network of sensors, to name a few non-limiting examples). The conversion and analysis component may be configured to execute an algorithm, such as a FFT algorithm, for transforming raw data from the time domain to the frequency domain. In some other cases, the conversion and analysis component may be used in concert with another device or software module that can perform FFT.
Using spectral analysis rather than static pressure-based sensing enables higher signal to noise ratios than traditional static pressure-based sensors. For instance, an increase in wireline friction or sticking would not be visible in static pressure measurements, but would cause an acoustic signal or vibration that could be detected by a dynamic acoustic sensor and seen via analysis of frequency components in the time or frequency domain. Such changes from the noise floor could be seen as frequency peaks in the frequency domain, as seen in
Since this disclosure looks at a spectral analysis of acoustic and vibration waves in well hole fluid, many different signatures can be identified and analyzed, each having their own benefits (e.g., detecting different down well phenomena, or where some signals might have a higher signal to noise ratio). For instance, in the frequency domain, the consistent pumping of pumping components, such as pump trucks, at an adjacent well may have a much greater amplitude than signals at other frequencies. Fluid, mud, and proppant flowing through the well holes, perforations in stage walls, and fractures may also have easily identifiable signatures in the frequency domain. For instance, sand moving along edges of a pipe, well hole, or fracture may generate acoustic or vibration waves at a unique frequency (e.g., at a different frequency or frequencies from signals generated by clean water moving through the same structure). Also, areas where fluids become turbulent may be identifiable in the frequency domain, as distinct from areas of laminar fluid flow.
Being able to distinguish between different processes or events during development of a well or offset well may allow the conversion and analysis component to identify signatures coming from specific activities as well as from specific locations in the observation or offset well. For instance, one adjacent well may be packing sand and gel into existing cracks while another adjacent well may be opening cracks with pumped fluid. In such cases, given knowledge of the timing of these processes at nearby wells, the conversion and analysis component may be able to identify which well is causing which signals based on which it may monitor changes in the signal from one well to the other (e.g., where one of the two wells is approaching a wireline sticking event). Alternatively, spectral analysis of acoustic or vibration signals in a well may help operators better understand drill bit torque and wear, or even predict an impending wireline sticking event. In another example, spectral analysis of acoustic or vibration signals in a well may help operators more quickly identify component failures or preemptively remove a component from operation before failure. In yet another example, analyzing acoustic or vibration data in the frequency domain may shed light on casing wear and allow operators to act in response to excessive wear.
In an embodiment, high frequency acoustic signals in a well can be matched with known signals indicating an increasing likelihood of a mishap, such as impending sticking of the coil tube, sticking of wireline, or stalling of a drill bit. This can allow preventive actions to be taken before a catastrophic event, such as coil tube sticking, occurs. Similarly, acoustic signatures of components that are nearing failure may be monitored and preventative maintenance may be carried out in response. For instance, it is well known in the prior art that paddle trucks include pumps that begin to disintegrate or “chunk” out prior to complete failure. However, the pressurized nature of these pumps prevents them from being visually monitored. According to aspects of the present disclosure, a conversion and analysis component (i.e., on the well head, or coupled to a fluid line of the pump truck, or on an offset well) may be used to detect the frequency signature of “chunking” and issue a warning.
Raw Time Domain Analysis and/or Time & Frequency Domain Analysis
In some cases, the acoustic or vibration data in the time domain may be analyzed, for instance, by a machine learning model without conversion. In such cases, the conversion and analysis component may be responsible for analysis, but not conversion, of the time domain data. It should be noted that, even though no conversion of time domain data into the frequency domain takes place, the model may still have access to frequency information associated with the measured signal. In some cases, knowledge of frequency space decomposition of a signal may be utilized to deconstruct a single waveform in time into a composite of simpler, underlying waveforms (e.g., sinusoidal waveforms). In some other cases, a Short-time Fourier transform (STFT) may be used to determine the sinusoidal frequency and phase content of local sections of a signal as it changes over time. STFT computation may involve dividing a longer time signal into shorter segments of equal length and then computing the Fourier transform separately on each shorter segment. In some cases, once the Fourier spectrum is revealed for each shorter segment, the changing spectra may be plotted as a function of time (i.e., also known as a spectrogram or waterfall plot).
Furthermore, while this disclosure has discussed use cases where a source of a frequency signal is constant, such as that of fracking fluid passing through a perforation in a casing or a pumping frequency of a pump truck, in other embodiments, the source frequency may vary in time. In some instances, frequency monitoring operations may be controlled and fine-tuned to assist in distinguishing a signal from background noise. In one example, operators may vary and control a generated source frequency signal (e.g., from a pump truck) and monitor changes in the observed signal in the frequency and/or time domain For instance, rather than simply monitoring a 33 Hz pump truck signal, operators could perform a frequency sweep for the generated pump truck signal (e.g., by gradually adjusting a pump truck frequency through a range, such as 20 Hz to 40 Hz). In this example, the conversion and analysis component may be used to not only pick up on these unique frequencies (e.g., between 20 and 40 Hz) over background frequencies, but also monitor and observe the signal changing in the frequency domain as a function of time. Additionally or alternatively, the source frequency may also be adjusted to optimize travel through a given medium. For instance, where a certain shale formation separates an observation well and an observation well coupled to pump trucks, the pump trucks' revolution per minute (RPMs) could be gradually adjusted until a highest amplitude signal (i.e., corresponding to an optimized frequency for travel through the shale formation between the wells) is observed by the conversion and analysis component. In this way, a source signal can be optimized for detection in an offset well.
It should be noted that, pump trucks are just one example of an acoustic or vibration source, and different acoustic/vibration sources may be utilized in different embodiments. For instance, surface vibrators or surface oscillators used for releasing stuck drill strings may be used as acoustic or vibration sources. In some other cases, surface vibrators used to impart vibratory seismic energy into the ground may be used as acoustic/vibration sources. In yet other cases, an acoustic transducer, ultrasound transducer, sonar transducer, etc., may be used to inject energy into the system.
Some embodiments of this disclosure pass acoustic or vibration data in the frequency domain to a machine learning model for analysis, labeling, and training of the model. In some embodiments, the model may be configured to use artificial intelligence based on, for example, a neural network or other type of machine learning algorithm. In some cases, the artificial intelligence algorithm or model may receive time domain data converted to a frequency domain, for instance, using a FFT algorithm or another algorithm for computing the discrete Fourier transform (DFT) of a sequence. A DFT may be obtained by decomposing a sequence of values into components of different frequencies. In some cases, a conversion and analysis component may be utilized to perform the conversion from time to frequency domain. In some other cases, the acoustic or vibration data in the time domain may be passed to a machine learning model without conversion. In such cases, the conversion and analysis component may be responsible for analysis, but not conversion, of the time domain data. It should be noted that, even though no conversion of time domain data into the frequency domain takes place, the model may still have access to frequency information associated with the measured signal. In some cases, the model may look at a window of data in one shot (or one local section of a signal as it changes over time) and learn to detect, for instance, high and low frequency waveforms and structures. The model or neural network may encompass knowledge of frequency space decomposition of a signal and may be configured to deconstruct a single waveform in time into a composite of simpler, underlying waveforms (e.g., sinusoidal waveforms). Thus, in some aspects, the model may be trained to perform something akin to Fourier analysis. In some other cases, the model may utilize a Short-time Fourier transform (STFT) to determine the sinusoidal frequency and phase content of local sections of a signal as it changes over time. STFT computation may involve dividing a longer time signal into shorter segments of equal length and then computing the Fourier transform separately on each shorter segment. In some cases, once the Fourier spectrum is revealed for each shorter segment, the changing spectra may be plotted as a function of time (i.e., also known as a spectrogram or waterfall plot).
In some embodiments, a plurality of distinct machine-learning algorithms may be operated in parallel, which may serve to enhance the accuracy of predicting future wireline sticking or jamming events. In some aspects, the use of multiple machine-learning algorithms may also decrease false positive indications as compared to the use of a single machine learning algorithm. In some cases, a combination of three or four machine learning algorithms may be operated in parallel, which may provide a balance of high accuracy versus system complexity. Some non-limiting examples of machine learning algorithms may include a neural network, a decision tree, a support vector machine, and Bayesian methods.
In some cases, a neural network may comprise a plurality of input nodes, where an input node refers to a point within the neural network to which a parameter (e.g., a drilling parameter) may be provided for further processing. Further, the neural network may comprise one or more output nodes, where each output node represents a calculated and/or predicted parameter based on the input data at the input nodes. In some cases, one or more layers of hidden nodes may lie between the input and output nodes, where the hidden nodes may be coupled to some or all of the input nodes and/or the output nodes. Each of the hidden nodes may be configured to perform a mathematical function that is determined or learned during a training phase of the neural network, where the mathematical function may be determined based on the data of the input nodes to which it is coupled. Likewise, the output nodes may perform mathematical functions based on data provided from the hidden nodes. In some embodiments, the neural network may be provided one or more drilling parameters in real-time, as well as one or more historical values of the drilling parameters based on preprocessing, for instance, by a wireline sticking event prediction software. In other words, the neural network may be trained using historical data from fracking and drilling operations where a wireline sticking event actually occurred. In such cases, the neural network may produce a value at an output node based on an input value provided to the input node, where the value may be a probability of occurrence of a wireline sticking event. Some non-limiting examples of drilling parameters may include a value indicative of weight-on-bit; a value indicative of hook load, wherein the hook may be used to control the ascent/descent of the wireline or a drill string in the borehole; a value indicative of rate-of-penetration; a value indicative of rotary speed of the drill pipe; a value indicative of torque applied to the drill pipe; a value indicative of drilling fluid pump pressure; a value indicative of inclination of the drill string; a value indicative of length of drill string; measurement-while-drilling data; logging-while-drilling data; and a value indicative of drilling fluid flow rate.
