Fluid pumping operations framework

Abstract

A method can include receiving data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; generating an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assessing the performance of the fluid pumping job based at least in part on the inference.

Description

FLUID PUMPING OPERATIONS FRAMEWORKRELATED APPLICATION

[0001] This application claims priority to and the benefit of a U.S. Provisional Application having Serial No. 63 / 657,825, filed 8 June 2024, which is incorporated herein in its entirety.BACKGROUND

[0002] Various types of equipment may be utilized to perform fluid pumping operations with respect to a borehole in a subterranean environment. As an example, a cementing job may involve fluid pumping operations for various types of fluids to deposit cement in a subterranean environment, for example, consider depositing cement between a casing string and a borehole wall.SUMMARY

[0003] A method can include receiving data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; generating an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assessing the performance of the fluid pumping job based at least in part on the inference. A system can include one or more processors; memory accessible to at least one of the one or more processors; and processor-executable instructions stored in the memory and executable to instruct the system to: receive data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; generate an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assess the performance of the fluid pumping job based at least in part on the inference. One or more computer-readable storage media can include processor-executable instructions to instruct a computing system to: receive data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; generate an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assess theperformance of the fluid pumping job based at least in part on the inference. Various other apparatuses, systems, methods, etc., are also disclosed.

[0004] This summary is provided to introduce a selection of concepts that are further described below in the detailed description. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to be used as an aid in limiting the scope of the claimed subject matter.BRIEF DESCRIPTION OF THE DRAWINGS

[0005] Features and advantages of the described implementations may be more readily understood by reference to the following description taken in conjunction with the accompanying drawings.

[0006] Fig. 1 illustrates examples of an environment, equipment and an assembly;

[0007] Fig. 2 illustrates an example of an assembly;

[0008] Fig. 3 illustrates examples of equipment;

[0009] Fig. 4 illustrates an example of a method;

[0010] Fig. 5 illustrates an example of a graphical user interface;

[0011] Fig. 6 illustrates an example of a graphical user interface;

[0012] Fig. 7 illustrates an example of a graphical user interface;

[0013] Fig. 8 illustrates an example of a plot;

[0014] Fig. 9 illustrates an example of a graphical user interface;

[0015] Fig. 10 illustrates an example of a method;

[0016] Fig. 11 illustrates an example of a machine learning model architecture;

[0017] Fig. 12 illustrates an example of a machine learning model architecture;

[0018] Fig. 13 illustrates an example of a method and an example of a system; and

[0019] Fig. 14 illustrates an example of a networked computational equipment that may form a networked computational system.DETAILED DESCRIPTION

[0020] The following description includes the best mode presently contemplated for practicing the described implementations. This description is not to be taken in alimiting sense, but rather is made merely for the purpose of describing the general principles of the implementations. The scope of the described implementations should be ascertained with reference to the issued claims.

[0021] Below, various examples of equipment are described along with examples of data, graphical user interfaces, methods, framework components, etc.

[0022] Figs. 1 and 2 show an example of an environment 100, an example of a portion of a completion 101 , an example of equipment 120 and examples of assemblies 150 and 250, which may be involved in one or more well completions operations. As an example, the equipment 120 may include a rig, a turntable, a pump, drilling equipment, pumping equipment, equipment for deploying an assembly, a part of an assembly, etc. (see, e.g., Fig 3). As an example, the equipment 120 may include one or more controllers 122. As an example, a controller may include one or more processors, memory and instructions stored in memory that are executable by a processor, for example, to control one or more pieces of equipment (e.g., motors, pumps, sensors, etc.). As an example, a controller may include and / or be operatively coupled to a framework, which may include one or more trained ML models. As an example, the equipment 120 may be deployed at least in part at a well site and, optionally, in part at a remote site.

[0023] Fig. 1 shows an environment 100 that includes a subterranean formation into which a bore 102 extends where a tool 112 such as, for example, a drill string is disposed in the bore 102. As an example, the bore 102 may be defined in part by an angle (□); noting that while the bore 102 is shown as being deviated, it may be vertical or may include one or more vertical sections along with one or more deviated sections. As shown in an enlarged view with respect to an r, z coordinate system (e.g., a cylindrical coordinate system), a portion of the bore 102 includes casings 104-1 and 104-2 having casing shoes 106-1 and 106-2. As shown, cement annuli 103-1 and 103-2 are disposed between the bore 102 and the casings 104-1 and 104-2. Cement such as the cement annuli 103-1 and 103-2 may support and protect casings such as the casings 104-1 and 104-2 and when cement is disposed throughout various portions of a wellbore such as the bore 102, cement may help achieve zonal isolation.

[0024] In the example of Fig. 1 , the bore 102 has been drilled in sections or segments beginning with a large diameter section (see, e.g., n) followed by an intermediate diameter section (see, e.g., t2) and a smaller diameter section (see, e.g., rs). As an example, a large diameter section may be a surface casing section, which may be three or more feet in diameter and extend down several hundred feet to several thousandteet. A surface casing section may aim to prevent washout of loose unconsolidated formations. As to an intermediate casing section, it may aim to isolate and protect high pressure zones, guard against lost circulation zones, etc. As an example, intermediate casing may be set at about 6000 feet (e.g., about 2000 m) and extend lower with one or more intermediate casing portions of decreasing diameter (e.g., in a range from about thirteen to about five inches in diameter). A so-called production casing section may extend below an intermediate casing section and, upon completion, be the longest running section within a wellbore (e.g., a production casing section may be thousands of feet in length). As an example, production casing may be located in a target zone where the casing is perforated for flow of fluid into a bore of the casing.

[0025] As shown in Fig. 1 , the assembly 150 may include a pump down plug 160, a setting ball 162, a handling sub with a junk bonnet and setting tool extension 164, a rotating dog assembly (RDA) 166, an extension(s) 168, a mechanical running tool 172, a hydraulic running tool 174, a hydromechanical running tool 176, a retrievable cementing bushing 180, a slick joint assembly 182 and / or a liner wiper plug 184.

[0026] As shown in Fig. 2, the assembly 250 may include a liner top packer with a polished bore receptacle (PBR) 252, a coupling(s) 254, a mechanical liner hanger 262, a hydraulic liner hanger 264, a hydraulic liner hanger 266, a liner(s) 270, a landing collar with a ball seat 272, a landing collar without a ball seat 274, a float collar 276, a liner joint or joints 278 and / or 280, a float shoe 282 and / or a reamer float shoe 284.

[0027] While various examples of equipment are shown and described in Figs. 1 and 2, an assembly, a system, etc., may include one or more alternative and / or additional types of equipment, which may be utilized in field operations that may involve performance of one or more alternative and / or additional actions.

[0028] Fig. 3 also shows an example of equipment 370 and an example of equipment 380. Such equipment, which may be systems of components, may be suitable for use in the geologic environment 320. While the equipment 370 and 380 are illustrated as land-based, various components may be suitable for use in an offshore system. As an example, one or more pieces of equipment as shown in Fig. 3 may be utilized in performing one or more field operations (e.g., field actions, etc.).

[0029] The equipment 370 includes a platform 371 , a derrick 372, a crown block 373, a line 374, a traveling block assembly 375, drawworks 376, a landing 377 (e.g., a monkeyboard), and one or more fluid pumps 378 (e.g., to pump one or more types oftluid to a downhole environment). As an example, the line 374 may be controlled at least in part via the drawworks 376 such that the traveling block assembly 375 travels in a vertical direction with respect to the platform 371. For example, by drawing the line 374 in, the drawworks 376 may cause the line 374 to run through the crown block 373 and lift the traveling block assembly 375 skyward away from the platform 371 ; whereas, by allowing the line 374 out, the drawworks 376 may cause the line 374 to run through the crown block 373 and lower the traveling block assembly 375 toward the platform 371. Where the traveling block assembly 375 carries pipe (e.g., casing, etc.), tracking of movement of the traveling block assembly 375 may provide an indication as to how much pipe has been deployed. As an example, the equipment 370 may include one or more pumps or pumping systems. For example, consider a pump that may pump drilling fluid (e.g., mud), cement, etc.

[0030] A derrick may be a structure used to support a crown block and a traveling block operatively coupled to the crown block at least in part via line. A derrick may be pyramidal in shape and offer a suitable strength-to-weight ratio. A derrick may be movable as a unit or in a piece-by-piece manner (e.g., to be assembled and disassembled).

[0031] As an example, drawworks may include a spool, brakes, a power source and assorted auxiliary devices. Drawworks may controllably reel out and reel in line. Line may be reeled over a crown block and coupled to a traveling block to gain mechanical advantage in a “block and tackle” or “pulley” fashion. Reeling out and in of line may cause a traveling block (e.g., and whatever may be hanging underneath it), to be lowered into or raised out of a bore. Reeling out of line may be powered by gravity and reeling in by a motor, an engine, etc. (e.g., an electric motor, a diesel engine, etc.).

[0032] As an example, a crown block may include a set of pulleys (e.g., sheaves) that may be located at or near a top of a derrick or a mast, over which line is threaded. A traveling block may include a set of sheaves that may be moved up and down in a derrick or a mast via line threaded in the set of sheaves of the traveling block and in the set of sheaves of a crown block. A crown block, a traveling block and a line may form a pulley system of a derrick or a mast, which may enable handling of heavy loads (e.g., drillstring, pipe, casing, liners, etc.) to be lifted out of or lowered into a bore. As an example, line may be about a centimeter to about five centimeters in diameter as, for example, steel cable. Through use of a set of sheaves, such line may carry loads heavier than the line could support as a single strand.

