A method, system, and one or more computer-readable storage media for burner control.

Abstract

BURNER CONTROL. The methods for controlling hydrocarbon burner systems described in this document include acquiring image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; processing the image data using at least one machine learning model trained to generate a flame mask and a smoke mask; characterizing the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and, based on the characterization, controlling the operation of the burner system.

Description

1 / 68 METHOD, SYSTEM AND ONE OR MORE COMPUTER-READABLE STORAGE MEDIA FOR BURNER CONTROL CROSS-REFERENCE TO RELATED APPLICATIONS

[001] This application claims priority over U.S. Provisional Application Serial No. 62 / 957648, filed January 6, 2020, which is incorporated herein by reference. FUNDAMENTALS FIELD

[002] The embodiments described here generally refer to burners for excess hydrocarbon. Specifically, the embodiments described here refer to combustion control in such burners. DESCRIPTION OF THE RELATED TECHNIQUE

[003] The global oil and gas industry is trending towards improved environmental safety and compliance across all the various phases of a well's lifecycle. Exploration and production involve dynamic testing of wells that can produce a large amount of hydrocarbons at the surface. Excess hydrocarbons cannot be stored, so the most economically viable option is often to dispose of the excess hydrocarbons by flaring. This is even more relevant for offshore operations.

[004] The combustion of hydrocarbons will typically result in some environmental impact, even for clean burner operation with no visible residue and smoke. Most environmental impacts are created by spills and residues. This can be due to incomplete combustion resulting from fluid changes, poor burner operating parameters, and / or poor monitoring. The start-up and shutdown phases Petition 870240067970, dated 09 / 08 / 2024, page 13 / 91 2 / 68 are critical and need to be closely monitored, which requires good communication and human interaction.

[005] Even the best burner needs constant monitoring and air supply adjustment during such operations to maintain acceptable combustion through variations in fluid properties, flow rates, and weather conditions.

[006] For the continuous burning phase, which can last for days, monitoring and regulating the air supply to the burner becomes difficult. Failure to monitor combustion and adjust the air supply according to the flame or smoke appearance will have an immediate impact on the quality and emissions of the burner's combustion. Improved methods for monitoring and controlling hydrocarbon burners are needed. SUMMARY

[007] The embodiments described in this document provide methods for controlling hydrocarbon burner systems, which include acquiring image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; processing the image data using at least one machine learning model trained to generate a flame mask and a smoke mask; characterizing the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and, based on the characterization, controlling the operation of the burner system. Several other examples of methods, systems, computer program products, etc., are also described in this document. Petition 870240067970, dated 09 / 08 / 2024, page 14 / 91 3 / 68 BRIEF DESCRIPTION OF THE DRAWINGS

[008] In order for the features of this disclosure mentioned above to be understood in detail, a more specific description of the disclosure, briefly summarized above, may be taken as a reference for the embodiments, some of which are illustrated in the accompanying drawings. It should be noted, however, that the accompanying drawings illustrate only exemplary embodiments and should not, therefore, be considered limiting of their scope, and may be admitted in other equally effective embodiments.

[009] Figure 1 is a series of diagrams of illustrative environments and an example of a burner system.

[010] Figure 2 is a diagram of an example of a burner system.

[011] Figure 3 is a diagram of an example of a burner rod.

[012] Figure 4 is a diagram of an example of a burner system.

[013] Figure 5 is a diagram of an example of a method.

[014] Figure 6 is a diagram of an example of a color map.

[015] Figure 7 is a diagram of an example of a smoke graph.

[016] Figure 8 is a diagram of an example of a method.

[017] Figure 9 is a diagram of an example of a machine learning model.

[018] Figure 10 is a diagram of an example of Petition 870240067970, dated 09 / 08 / 2024, page 15 / 91 4 / 68 a computational structure and some examples of models.

[019] Figure 11 is a diagram of an example of a method.

[020] Figure 12 is a diagram of an example of a method and an example of a system.

[021] Figure 13 is a diagram of an example of a computer system.

[022] Figure 14 is an example component diagram of a system and a networked system.