With regards to fracking and drilling operations, a decision tree machine learning algorithm may be an example of a predictive model comprising a plurality of interior nodes that may be traversed based on a set of input parameters (e.g., drilling parameters, such as drill string RPM, torque, etc.). In such cases, the predicted value (e.g., of a wireline sticking event) may be based on arriving at an end node following transitioning from node to node, where the transitioning may be based on the set of input parameters. In such cases, the end node may be dictated by the input parameters. It should be noted that, in some cases, decision trees may also be referred to as classification or regression trees.
In some cases, support vector machines are a class of machine-learning algorithms that perform classifications of data into groups. In particular, support vector machines can be thought of as performing classification by analysis of the data in a multidimensional space. Training data for support vector machines may be “plotted” or “mapped” into the multidimensional space and classified or grouped spatially. It should be noted that the plotting or mapping need not be a true physical plotting, but a conceptual operation. After the training phase, data to be analyzed may be plotted or mapped into the multidimensional space. Further, the support vector machine may be configured to determine the most likely classification of the data. In some cases, the classification of the data to be analyzed may be a “distance” calculation between the spatial location of the data to be analyzed in the mappings and the “nearest” classification. In one non-limiting example, the support vector machine may be provided one or more drilling parameters from drilling and fracking operations where wireline sticking or jamming took place, as well as operations where no sticking event occurred. In this case, the support vector machine may be configured to plot the data in a multidimensional space and classify the data. During actual drilling and fracking operations (i.e., when real-time drilling parameters are provided to the support vector machine), the support vector machine may plot a data point under test in the multidimensional space, and predict a result (i.e., a probability of a wireline sticking event) based on the spatial position of the plotted point relative to a spatial delineation (or classification line) between data with wireline sticking events and those without.
In yet other cases, the machine learning algorithm may comprise the use of Bayesian methods. Bayesian methods represent a logically different view of data and probabilities and may be thought of as testing the plausibility of a hypothesis (e.g., a wireline sticking event will occur in the future) based on a previous set of data. In some aspects, Bayesian methods may be considered non-deterministic since they generally assume the plausibility of a hypothesis is based on unknown or unknowable underlying data or assumptions. In some embodiments, a value indicative of plausibility of a hypothesis may be determined based on the previous data (e.g., the training data), following which plausibility may be tested again in view of new data (i.e., with the drilling parameters applied). From the evaluation, a plausibility of the truth of the hypothesis may be determined.
In some cases, the on-site or cloud-based storage and analysis unit 112 may include a trained model (e.g., as part of a machine-learning system) based on previous wireline operations and their frequency signatures (and optionally previously classified by the machine-learning system). For instance, the model may have been trained using acoustic or vibration data from previous drilling events, for instance, an event that led to a falloff in production. In one example, frequency signatures for previous sticking events or wireline jamming events could also be used to train the model to detect acoustic or vibration frequency signatures that suggest an eminent wireline jamming event. In some embodiments, the on-site or cloud-based storage and analysis unit 112 may be configured to take action in response to the on-site or cloud-based storage and analysis unit 112 identifying a threshold increase in wireline friction or an imminent sticking event, for instance, by providing automated feedback control to the well. More specifically, the automated feedback may include reducing a speed of descent of the wireline, stopping motion of the wireline, reversing a direction of the wireline, increasing a flush time, performing a dedicated flush, or using a smaller plug, or perform another applicable action.
In some embodiments, the on-site or cloud-based storage and analysis 112 may monitor a signature of pump trucks 122 pumping fluids into the offset well head 120. These trucks may be operating pumps at around 33 Hz. In such cases, the frequency signature (i.e., at 33 Hz) generated by the pump truck may be of a larger amplitude than other frequency components generated by the illustrated drilling system 100.
By better understanding wireline friction and sticking and providing earlier predictions of such events, the herein disclosed systems, methods, and apparatus may not only help reduce downtime, but also minimize the abandonment of stages during a fracking operation.
The signals can either be sourced at the observation well (e.g., acoustic waves from a fracture initiation) or an adjacent or offset well. In some embodiments, the sensor(s) may be configured to couple to processors (e.g., Raspberry Pi) located in the spoke computers 208-a and/or 208-b. In some cases, a spoke computer 208 may comprise one or more processors for each well pad 202 in electronic communication with the respective spoke computer. In some cases, in addition to reading the acoustic or vibration signal measurements from the one or more sensors, the computer system (e.g., spoke computers 208) may also be configured to read one or more surface-based parameters directly or indirectly. A non-limiting list of surface-based parameters that may be directly or indirectly read by the computer systems comprises: a hook load; RPM of the drill string at the surface; torque applied to the drill string at the surface; pressure of the drilling fluid as the drilling fluid is pumped into the drill string; pressure of the drilling fluid returning to the surface; and standpipe pressure of the drilling fluid.
In some embodiments, the one or more processors of the spoke computers 208 may be coupled to an antenna system 212. In some cases, the antenna system 212 may comprise an omnidirectional antenna, although other types of antennas are contemplated in different embodiments. Each antenna system 212 may be in communication with a wide area network (WAN), such as a 4G or 5G network. In another embodiment, the antennas of the antenna system 212 may form a local area wireless network wherein one of the antennas may be configured as an interface (e.g., a gateway) between the local area wireless network and a wide area network. In some embodiments, cellular (e.g., multi-beam antennas, sector antennas) or satellite (e.g., dish) antennas may be deployed for communication with a wide area network, to name a few non-limiting examples. Further, omnidirectional or Yagi type antennas, to name two non-limiting examples, may be utilized for local area network communication.
In some cases, the remote hub 214 may be in communication with the antenna systems 212 and the spoke computers 208. Further, the remote hub 214 may be configured to contact an insight program 226 via an Application Programming Interface (API) 224. In some examples, this communication may involve a local area network or a wide area network. Insight 226 may be configured to store data for a training model in the database 230, as well as continually train the model using new data acquired from the acoustic sensors at the well heads. In some cases, the drilling system 200 may also support a web app 228 to provide one or more insights, warnings, feedback, and/or instructions to pad operators. In some examples, the web app 228 may be accessible via a user interface displayed on a user device (e.g., laptop, smartphone, tablet, etc.).
In some embodiments, the processors may comprise (or may be coupled to) a conversion and analysis component. In other embodiments, the processors may send their data through the network(s) to a centralized conversion and analysis component. In some cases, the centralized conversion and analysis component may or may not be located near the well pads 202. For instance, the centralized conversion and analysis component may be located off-site in some embodiments.
As illustrated, the drilling system 200 may further comprise one or more additional components, modules, and/or sub-systems, including, but not limited to, a Data Acquisition and Control System (DASTrac 216), a fracking client 218, a Coiled Tubing (CT) Data Acquisition module 220, and a CT client 222. In some cases, the DASTrac 216 may comprise a data acquisition and control program for acquiring fracking operations data from wellsite process control units and other sensors. Further, DASTrac 216 may be configured to display the acquired data from the data acquisition system in both numeric and graphical form in real time, which may enable operators to change job profiles, scale parameters, advance stages, change stages, and hold stages in response to seeing fracture scores, to name a few non-limiting examples. In some cases, the CT Data Acquisition module 220 may be configured to measure and control technological parameters of coiled tubing units during repair and stimulation operations of oil and gas wells. The CT Data Acquisition module 220 may also be configured to record the measured technological parameters on electronic media, and optionally display and visualize them on an operator's computer display. In some cases, the CT client 222 may be configured to access coiled tubing data from the CT Data Acquisition module 220, for instance, directly via the API 224. In the oil and gas industry, coiled tubing may refer to a long metal pipe, usually anywhere between 1 to 3.25 inches in diameter (although other diameters are contemplated in different embodiments), which is supplied spooled on a reel. In some cases, coiled tubing may be used for interventions in oil and gas wells, as production tubing in depleted gas wells, and/or as an alternative to a wireline (i.e., the coiled tubing may be used to carry out operations similar to a wireline). In some embodiments, coiled tubing may be configured to perform open hole drilling and milling operations. Further, due to their high pressure tolerance abilities (e.g., ranging from 55,000 PSI to 120,000 PSI), they may also be utilized to fracture a reservoir. In some cases, one or more sensors (not shown) may be coupled to the coiled tubing and sent downhole. The CT Data Acquisition module 220 may collect real-time downhole measurements from the sensors, where the measurements may be used to model the fatigue on the coiled tubing, predict coiled tubing performance, fluid behavior at modeled downhole well conditions, to name a few non-limiting examples. In some cases, the real-time downhole measurements collected by the CT Data Acquisition module 220 may also be used to optimize treatments, for instance, during interventions (i.e., when the well is taken offline).