[0033] As an example, a derrickman may be a rig crew member that works on a platform attached to a derrick or a mast. A derrick may include a landing on which a derrickman may stand. As an example, such a landing may be about 10 meters or more above a rig floor. In an operation referred to as trip out of the hole (TOH), a derrickman may wear a safety harness that enables leaning out from the work landing (e.g., monkeyboard) to reach pipe in located at or near the center of a derrick or a mast and to throw a line around the pipe and pull it back into its storage location (e.g., fingerboards), for example, until it a time at which it may be desirable to run the pipe back into the bore. As an example, a rig may include automated pipe-handling equipment such that the derrickman controls the machinery rather than physically handling the pipe.

[0034] As an example, a trip may refer to the act of pulling equipment from a bore and / or placing equipment in a bore. As an example, equipment may include a drillstring that may be pulled out of a hole and / or placed or replaced in a hole. As an example, a pipe trip may be performed where a drill bit has dulled or has otherwise ceased to drill efficiently and is to be replaced.

[0035] Equipment may be instrumented with various sensors, for example, consider a hook load sensor, a standpipe pressure sensor, a block position sensor and a rotational sensor for sensing rotations per minute (e.g., RPM). Such sensors may be at surface and generate surface data. As to standpipe pressure, it may be pressure in a standpipe such as a standpipe that receives fluid from a vibrating hose where a kelly hose may receive the fluid from the standpipe. Fluid may be an appropriate fluid for performing one or more types of operations (e.g., drilling fluid, cement, pills, etc.). As explained, equipment may include one or more pumps or pumping system suitable for pumping one or more types of fluids (see, e.g., the one or more pumps 378 of Fig. 3, etc.).

[0036] Fig. 4 shows an example of a method 400 for cementing, for example, to create a cement annulus between a casing and a formation. As shown, the method 400 can include a pre-job block 410 for performing pre-cementing job tasks, a pump spacer block 420 for pumping a spacer (e.g., a spacer fluid), a pump cement block 430 for pumping cement (e.g., a lead slurry, a tail slurry, etc.), a displace plug block 440 for displacing a plug, and a test pressure block 450 for testing pressure. In the example of Fig. 4, the method 400 is illustrated using simplified representations of equipment. For example, consider a borehole 412 into which a casing string 422 may be inserted where the casing string 422 may include one or more components 424 and 426 positioned at ornear a distal end. As an example, the one or more components 424 and 426 may be or include a guide shoe and / or a float shoe and / or a float collar. As an example, the method 400 may utilize a number of plugs, which may include a bottom plug 432 and a top plug 442, which, as shown, may be pumped into position within the casing string 422. Various cementing operations may utilize pressure measurements as may be acquired by one or more pressure sensors. For example, the test pressure block 450 may include utilizing a pressure sensor 452. As an example, volumes of fluids pumped into the casing string 422 may be tracked, for example, as part of a pre-job plan per the pre-job block 410.

[0037] A completions process known as primary cementing involves creating a cement sheath that can serve one or more functions. For example, consider cement sheath (e.g., annulus) that may provide a hydraulic seal that establishes zonal isolation, preventing fluid communication between producing zones in the borehole and blocking the escape of fluids to the surface. As to another function, a cement sheath may provide for anchoring and supporting a casing string and protecting casing material (e.g., steel, etc.) against corrosion by formation fluids. In various instances, failure to achieve these functions may limit a well and / or use.

[0038] In various instances, a primary cementing job may employ a two-plug cement placement method (e.g., consider a bottom plug and a top plug). For example, consider a workflow where, after drilling through an interval to a desired depth of a borehole, a drilling crew removes the drillstring, leaving the borehole filled with drilling fluid (e.g., mud). The crew may then lower a casing string to the bottom of the borehole. As explained, the bottom end of the casing string may be protected by a guide shoe or a float shoe. Such types of shoes tend to be tapered, for example, as bullet-nosed devices that may guide a casing string toward the center of a borehole to minimize contact with rough edges or washouts during installation. A guide shoe differs from a float shoe in that the former lacks a check valve. Such a check valve aims to hinder reverse flow, or U- tubing, of fluids from an annulus into a bore of a casing string. As an example, centralizers may be placed along various casing sections to help reduce risk of the casing string sticking while it is lowered into a borehole. In addition, centralizers may help to keep a casing string in the center of a borehole to help ensure placement of a uniform cement sheath in an annulus between the casing string and the borehole wall.

[0039] As a casing string is lowered into a borehole, an interior space (e.g., a bore) of the casing string may fill with drilling fluid. One aim of primary cementing may be to remove drilling fluid from the interior space and the borehole, place a cement slurry in anannulus and till the interior space with a displacement fluid such as drilling fluid, brine or water.

[0040] Cement slurries and drilling fluids tend to be chemically incompatible. Commingling them may result in a thickened or gelled mass at an interface that may be difficult to remove from borehole, possibly confounding placement of a substantially uniform cement sheath throughout an annulus. Therefore, chemical and physical techniques may be employed to maintain fluid separation. For example, chemical washes and spacer fluids may be pumped after the drilling fluid and before the cement slurry. These fluids may have an added benefit of cleaning the casing and formation surfaces, which may help to achieve suitable cement bonding.

[0041] As an example, one or more wiper plugs may be utilized. A wiper plug may be an elastomeric device that provides a physical barrier between fluids pumped inside a casing string. As an example, a bottom plug may provide for separating a cement slurry from drilling fluid, and a top plug may provide for separating a cement slurry from a displacement fluid. As an example, a bottom plug may include a membrane that ruptures when it lands at a bottom position of a casing string, creating a pathway through which a cement slurry may flow output of the casing string and into an annulus. As an example, a top plug may be without a membrane such that, when landed on top of a bottom plug, hydraulic communication is severed between the interior space of the casing string and the annulus between the casing string and the borehole wall.

[0042] After a cementing operation, a wait period allows for cement to cure, set and develop strength (e.g., waiting on cement (WOC)). After the WOC period, which may be less than approximately 24 hours, additional drilling, perforating or other operations may commence.

[0043] As explained, the method 400 can include various operations where different fluids, different devices, etc., are employed. As explained, fluids may include chemical washes, spacer fluids, etc., which may be pumped down a bore of a casing string to displace drilling fluid (e.g., mud). In the example of Fig. 4, the pump spacer block 420 may represent such a process that employs a spacer (e.g., a spacer fluid). As shown in the pump cement block 430, the bottom plug 432 is in position as having been pumped down and its membrane ruptured such that cement (e.g., cement slurry) can flow through the casing string 422 and into the annular region between the casing string 422 and the borehole wall. The pump cement block 430 may be broken into two phases where a first phase involves pumping cement slurry to push the bottom plug 432 down the casingstnng 422 into a desired position by filling the casing string 422 with a first volume of cement slurry that also acts to push out drilling fluid that may be in the interior space (e.g., bore) of the casing string 422. In such an example, once the bottom plug 432 is seated in position, further pumping of cement slurry can cause a membrane of the bottom plug 432 to rupture (e.g., upon building of sufficient pressure as may be specified by a rupture pressure) such that an additional volume of cement slurry can be pumped through the casing string 422 such that cement slurry fills the annular region to a desired level, which may be to an intermediate level, to surface, etc. As indicated in the displace plug block 440, the top plug 442 may be utilized after the cement slurry has been pumped where a displacement fluid is pumped to move the top plug 442 into position (e.g., on top of the bottom plug 432). For example, pumping displacement fluid can force the top plug 442 downward until it lands on the bottom plug 432, thereby isolating the interior space of the casing string 422 from the annulus. During such a process, cement slurry that is present within the interior space of the casing string 422 is evacuated and wiped from the interior wall where such evacuated cement slurry can be the last amount of cement slurry to exit the casing string 422 to fill the annulus.

[0044] As to the test pressure block 450, it may involve pressure testing as a hydraulic testing process. Such testing may be conducted after each surface- or intermediate-casing cement job. As an example, a driller may first perform a casing pressure test in an effort to verify mechanical integrity of a tubular string. The driller may then insert a drillstring into the casing string 422, as secured by cement, and drill out one or more components (e.g., consider the plugs 442 and 432 and the one or more components 426 and 424). Next, the driller may perform a pressure integrity test by increasing an internal casing pressure until it exceeds a pressure that may be applied during a next drilling phase (e.g., drilling of another section, segment, etc.). In such an example, if leakage is not detected, the cement job may be deemed to have successfully formed a cement seal.

[0045] Fig. 5 shows an example of a graphical user interface 500 with example plots that include stages of a cementing workflow with respect to time and various data- based features with respect to time. As shown, the cementing workflow can include stages that progress from a Mud1 stage to a Spaced stage, to a Tail stage, to a PlugDisplacement stage and to a Mud2 stage. Such stages may be, at times, challenging to discern using various data channels.