[023] To facilitate understanding, identical reference numerals have been used, where possible, to designate identical elements that are common to the Figures. It is contemplated that the elements and characteristics of one modality may be beneficially incorporated into other modalities without additional recitation. DETAILED DESCRIPTION

[024] The following description includes the best currently contemplated method for carrying out the implementations described. This description should not be taken in a limiting sense, but is produced merely for the purpose of describing the general principles of the implementations. The scope of the implementations described should be determined with reference to the published claims.

[025] Figure 1 shows examples of environments 101, which include a marine environment 102 and a terrestrial environment 104, wherein the marine environment 102 includes various equipment and wherein the terrestrial environment 104 includes various equipment. As shown, each of the environments 101 may include one or more wellheads 106 (e.g., wellhead equipment). A wellhead may be a termination Petition 870240067970, dated 09 / 08 / 2024, page 16 / 91 5 / 68 of the surface of a wellbore that may include a system of coils, valves, and various adapters that, for example, can provide pressure control of a production well. A wellhead may be on a land surface, a subsea surface (e.g., an ocean floor, etc.), etc. As an example, conduits from multiple wellheads may be joined into one or more pipelines so that fluid from multiple wells can flow into a common conduit.

[026] At various times, a well may be tested using a process called well testing. Well testing may include one or more of a variety of well testing operations. In many cases, fluid may flow from a well or wells to the surface where the fluid is subjected to one or more well testing operations and generates waste (e.g., refuse fluid), which must be handled appropriately, for example, according to circumstances, regulations, etc. For example, consider loading refuse fluid onto a tanker for transport to a facility that can dispose of the refuse fluid. Another way to handle refuse fluid may be through combustion, which may be termed burning. As an example, burning may be part of a well testing process, either burning to handle refuse fluid and / or to analyze one or more aspects of how one or more refuse fluids burn.Regarding the latter, the burn-up can optionally provide data regarding one or more well fluid characteristics (e.g., a component thereof, etc.).

[027] As an example, well testing can be carried out during one or more phases, such as during exploration and Petition 870240067970, dated 09 / 08 / 2024, page 17 / 91 6 / 68 assessment, in which hydrocarbon production is tested using a temporary production facility that can provide fluid sampling, flow rate analysis, and pressure information generation, for example, to help characterize a reservoir. Several decisions can be made based on well testing, such as, for example, decisions regarding production methods, facilities, and potential well productivity improvements.

[028] As an example, disposal of hydrocarbons produced during well testing can be by flaring, which may include on-site flaring and / or off-site flaring. Flaring may be particularly suitable when storage facilities are not available (e.g., consider mobile offshore drilling rigs, remote onshore locations, etc.).

[029] With regard to the exemplary environments 101 in Figure 1, well testing can be performed, for example, using equipment shown in the marine environment 102 and / or using equipment shown in the terrestrial environment 104. As an example, an environment may be under exploration, development, appraisal, etc., and such environment includes at least one well where well fluid can be produced (e.g., by natural pressure, by fracturing, by artificial lift, by pumping, by flooding, etc.). In such an environment, various types of equipment may be on site, which can be operationally coupled to the well testing equipment.

[030] Figure 1 shows an example of a system 110 that can be operationally coupled to one or more Petition 870240067970, dated 09 / 08 / 2024, page 18 / 91 7 / 68 conduits that can transport well fluid, for example, from one or more wellheads. As shown, the system 110 may include a computing system 111 (CS), which may include one or more processors 112, memory 114 accessible to at least one of the one or more processors 112, instructions 116 that can be stored in memory 114 and executable by at least one of the one or more processors 112, and one or more interfaces 118 (for example, wired, wireless, etc.). In the example in Figure 1, system 110 is shown including various communication symbols, which may be for transmitting and / or receiving information (e.g., data, commands, etc.), for example, to and / or from computer system 111. As an example, computer system 111 may be a controller that can issue control instructions to one or more pieces of equipment in an environment, such as, for example, the marine environment 102 and / or the terrestrial environment 104.For example, computer system 111 can be local, remote, or distributed in such a way that it is partly local and partly remote.