The spoke computers can include memory for storing electrical signals, a current frequency domain spectrum, or both, measured by sensors at one or more well heads, circulating fluid lines, or standpipes at the well pads 202-a and 202-b. The database 230 can also include memory for storing electrical signals, a current frequency domain spectrum, or both, measured by sensors at one or more well heads, circulating fluid lines, or standpipes at the well pads 202-a and 202-b. The database 230 can also be configured to store frequency domain spectra measured during previous wireline operations. The database 230 can also include previous classifications or identifications of wireline events associated with the previous frequency domain spectra. This may include a mapping between events (e.g., increased wireline friction or wireline sticking) and previous frequency domain spectra.
In some cases, one or more user/operator devices 305, such as user/operator devices 305-a, 305-b, 305-c, and/or 305 may be in communication with a configuration app 328. The configuration app (also referred to as config app 328) may be in communication with the hub 314 and may be used to assign sensors to particular wells 302 and/or spoke computers 308, for instance. The config app 328 may also be used for configuring one or more of the hub 314, the sensors, and the spoke computers 308.
As shown, the intermediate portion of the wireline 430 may pass over a pulley system 410 located above the wellhead 425 of the drilling rig 401. The drilling system 400-a may further include a line 435 spooled on a winch (not shown) adjacent a first side of the drilling rig. The line 435 may pass from the winch to a block located near a top of the drilling rig 401, down through a second pulley system 405, back up to the block (near the top of the rig), and back down to an anchor (not shown) adjacent a second side of the drilling rig. In other words, the line 435 may be winched and anchored, respectively, on opposing sides of the drilling rig 401. In some cases, the line 435 may facilitate in raising and lowering the wireline 430 in the borehole. In some aspects, the pulley systems 405 and 410 may form a double pulley system, wherein the line 435 may be used to control the ascent and/or descent of the wireline 430 within the borehole (not shown). In some other cases, the second pulley system 405 may be replaced by a traveling block and a hook, where the hook couples the traveling block to the wireline 430. Similar to the embodiment with the second pulley system 405, a line 435 may be passed down to the traveling block and hook system for raising and lowering the wireline 430. Thus, the block located near the top of the drilling rig and the traveling block may act as a block-and-tackle device to provide mechanical advantage in raising the lowering the wireline 430 coupled to the wireline truck 420.
In one or more embodiments, the line 435 may include a fast line that extends from the winch (i.e., adjacent a first side of the drilling rig) to the block at the top of the drilling rig and a deadline that extends from the same block to the anchor (i.e., adjacent a second side of the drilling rig). In one or more embodiments, a supply spool may store additional line 435 that can be used when the line 435 has been in use for some time and is considered worn. In one or more embodiments, a pulley or hookload sensor may provides signals representative of the load imposed by the wireline 430 on the second pulley system 405, or alternatively, the hook. In one or more embodiments, the pulley or hookload sensor may be coupled to the deadline to measure the tension in the line 435.
As discussed above, in some circumstances, the wireline 430 may become stuck in the well or borehole for a variety of reasons, including a collapse of the borehole, differential sticking in which the pressure exerted by fluids overcomes formation pressures causing the wireline 430 to stick to the wall of the borehole, swelling of the borehole, etc. Once the wireline is stuck, pulling on the wireline with a pressure beyond a safe limit may damage the wireline 430 or other tools/equipment in the borehole.
In some cases, one or more drilling parameters may be monitored and recorded and passed on to a machine learning algorithm for analysis and classification. Some non-limiting examples of drilling parameters may include a value indicative of weight-on-bit; a value indicative of hook load; a value indicative of rate-of-penetration; a value indicative of rotary speed of the drill pipe; a value indicative of torque applied to the drill pipe; a value indicative of drilling fluid pump pressure; a value indicative of inclination of the drill string; a value indicative of length of drill string; measurement-while-drilling data; logging-while-drilling data; and a value indicative of drilling fluid flow rate. In some cases, the likelihood of an event (e.g., wireline sticking) may be determined based on comparing acoustic or vibration signatures given the applied drilling parameters for known wireline sticking events and those without, described in more detail below. In some embodiments, one or more surface-based parameters may be directly or indirectly read by a computer system, including, but not limited to, a hook load; RPM of the drill string at the surface; torque applied to the drill string at the surface; pressure of the drilling fluid as the drilling fluid is pumped into the drill string; pressure of the drilling fluid returning to the surface; and standpipe pressure of the drilling fluid. These surface-based parameters may also be used to predict events, such as wireline sticking.
After hub 414 receives the data (e.g., raw data, or frequency domain sensor data) from spoke computer 408, it may further relay said data on to a frequency spectral analysis module 401 via antenna system 412-b and/or API 424. The API 424 may implement one or more aspects of API 224 discussed in relation to
As illustrated, the frequency spectral analysis module 401 may be electronically and communicatively coupled to a classification and prediction module 402. The classification and prediction module 402 may be configured to determine if the frequency spectrum of the raw data aligns with signatures for known failures, such as an equipment failure, performance degradation, or a specific equipment condition, to name three non-limiting examples. If such a classification occurs, then an events and notification module 403 may be activated. In some circumstances, the events and notification module 403 may be configured to issue a warning to an operator of the pump truck 420. Alternatively, the classification and prediction module 402 may be configured to analyze raw time series data and determine if this raw data aligns with known time series signatures (e.g., for wireline friction and/or sticking). If such a classification occurs, the events and notification module 403 may also be activated to send a warning to an operator.
It should be noted that, the specifics of the
Additionally or alternatively, a vibration sensor (not shown) may be attached to a component (e.g., metal component, such as a pipe) of the wellhead. In such cases, vibrations felt through the metal component may also be measured and recorded. Similar to
The spectral artifacts seen in
In general, Artificial intelligence (AI) models aim to learn a function (f(X)) which provides the most precise correlation between input values (X) and output values (Y), such that Y=f(X). In some embodiments, AI models may be used to monitor and preemptively detect impending wireline friction and sticking events, further described below. Further, the AI models described in this disclosure may be of a variety of types, for example linear regression models, logistic regression models, linear discriminant analysis models, decision tree models, naïve bayes models, K-nearest neighbors models, learning vector quantization models, support vector machines, bagging and random forest models, and deep neural networks.
In some embodiments, an AI model may be trained to identify time series signatures or spectral signatures of wireline friction, especially, an amount or type of friction that is likely to lead to wireline sticking. In some cases, the AI model may be trained via a time sequence analysis or spectral analysis of acoustic or vibration data in a well, wellhead, circulating fluid line, pump line, standpipe, or other component in fluid communication with the well, to name a few non-limiting examples. In these embodiments, an acoustic or vibration sensor can be arranged to monitor signals from components within the same well being monitored.
For instance,
In some circumstances, the machine learning component 1712 may observe that the time series data matches (or resembles) a time series signature previously correlated to friction that preceded a sticking event(s). For instance, wireline friction can lead to increased variations in signal amplitudes over time as compared to amplitude variations when there is less friction. In some cases, an increase in frequency and/or amplitude of spikes in the time domain may be indicative of a greater amount of friction occurring, and thus, a higher likelihood of a sticking event. In such cases, the machine learning component 1712 may identify a time series signature (e.g., threshold increase in variation over time) that corresponds to a known time series signature (i.e., a time series signature discovered in a trained model of the machine learning component 1712) associated with a particular state or event (e.g., wireline friction and sticking). In some embodiments, an average acoustic vibration can be measured during wireline descent and a current acoustic vibration can be compared to this average. If the current acoustic vibration exceeds the average by a threshold, then the machine learning component 1712 can classify the current acoustic vibration as associated with excessive wireline friction.
When the optional converter 1710 is used, the machine learning component 1712 may observe that the frequency domain version of the data matches frequency signatures that have previously been correlated to increased friction preceding a sticking event. Specifically, a wireline may begin to emit particular frequencies of sound (or other signals) when excessive friction in the wellbore occurs. However, once jamming or sticking occurs, the frequencies of sound (or signals) emitted may shift, or even degrade/decrease. In other words, the frequencies associated with wireline jamming may be different from the frequencies associated with increasing friction in the wellbore. For instance, in some cases, the plug may be formed of a different material than the perforation gun (e.g., composite versus metal). In such cases, a rubbing of these components on walls of the wellbore may be associated with different and unique spectral signals since one of these two components may experience friction sooner than the other in a pre-jamming situation. In some cases, the degree or amount of friction may be associated with different frequency signals (e.g., increasing friction may result in an increasing amplitude of acoustic or vibration signal(s) up the wellbore for certain frequencies, or increasing friction may result in decreasing frequencies of spectral peaks, or increasing friction may result in decreasing bandwidth of one or more spectral peaks, to name a few non-limiting examples). Thus, the machine learning component 1712 may identify frequency signatures that resemble or correspond to frequency signatures used to train the model, where the frequency signatures used to train the model correspond to particular states or events (e.g., wireline friction and sticking).