[0046] As an example, a framework may provide for implementation of a real-time and / or post-job automatic cementing job method that may generate one or more diagnostics, performance metrics, control instructions, etc. As an example, such a framework may provide for control of one or more cementing jobs (e.g., cementing method, workflow, etc.). As an example, such a framework may utilize one or more machine learning models (ML models) where feature engineering may be employed, for example, to arrive at a set of features that may provide for improved stage detection in a data-based manner. As an example, an ML model may include features such as the features shown in the example of Fig. 5. As shown, the various data-based features can include a surface pressure feature (CMT_PRESS), a pump rate feature (CMT_RATE), a fluid density feature (CMT_DENS), a cumulative pumped volume feature (cumPumpVol), and a downhole density feature (downholeDensity).

[0047] As an example, one or more features may be generated through a process of feature engineering. In various instances, feature engineering may provide for improved learning of a machine learning model, improved performance of a trained machine learning model, etc. In various instances, engineered features may differ from raw data. For example, an engineered feature may be defined and generated using one or more types of data, one or more data analytic techniques, one or more statistical techniques, etc. As an example, consider, as an analogy, an ML model to distinguish dogs and cats in images (e.g., as a classifier, etc.). While the dogs and cats may each have eyes, ears and a nose as raw features, an engineered feature, for example, may be a ratio of a distance between eyes and a distance from eyes to end of a nose. As explained, feature engineering may provide for improved accuracy, robustness, reduced training demands, improved hyperparameter tuning, etc. As an example, feature engineering may provide for engineering one or more features and / or defining or selecting a group of features. As an example, where desirable, feature engineering may result in fewer features, which may provide for more rapid execution and / or execution with fewer computational resources (e.g., using an edge framework, etc.). In various instances, an engineered feature may have real-world physical significance, which may also provide for an improved understanding of output from a machine learning model, which may facilitate control of a real-world field operation (e.g., to improve operation, mitigate issues, etc.), planning, system design, system modification, etc.

[0048] As an example, a data processing workflow can provide for stage detection and tagging (e.g., of field operational stages, etc.) followed by job diagnostics andperformance evaluation. As an example, such a workflow can include tagging job stages with a multiclass classification neural network pre-trained using features including surface pressure, pump rate and density time series data that serve as inputs and stage and event labels serving as an output sequence. In such an example, job parameters within the tagged stages may be compared with ones defined by cementing job designs to detect potential deviations and to allow for taking of one or more adjustive actions promptly, as appropriate.

[0049] As explained, a framework may be an automatic real-time cementing job framework that can provide for diagnostics and performance evaluation. As explained, such a framework may implement an ML model for stage tagging and event detection.

[0050] As explained, well cementing operations can involve various stages related to pumping different fluids, cement plug launching, and landing events. Maintaining control of these stages during operation can be improved through use of a framework that integrates an automatic ML model-based approach to stage tagging and event detection, which may be utilized to generate indicates of deviations from one or more design ranges in one or more pumping stages, as well as, for example, one or more changes in surface pressure, density and / or flow rate. Such a framework may facilitate interpretation during real-time monitoring and promptly allows for implementation of one or more types of control actions, as may be appropriate and / or desired.

[0051] As an example, an approach to stage tagging may utilize one or more ML model-based natural language processing (NLP) techniques. As an example, a multiclass classification neural network may be trained using surface pressure, pump rate, and density time series data serving as inputs and stage and event labels serving as output sequences.

[0052] In an example trial, a training dataset including data for 1 ,024 cementing jobs was generated synthetically based on statistical parameters of data acquired during 23 actual jobs. In the example trial, model hyperparameters were fine-tuned using a k- fold technique in which the entire dataset was separated into k groups where each of these groups iteratively served as a validation dataset, while the remainder of the data was used to train the ML model.

[0053] The dataset was utilized for assessing a number of ML models, which included neural network architectures, including recurrent architectures (e.g., LSTM and GRU) and transformer architectures. In assessing results, the LSTM architecturedemonstrated a higher l-l score (e.g., used to characterize predictive performance) as a metric for multiclass classification. The assessment for the LSTM architecture demonstrated an average stage time detection error less than one minute for the most optimal architecture. Deviations from cementing job designs were identified for several jobs and adjustments for subsequent job designs were generated for implementation (e.g., job planning and / or control).

[0054] As an example, a framework for cement job stages can be implemented to effectively monitor and identify a sequence of events, allowing timely intervention, and to predict one or more issues where prediction results may be enhanced by considering monitored and computed variables in a real-time mode and by integrating a reasoning tool for automatic detection and diagnosis of events such as changing the pump rate, beginning and end of U-tubing, or new fluid in an annulus. Applying such a level of automation to an interpretation process provides a rapid and repeatable approach to assessing cement jobs that also acts to minimize occurrences of individual interpretation biases.

[0055] As explained, well cementing operations encompass various stages, which can include pumping of diverse fluids and precise launching and landing of one or more plugs (e.g., consider a bottom plug and a top plug, etc.). In a pre-job stage, rigorous simulations may be conducted in an effort to ensure fulfillment of a job’s operations and objectives. As cementing can involve various relatively complex and sometimes “blind” operations, where the inability to directly monitor downhole conditions may pose a substantial challenge. In various instances, human supervision has to rely on interpretation of surface sensor data (e.g., gauge readings, etc.), which places demands on one or more humans as to meticulous management and control of such stages. Such demands may be placed upon an engineer or a job supervisor that is to directly oversees field operations, which renders cementing vulnerable to errors stemming from human oversight. As explained, a framework may be implemented to provide for automatic event detection using data acquired during a cementing job. Such a framework may improve cementing and provide for at least some level of automation to one or more data interpretation processes.

[0056] As to human monitoring of cementing operations for purposes of compliance with planned operations, a volume-based approach may be utilized to detect stages indirectly. For example, as explained, various different fluids may be pumped as may be associated with different stages. However, data channels such as rate, densityand pressure may not indicate wnat fluid is being pumped. Hence, such data channels, as time series data, may lack a direct indication of fluid type. Additionally, at times, such data channels may include gaps, signal-to-noise issues, etc. For various reasons, human interpretation based on data channel data may be challenging and subject to human interpretation errors. Human interpretation for monitoring may resort to a plan that outlines various stages and utilization of the planned stages, as a sequence, to be aligned with time series data. Such an approach may be subjective as to stage detection, particularly start and / or end times of one or more stages.

[0057] In various instances, compliance with design may be subject to regulation. For example, a regulatory authority may demand that actual cementing operations comply with designed operations. In various instances, a design (e.g., a plan) is subject to approval and is expected to be followed where proof of compliance is required.

[0058] As explained, a framework may integrate an automatic ML model-based approach to stage tagging and event detection that may aim to indicate one or more deviations from one or more design ranges in one or more pumping stages, as well as, for example, changes in surface pressure, density, and flow rate. As explained, a framework may provide for identification of one or more stages in real-time and / or after one or more operations have been terminated. Such a framework may enhance an ability to interpret a job during real-time monitoring, enabling prompt control action. As explained, a framework may provide for categorizing previously completed jobs, which may, thereby, help to enrich understanding of past operations.

[0059] As explained, a framework can provide for real-time and / or post-job automatic cementing job diagnostics and performance evaluation. As an example, a data processing workflow can include stage detection and tagging followed by job diagnostics and performance evaluation.

[0060] As an example, a framework may provide for classification and / or regression (e.g., prediction). For example, classification may be implemented to detect one or more stages whereas regression may provide for prediction of an end time, a start time, etc., as to one or more stages. As explained, such a framework may be applied for cementing operations, which may be performed as a primary cementing job, a secondary cementing job, etc. As an example, a framework may be applied for one or more types of field operations that involve pumping of one or more fluids.

[0061] As an example, a framework may be operatively coupled to a cementing model framework such as, for example, the CEMENTICS framework (SLB, Houston, Texas). The CEMENTICS framework provides for modeling of zonal isolation, which may be via a simulation engine that can generate simulation results that may be utilized to achieve improved cementing design and evaluation, particularly in challenging environments. The CEMENTICS framework can provide for simulation of hydraulic and temperature changes in various configurations, including deepwater, highly deviated, and horizontal wells, enabling operators to examine effects of changing drilling fluid properties as pressure and temperature vary. The CEMENTICS framework may be implemented to help design optimal primary cement jobs and cement plug placement, minimizing contamination risk, maximizing zonal isolation, and increasing cement job success. The CEMENTICS framework may also include the WELLCLEAN III pipe and annular mud displacement simulator (SLB, Houston, Texas), which may provide for coupling of pipe and annulus simulation, for example, to provide a better understanding of mud displacement in an annulus. The CEMENTICS framework may provide for simulation of mud displacement in a three-dimensional wellbore, enabling examination of cement coverage across targeted zones.

[0062] Fig. 6 shows an example of a graphical user interface 600 with example graphics as to input data and tags as part of an ML model-based workflow. As shown, tags may be defined such as, for example, pre-job, pumping spacer, pumping cement, plug displacement and pressure test where such tags represent stages of a cementing workflow where the stages are to be performed in a sequence (e.g., a sequence with respect to time). As explained, a cementing job or cement job may pertain to casing where a casing tally may be characterized using cumulative volume versus casing inner diameter (ID).