[031] With reference, again, to wellhead 106, it may include various types of wellhead equipment, such as, for example, casing heads and tubing, a production tree, a blowout preventer, etc. Fluid produced from a well may be routed through wellhead 106 and into system 110, which may be configured with various features for well testing operations.

[032] In the example in Figure 1, the 110 system is shown to include multiple segments, which can be Petition 870240067970, dated 09 / 08 / 2024, page 19 / 91 8 / 68 categorized operationally. For example, consider a well control segment 120, a separation segment 122, a fluid management segment 124, and a flare segment 126.

[033] As shown in the example in Figure 1, the well control segment 120 is an assembly of several components, such as a pipe 130, a choke pipe 132, a pipe 134, a heat exchanger 136 and a meter 138; the separation segment 122 includes a separator 142; the fluid management segment 124 is an assembly of several components, such as pipes and pumps 144, a pipe 146-1, a pipe 146-2, a tank 148-1 and a tank 148-2; and the flare segment 126 includes a burner 152 and one or more chambers 154.

[034] As mentioned, in the example in Figure 1, system 110 includes several features for one or more aspects of well testing operations; noting that system 110 may include fewer features, more features, alternative features, etc. For example, consider one or more of a specific gas gravity meter, a water content meter, a gas-to-oil ratio sensor, a carbon dioxide sensor, a hydrogen sulfide sensor, or a shrinkage measuring device. Several features may be upstream and / or downstream of a separator segment or a separator.

[035] With regard to fluid flow from a well or wells, such fluid may be received by the well control segment 120 and then routed through one or more conduits to the separation segment 122. In the example of Petition 870240067970, dated 09 / 08 / 2024, page 20 / 91 9 / 68 Figure 1, the well control segment 120 and the heat exchanger 136 can be supplied as a steam heat exchanger and the meter 138 to measure fluid flow through the well control segment 120.

[036] As mentioned, well control segment 120 can transport fluid received from one or more wells to separator 142. As an example, separator 142 can be a horizontal separator or a vertical separator, and can be a two-phase separator (e.g., to separate gas and liquids) or a three-phase separator (e.g., to separate gas, oil, and water). A separator may include various features to facilitate the separation of incoming fluid components (e.g., diffusers, mist extractors, vanes, deflectors, precipitators, etc.).

[037] As an example, fluid can be single-phase or multi-phase fluid, where phase refers to an immiscible component (for example, consider two or more from among oil, water and gas).

[038] As an example, separator 142 can be used to substantially separate multi-phase fluid into its oil, gas, and water phases, as appropriate and as present, wherein each phase emerging from separator 142 can be termed a separated fluid. Such separated fluids can be routed in the opposite direction from separator 142 to fluid management segment 124. In several cases, the separated fluids may not be completely homogeneous. For example, separated gas leaving separator 142 may include some residual amount of water or oil, and separated water leaving separator 142 Petition 870240067970, dated 09 / 08 / 2024, page 21 / 91 10 / 68 may include some amount of entrained oil or gas. Similarly, separated oil leaving separator 142 may include some amount of entrained water or gas.

[039] As shown in the example in Figure 1, the fluid management segment 124 includes flow control equipment, such as various pipes and pumps (generally represented by block 144) to receive fluids from separator 142 and transport the fluids to other destinations, as well as additional pipes 146-1 and 146-2 to route fluid to and from fluid tanks 148-1 and 148-2. Although two pipes 146-1 and 146-2 and two tanks 148-1 and 148-2 are represented in Figure 1, note that the number of pipes and tanks can vary. For example, in one embodiment, the fluid management segment 124 may include a single pipe and a single tank, while in other embodiments, the fluid management segment 124 may include more than two pipes and more than two tanks.