In some embodiments, the machine learning component 1712 may use new insights from the time series data or from the converted frequency domain version of the data, to further train the model 1714. Additionally or alternatively, the machine learning component 1712 may also send a signal or indication to an operator user interface or an instruction to wireline equipment in response to labeling the new time series signature or frequency signature. For instance, a warning may be sent to an optional operator computer 1716 in response to the machine learning component 1712 applying a jammed state label to the new frequency or time series signature. In some other cases, the machine learning component 1712 may also send instructions to the optional operator computer 1716 to present a visual representation of an amount of friction on the computer's 1716 display. An operator can view the computer's 1716 display and in response, control the controllers 1718 and 1720. Alternatively, the machine learning component 1712 can directly control the controllers 1718 and 1720 as part of a closed feedback loop with or without involving the operator computer 1716. In some embodiments, the optional operator computer 1716 may comprise a visual display, such as an LCD screen, a LED screen, a plasma screen, to name a few non-limiting examples.
In some embodiments, the machine learning component 1712 can provide a warning and/or instructions to the operator computer 1716, and/or to the controllers 1718, 1720. The instructions, or suggestion (with or without a warning), may include: (1) a decrease in wireline speed; (2) stoppage of wireline movement; (3) a reversal of wireline speed; (4) a lengthening of flush time; (5) a dedicated flush; and (6) use of a smaller plug, to name some non-limiting examples.
In some embodiments, the first controller 1718 can be configured to control a direction and speed of the wireline 1718. Furthermore, the second controller 1720 can be configured to control flushing of the well 1702. In some embodiments, the first and second controllers 1718, 1720 may be deployed as a single controller or unit. For instance, the first and second controllers 1718, 1720 may be installed within a single housing or enclosure and may be in electronic communication with each other. In other cases, a single controller (not shown) may be configured to perform the functions of both the first and second controllers 1718, 1720.
In some embodiments, an average acoustic vibration can be measured during wireline descent and a current acoustic vibration can be compared to this average. If the current acoustic vibration exceeds the average by a threshold, then the machine learning component 1812 can classify the current acoustic vibration as associated with excessive wireline friction.
In some implementations, method 2000 may be implemented in one or more processing devices (e.g., a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanism for electronically processing information). The one or more processing devices may include one or more devices executing some or all of the operations of method 2000 in response to instructions stored electronically on an electronic storage medium. The one or more processing devices may include one or more devices configured through hardware, firmware, and/or software to be specifically designed for execution of one or more of the operations of method 2000.
An operation 2002 may include providing an acoustic or vibration sensor (e.g., acoustic or vibration sensor 1704 or 1804 in
An operation 2004 may include acquiring acoustic or vibration data in the fluid via the acoustic or vibration sensor in a time domain. Operation 2004 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to vibration data acquiring module 1910 (as shown in
In some embodiments, method 2000 may turn to operations 2006 and 2008 (shown as optional by the dashed lines), for instance, if frequency spectrum analysis is to be performed on the data. Alternatively, operations 2006 and 2008 may be skipped or eliminated if time series data is to be analyzed. In some embodiments, an optional operation 2006 may include transferring the acoustic or vibration data to a converter such as a spectrum analyzer (e.g., spectrum analyzer 1810 in
In some cases, a converter (e.g., spectrum analyzer) may optionally convert the acoustic or vibration data from the time domain to a frequency domain (Operation 2008). Optional operation 2008 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to acoustic vibration data converting module 1914 (as shown in
In some cases, method 2000 may further include performing an operation 2010. Operation 2010 may include comparing (e.g., via machine learning component 1812 described in relation to
As shown, method 2000 may further include performing an operation 2012, where operation 2012 may include assigning one of a plurality of labels to the acoustic or vibration data in the time domain (or frequency domain) based on the comparing. The plurality of labels may include an excessive wireline friction state (where wireline friction exceeds a threshold) and/or a sticking state, to name two non-limiting examples. Operation 2012 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to label assignment module 1918 (as shown in
In some cases, for instance, if the friction state label is assigned, an operation 2014 may include generating a first indication on an operator display (e.g., a display of operator computers 1716 and/or 1816 shown in
In some embodiments, if the sticking state label is assigned, an operation 2016 may include generating a second indication on an operator display (e.g., display of operator computers 1716 and/or 1816 described in relation to
While not shown in
As shown, at Block 2104, the data from the acoustic sensor can be fed into the model, wherein the model may be configured to identify time series or spectral signatures in the data that match known acoustic (or vibration) behavior of an event, such as wireline friction preceding a sticking event or a jamming event. In other words, prior to the start of the method in flowchart 2100, the model may be trained to recognize a time series signal or spectral signal associated with certain wireline events, including a wireline sticking event. This training can be based on previous frequency spectra associated with previous excessive wireline friction events or wireline sticking events. In some embodiments, at Block 2106, the model may then classify or identify the event, for instance, by matching the sensor data with one or more categories of events.
At Block 2108, the results data associated with the event may be collected, where the results data may include any data type that is related to (or a consequence of) the event classified at Block 2106. For instance, knowledge of a wireline getting stuck may be one example of results data for acoustic or vibration data that preceded or occurred at the time of the sticking. In some embodiments, at Block 2110, the model may find correlations between the results data and the classified event data. In some cases, multiple sets of results data may be correlated to a single classified event. After one or more correlations (if any) have been made, the model may be configured to learn from (i.e., be trained using) the correlations at Block 2112. As shown, after Block 2112, the method may return to Block 2102 and restart. In other words, the flowchart 2100 may be an example of a closed-loop feedback system, where the model is continuously trained using newly acquired sensor data. In some cases, one or more of the blocks (2102, 2104, 2106, 2108, 2110, 2112) may be repeated to continually update the model. In this way, the training model described in this disclosure may refine its ability to accurately detect and model events. In some cases, the model can then be used to identify events and states based on acquired acoustic or vibration data in the time domain (or alternatively in the frequency domain where a time-to-frequency conversion has taken place as discussed earlier).
In some embodiments, the model can be trained on data unrelated to wireline events. This could be performed to help the model identify spectral signatures unrelated to wireline events and thereby more quickly and more accurately identify spectral signatures that are related to wireline events. For instance, such training may be used to form a model that knows that spectral indicators below a certain threshold frequency are primarily related to movement of water through a formation and not wireline events. Accordingly, the model may be able to focus on higher frequency data when looking for wireline events in the future. This is just an illustrative example, and not intended to limit the scope of the disclosure.
At Block 2204, the method 2200 may comprise monitoring the acoustic or vibration signal via the sensor, as previously described in relation to
In some cases, at Block 2210, the method 2200 may comprise analyzing the time series (or optionally the frequency spectrum of the transformed acoustic or vibration data) using a model to identify known activities and states such as wireline friction or wireline sticking, to name two non-limiting examples. While not shown, in some cases, the analysis may comprise labeling or categorizing the data, where the labeled or categorized data may be used to further train the model, as previously described in relation to
In some examples, once the acoustic signal in the frequency domain or time domain has been classified or labeled, the method 2200 may comprise automatically controlling a fracking operation, such as a direction of travel of the wireline, to adjust a state of the fracking operation at Block 2212. For instance, in response to an excessive friction state (i.e., friction exceeding a threshold) of the wireline, the method 2200 may comprise instructing the well (e.g., through first controller 1718) to reverse a direction of the wireline or decrease a lowering speed. Alternatively, or in parallel, the method 2200 may comprise providing feedback to an operator computer (e.g., operator computer 1816 in
Computing platform(s) 1902 may be configured by machine-readable instructions 1906. Machine-readable instructions 1906 may include one or more instruction modules. The instruction modules may include computer program modules. The instruction modules may include one or more of acoustic/vibration data acquiring module 1910, acoustic/vibration data transferring module 1912, optional acoustic/vibration data converting module 1914, data comparing module 1916, label assignment module 1918, indication generating module 1920, wireline direction adjusting module 1922, wireline speed adjusting module 1924, model training module 1930, and/or other instruction modules.
In some embodiments, acoustic or vibration sensor(s) 1908 may be in communication with the computing platform(s) 1902 and may be configured to provide raw data to the processor(s) 1938. Furthermore, the acoustic or vibration sensor(s) 1908 may be configured to be in direct physical contact with fluid within a well (or direct physical contact with a component of the well, such as a pipe). In some cases, the acoustic or vibration sensor(s) 1908 may be designed for high frequency applications (e.g., having greater than 1000 sample/second rate). In an embodiment, the acoustic or vibration sensor 1908 may include a piezoelectric material configured to generate a current or voltage proportional to an amplitude of vibration of the piezoelectric material.
Acoustic/vibration data acquiring module 1910 may be configured to acquire acoustic or vibration data in a time domain from the sensor(s) 1908.
Acoustic/vibration data transferring module 1912 may be configured to transfer the acoustic or vibration data to a data comparing module, for instance, when time series data is analyzed. Alternatively, the acoustic or vibration data transferring module 1912 may be configured to transfer the data to a spectrum analyzer (i.e., when frequency spectrum data is used for analysis). It should be noted that, the spectrum analyzer may or may not be part of the same computing platform that various other modules in
In some embodiments, the computing system 1900 may comprise an optional acoustic/vibration data converting module 1914, wherein the acoustic/vibration data converting module 1914 may be configured to convert the acoustic or vibration data from the time domain to a frequency domain, for instance, via the spectrum analyzer.