[0063] In the example of Fig. 6, the graphics include a data plot of rate and density data with respect to time and a data plot of pressure with respect to time. Additionally, a plot of volumes versus time of various fluids is shown where patterns may be utilized to indicate different types of fluid within a borehole at various points in time. In the example of Fig. 6, four different hatching patterns are utilized to represent four different fluids. As shown, one pattern is utilized to represent a first fluid that eventually disappears from within the bore of the casing string (e.g., consider mud), another pattern is utilized to represent a second fluid that appears at a point in time and that eventually disappears from within the bore of the casing string (e.g., consider spacer fluid), yet another patternis utilized to represent a third fluid that appears at a point in time and that eventually disappears from within the bore of the casing string (e.g., consider cement slurry), while a fourth pattern is utilized to represent a fourth fluid that appears at a point in time and that does not disappear from within the bore of the casing string (e.g., consider a Newtonian fluid, etc.). In the plot shown, there are three interfaces between the different patterns, each of which reaches the bottom of the plot at certain point in time. Various patterns (e.g., fluids) may provide for times that correspond to a stage or stages. For example, one pattern extends across a pre-job stage where the pumping of spacer fluid, as may be represented by another pattern, may indicate pumping of spacer fluid, which is shown to be the second stage. As fluids are introduced at surface via a bore of a casing string, these fluids may progress downhole, out of the bore of the casing string and into an annulus. While patterns are mentioned, a graphical user interface may employ colors, gradients, shading, etc., to represent different fluids, as may be relevant to one or more field operations.

[0064] In the example of Fig. 6, it may be noted that fluid introduced flows into a bore of a casing string, which may be a considerable length as it may extend to or near a bottom of a borehole. Hence, a bore of a casing string may accommodate a substantial volume of fluid or fluids. As an example, cumulative volume may be a proxy for depth (e.g., measured depth). While various examples pertain to casing string volumes with respect to time, a framework may utilize one or more of casing string volumes and / or annulus volumes with respect to time; noting that pumping of fluid generally involves pumping fluid through a casing string to reach an annulus and / or to cause fluid to exit the casing string and move into the annulus (e.g., consider fluid separated by a plug).

[0065] As shown in the example of Fig. 6, various fluids may be utilized where rate, density and pressure data may be acquired. As shown, a plot or other graphic may provide for tracking volume and / or movement of fluid downhole. As an example, a framework may provide for relating rate and density to pressure. As an example, where pressure may deviate from a specification in a plan, a framework may provide for issuing a signal, which may be a notification signal, a control signal, etc., such that action may be taken in the field to adjust for the deviation. As explained, a framework may operate in real-time such that real-time deviations from plan (e.g., design) to be detected such that one or more actions may be taken (e.g., to address a deviation, etc.).

[0066] In the example of Fig. 6, an example of an input word is shown as a vector that includes three elements, as derived from the channels of data for rate, density andpressure. in a natural language processing (NLP) approach, the input word of three elements may be interpreted as having meaning, for example, as would words that find definitions in a dictionary, etc. Hence, where one or more of the elements of the vector change, the vector itself will have a different meaning. As to possible meanings for the vector and its possible elements, consider translation to stages, which, as explained, may include five stages that may be represented using appropriate indicators (e.g., zeros and ones) in a translation vector. For example, a five-element output vector, or translation vector, is shown in the example of Fig. 6 where such a vector may be expected to have a single “1” (e.g., or“0”, etc.) to indicate a single one of the stages as being the translation of the input word (e.g., input vector). While a discrete approach is illustrated in the example of Fig. 6, a fuzzy or other approach may be utilized.

[0067] As an example, a framework may provide for normalization of data. For example, consider scaling data, making data dimensionless, etc., which may help to balance impact of data and / or changes in data, etc. Such an approach may provide for improved model stability, etc. As an example, feature engineering may provide for normalization of data, processed data, combinations of data, etc.

[0068] As an example, input may be provided as an instantaneous input, which may be provided with prior input (e.g., corresponding to prior instances in time). As explained, as output is often sequential and time-dependent, prior input can be relevant for determining the stage at a specific point in time. For example, consider a recurrent neural network approach that applies to time series data such that prior instances in time are instructive to determining output for a current instant in time (e.g., and / or for one or more future times for a predictive model, etc.). Depending on type of model or models employed, a framework may provide for generation of output using input for a snapshot in time such as a current time or a relatively small time window (e.g., minutes, etc.) that includes both the current and previous time(s).

[0069] As an example, a framework may provide for generation of probabilities such that, for example, output as to a stage or stages may be provided as a probability and / or with respect to one or more probability metrics. For example, consider the translation vector as including a number of values ranging from 0 to 1 where a value of 1 with four values of 0 indicates a 100 percent probability that a stage is detected. As an example, in a real-time approach, an output vector may be plotted where values may change as input indicates a change or a trend toward a change in meaning, as would be interpreted to mean a change from one stage to another stage. As explained, stages maybe expected to occur in sequence such that knowledge of the sequence may facilitate stage detection.

[0070] As an example, an approach to training (e.g., learning) can involve supervised learning. For example, actual data may be labeled, which may be via human interpretation and / or assisted by one or more digital tools. As explained, synthetic data may be generated using one or more simulators, which may provide for introducing noise, sensor issues, etc. Such synthetic data may be considered to be inherently labeled (e.g., as a simulator progresses from simulating one stage to another stage of a cementing process, etc.).

[0071] As explained, features may be data-based features that may include direct data channel data (e.g., as may be scaled, normalized, etc.) and / or that may include formula-based features that depend on one or more formulas that include data channel data as one or more formula variables. For example, consider a downhole density distribution feature that may be based on the following formula:where (t) is the surface densitometer reading during a cement job, Q(t) is a surface flowmeter reading during a cement job, and V(h) is a cumulative casing volume as a function of measured depth h.

[0072] The foregoing formula provides for computation of a downhole density, which may depend on different fluids. As an example, the downhole density may be an engineered feature, for example, one that is engineered from one or more types of available data (e.g., actual measured data, synthetic data, etc.). As explained, a pumped fluid can progress downward within a bore of a casing string where, for example, the pumped fluid may be to move a plug. Hence, in various instances, when a pumped fluid reaches a distal end of a bore of a casing string, it may be expected that a plug has also reached a proper seating position in the bore of the casing string (e.g., at or near the distal end of the bore of the casing string). Such an occurrence may signal the end of a stage and / or the beginning of another stage (e.g., in a sequence of stages). While downhole density is mentioned, an ML model may be operable with or without downhole density as an input (e.g., a feature). As an example, a feature may be stage dependent in that it helps to detect one or more stages but not one or more other stages; noting that various features may help in determining what stage is not occurring. As an example,maintaining a fixed number of features may provide for simplicity in coding, etc. (e.g., as in PYTHON, etc.). As an example, features relevant to past stages may be relevant to detection of future stages.

[0073] As explained, input to an ML model can include measurements that may be surface channel measurements (e.g., available at surface) and / or computed measurements that may be for surface and / or for downhole. In the example of downhole density, a downhole density sensor may not be available due to impracticality or due to concept, hence, an engineered feature of an ML model may provide for computation thereof using a data-based formula.

[0074] Fig. 7 shows an example of a graphical user interface 700 that includes a graphical representation of a workflow that includes a number of features as input to one or more time series ML models. As to data, as explained, pressure, rate and density may be utilized as input features. As an example, such a set of features may be considered to be a minimum set of features. As an example, cumulative pumped volume may be tracked, for example, given density and rate. As such, cumulative pumped volume may be derived from measured data (e.g., density and rate over a period of time, etc.). As an example, additional and / or alternative features may be utilized. As an example, section type may be included as an input (e.g., surface, intermediate, etc.), and design data with planned stages may also be used as input. As an example, one or more field specific inputs may be utilized, for example, depending on types of completions utilized for wells of a field.

[0075] In the example of Fig. 7, the ML models may include one or more of RNN, GRU, LSTM, transformer, etc., types of model architectures.

[0076] In the example of Fig. 7, the output is shown as a number of tags, which can be a sequence of stages for field operations. As shown, examples of tags can include pressure test, spacer, cement, displacement, and casing test. As an example, tags may be adjusted, selected, etc., depending on desire use of the output. For example, cement may be relevant to reporting in terms of compliance with one or more regulatory standards. As an example, additional and / or alternative tags may be utilized.

[0077] Fig. 8 shows an example plot 800 that shows losses versus epoch with respect to training and validation. The plot 800 provides for a visualization of how well a model’s predictions match actual data. As shown, losses begin to tail off at approximately 25 to 40 epochs.

[0078] big. 9 shows an example of a graphical user interface 900 with example plots as to labels and features with respect to time. In the example plots, input is available until a time of approximately 14000 seconds such that input values are zero, which provide for spurious output; whereas, for input available from 0 seconds to 14000 seconds, output as generated (e.g., classified stages) match the actual stages. Hence, the plots demonstrate that where input is available, stage detection is accurate; whereas, where input may be zero or other non-sensical values, a framework may provide nonmeaningful output (e.g., or an indication that an issue may exist for input, etc.).

[0079] Fig. 10 shows an example of a method 1000 that includes reception blocks 1010 and 1020 for receiving cementing design parameters input and cementing unit input, which may be provided to an ML model block 1040 where output generated by an ML model can provide for assessing a cementing job with respect to design. For example, the method 1000 may be utilized for forensic and / or real-time jobs. As shown in the example of Fig. 10, the method may include a cementing stage tags block 1060 that may include tags as labels such as classification labels as to one or more stages of a cementing job. In such an example, various aspects of a planned fluid pumping job may be compared to various aspects of an actual fluid pumping job. Such an approach may be performed forensically and / or during job performance, which may provide for control of job performance (e.g., equipment control, fluid control, plug control, etc.).