[040] With respect to the piping and pumps 144, these may include a variety of piping and pumps, such as a gas piping, an oil piping, an oil transfer pump, a water piping, and a water transfer pump. In at least some embodiments, the piping and pumps 144 may be used to route fluids received from the separator 142 to one or more of the fluid tanks 148-1 and 148-2 by means of one or more of the additional piping 146-1 and 146-2, and to route fluids between tanks 148-1 and 148-2. As an example, the piping and pumps 144 may include features for routing fluids received from the separator 142 Petition 870240067970, dated 09 / 08 / 2024, page 22 / 91 11 / 68 directly to one or more burners 152 to burn gas and oil (for example, bypassing tanks 148-1 and 148-2) or to route fluids from one or more tanks 148-1 and 148-2 to one or more burners 152.

[041] As noted above, components of system 110 may vary between different applications. As an example, equipment within each functional group of system 110 may also vary. For example, heat exchanger 136 could be supplied as part of separation segment 122, rather than well control segment 120.

[042] In certain embodiments, system 110 may be a surface well testing system that can be remotely monitored and controlled. Remote monitoring may be effected with sensors installed on various components. In some cases, a monitoring system (e.g., sensors, communication systems, and human-machine interfaces) may enable the monitoring of one or more of segments 120, 122, 124, and 126. As shown in the example in Figure 1, one or more cameras 154 may be used to monitor one or more firing operations of one or more burners 152, which may seek to facilitate the control of such one or more firing operations, at least in part, through the analysis of image data acquired by at least one of the one or more cameras 154.

[043] Figure 2 shows an example of a 250 system, which can be called a surface well test system. The 250 system may include several features of the 110 system in Figure 1.

[044] In Figure 2, a multiphase fluid Petition 870240067970, dated 09 / 08 / 2024, page 23 / 91 12 / 68 (represented here by arrow 252) enters a test head 254 and is routed to a separator 270 via a surface safety valve 256, a steam heat exchanger 260, a throttling pipe 262, a flow meter 264, and an additional pipe 266. In the example in Figure 2, the system 250 includes a chemical injection pump 258 to inject chemicals into the multiphase fluid flowing towards the separator 270.

[045] In the embodiment shown in Figure 2, separator 270 is a three-phase separator that generally separates the multi-phase fluid 252 into gas, oil, and water components. The separated gas is routed downstream of separator 270 through a gas pipeline 274 to either of the burners 276-1 and 276-2 for gas and oil combustion. The gas pipeline 274 includes valves that can be actuated to control the gas flow from the gas pipeline 274 to either of the burners 276-1 and 276-2. Although shown side-by-side in Figure 2, for clarity, burners 276-1 and 276-2 can be positioned separately from each other, as well as on opposite sides of a platform, etc.

[046] As shown, the oil separated from separator 270 can be routed downstream to an oil pipeline 280. Valves on the oil pipeline 280 can be operated to allow oil flow to either of the burners 276-1 and 276-2 or either of the tanks 282 and 284. Tanks 282 and 284 can be of a suitable shape, however, they are represented in Figure 2 as vertical compensating tanks, each having two fluid compartments. This allows each tank Petition 870240067970, dated 09 / 08 / 2024, page 24 / 91 13 / 68 simultaneously retain different fluids, such as water in one compartment and oil in another compartment. An oil transfer pump 286 can be operated to pump oil through the well test system 250 downstream of the separator 270. The water separated from the separator 270 can be similarly routed to a water pipe 290. Like the oil pipe 280, the water pipe 290 includes valves that can be opened or closed to allow water to flow to either of the tanks 282 and 284 or to a water treatment and disposal device 294. A water transfer pump 292 can be used to pump water through the system.

[047] A well test area in which the 250 well test system (or other embodiments of a well test system) is installed may be classified as a hazardous area. In some embodiments, the well test area is classified as a Zone 1 hazardous area according to the International Electrotechnical Commission (IEC) standard 60079-10-1:2015.