Data comparing module 1916 may be configured to compare the acoustic or vibration data in the time series (or alternatively, in the frequency domain) to a model trained on time series signatures (or frequency signatures) corresponding to known incidences of wireline friction and sticking. In some examples, training the model may include building or updating a wireline friction curve. By way of non-limiting example, the comparing may comprise considering a recorded/measured amplitude versus a baseline or steady state amplitude, and/or evaluating a variation in frequency components of the acoustic or vibration data. In another non-limiting example, the comparing may consider a number of frequency spikes, a width of the frequency spikes, and an amplitude of the frequency spikes in the frequency signatures. In other words, under normal operation, the acoustic or vibration data may have a consistent set of one or more frequency components in the spectral regime. However, when friction begins to build, the one or more frequency components may change or increase in number, which may be indicative of increased friction. Increased amplitude, rate of amplitude spikes, or variation in amplitudes over time in the time domain may also be analyzed in the comparing.
Label assignment module 1918 may be configured to assign one of a plurality of labels to the acoustic or vibration data in the time series (or frequency domain) based on the comparing. By way of non-limiting example, the plurality of labels may include an excessive wireline friction state or a wireline sticking state.
In some cases, for instance, if the excessive friction state label is assigned, an indication generating module 1920 may be configured to generate and display a first indication on an operator display suggesting a reduction in wireline speed, a reversal of wireline direction (e.g., up instead of downhole), or use of a smaller plug. Additionally, if a sticking state is assigned, the indication generating module 1920 may be configured to generate and display a second indication on an operator display, wherein the second indication may suggest initiation of fishing procedures.
In some embodiments, in response to the excessive friction state label, a wireline direction adjusting module 1922 may be configured to reverse a direction of the wireline until new acoustic or vibration data in the time series (or frequency domain) fall within parameters of an optimal set of known time series signatures or frequency domain signatures, or until the excessive friction state label is no longer assigned.
In some cases, in response to the excessive friction state label, wireline speed adjusting module 1924 may be configured to decrease a lowering speed of the wireline until new acoustic or vibration data in the time series (or frequency domain) fall within parameters of an optimal set of known time series signatures (or frequency signatures), or until the excessive friction state label is no longer assigned.
While not shown, the machine-readable instructions 1906 may also include, or may replace one or more of wireline direction adjusting module 1922 and wireline speed adjusting module 1924 with: an increased flush time module; a perform dedicated flush module; and/or a smaller plug module. Any one or more of these optional modules may be called on to control the wireline, which may further assist in minimizing the risk of excessive friction and/or sticking.
In some embodiments, model training module 1930 may be configured to train the model of acoustic or vibration data in the time series (or frequency domain) using the one of the plurality of labels and the acoustic or vibration data in the time series (or frequency domain).
In some implementations, computing platform(s) 1902, remote platform(s) 1904, and/or external resources 1934 may be operatively linked via one or more electronic communication links. For example, such electronic communication links may be established, at least in part, via a network such as the Internet and/or other networks. It will be appreciated that this is not intended to be limiting, and that the scope of this disclosure includes implementations in which computing platform(s) 1902, remote platform(s) 1904, and/or external resources 1934 may be operatively linked via some other communication media.
A given remote platform 1904 may include one or more processors configured to execute computer program modules. The computer program modules may be configured to enable an expert or user associated with the given remote platform 1904 to interface with system 1900 and/or external resources 1934, and/or provide other functionality attributed herein to remote platform(s) 1904. By way of non-limiting example, a given remote platform 1904 and/or a given computing platform 1902 may include one or more of a server, a desktop computer, a laptop computer, a handheld computer, a tablet computing platform, a NetBook, a Smartphone, a gaming console, and/or other computing platforms.
External resources 1934 may include sources of information outside of system 1900, external entities participating with system 1900, and/or other resources. For instance, external data may be fed into the model to help with initial training. In some implementations, some or all of the functionality attributed herein to external resources 1934 may be provided by resources included in system 1900.
Computing platform(s) 1902 may include electronic storage 1936, one or more processors 1938, and/or other components. Computing platform(s) 1902 may include communication lines, or ports to enable the exchange of information with a network and/or other computing platforms. Illustration of computing platform(s) 1902 in
Electronic storage 1936 may comprise non-transitory storage media that electronically store information. The electronic storage media of electronic storage 1936 may include one or both of system storage that is provided integrally (i.e., substantially non-removable) with computing platform(s) 1902 and/or removable storage that is removably connectable to computing platform(s) 1902 via, for example, a port (e.g., a USB port, a firewire port, etc.) or a drive (e.g., a disk drive, etc.). Electronic storage 1936 may include one or more of optically readable storage media (e.g., optical disks, etc.), magnetically readable storage media (e.g., magnetic tape, magnetic hard drive, floppy drive, etc.), electrical charge-based storage media (e.g., EEPROM, RAM, etc.), solid-state storage media (e.g., flash drive, etc.), and/or other electronically readable storage media. Electronic storage 1936 may include one or more virtual storage resources (e.g., cloud storage, a virtual private network, and/or other virtual storage resources). Electronic storage 1936 may store software algorithms, information determined by processor(s) 1938, information received from computing platform(s) 1902, information received from remote platform(s) 1904, and/or other information that enables computing platform(s) 1902 to function as described herein.
Processor(s) 1938 may be configured to provide information processing capabilities in computing platform(s) 1902. As such, processor(s) 1938 may include one or more of a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanisms for electronically processing information. Although processor(s) 1938 is shown in
It should be appreciated that although modules 1910, 1912, 1914, 1916, 1918, 1920, 1922, 1924, and/or 1930 are illustrated in
In some cases, rotary and coiled tubing drilling often uses a positive displacement motor (PDM) to rotate the drill bit. A PDM is a downhole tool that uses hydraulic power from fluid flowing therethrough to drive a drill bit. During drilling operations, the unloaded PDM rotates at a constant RPM and achieves a “freespin” motor pressure, with respect to the fluid flow rate. As the drill bit encounters the bottom of the hole and force is transferred to the bit, referred to as weight-on-bit (WOB), the motor will sense an increase in torque. The increase in torque is a result of increased resistance to rotating at the constant RPM (assuming a constant flow rate). In turn, the PDM requires additional pressure to turn the motor at the constant RPM while under increased resistance. If the resistance increases to a condition which prohibits the PDM from rotating (i.e., excessive WOB), a motor stall is encountered. Stalling usually occurs when the application of excessive weight on bit or hole sloughing stops the bit from rotating and when the power section of the drilling motor is not capable of providing enough torque to power through. During a motor stall, the motor stops turning, the downhole fluid path is severely restricted, and the surface pump pressure dramatically increases. This pressure increase is developed because the rotor is no longer able to rotate inside the stator, forming a long seal between the two. If fluid circulation continues during a stall, the drilling fluid forces its way through the power section by deflecting the stator rubber. Drilling fluid will still circulate through the motor, but the bit will not turn. Operating in this state will erode and possibly chunk the stator in a very short period of time, resulting in extensive damage. If the PDM is damaged, the drilling process will be stopped, and the coiled tubing will be fatigue-cycled as the bit is pulled off bottom and run back into the hole to start drilling again.
To avoid stalls and damage to the drill bits or PDM, each PDM has a specification sheet that provides a user with information about the operation of the PDM, and sometimes includes a torque curve to help operators to maximize drilling effectiveness while avoiding damage to the motor or drill bit. The torque curve may identify a differential pressure versus rotations per minute (“RPM”) and/or torque curve for the PDM at a given flow rate through the PDM. As the differential pressure increases, the RPM generally decrease toward zero, at which point the PDM stalls.
Since it can be difficult to determine when a PDM is near stalling or when one has pushed past the Full Load line on a torque curve, accurate understanding of PDM operation and location on torque may serve to optimize drilling operations. In some circumstances, incomplete knowledge about the PDM and drill head may also lead to drill bit wear/damage and PDM wear/damage. For instance, inefficient bottom hole cleaning may lead to wear on the drill bit and vibrations that decrease the efficiency of drilling. Likewise, after a certain value of rotary speed has been met, Rate of Penetration (ROP) decelerates as the bit starts skating on top of the rock rather than getting good penetration of the cutting structure. This can be caused by instability of the drilling assembly in the wellbore.
Via spectral analysis of acoustic or vibrational data in a well, pump line, or other component in fluid communication with the well, one can use a trained model to identify spectral signatures of activities and conditions to help peer through the “fog” that traditional operators face while evaluating motor torque and drill bit wear. In these embodiments, an acoustic or vibrational sensor can be arranged to monitor signals from components within the same well being monitored.