[0080] As explained, a method may provide for automatic fluid pumping job diagnostics and performance evaluation in real-time, which may provide for control of a fluid pumping job. As an example, a framework may provide for implementation of realtime and / or post-job automatic cementing job diagnostics and performance evaluation. As an example, a workflow may include a phase of detection and tagging followed by a phase of job diagnostics and performance evaluation. In a detection and tagging phase, job stages may be tagged with a multiclass classification machine learning model as may be pre-trained using surface pressure, pump rate and density time series data serving as inputs and stage and event labels serving as an output sequence (e.g., a sequence of stages and / or events). As explained, a phase may provide for assessment, for example, where job parameters within tagged stages may be compared with ones defined by a fluid pumping job design (e.g., a digital plan, etc.), which may provide for detection of one or more potential deviations between actual and design. In such an example, a framework may provide for issuing one or more control signals for controlling a fluid pumping job, for example, to help assure compliance with a design (e.g., a plan, etc.) and / or to addressone or more issues (e.g., consider mitigation of an issue, reduced risk of occurrence of an issue, etc.).

[0081] Fig. 11 shows an example of an architecture 1100 where input sizes are indicated, dropout, linear layers, GRU or LSTM or RNN + ReLU layers, dropout, and output sizes. As an example, the sizes, layers, etc., may be adjusted to accommodate one or more aspects of field operational stages that involve pumping of one or more fluids.

[0082] Fig. 12 shows an example of an architecture 1200 where inputs and outputs may be processed in a transformer approach. Such an architecture may be suitable for used in instances where position as in a sequence is relevant, which may occur in language models (e.g., consider large language models (LLMs), etc.). In particular, positional encoding may be utilized as a technique used in a transformer architecture and other sequence-to-sequence models to provide information about the order and position of elements in an input sequence. Positional encoding may be contrasted to standard embeddings (e.g., word embeddings) that may not inherently contain information about position of elements. Unlike RNNs, a transformer architecture can process input tokens in parallel; however, a transformer architecture may not inherently possess an ability to handle sequences (e.g., positional information, etc.), as may RNNs, which find use in handling time series input. Without positional information, the input tokens may be treated as a bag-of-words, thereby making it difficult for a model to understand the sequential nature of the input. Therefore, positional encoding may be added to input embeddings to help a model understand the sequential structure of data (e.g., features) and differentiate between elements in different positions. For example, in a transformer architecture positional encoding may be added to input embeddings before feeding data (e.g., features) into encoder and decoder stacks. Such an approach may help to allow a model to understand sequential relationships between tokens in an input sequence and generate coherent output sequences. As explained, a framework may utilize a language type of approach whereby an input vector may be “translated” to output as may be via a classification scheme (e.g., to indicate a detected stage in a sequence of stages).

[0083] As explained, a stage detection technique may utilize one or more types of ML NLP techniques, for example, where input can be imagined as “input words” and where a cementing job has stage tags as corresponding “translations”.

[0084] As an example, a model may include one or more features of a transformer type of model. As an example, a foundational generative pre-trained transformer (GPT)model may be adapted to produce more targeted systems directed to specific tasks and / or subject-matter domains. Techniques for such adaptation may include additional fine-tuning (e.g., beyond tuning of a foundation model, etc.), certain forms of prompt engineering, etc. As an example, an LLM may be a chatbot type of LLM. For example, consider the OpenAI ChatGPT LLM, which is an online chat interface powered by an instruction-tuned language model trained in a similar fashion to InstructGPT. Other chatbots may include features of GPT-4 (OpenAI), Bard (e.g., LaMDA family of conversation-trained language models, PaLM, etc.) (Google, Mountain View, California), etc.

[0085] As an example, a LLM Meta Al (LLaMA) LLM may be utilized, which includes a transformer architecture; noting some architectural differences compared to GPT-3. For example, LLaMA utilizes the SwiGLU activation function rather than ReLU, uses rotary positional embeddings rather than absolute positional embedding, and uses root-mean-squared layer-normalization rather than standard layer-normalization. Further, there may be an increase in context length from 2K (Llama 1 ) tokens to 4K (Llama 2) tokens between.

[0086] As an example, a framework may utilize one or more features of the DALL E family of text-to-image models (OpenAI, San Francisco, California), which may utilize one or more deep learning techniques to generate digital images from natural language descriptions (e.g., prompts, etc.).

[0087] A DALL E model may be provided as a multimodal implementation of a GPT with billions of parameters that effectively swap text for pixels. Such a model may be trained using text-image pairs from one or more databases. As an example, input to a transformer model may be a sequence of tokenized image captions followed by tokenized image patches. As an example, an image caption may be in a particular language (e.g., a corpus), tokenized by byte pair encoding, for example, up to 256 tokens long. In an example where 256 is given as a limit, each image may be a 256x256 RGB image (e.g., or other color scheme), divided into 32x32 patches of 4x4 each. In such an approach, each patch may then be converted by a discrete variational autoencoder to a token. As an example, Contrastive Language-Image Pre-training (CLIP) may be utilized as a technique for training a pair of models where one model takes in a piece of text and outputs a single vector and another model takes in an image and outputs a single vector. To train such a pair of models, as an example, a method may include preparing a large dataset of image-caption pairs followed by sampling of batches.

[0088] As an example, one or more neural networks may be implemented as part of a supervised machine learning approach where that trains the one or more neural networks by examples, each of which contains pairs of a known input (features) with desired output (labels or tags). Machine learning involves adjusting weights of a network to improve accuracy of a result. For example, learning may involve minimizing observed errors. Weights may be adjusted in iterations, which may be referred to as epochs. Learning may be deemed complete when examining additional observations does not usefully reduce error rate.

[0089] As explained, examples used for learning can be or include historical cementing job data recorded with real surface sensors for real casing configurations, or can be or include synthetic data, for example, obtained using one or more numerical techniques (e.g., numerical simulation models, etc.). As an example, a dataset may be divided into three non-intersecting subsets: training dataset used for adjusting the weights, validation dataset used for determination of optimal epoch and testing dataset used for overall model performance validation.

[0090] As an example, a trained model may be deployed as part of a framework to output labels from sample features. For example, a cementing job tagging model may output cementing job stage tags serving as model labels from sensor measurements and well casing configuration parameters serving as model features. As to how a model operates to generate output, one or more of classification and regression may be utilized; noting that various problems may be cast as classification and / or regression problems.

[0091] As an example, an architecture of a neural network may be determined by type of data and / or engineered features. As explained, surface sensor measurements are intrinsically time series. These time series may be grouped into independent data blocks associated with cementing jobs, where each job belongs to a particular well with its own completion parameters and borehole configuration that may affect relationship(s) between surface measurements and assigned tags.

[0092] Some examples of cementing job features and labels are listed in Table 1 , below.Table 1. Example Features and Labels (e.g., tags)

[0093] To help a neural network to reveal relationships between features and labels more easily, model features may be represented in non-dimensional form. For example, consider features normalized as follows:where, each normalized feature anrmvaries from O to 1 and amhland amnxare minimum and maximum values taken according to a training dataset.

[0094] As data may naturally be combined into data blocks by cementing jobs that do not have mutual temporal dependence, these treatment stages may serve as batches when training a model. As an example, a collection of batches may be represented as a three-dimensional array shaped as N}obsx NMaxSamplesx NPeatures, where N}obsis number of cementing jobs, NMaxSampiesis length of the longest job (e.g., where the rest of the jobs are padded to this length with zeros) and NFeaturesis the number of features.

[0095] As an example, temporal patterns may be captured using one or more RNNs. As an example, to help reduce gradients decay, one or more of GRU (Gated Recurrent Units) and / or LSTM (Long Short-Teem Memory) architectures may be utilized. As explained, a transformer architecture as may find use in NLP (e.g., natural language translation) may be utilized.

[0096] As explained, to train a neural network model, stages of several cementing job datasets may be manually tagged and appropriate features computed, if desired or as appropriate. As an example, labelled real data may be used to extract different statistical parameters of the data. Examples of such parameters may include mean and standard deviations of stage lengths, pressures, and pump rates. In a trial example, such parameters were used as inputs to generate 1024 synthetic datasets, that were split into training, validation and testing subsets.

[0097] As explained with respect to the plot 800 of Fig. 8, training and validation losses obtained during the training may be plotted. As to the plot 800, the results therein correspond to a transformer architecture. As indicated, the losses do not substantially change after approximately 150 epochs meaning that the ability of the model has effectively stopped improving such that a training loop may be interrupted (e.g., terminated); noting, again, that most of the losses were reduced after approximately 40 epochs. As an example, a loss assessment may be implemented as part of a training and / or validation process to automatically determine when an iterative workflow may be suitably terminated.

[0098] As explained with respect to Fig. 9, the plots 900 show a comparison of original and output tags for a synthetic cementing job example, where full agreement exists between original and class tags during the job execution (e.g., for a real-time implementation, etc.).

[0099] As explained, workflow may involve performance evaluation as to a trained ML model. For example, job parameters within tagged stages may be compared with the ones defined by a cementing job design to detect potential deviations and to allow for taking appropriate action promptly.