[048] In the example in Figure 2, a cabin 296 at a well site may include various types of equipment for acquiring data from the well test system 250. This acquired data may be used to monitor and control the well test system 250. In at least some cases, the cabin 296 may be separated from the well test area with the well test system 250 in a non-hazardous area. This is represented by the dashed line 298 in Figure 2, which generally serves as a demarcation between the hazardous area containing the well test system 250 and the non-hazardous area of ​​the cabin 296. Petition 870240067970, dated 09 / 08 / 2024, page 25 / 91 14 / 68

[049] The equipment of a well testing system can be monitored during a well testing process to verify proper operation and facilitate process control. Such monitoring may include taking various measurements during a well test, examples of which may include temperature and pressures of choke piping (upstream and downstream), heat exchanger temperature and pressure, separator temperature and pressures (static and differential), separator oil flow rate and volume, separator water flow rate and volume, and fluid levels in tanks of a system.

[050] As an example, a mobile monitoring system can be fitted to a surface well testing facility. In such an example, monitoring of a well testing process can be carried out on a mobile device (e.g., a mobile device suitable for use in the hazardous area of ​​Zone 1, such as the well testing area). Various types of information can be automatically acquired by sensors and then presented to an operator via the mobile device. The mobile monitoring system can provide various functions, such as a sensor data display, video display, interpretation of sensor or video information for quality assurance and quality control purposes, and a manual input screen (e.g., for a digital logbook for recording measurements taken by the operator).

[051] Figure 3 shows an example of a 300 burner rod, which can be configured for horizontal mounting, angled mounting, vertical mounting, etc. For example, the 300 burner rod can be mounted in Petition 870240067970, dated 09 / 08 / 2024, page 26 / 91 15 / 68 a platform with a rotating base plate and stays. In such an example, horizontal stays can help guide the burner rod 300; vertical stays, which are fixed to the main structure of the platform, can support the burner rod 300. A rotating base can enable horizontal and vertical positioning of the burner rod 300 and its burner. As an example, the burner rod 300 can be positioned slightly above horizontal so that the oil left in the piping after combustion operations does not leak.

[052] For example, a burner can be mounted on the rod or mounted on another type of support structure. For example, a structure can support several conduits that supply fluid, such as, for example, one or more of air, water, oil and propane.

[053] In the example in Figure 3, conduits or lines include an additional gas line 310, pilot line cables 320, an oil line 330, a water line 340, a water wall screen line 350, an air line 360 ​​and a main gas line 370.

[054] As an example, a burner can be configured and controlled to perform in a desirable manner. For example, it may be desirable to burn in a precipitation-free and smokeless manner for combustion of liquid hydrocarbons produced during well testing. As an example, a burner geometry may utilize pneumatic atomization and intensified air induction. As an example, a burner may be equipped with one or more pilots, a front flame ignition system (BRFI), and a built-in water screen to reduce heat radiation. As Petition 870240067970, dated 09 / 08 / 2024, page 27 / 91 16 / 68 For example, a burner can be fitted with an automatic shut-off valve that prevents oil spillage at the beginning and end of a burn run.

[055] As an example, a burner may include a high derating feature (e.g., 1:5), which may optionally be further extended to a higher ratio (e.g., 1:30) with the use of a multi-rate kit option (BMRK), which allows selection of the number of operating nozzles. For onshore operations, a slider may be used for slide mounting.

[056] As an example, a burner may be suitable for high-efficiency burning with one or more types of oil (for example, including, in particular, heavy and waxy oils). As an example, a burner may operate effectively down to a water content rating (for example, up to 25 percent water content), which may be desirable for various types of cleaning operations.

[057] As a burner can be operated in a manner that provides substantially no liquid precipitation and substantially no visible smoke emission, such a burner may be particularly suitable for operations in environmentally sensitive areas.