The machine learning component 2312 may be configured to use any new insights from the converted frequency domain version of the data to further train the model 2314. It may be also be configured to send a signal or instruction in response to labeling of the new frequency signature. For instance, if it applies a stalled state label to the new frequency signature, then a warning may be sent to an optional operator computer 2316. In some other cases, the machine learning component 2312 may also send instructions to the optional operator computer 2316 to visualize a point on a torque curve on the optional computer's 2316 display. For instance, as the new frequency signature changes (e.g., as a PDM moves from optimum, to stalling, to stalled), the machine learning component 2312 may instruct the optional operator computer 2316 to visualize said changes on the displayed torque curve. In response, the operator may instruct the well to adjust fluid flow through a fluid flow controller 2318 and/or to adjust weight on bit via a weight on bit controller 2320. Alternatively, the machine learning component 2312 may be configured to directly control the controllers 2318 and 2320, with or without involving the optional operator computer 2316, as a closed feedback loop.
In an embodiment, the model 2314 can be trained to recognize spectral signatures corresponding to one or more states of the drill bit, including, but not limited to, a stalled state, a stalling state, an optimal state, a sub-optimal state, and a zero-weight state. Once acoustic or vibrational data is converted to the frequency domain, the model 2314 can classify the spectral signature according to one or more labels including, but not limited to, a stalled state label, a stalling state label, an optimal state label, a sub-optimal state label, and a zero-weight state label.
Although this discussion has often used operating states of a PDM as illustrative examples, one of skill in the art will appreciate that these systems, methods, and apparatus can be applied to a variety of different diagnostics within the drilling space. For instance, but not to limit this disclosure, the herein disclosed systems, methods, and apparatus for spectral analysis and machine learning of acoustic signatures can be applied to the following: drill bit wear; drill bit damage; PDM damage; PDM lifetime; casing wear; drill string buckling; drill string centering in the bore; drill bit stick slip condition; drill string tension; weight on bit; type of formation that bit is drilling through; relative hardness of formation that bit is drilling though; effectiveness of debris removal from drill bit; effectiveness of debris removal from entire wellbore; pump chunking; pump truck damage; pump truck failure; drill string sticking; coiled tubing sticking; coiled tubing collapse; coiled tubing pin hole damage; excessive coiled tubing pressure; corrosion; hydraulic fracturing screenout; perforation effectiveness; perforation cluster efficiency; frack pumping effectiveness; and proppant effectiveness.
In some implementations, method 2500 may be implemented in one or more processing devices (e.g., a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanisms for electronically processing information). The one or more processing devices may include one or more devices executing some or all of the operations of method 2500 in response to instructions stored electronically on an electronic storage medium. The one or more processing devices may include one or more devices configured through hardware, firmware, and/or software to be specifically designed for execution of one or more of the operations of method 2000.
An operation 2502 may include providing an acoustic or vibrational sensor configured for direct physical contact with fluid within a well.
An operation 2504 may include acquiring acoustic or vibration data in the fluid via the acoustic or vibrational sensor in a time domain. Operation 2504 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to vibration data acquiring module, in accordance with one or more implementations.
An operation 2506 may include transferring the acoustic or vibration data to a spectrum analyzer. Operation 2506 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to acoustic vibration data transferring module 2612, shown in
An operation 2508 may include converting the acoustic or vibration data from the time domain to a frequency domain via the spectrum analyzer. Operation 2508 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to acoustic vibration data converting module 2614, shown in
An operation 2510 may include comparing the acoustic or vibration data in the frequency domain to a model trained on frequency signatures corresponding to known incidences of positive displacement motor stalling and/or low positive displacement motor torque. Operation 2510 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to data comparing module 2616, shown in
An operation 2512 may include assigning one of a plurality of labels to the acoustic or vibration data in the frequency domain based on the comparing. The plurality of labels may include a stalled state label, an optimal state label, a sub-optimal state label, and a zero-weight state label. Operation 2512 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to label assignment module 2618, shown in
An operation 2514 may include if the stalled state label is assigned, then generating a first indication on an operator display suggesting a reduction in fluid flow to the drill head and/or weight on bit. Operation 2514 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to indication generating module 2620, shown in
An operation 2516 may include if the sub-optimal state label or zero-weight state label is assigned, then generating a second indication on the operator display suggesting an increase in fluid flow to the drill head and/or weight on bit. Operation 2516 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to indication generating module 2620, adjusting module 2622, flow decrease module 2626 (i.e., for controlling fluid flow) and/or a weight on bit increasing module, in accordance with one or more implementations.
The method 2500 may further include, in response to the stalled state label, the sub-optimal state label, or the zero-weight state label, adjusting weight on bit and/or fluid flow to the drill head until new acoustic or vibration data in the frequency domain falls within parameters of an optimal set of known frequency signatures or is assigned the optimal state label. Operation 2518 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to weight adjusting module 2622, weight decrease module 2624, and/or a flow decrease module 2626 in accordance with one or more implementations. Additionally or alternatively, a fluid flow and/or a weight on bit increasing module may be utilized to perform operation 2518.
The method 2500 may further include, in response to the stalled state label, decreasing a weight on bit. Operation 2520 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to adjusting module 2622 or weight decrease module 2624, in accordance with one or more implementations.
The method 2500 may further include, in response to the stalled state label, decreasing fluid flow to the drill head. Operation 2522 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to adjusting module 2622 or flow decrease module 2626, in accordance with one or more implementations.
The method 2500 may further include, in response to the sub-optimal state label or zero-weight state label, increasing weight on bit and/or fluid flow to the drill head. Operation 2524 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to adjusting module 2622, in accordance with one or more implementations. Alternatively, a fluid flow and/or a weight on bit increasing module may be utilized to perform 2524.
The method 2500 may further include training the model of acoustic or vibration data in the frequency domain using the one of the plurality of labels and the acoustic or vibration data in the frequency domain Operation 2526 may be performed by one or more hardware processors configured by machine-readable instructions including a module that is the same as or similar to model training module 2630, in accordance with one or more implementations.
Computing platform(s) 2602 may be configured by machine-readable instructions 2606. Machine-readable instructions 2606 may include one or more instruction modules. The instruction modules may include computer program modules. The instruction modules may include one or more of vibration data acquiring module 2610, acoustic vibration data transferring module 2612, acoustic vibration data converting module 2614, data comparing module 2616, label assignment module 2618, indication generating module 2620, weight adjusting module 2622, weight decrease module 2624, flow decrease module 2626, drill head increasing module 2628, model training module 2630, and/or other instruction modules, such as a fluid flow and/or a weight on bit increasing module.
Acoustic or vibrational sensor(s) 2608 can be in communication with the computing platform(s) 2602 and provide raw data to the processor(s) 2638. The acoustic or vibration sensor(s) 2608 can be configured for direct physical contact with fluid within a well. The sensor(s) 2608 can be high frequency (e.g., having greater than 1000 sample/second rate, although other sampling rates are contemplated in different embodiments). The acoustic or vibrational sensor 2608 may include a piezoelectric material configured to generate a current or voltage proportional to an amplitude of vibration of the piezoelectric material.
Vibration data acquiring module 2610 may be configured to acquire acoustic or vibration data in a time domain from the sensor(s) 2608.
Acoustic vibration data transferring module 2612 may be configured to transfer the acoustic or vibration data to a spectrum analyzer. The spectrum analyzer may or may not be part of the same computing platform that various other modules in
Acoustic vibration data converting module 2614 may be configured to convert the acoustic or vibration data from the time domain to a frequency domain via the spectrum analyzer.
Data comparing module 2616 may be configured to compare the acoustic or vibration data in the frequency domain to a model trained on frequency signatures corresponding to known incidences of various positive displacement motor states such as stalling, optimum torque, sub-optimum torque, and zero-weight conditions. Training the model may include building or updating a torque curve. By way of non-limiting example, the comparing may consider a number of frequency spikes, a width of the frequency spikes, and an amplitude of the frequency spikes in the frequency signatures.
Label assignment module 2618 may be configured to assign one of a plurality of labels to the acoustic or vibration data in the frequency domain based on the comparing. By way of non-limiting example, the plurality of labels may include a stalled state label, an optimal state label, a sub-optimal state label, and a zero-weight state label.
Indication generating module 2620 may be configured to, if the stalled state label is assigned, then generate a first indication on an operator display suggesting a reduction in fluid flow to the drill head and/or reduction in weight on bit.
Indication generating module 2620 may be configured to, if the sub-optimal state label or zero-weight state label is assigned, then generate a second indication on the operator display suggesting an increase in fluid flow to the drill head and/or weight on bit.
Adjusting module 2622 may be configured to, in response to the stalled state label, the sub-optimal state label, or the zero-weight state label, adjust weight on bit and/or fluid flow to the drill head until new acoustic or vibration data in the frequency domain falls within parameters of an optimal set of known frequency signatures or is assigned the optimal state label.
Weight decrease module 2624 may be configured to, in response to the stalled state label, decrease a weight on bit.
Flow decrease module 2626 may be configured to, in response to the stalled state label, decrease fluid flow to the drill head.
Drill head increasing module 2628 may be configured to, in response to the sub-optimal state label or zero-weight state label, increase weight on bit and/or fluid flow to the drill head. In some other cases, one or more of a fluid flow increasing module and a weight on bit increasing module may be configured to perform a same or similar function as the drill head increasing module 2628.
Model training module 2630 may be configured to train the model of acoustic or vibration data in the frequency domain using the one of the plurality of labels and the acoustic or vibration data in the frequency domain.