[0100] As explained with respect to the method 1000 of Fig. 10, a framework may provide for a method where a machine and / or a user inputs cement design parameters (e.g., or other fluid pumping job design parameters). Where, subsequently, cement unit acquisition data are fed into the framework. In such an example, a framework model can be executed for conducting an analysis for generation of output identifying various stages of the cementing job over time. As explained, in real-time operations, such output may facilitate a comparison between a planned design and actual execution, highlighting one or more deviations. As explained, a framework may be utilized in real-time and / or for post-job analysis. As an example, identifying deviations may enhance understanding of a job’s performance. As explained, a framework may enable on-site cement operators to take appropriate actions if it is determined that an operation is not proceeding as intended.

[0101] As an example, during a displacement stage, a framework may compare the cumulative displacement volume computed by pump rate integration over a time for the time interval labelled with “Displacement” tag with the cumulative displacement volume required by a job design for the displacement stage. In such an example, if the measured displacement volume exceeds the designed displacement volume, anotitication signal and / or a control signal may be issued, for example, to provide for interruption of an operation and / or design modification. As an example, a framework may be utilized in a planning and / or re-planning workflow. As explained, a framework may be operatively coupled to a framework such as the CEMENTICS framework.

[0102] As an example, a framework may be utilized for comparison of a slurry density during a time interval labelled with a “Cement” tag. In such an example, if the measured density substantially differs from the density specified by a job design for a cement pumping stage, a notification signal and / or a control signal may be issued, for example, for checking a densitometer sensor, visual fluid inspection, density control, etc.

[0103] As an example, a framework may provide for implementing a method for real-time automatic cementing job diagnostics and performance evaluation, which can provide for control of one or more aspects of a cementing job. As an example, a framework may provide for defining a cementing job design based on modeling for a particular borehole completion configuration, obtaining a machine learning cement stage tagging model, pre-trained with cementing job data, generation of current stage tag predictions with the model in real-time, comparison of measured job parameters with the parameters required by the design, and issuance of one or more signals (e.g., notifications, control commands, etc.), for example, if one or more measured parameters differ from one or more parameters required by the design.

[0104] As an example, a data used to train a machine learning model can be or include real historical data and / or data that are generated synthetically.

[0105] As an example, a computed parameter may be cumulative volume. As an example, a measured parameter may be cementing head pressure. As an example, a measured parameter may be pump rate. As an example, a measured parameter may be density.

[0106] As an example, a framework may include and / or be operatively coupled to a gateway such as, for example, an AGORA gateway (e.g., consider one or more processors, memory, etc., which may be deployed as a “box” that may be locally powered and that may communicate locally with other equipment via one or more interfaces). As an example, one or more pieces of equipment may include computational resources that may be akin to those of an AGORA gateway or more or less than those of an AGORA gateway. As an example, an AGORA gateway may be a network device.

[0107] As an example, a gateway may include one or more features of an AGORA gateway (e.g., v.202, v.402, etc.) and / or another gateway. For example, consider an INTEL ATOM E3930 or E3950 Dual Core with DRAM and an eMMC and / or SSD. Such a gateway may include a trusted platform module (TPM), which may provide for secure and measured boot support (e.g., via hashes, etc.). A gateway may include one or more interfaces (e.g., Ethernet, RS485 / 422, RS232, etc.). As to power, a gateway may consume less than about 100 W (e.g., consider less than 10 Wor less than 20 W). As an example, a gateway may include an operating system (e.g., consider LINUX DEBIAN LTS). As an example, a gateway may include a cellular interface (e.g., 4G LTE with Global Modem I GPS, etc.). As an example, a gateway may include a WIFI interface (e.g., 802.11 a / b / g / n). As an example, a gateway may be operable using AC 100-240 V, 50 / 60 Hz or 24 VDC. As to dimensions, consider a gateway that has a protective box with dimensions of approximately 10 in x 8 in x 4 in.

[0108] Fig. 13 shows an example of a method 1300 and an example of a system 1390. As shown, the method 1300 may include a reception block 1310 for receiving data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; a generation block 1320 for generating an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and an assessment block 1330 for assessing the performance of the fluid pumping job based at least in part on the inference. In such an example, a control block may be included for controlling performance of the fluid pumping job based at least in part on the inference, which may include controlling based at least in part on one or more assessments as to performance.

[0109] The method 1300 is shown in Fig. 13 in association with various computer- readable media (CRM) blocks 1311 , 1321 and 1331. Such blocks generally include instructions suitable for execution by one or more processors (or processor cores) to instruct a computing device or system to perform one or more actions. While various blocks are shown, a single medium may be configured with instructions to allow for, at least in part, performance of various actions of the method 1300. As an example, a computer-readable medium (CRM) may be a computer-readable storage medium that is non-transitory and that is not a carrier wave. As an example, one or more of the blocks 1311 , 1321 and 1331 may be in the form processor-executable instructions.

[0110] in the example ot big. 13, the system 1390 includes one or more information storage devices 1391 , one or more computers 1392, one or more networks 1395 and instructions 1396. As to the one or more computers 1392, each computer may include one or more processors (e.g., or processing cores) 1393 and memory 1394 for storing the instructions 1396, for example, executable by at least one of the one or more processors 1393 (see, e.g., the blocks 1311 , 1321 and 1331 ). As an example, a computer may include one or more network interfaces (e.g., wired or wireless), one or more graphics cards, a display interface (e.g., wired or wireless), etc.

[0111] As an example, a fluid pumping job may be one or more types of fluid pumping jobs. A fluid pumping job may include pumping of a fluid or fluids according to a sequence, which may be a sequence of stages. As an example, one or more types of fluid may be pumped, one or more types of devices may be pumped (e.g., plugs, etc.), etc. In various instances, a device may be a degradable device that may degrade upon exposure to a particular fluid, etc. As an example, a fluid pumping job may be a type of completions job, a type of drilling job, a type of treatment job, etc. As an example, a treatment job may involve pumping of fluid to condition a bore wall (e.g., consider acid, etc.), to fracture a formation, etc.

[0112] As an example, a hydraulic fracturing job may be a type of fluid pumping job where a fluid that may include chemicals and proppant may be pumped into a borehole with suitable pressure to cause fracturing of reservoir rock to effectively increase drainage of reservoir fluid from the reservoir rock. Hydraulic fracturing may be considered to be a type of treatment such as a stimulation treatment that may be performed on oil and / or gas wells in relatively low-permeability reservoirs. As an example, one or more engineered fluids may be pumped at relatively high pressure and rate into a reservoir interval to be treated, causing a fracture to open. In such an example, wings of the fracture may extend away from a wellbore in opposing directions according to natural stresses within a formation. As an example, proppant, such as grains of sand of a particular size, may be mixed with a treatment fluid to flow into a fracture to help keep the fracture open when the treatment is complete. As an example, hydraulic fracturing may create a relatively high-conductivity communication region withing an area of formation and / or may bypass damage that may exist in a near-wellbore area.

[0113] As an example, a hydraulic fracturing job may involve utilization one or more multistage plug-and-perf completion frac plugs, which may provide for isolation of one or more sections (e.g., intervals). As an example, a frac plug may be designed tooptimize production from cased hole wells where plug-and-perf completions may enable performing perforation and treatment in each stage optimally, for example, by applying knowledge gained from one or more prior stages, for example, to modify and improve treatment. As an example, a framework may provide for assessing one stage for improving a subsequent stage or, for example, may provide for assessing a current stage for improving that current stage. As an example, a frac plug may be a dissolvable frac plug that dissolve in the presence of one or more wellbore fluids, for example, after fracturing is complete, which may provide for simplifying cleanout operations (e.g., reduced demand for milling, etc.) and reducing time to production. As an example, time savings may result in reduced diesel consumption and hence, lower CO2 emissions.

[0114] As explained, one or more types of fluid pumping jobs may involve pumping of one or more plugs. As to cementing, as explained, a rubber plug may be used to separate a cement slurry from one or more other fluids, for example, to help reduce contamination and maintain predictable slurry performance. As an example, a bottom plug may be launched ahead of a cement slurry to minimize contamination by fluid(s) inside a casing prior to cementing. In such an example, a diaphragm in the plug body may ruptures to allow cement slurry to pass through after the plug reaches a landing collar. As an example, a top plug may have a solid body that provides positive indication of contact with a landing collar and a bottom plug through an increase in pump pressure (see, e.g., the block 450 of Fig. 4); noting that one or more types of pressure tests, pressure measurements, etc., may be performed.

[0115] As an example, a treatment fluid may be a fluid designed and prepared to resolve a specific wellbore or reservoir condition. Various types of treatment fluids may be prepared at a wellsite for one or more of a wide range of purposes, such as stimulation, isolation or control of reservoir gas or water, etc. A treatment fluid may be intended for specific conditions and be prepared and used as directed (e.g., according to a plan), for example, to help ensure reliable and predictable performance. As an example, a fluid pumping job may utilize one or more types of treatment fluids, which may, for example, be utilized in a particular order (e.g., a sequence). As explained, a method may provide for assessing and / or controlling a fluid pumping job.

[0116] As an example, a fluid pumping job may utilize one or more types of pumps, which may include surface pumps and / or downhole pumps. As to downhole pumps, consider an electric submersible pump (ESP) that may provide for pumping fluid in oneor more directions in a borenole. For example, an ESP may be utilized for injection of fluid and / or for production of fluid.