[058] Figure 4 is a system diagram of an example of a burner control system 400 according to one embodiment. The system 400 includes at least one camera 407-1 and 407-2 positioned to capture one or more images 402-1 and 402-2 of a flame emitted by a burner 401. In the example of Figure 4, two cameras 407-1 and 407-2 are shown capturing images 402-1 and 402-2 from different locations to acquire image data from more than Petition 870240067970, dated 09 / 08 / 2024, page 28 / 91 17 / 68 a plane image of the flame. Burner 401 includes a fuel feed 403 that flows fuel to burner 401 (see, for example, burners 2761 and 276-2 in Figure 2, burner 300 in Figure 3, etc.). Burner 401 also includes an air supply 405 that flows air to burner 401. The flow rate of the air supply 405 is controlled by a control valve 408, where an air flow sensor 411 detects the air flow rate in burner 401. A fuel flow sensor 413 detects the fuel flow rate in burner 401. Other sensors 404, along with at least one camera 407-1 and 407-2, are operationally coupled to the local electronic equipment 420. The sensors 404 can detect and produce signals representing effective combustion parameters such as temperature, wind speed, and ambient humidity.Sensors 404, 411, and 413, and cameras 407-1 and 407-2 send data, which includes data representing images 402-1 and 402-2, along with data representing readings from sensors 404, 411, and 413, directly and / or indirectly, to local electronic equipment 420. The data sent to local electronic equipment 420 may represent a state of combustion occurring in burner 401.

[059] As shown in the example in Figure 4, the local electronic equipment 420 can be in communication with the remote electronic equipment 440. For example, consider the use of one or more wired and / or wireless interfaces that allow communication between the local and remote electronic equipment 420 and 440. In such an example, various computational tasks can be performed locally and / or remotely. Petition 870240067970, dated 09 / 08 / 2024, page 29 / 91 18 / 68 For example, consider a local computing device that may include an application that can render a graphical user interface (GUI) on a display. In such an example, the GUI may include control graphics that are selectable to issue instructions, such as, for example, one or more application programming interface (API) calls, which may be directed to the local electronic equipment 420, the remote electronic equipment 440, etc., to cause one or more actions to occur, such as formulating a response in an API call.

[060] As an example, local and remote electronic equipment 420 and 440 can be configured in a client-server arrangement where local electronic equipment 420 operates as a client and remote electronic equipment 440 operates as a server. As an example, data acquired by local electronic equipment 410 (e.g., as part of system 400) can be processed for local control and / or transmitted (e.g., as raw and / or processed data) for processing, storage, etc., by remote electronic equipment 440. As an example, remote electronic equipment 440 can include one or more cloud-based resources. In such an example, remote electronic equipment 440 can provide services, such as, for example, software as a service.As an example, control performed by system 400 to control burner 401 can be based on local and / or remote computing (for example, using local electronic equipment 420, remote electronic equipment 440, etc.).

[061] As an example, control can be model-based, where such a model can be a model based on Petition 870240067970, dated 09 / 08 / 2024, page 30 / 91 19 / 68 laws of physics, a machine learning model, a hybrid model, etc. As an example, a model-based approach can allow the prediction of various parameters, such as, for example, air control parameters based on data from sensors 404, 411, and 413, and at least one of one or more cameras 407-1 and 407-2. As an example, one or more air control parameters can be applied to control valve 408 to control the air supply to burner 401, which may be part of a combustion process that generates a flame that can be captured, as represented, in one or more images 402-1 and 402-2.

[062] Camera, as used in this document, means an imaging device. A camera can capture an image of electromagnetic (EM) radiation on a medium that can be converted into data for use in digital processing. The conversion can occur within the camera or in a separate processor. The camera can capture images at a wavelength or across a spectrum (e.g., or spectra), which may encompass the ultraviolet (UV) spectrum, the visible spectrum, and / or the infrared spectrum. For example, a camera can capture an image of wavelengths from approximately 350 nm to approximately 1500 nm.For example, one or more of a wide-spectrum imaging device, such as a LIDAR-type detector, a narrower-spectrum detector, such as a charged-coupled device (CCD) array, and a short-wave infrared detector, may be used as an imaging device or as imaging devices. Cameras may be monovision cameras or stereo cameras. Petition 870240067970, dated 09 / 08 / 2024, p. 31 / 91 20 / 68

[063] In the example of Figure 4, the local electronic equipment 420 is shown including an image processing unit 410, which may include and / or be operationally coupled to a model or models for the purpose of processing one or more of the images 402-1 and 402-2, as captured by at least one camera 407-1 and 407-2. As an example, a dataset, together with sensor data representing oil flow rate, gas flow rate, water or steam flow rate, air flow rate, pressure, temperature, wind speed, ambient humidity and other effective combustion parameters, may be considered different types of inputs. As an example, a model may receive input or inputs and may output one or more air control parameters, such as flow rate, pressure and / or temperature, to the burner 401 (for example, or burners) controlled by the system 400.