In some implementations, computing platform(s) 2602, remote platform(s) 2604, and/or external resources 2634 may be operatively linked via one or more electronic communication links. For example, such electronic communication links may be established, at least in part, via a network such as the Internet and/or other networks. It will be appreciated that this is not intended to be limiting, and that the scope of this disclosure includes implementations in which computing platform(s) 2602, remote platform(s) 2604, and/or external resources 2634 may be operatively linked via some other communication media.
A given remote platform 2604 may include one or more processors configured to execute computer program modules. The computer program modules may be configured to enable an expert or user associated with the given remote platform 2604 to interface with system 2600 and/or external resources 2634, and/or provide other functionality attributed herein to remote platform(s) 2604. By way of non-limiting example, a given remote platform 2604 and/or a given computing platform 2602 may include one or more of a server, a desktop computer, a laptop computer, a handheld computer, a tablet computing platform, a NetBook, a Smartphone, a gaming console, and/or other computing platforms.
External resources 2634 may include sources of information outside of system 2600, external entities participating with system 2600, and/or other resources. For instance, external data may be fed into the model to help with initial training. In some implementations, some or all of the functionality attributed herein to external resources 2634 may be provided by resources included in system 2600.
Computing platform(s) 2602 may include electronic storage 2636, one or more processors 2638, and/or other components. Computing platform(s) 2602 may include communication lines, or ports to enable the exchange of information with a network and/or other computing platforms. Illustration of computing platform(s) 2602 in FIG. 26 is not intended to be limiting. Computing platform(s) 2602 may include a plurality of hardware, software, and/or firmware components operating together to provide the functionality attributed herein to computing platform(s) 2602. For example, computing platform(s) 2602 may be implemented by a cloud of computing platforms operating together as computing platform(s) 2602.
Electronic storage 2636 may comprise non-transitory storage media that electronically stores information. The electronic storage media of electronic storage 2636 may include one or both of system storage that is provided integrally (i.e., substantially non-removable) with computing platform(s) 2602 and/or removable storage that is removably connectable to computing platform(s) 2602 via, for example, a port (e.g., a USB port, a firewire port, etc.) or a drive (e.g., a disk drive, etc.). Electronic storage 2636 may include one or more of optically readable storage media (e.g., optical disks, etc.), magnetically readable storage media (e.g., magnetic tape, magnetic hard drive, floppy drive, etc.), electrical charge-based storage media (e.g., EEPROM, RAM, etc.), solid-state storage media (e.g., flash drive, etc.), and/or other electronically readable storage media. Electronic storage 2636 may include one or more virtual storage resources (e.g., cloud storage, a virtual private network, and/or other virtual storage resources). Electronic storage 2636 may store software algorithms, information determined by processor(s) 2638, information received from computing platform(s) 2602, information received from remote platform(s) 2604, and/or other information that enables computing platform(s) 2602 to function as described herein.
Processor(s) 2638 may be configured to provide information processing capabilities in computing platform(s) 2602. As such, processor(s) 2638 may include one or more of a digital processor, an analog processor, a digital circuit designed to process information, an analog circuit designed to process information, a state machine, and/or other mechanisms for electronically processing information. Although processor(s) 2638 is shown in
It should be appreciated that although modules 2608, 2610, 2612, 2614, 2616, 2618, 2620, 2622, 2624, 2626, 2628, 2630, and/or 2632 are illustrated in
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein, the acoustic vibrations are caused by the wireline rubbing against walls of a borehole.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the sensor samples at greater than 1 kHz. In some examples of the method, system, and non-transient computer-readable storage medium described herein the sensor is an acoustic sensor.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the converter uses a Fast-Fourier transform to convert the electrical signal in a window of time into a current frequency domain spectrum.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system considers one or more of a number of frequency spikes in the current frequency domain spectrum, a width of one or more frequency spikes in the current frequency domain spectrum, and an amplitude of the one or more frequency spikes in the current frequency domain spectrum.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system is trained on frequency domain spectra measured during previous hydraulic fracturing operations as a machine-learning input and associated well outcomes as machine-learning outputs. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system is trained on previous frequency domain spectra as a machine-learning input and associated wireline sticking events as a machine-learning output. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system is trained to classify the current frequency domain spectrum on a grouping of previous frequency domain spectra measured during previous wireline operations that most closely matches the current frequency domain spectra
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the associated wireline sticking events comprise a full sticking event. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system is trained on wireline sticking events as a machine-learning input and associated previous frequency domain spectra as a machine-learning output. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system is configured to classify based on a grouping of frequency domain spectra measured during previous wireline operations that most closely match the current frequency domain spectra. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system is further configured to analyze the electrical signal for the window of time in the time domain in conjunction with analyzing the current frequency domain spectrum to classify the current frequency domain spectrum
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the sensor is configured to be in contact with the fluid in the well or with a surface of a circulating fluid line or standpipe at the wellhead. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the sensor is configured to be in contact with the fracking fluid in the well or with a surface of the circulating fluid line or the standpipe at the wellhead.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein: one or more of the sensor and the machine-learning system is configured to measure and analyze pressure sensor data during the window of time. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the machine-learning system is configured to classify the current frequency domain spectrum as associated with a start or end of plug transport down the well. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the converter is on a local converter and the current frequency domain spectrum is transported via a large area network to a remote server hosting the machine-learning system. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the electrical signal is transported via a large area network to a remote converter configured to convert the electrical signal in the window of time into the current frequency domain spectrum.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein wireline friction is identified by an increase in a number or width of frequency peaks in the current frequency domain spectrum, or alternatively, a decrease in the number or width of frequency peaks in the current frequency domain spectrum.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein: the converter is a spectrum analyzer. In some embodiments, the system may further comprise a wellbore with a casing; and a fracking pump.
In some examples of the method and non-transient computer-readable storage medium described herein the analyzing considers a number of frequency spikes in the current frequency domain spectrum, a width of one or more frequency spikes in the current frequency domain spectrum, and/or an amplitude of the frequency spikes in the current frequency domain spectrum. In some examples of the method and non-transient computer-readable storage medium described herein the converting comprises a Fast-Fourier transform.
In some examples of the method and non-transient computer-readable storage medium described herein the machine-learning system is trained on previous frequency domain spectra as a machine-learning input and associated wireline sticking events as a machine-learning output. In some examples of the method and non-transient computer-readable storage medium described herein the classifying is based on a grouping of frequency domain spectra measured during previous wireline operations that most closely matches the current frequency domain spectra. In some examples of the method and non-transient computer-readable storage medium described herein the machine-learning system is trained on wireline sticking events as a machine-learning input and associated previous frequency domain spectra as a machine-learning output.
In some examples of the method and non-transient computer-readable storage medium described herein the classifying is based on a grouping of frequency domain spectra measured during previous wireline operations that most closely match the current frequency domain spectra.
In some examples of the method and non-transient computer-readable storage medium described herein, the method further comprises analyzing the electrical signal for the window of time in the time domain in conjunction with analyzing the current frequency domain spectrum to perform the classifying. In some examples of the method and non-transient computer-readable storage medium described herein, the method further comprises analyzing the electrical signal for the window of time in the time domain and using this in addition to the analyzing the current frequency domain spectrum to perform the classifying. In some examples of the method and non-transient computer-readable storage medium described herein, the performing the classifying is based at least in part on an analysis of pressure sensor data during the window of time.
In some examples of the method and non-transient computer-readable storage medium described herein, the method further comprises classifying the current frequency domain spectrum as associated with a start or end of plug transport down the well. In some examples of the method and non-transient computer-readable storage medium described herein, the converting is performed on a local converter and the current frequency domain spectrum is transported via a large area network to a remote server hosting the machine-learning system. In some examples of the method and non-transient computer-readable storage medium described herein the electrical signal is transported via a large area network to a remote converter for performing the converting.
In some examples of the method and non-transient computer-readable storage medium described herein, the method further comprises adjusting parameters of subsequent wireline operations to change how a subsequent frequency domain spectrum is classified. In some examples of the method and non-transient computer-readable storage medium described herein the classifying is based on the acoustic vibrations increasing a threshold level above an average of acoustic vibrations during wireline descent.
In some examples of the method and non-transient computer-readable storage medium described herein, increased wireline friction is identified by an increase in a number or width of frequency peaks in the current frequency domain spectrum. For instance, the increased wireline friction may be identified based at least in part on identifying the increase in the number or width of frequency peaks in the current frequency domain spectrum.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the parameter of the wireline operation is perforation gun pressure.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the parameter of the wireline operation is fracking stage duration.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the parameter of the wireline operation is a pressure of fluid forced into a subterranean formation.
In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the parameter of the wireline operation is pH of the fracking fluid pumped into the well. In some examples of the method, system, computing platform, and non-transient computer-readable storage medium described herein the parameter of the wireline operation is a length of flush time.