[0117] As to types of machine learning models, consider one or more of a support vector machine (SVM) model, a k-nearest neighbors (KNN) model, an ensemble classifier model, a neural network (NN) model, etc. As an example, a machine learning model may be a deep learning model (e.g., deep Boltzmann machine, deep belief network, convolutional neural network, stacked auto-encoder, etc.), an ensemble model (e.g., random forest, gradient boosting machine, bootstrapped aggregation, AdaBoost, stacked generalization, gradient boosted regression tree, etc.), a neural network model (e.g., radial basis function network, perceptron, back-propagation, Hopfield network, etc.), a regularization model (e.g., ridge regression, least absolute shrinkage and selection operator, elastic net, least angle regression), a rule system model (e.g., cubist, one rule, zero rule, repeated incremental pruning to produce error reduction), a regression model (e.g., linear regression, ordinary least squares regression, stepwise regression, multivariate adaptive regression splines, locally estimated scatterplot smoothing, logistic regression, etc.), a Bayesian model (e.g., naive Bayes, average on-dependence estimators, Bayesian belief network, Gaussian naive Bayes, multinomial naive Bayes, Bayesian network), a decision tree model (e.g., classification and regression tree, iterative dichotomiser 3, C4.5, C5.0, chi-squared automatic interaction detection, decision stump, conditional decision tree, M5), a dimensionality reduction model (e.g., principal component analysis, partial least squares regression, Sammon mapping, multidimensional scaling, projection pursuit, principal component regression, partial least squares discriminant analysis, mixture discriminant analysis, quadratic discriminant analysis, regularized discriminant analysis, flexible discriminant analysis, linear discriminant analysis, etc.), an instance model (e.g., k-nearest neighbor, learning vector quantization, self-organizing map, locally weighted learning, etc.), a clustering model (e.g., k-means, k-medians, expectation maximization, hierarchical clustering, etc.), etc.

[0118] As an example, a machine model, which may be a machine learning model (ML model), may be built using a computational framework with a library, a toolbox, etc., such as, for example, those of the MATLAB framework (MathWorks, Inc., Natick, Massachusetts). The MATLAB framework includes a toolbox that provides supervised and unsupervised machine learning algorithms, including support vector machines (SVMs), boosted and bagged decision trees, k-nearest neighbor (KNN), k-means, k- medoids, hierarchical clustering, Gaussian mixture models, and hidden Markov models.Another MA I LAB framework toolbox is the Deep Learning Toolbox (DLT), which provides a framework for designing and implementing deep neural networks with algorithms, pretrained models, and apps. The DLT provides convolutional neural networks (ConvNets, CNNs) and long short-term memory (LSTM) networks to perform classification and regression on image, time-series, and text data. The DLT includes features to build network architectures such as generative adversarial networks (GANs) and Siamese networks using custom training loops, shared weights, and automatic differentiation. The DLT provides for model exchange various other frameworks.

[0119] As an example, the TENSORFLOW framework (Google LLC, Mountain View, CA) may be implemented, which is an open-source software library for dataflow programming that includes a symbolic math library, which may be implemented for machine learning applications that may include neural networks. As an example, the CAFFE framework may be implemented, which is a DL framework developed by Berkeley Al Research (BAIR) (University of California, Berkeley, California). As another example, consider the SCIKIT platform (e.g., scikit-learn), which utilizes the PYTHON programming language. As an example, a framework such as the APOLLO Al framework may be utilized (APOLLO. Al GmbH, Germany). As an example, a framework such as the PYTORCH framework may be utilized (Facebook Al Research Lab (FAIR), Facebook, Inc., Menlo Park, California).

[0120] As an example, a training method may include various actions that may operate on a dataset to train an ML model. As an example, a dataset may be split into training data and test data where test data may provide for evaluation. A method may include cross-validation of parameters and best parameters, which may be provided for model training.

[0121] The TENSORFLOW framework may run on multiple CPUs and GPUs (with optional CUDA (NVIDIA Corp., Santa Clara, California) and SYCL (The Khronos Group Inc., Beaverton, Oregon) extensions for general-purpose computing on graphics processing units (GPUs)). TENSORFLOW is available on 64-bit LINUX, MACOS (Apple Inc., Cupertino, California), WINDOWS (Microsoft Corp., Redmond, Washington), and mobile computing platforms including ANDROID (Google LLC, Mountain View, California) and IOS (Apple Inc.) operating system-based platforms.

[0122] TENSORFLOW computations may be expressed as stateful dataflow graphs; noting that the name TENSORFLOW derives from the operations that suchneural networks perform on multidimensional data arrays. Such arrays may be referred to as “tensors”.

[0123] As an example, a device may utilize TENSORFLOW LITE (TFL) or another type of lightweight framework. For example, consider a gateway that may be in the field (e.g., on-site) and that may utilize the TFL and / or one or more other types of lightweight frameworks. The TFL framework is a set of tools that enables on-device machine learning where models may run on mobile, embedded, and loT devices. The TFL framework is optimized for on-device machine learning, by addressing latency (no round-trip to a server), privacy (no personal data leaves the device), connectivity (Internet connectivity is demanded), size (reduced model and binary size) and power consumption (e.g., efficient inference and a lack of network connections). The TFL framework offers multiple platform support, covering ANDROID and iOS devices, embedded LINUX, and microcontrollers. The TFL framework offers diverse language support includes JAVA, SWIFT, Objective-C, C++, and PYTHON. The TFL framework may provide high performance via hardware acceleration and model optimization.

[0124] As an example, a gateway device may be operatively coupled to one or more sensors for receipt of data where an Al engine may generate one or more inferences using at least a portion of such data. In such an example, the gateway may include an interface for outputting one or more outputs to one or more display devices, controllers, network devices, etc., where, for example, one or more job operations may be controlled based at least in part on such one or more outputs.

[0125] As an example, a method may include receiving data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; generating an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assessing the performance of the fluid pumping job based at least in part on the inference. In such an example, multiple stages may be stages of a sequence that may be expected to be followed during performance of a fluid pumping job. As an example, a method may include controlling a fluid pumping job based at least in part on an inference of a machine learning model. For example, consider controlling a fluid pumping job based at least in part on an assessment or assessments.

[0126] As an example, a fluid pumping job may involve pumping different fluids. For example, consider different fluids that may include a cement slurry. As an example, consider different fluids that may include one or more of a spacer fluid and a displacementfluid. As an example, different fluids may include a cement slurry and one or more of a spacer fluid and a displacement fluid. As an example, a fluid pumping job may be performed in a sequence where one or more stages in the sequence involve pumping one of a number of different fluids.

[0127] As an example, a fluid pumping job may include pumping one or more plugs. For example, one or more plugs may include one or more of a bottom plug and a top plug. As an example, a plug may include a membrane where application of pressure (e.g., fluid pressure) may cause rupturing of the membrane. As an example, a plug may act to clean a surface of tubing such as a casing surface. For example, a plug may move within a bore of a tube where the plug effectively wipes a bore wall that defines the bore of the tube (e.g., consider a casing wall of casing as being a bore wall of casing as a type of tube, etc.). As an example, a plug may be a frac plug, which may be a dissolvable plug that dissolves upon exposure to fluid of a particular character or may be a non-dissolvable plug that may be subject to milling (e.g., to mill out or drill out). As to a dissolvable plug, as an example, a sequence of events may involve setting a dissolvable plug (e.g., via pumping fluid) and dissolving a dissolvable plug upon exposure to a particular fluid. As an example, a framework may provide for classification of an event or events in a sequence during performance of a job that involves fluid pumping, which may include pumping to position a plug, to dissolve a plug, to rupture a component of a plug, etc.

[0128] As an example, a machine learning model may be or may include a time series model. For example, consider a time series model that may include one or more recurrent structures, a time series model that may be or may include a recurrent neural network model, a time series model that may include long short-term memory, etc.

[0129] As an example, a machine learning model may be or may include a transformer model. In such an example, the transformer model may include positional encoding structures. In such an example, positional encoding structures may provide for training a machine learning model to generate a trained machine learning model that can account for a sequence of events such as, for example, stages of a fluid pumping job that are expected to be performed according to a sequence.

[0130] As an example, a machine learning model may include features that depend on the data from field equipment. In such an example, features may include a density feature based on density data from the field equipment, a rate feature based on rate data from the field equipment, and a pressure feature based on pressure data from the field equipment. As an example, a machine learning model may generate output usingdesign data, where, tor example, the output is for one or more planned pumping stages (e.g., lead slurry, displacement, etc.) and / or for one or more casing configurations (e.g., surface, intermediate, etc.).

[0131] As an example, a machine learning model may include one or more of a downhole density feature and a cumulative volume feature. In such an example, the downhole density feature and the cumulative volume feature may be computed features, for example, according to one or more formulas, which may be based at least in part on measured data (e.g., sensor data, etc.).

[0132] As an example, a method may include comparing the performance of a fluid pumping job to a fluid pumping job plan. For example, consider a method that may include receiving a digital plan and comparing performance to a sequence, metrics, parameters, etc., specified in the digital plan. As an example, such a method may perform a comparison automatically where performance of a fluid pumping job may be assessed using a machine learning model (e.g., for stage detection, etc.). As an example, a method may include controlling performance of a fluid pumping job based at least in part on comparing the performance of a fluid pumping job to a fluid pumping job plan (e.g., a digital plan). As an example, such controlling may be performed automatically and / or semi-automatically, where, for example, a comparison or comparisons may be performed automatically and / or semi-automatically.