[064] With respect to a machine learning model (ML model), such a model may be a neural network model (NN model). As an example, a trained ML model may be used to control one or more burners. As an example, a trained ML model may be trained with respect to a particular burner and / or burner type. In such an approach, a trained ML model may be selected based, at least in part, on burner specifications (e.g., manufacturer, model, features, history, etc.).

[065] Various types of data can be acquired and optionally stored, which can provide training for one or more ML models and / or offline analysis, Petition 870240067970, dated 09 / 08 / 2024, page 32 / 91 21 / 68 etc. For example, air control parameters emitted by a trained NN model can be stored in digital storage for later analysis, which may include additional training, training of a different ML model, etc. During the control of an ongoing combustion operation, one or more air control parameters can be transmitted to one or more control valves that control air supply to one or more burners to the extent that they can be operationally coupled to the system controlled by the control system. Subsequent image acquisitions and sensor data can be captured, and the control cycle repeated as many times as desired. Repetition frequency may depend on various temporal constants of the 400 system. As an example, a cycle can be as short as a fraction of a second or as long as five to ten minutes.As an example, consider an operating frequency of 1 Hz where multiple images are captured at one-second intervals, such as in a video feed where the computation of air control parameters and the application of the computed air control parameters to a control valve that controls the air supply to the burner are based on the images at one-second intervals. For example, the video could be live, with some amount of latency due to transmission and processing time, or the video could be deliberately delayed by a certain amount of delay.

[066] The image processing unit 410 converts signals derived from photons received by one or more cameras 407-1 and 407-2 into data. The image processing unit 410 may be inside or separate from Petition 870240067970, dated 09 / 08 / 2024, page 33 / 91 22 / 68 a camera. As an example, the 410 image processing unit can convert signals received from one or more 407-1 and 407-2 cameras into digital data representing photointensity in defined areas of the image and can assign position information to each digital data value. As an example, photointensity can be deconvolved into constituent wavelengths to produce a spectrum for each pixel. Such a spectrum can be sampled into defined boxes, and the data from such sampling structured into a dataset representing the spectral intensity of the received image, for example, as a function of xy position in the image. As an example, a time stamp can be added.

[067] As an example, an image can be a pixelated image with pixel position coordinates, a pixel intensity, and a time record. Regarding depth, various conventions can be used and depend on the equipment and / or processing. When color is used, a color intensity can be mentioned. For example, 8-bit color and 24-bit color can be the same when, in an RGB color space, 8 bits refers to each R, G, and B (e.g., subpixel), while 24 bits is a sum of the three 8-bit channels (e.g., 3x8 = 24). Standards can include, for example, monochrome (e.g., 1 bit) up to 4K (e.g., 12-bit color, which provides 4096 colors), etc.

[068] With regard to color models, RGB can be mapped to a cube. For example, a horizontal x-axis can be a red (R) geometric axis for red values, a y-axis can be a blue (B) geometric axis for blue values, and an axis Petition 870240067970, dated 09 / 08 / 2024, page 34 / 91 23 / 68 geometric z can be a geometric axis of green (G) for values ​​of green. The origin of such a cube (e.g., 0, 0, 0) can be black and an opposite point can be white (e.g., 1, 1, 1).

[069] Another type of color model is the Y'UV model, which defines a color space in terms of a luma component (Y') and two chrominance components, called U (blue projection) and V (red projection), respectively. The Y'UV color model is used in the PAL composite color video standard (except PAL-N).