One aspect of the present disclosure relates to a system configured for optimizing drill head positive displacement motor torque. The system may include one or more hardware processors configured by machine-readable instructions. The processor(s) may be configured to provide an acoustic or vibration sensor configured for direct fluid communication with fluid within a well. The processor(s) may be configured to acquire acoustic or vibration data in the fluid via the acoustic or vibration sensor in a time domain. The processor(s) may be configured to transfer the acoustic or vibration data to a spectrum analyzer. The processor(s) may be configured to convert the acoustic or vibration data from the time domain to a frequency domain via the spectrum analyzer. The processor(s) may be configured to compare the acoustic or vibration data in the frequency domain to a model trained on frequency signatures corresponding to known incidences of positive displacement motor stalling and/or low positive displacement motor torque. The processor(s) may be configured to assign one of a plurality of labels to the acoustic or vibration data in the frequency domain based on the comparing. The plurality of labels may include a stalled state label, an optimal state label, a sub-optimal state label, and a zero-weight state label. The processor(s) may be configured to, if the stalled state label is assigned, then generate a first indication on an operator display suggesting a reduction in fluid flow to the drill head and/or weight on bit. The processor(s) may be configured to, if the sub-optimal state label or zero-weight state label is assigned, then generate a second indication on the operator display suggesting an increase in fluid flow to the drill head and/or weight on bit.
Another aspect of the present disclosure relates to a method for optimizing drill head positive displacement motor torque. The method may include providing an acoustic or vibration sensor configured for direct fluid communication with fluid within a well. The method may include acquiring acoustic or vibration data in the fluid via the acoustic or vibration sensor in a time domain. The method may include transferring the acoustic or vibration data to a spectrum analyzer. The method may include converting the acoustic or vibration data from the time domain to a frequency domain via the spectrum analyzer. The method may include comparing the acoustic or vibration data in the frequency domain to a model trained on frequency signatures corresponding to known incidences of positive displacement motor stalling and/or low positive displacement motor torque. The method may include assigning one of a plurality of labels to the acoustic or vibration data in the frequency domain based on the comparing. The plurality of labels may include a stalled state label, an optimal state label, a sub-optimal state label, and a zero-weight state label. The method may include, if the stalled state label is assigned, then generating a first indication on an operator display suggesting a reduction in fluid flow to the drill head and/or weight on bit. The method may include, if the sub-optimal state label or zero-weight state label is assigned, then generating a second indication on the operator display suggesting an increase in fluid flow to the drill head and/or weight on bit.
Yet another aspect of the present disclosure relates to a non-transient computer-readable storage medium having instructions embodied thereon, the instructions being executable by one or more processors to perform a method for optimizing drill head positive displacement motor torque. The method may include providing an acoustic or vibration sensor configured for direct fluid communication with fluid within a well. The method may include acquiring acoustic or vibration data in the fluid via the acoustic or vibration sensor in a time domain. The method may include transferring the acoustic or vibration data to a spectrum analyzer. The method may include converting the acoustic or vibration data from the time domain to a frequency domain via the spectrum analyzer. The method may include comparing the acoustic or vibration data in the frequency domain to a model trained on frequency signatures corresponding to known incidences of positive displacement motor stalling and/or low positive displacement motor torque. The method may include assigning one of a plurality of labels to the acoustic or vibration data in the frequency domain based on the comparing. The plurality of labels may include a stalled state label, an optimal state label, a sub-optimal state label, and a zero-weight state label. The method may include, if the stalled state label is assigned, then generating a first indication on an operator display suggesting a reduction in fluid flow to the drill head and/or weight on bit. The method may include, if the sub-optimal state label or zero-weight state label is assigned, then generating a second indication on the operator display suggesting an increase in fluid flow to the drill head and/or weight on bit.
Still another aspect of the present disclosure relates to a computing platform configured for optimizing drill head positive displacement motor torque. The computing platform may include a non-transient computer-readable storage medium having executable instructions embodied thereon. The computing platform may include one or more hardware processors configured to execute the instructions. The processor(s) may execute the instructions to provide an acoustic or vibration sensor configured for direct fluid communication with fluid within a well. The processor(s) may execute the instructions to acquire acoustic or vibration data in the fluid via the acoustic or vibration sensor in a time domain. The processor(s) may execute the instructions to transfer the acoustic or vibration data to a spectrum analyzer. The processor(s) may execute the instructions to convert the acoustic or vibration data from the time domain to a frequency domain via the spectrum analyzer. The processor(s) may execute the instructions to compare the acoustic or vibration data in the frequency domain to a model trained on frequency signatures corresponding to known incidences of positive displacement motor stalling and/or low positive displacement motor torque. The processor(s) may execute the instructions to assign one of a plurality of labels to the acoustic or vibration data in the frequency domain based on the comparing. The plurality of labels may include a stalled state label, an optimal state label, a sub-optimal state label, and a zero-weight state label. The processor(s) may execute the instructions to, if the stalled state label is assigned, then generate a first indication on an operator display suggesting a reduction in fluid flow to the drill head and/or weight on bit. The processor(s) may execute the instructions to, if the sub-optimal state label or zero-weight state label is assigned, then generate a second indication on the operator display suggesting an increase in fluid flow to the drill head and/or weight on bit.
Even another aspect of the present disclosure relates to a system configured for optimizing drill head positive displacement motor torque. The system may include means for providing an acoustic or vibration sensor configured for direct fluid communication with fluid within a well. The system may include means for acquiring acoustic or vibration data in the fluid via the acoustic or vibration sensor in a time domain. The system may include means for transferring the acoustic or vibration data to a spectrum analyzer. The system may include means for converting the acoustic or vibration data from the time domain to a frequency domain via the spectrum analyzer. The system may include means for comparing the acoustic or vibration data in the frequency domain to a model trained on frequency signatures corresponding to known incidences of positive displacement motor stalling and/or low positive displacement motor torque. The system may include means for assigning one of a plurality of labels to the acoustic or vibration data in the frequency domain based on the comparing. The plurality of labels may include a stalled state label, an optimal state label, a sub-optimal state label, and a zero-weight state label. The system may include means for, if the stalled state label is assigned, then generating a first indication on an operator display suggesting a reduction in fluid flow to the drill head and/or weight on bit. The system may include means for, if the sub-optimal state label or zero-weight state label is assigned, then generating a second indication on the operator display suggesting an increase in fluid flow to the drill head and/or weight on bit.
Some portions of the present disclosure are presented in terms of algorithms or symbolic representations of operations on data bits or binary digital signals stored within a computing system memory, such as a computer memory. These algorithmic descriptions or representations are examples of techniques used by those of ordinary skill in the data processing arts to convey the substance of their work to others skilled in the art. An algorithm is a self-consistent sequence of operations or similar processing leading to a desired result. In this context, operations or processing involves physical manipulation of physical quantities. Typically, although not necessarily, such quantities may take the form of electrical or magnetic signals capable of being stored, transferred, combined, compared or otherwise manipulated. It has proven convenient at times, principally for reasons of common usage, to refer to such signals as bits, data, values, elements, symbols, characters, terms, numbers, numerals or the like. It should be understood, however, that all of these and similar terms are to be associated with appropriate physical quantities and are merely convenient labels. Unless specifically stated otherwise, it is appreciated that throughout this specification discussions utilizing terms such as “processing,” “computing,” “calculating,” “determining,” and “identifying” or the like refer to actions or processes of a computing device, such as one or more computers or a similar electronic computing device or devices, that manipulate or transform data represented as physical electronic or magnetic quantities within memories, registers, or other information storage devices, transmission devices, or display devices of the computing platform.
As will be appreciated by one skilled in the art, aspects of the present disclosure may be embodied as a system, method or computer program product. Accordingly, aspects of the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.), or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “circuit,” “module” or “system.” Furthermore, aspects of the present disclosure may take the form of a computer program product embodied in one or more computer readable medium(s) having computer readable program code embodied thereon.
As used herein, the recitation of “at least one of A, B and C” is intended to mean “either A, B, C or any combination of A, B and C.” The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present disclosure. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the disclosure. Thus, the present disclosure is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
The present application is a National Phase application based on PCT/US20/64303 filed on Dec. 10, 2020 which claims priority to U.S. Provisional Application Nos. 63/058,548, 62/945,929, 62/945,949, 63/058,534, 62/945,953 and 62/945,957 entitled “Spectral Analysis, Machine Learning, and Frac Score Assignment to Acoustic Signatures of Fracking Events”, “Spectral Analysis and Machine Learning to Detect Offset Well Communication Using High Frequency Acoustic or Vibration Sensing”, “Acoustic and Vibrational Sensor Based Micro-Seismic Analysis”, “Spectral Analysis and Machine Learning of Acoustic Signature of Wireline Sticking”, “Spectral Analysis and Machine Learning of Well Activity Using High Frequency Pressure Sensing of Phase-Locked Stimulation”, and “Spectral Analysis and Machine Learning of Acoustic Signature of Drill Bit Positive Displacement Motor Torque and Drill Bit Wear”, respectively, each of which are assigned to the assignee hereof and hereby expressly incorporated by reference herein.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US20/64303 | 12/10/2020 | WO |
| Number | Date | Country | |
|---|---|---|---|
| 63058548 | Jul 2020 | US | |
| 62945929 | Dec 2019 | US | |
| 62945949 | Dec 2019 | US | |
| 63058534 | Jul 2020 | US | |
| 62945953 | Dec 2019 | US | |
| 62945957 | Dec 2019 | US |