[0133] As an example, a system may include one or more processors; memory accessible to at least one of the one or more processors; and processor-executable instructions stored in the memory and executable to instruct the system to: receive data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; generate an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assess the performance of the fluid pumping job based at least in part on the inference. As an example, such a system may include instructions to instruct the system to control a fluid pumping job and / or to issue one or more control commands to one or more pieces of equipment for control of a fluid pumping job.

[0134] As an example, one or more computer-readable storage media may include processor-executable instructions to instruct a computing system to: receive data from field equipment during performance of a fluid pumping job that includes multiple stages at a wellsite; generate an inference as to an occurrence of one of the multiple stagesassociated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assess the performance of the fluid pumping job based at least in part on the inference. As an example, instructions may be included to instruct a computing system to control a fluid pumping job and / or to issue one or more control commands to one or more pieces of equipment for control of a fluid pumping job.

[0135] As an example, a computer program product may include one or more computer-readable storage media that may include processor-executable instructions to instruct a computing system to perform one or more methods and / or one or more portions of a method.

[0136] According to an embodiment, one or more computer-readable media may include computer-executable instructions to instruct a computing system to output information for controlling a process. For example, such instructions may provide for output to sensing process, an injection process, drilling process, an extraction process, an extrusion process, a pumping process, a heating process, etc.

[0137] In some embodiments, a method or methods may be executed by a computing system. Fig. 14 shows an example of a system 1400 that may include one or more computing systems 1401 -1 , 1401 -2, 1401 -3 and 1401 -4, which may be operatively coupled via one or more networks 1409, which may include wired and / or wireless networks.

[0138] As an example, a system may include an individual computer system or an arrangement of distributed computer systems. In the example of Fig. 14, the computer system 1401 -1 may include one or more modules 1402, which may be or include processor-executable instructions, for example, executable to perform various tasks (e.g., receiving information, requesting information, processing information, simulation, outputting information, etc.).

[0139] As an example, a module may be executed independently, or in coordination with, one or more processors 1404, which is (or are) operatively coupled to one or more storage media 1406 (e.g., via wire, wirelessly, etc.). As an example, one or more of the one or more processors 1404 may be operatively coupled to at least one of one or more network interface 1407. In such an example, the computer system 1401 -1 may transmit and / or receive information, for example, via the one or more networks 1409 (e.g., consider one or more of the Internet, a private network, a cellular network, a satellitenetwork, etc.). As shown, one or more other components 1408 may be included in the computer system 1401-1.

[0140] As an example, the computer system 1401-1 may receive from and / or transmit information to one or more other devices, which may be or include, for example, one or more of the computer systems 1401 -2, etc. A device may be located in a physical location that differs from that of the computer system 1401 -1 . As an example, a location may be, for example, a processing facility location, a data center location (e.g., server farm, etc.), a rig location, a wellsite location, a downhole location, etc.

[0141] As an example, a processor may be or include a microprocessor, microcontroller, processor module or subsystem, programmable integrated circuit, programmable gate array, or another control or computing device.

[0142] As an example, the storage media 1406 may be implemented as one or more computer-readable or machine-readable storage media. As an example, storage may be distributed within and / or across multiple internal and / or external enclosures of a computing system and / or additional computing systems.

[0143] As an example, a storage medium or storage media may include one or more different forms of memory including semiconductor memory devices such as dynamic or static random access memories (DRAMs or SRAMs), erasable and programmable read-only memories (EPROMs), electrically erasable and programmable read-only memories (EEPROMs) and flash memories, magnetic disks such as fixed, floppy and removable disks, other magnetic media including tape, optical media such as compact disks (CDs) or digital video disks (DVDs), BLUERAY disks, or other types of optical storage, or other types of storage devices.

[0144] As an example, a storage medium or media may be located in a machine running machine-readable instructions, or located at a remote site from which machine- readable instructions may be downloaded over a network for execution.

[0145] As an example, various components of a system such as, for example, a computer system, may be implemented in hardware, software, or a combination of both hardware and software (e.g., including firmware), including one or more signal processing and / or application specific integrated circuits.

[0146] As an example, a system may include a processing apparatus that may be or include a general-purpose processors or application specific chips (e.g., or chipsets), such as ASICs, FPGAs, PLDs, or other appropriate devices.

[0147] As an example, a device may be a mobile device that includes one or more network interfaces for communication of information. For example, a mobile device may include a wireless network interface (e.g., operable via IEEE 802.11 , ETSI GSM, BLUETOOTH, satellite, etc.). As an example, a mobile device may include components such as a main processor, memory, a display, display graphics circuitry (e.g., optionally including touch and gesture circuitry), a SIM slot, audio / video circuitry, motion processing circuitry (e.g., accelerometer, gyroscope), wireless LAN circuitry, smart card circuitry, transmitter circuitry, GPS circuitry, and a battery. As an example, a mobile device may be configured as a cell phone, a tablet, etc. As an example, a method may be implemented (e.g., wholly or in part) using a mobile device. As an example, a system may include one or more mobile devices.

[0148] As an example, a system may be a distributed environment, for example, a so-called “cloud” environment where various devices, components, etc. interact for purposes of data storage, communications, computing, etc. As an example, a device or a system may include one or more components for communication of information via one or more of the Internet (e.g., where communication occurs via one or more Internet protocols), a cellular network, a satellite network, etc. As an example, a method may be implemented in a distributed environment (e.g., wholly or in part as a cloud-based service).

[0149] As an example, information may be input from a display (e.g., consider a touchscreen), output to a display or both. As an example, information may be output to a projector, a laser device, a printer, etc. such that the information may be viewed. As an example, information may be output stereographically or holographically. As to a printer, consider a 2D or a 3D printer. As an example, a 3D printer may include one or more substances that may be output to construct a 3D object. For example, data may be provided to a 3D printer to construct a 3D representation of a subterranean formation. As an example, layers may be constructed in 3D (e.g., horizons, etc.), geobodies constructed in 3D, etc. As an example, holes, fractures, etc., may be constructed in 3D (e.g., as positive structures, as negative structures, etc.).

[0150] Although only a few examples have been described in detail above, those skilled in the art will readily appreciate that many modifications are possible in the examples. Accordingly, all such modifications are intended to be included within the scope of this disclosure as defined in the following claims. In the claims, means-plus- function clauses are intended to cover the structures described herein as performing therecited function and not only structural equivalents, but also equivalent structures. Thus, although a nail and a screw may not be structural equivalents in that a nail employs a cylindrical surface to secure wooden parts together, whereas a screw employs a helical surface, in the environment of fastening wooden parts, a nail and a screw may be equivalent structures.

Claims

CLAIMSWhat is claimed is:1 . A method comprising: receiving data from field equipment during performance of a fluid pumping job that comprises multiple stages at a wellsite; generating an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assessing the performance of the fluid pumping job based at least in part on the inference.

2. The method of claim 1 , wherein the fluid pumping job comprises pumping different fluids.

3. The method of claim 2, wherein the different fluids comprise a cement slurry.

4. The method of claim 2, wherein the different fluids comprise one or more of a spacer fluid and a displacement fluid.

5. The method of claim 1 , wherein the fluid pumping job comprises pumping one or more plugs.

6. The method of claim 5, wherein the one or more plugs comprise one or more of a bottom plug and a top plug.

7. The method of claim 1 , wherein the machine learning model comprises a time series model.

8. The method of claim 7, wherein the time series model comprises one or more recurrent structures.

9. The method of claim 7, wherein the time series model comprises a recurrent neural network model.

10. I he method ot claim ( , wherein the time series model comprises long short-term memory.

11. The method of claim 1 , wherein the machine learning model comprises a transformer model.

12. The method of claim 11 , wherein the transformer model comprises positional encoding structures.

13. The method of claim 1 , wherein the machine learning model comprises features that depend on the data from field equipment.

14. The method of claim 13, wherein the features comprise a density feature based on density data from the field equipment, a rate feature based on rate data from the field equipment, and a pressure feature based on pressure data from the field equipment.

15. The method of claim 13, wherein the machine learning model generates output using design data, wherein the output is for one or more planned pumping stages and / or for one or more casing configurations.

16. The method of claim 13, wherein the features comprise one or more of a downhole density feature and a cumulative volume feature.

17. The method of claim 1 , comprising comparing the performance of the fluid pumping job to a fluid pumping job plan.

18. The method of claim 17, comprising controlling the performance of the fluid pumping job based at least in part on the comparing.

19. A system comprising: one or more processors; memory accessible to at least one of the one or more processors; and processor-executable instructions stored in the memory and executable to instruct the system to:receive data from field equipment during performance of a fluid pumping job that comprises multiple stages at a wellsite; generate an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assess the performance of the fluid pumping job based at least in part on the inference.

20. One or more computer-readable storage media comprising processor-executable instructions to instruct a computing system to: receive data from field equipment during performance of a fluid pumping job that comprises multiple stages at a wellsite; generate an inference as to an occurrence of one of the multiple stages associated with the performance of the fluid pumping job based on at least a portion of the data using a machine learning model; and assess the performance of the fluid pumping job based at least in part on the inference.

Images