[070] Yet another type of color model is the HSV color model. The RGB color model can define a color as percentages of shades of red, green, and blue (e.g., as mixed together), while the HSV color model can define color with respect to hue (H), saturation (S), and value (V). For the HSV color model, as hue ranges from 0 to 1.0, corresponding colors range from red to yellow, green, cyan, blue, magenta, and back to red (e.g., values ​​of red exist at both 0 and 1.0); as saturation ranges from 0 to 1.0, corresponding colors (hues) range from unsaturated (e.g., shades of gray) to fully saturated (e.g., no white component); and as value, or brightness, ranges from 0 to 1.0, corresponding colors become increasingly brighter.

[071] Saturation can be described as, for example, representing the purity of a color, where colors with higher saturation can have higher values ​​(for example, represented as white in terms of saturation) and when color mixtures are represented Petition 870240067970, dated 09 / 08 / 2024, page 35 / 91 24 / 68 as shades of gray (for example, cyans, greens, and shades of yellow are mixtures of true colors). As an example, saturation can be described as the representation of the “color intensity” of a stimulus relative to its own brightness; wherein “color intensity” is an attribute of a visual sensation, according to this, the perceived color of an area appears to be more or less chromatic, and wherein “brightness” is an attribute of a visual sensation, according to this, an area appears to emit more or less light.

[072] As an example, a method might include accessing one or more resources with respect to color models (e.g., as a plug-in, external executable code, etc.). For example, consider a method that includes instructions to access an algorithm from a package, a computing environment, etc., such as, for example, the MATLAB® computing environment (marketed by MathWorks, Inc., Natick, MA). The MATLAB® computing environment includes an image processing toolbox, for example, with algorithms for color space conversions (e.g., color model), transformations, etc. As an example, the MATLAB® computing environment includes functions rgb2hsv” and hsv2rgb” for converting images between the RGB and HSV color spaces (see, for example, http: / / www.mathworks.com / help / images / converting-color-databetween-color-spaces.html).

[073] Figure 5 shows an example of a 500 method along with an example of an RGB 502 color model and an HSV 504 color model (for example, when colors are represented by their first letter). In the example in Figure 5, RGB 522, 524, and 526 data are provided from Petition 870240067970, dated 09 / 08 / 2024, page 36 / 91 25 / 68 one or more sources. As shown, a 550 transformation block can transform the continuous RGB color model representation of the 522, 524, and 526 data into 562 H-band data, 564 V-band data, and / or 504 S-band data to HSV color model bands. As an example, an algorithm that provides conversion of RGB data to HSV data can be implemented, optionally with zero padding, etc., for example, when fewer than three data sets are provided (e.g., two of the 522, 524, and 526 data sets).

[074] Figure 6 shows an example of a 610 color map, which includes hue (H) and saturation (S) coordinates (e.g., from an HSV color model). As shown, the hue includes a range of red and transitions to red / orange, orange, orange / yellow, and yellow.

[075] Flame colors can depend on the flame temperature and the material being burned. Red can be observed, for example, in a temperature range from approximately 525 °C (977 °F) to approximately 1000 °C (1830 °F); orange can be observed in a range from approximately 1100 °C (2010 °F) to approximately 1200 °C (2190 °F) and white can be observed in a range from approximately 1300 °C (2370 °F) to approximately 1500 °C (2730 °F).

[076] The orange color in a campfire depends on the temperature and sodium in the wood, while blue streaks may come from carbon and hydrogen in the wood. The colors of stars can also indicate their temperatures. The closest star to Earth, the sun, has a surface temperature of approximately 5538 °C (10000 °F). Petition 870240067970, dated 09 / 08 / 2024, page 37 / 91 26 / 68

[077] Figure 7 shows an example of a Ringelmann smoke chart 700 which includes four different graduated maps 710, 720, 730 and 740. The Ringelmann system may include graduated shades of gray, for example, considering five equal steps between white and black, which may be represented by maps 710, 720, 730 and 740 as rectangular grids of black lines of defined width and spacing on a white background; note that white and black are not shown in Figure 7.

[078] In terms of numbered cards, Ringelmann's charts can be reproduced as follows: Card 0, white; Card 1, black lines 1 mm thick, separated by 10 mm, leaving square white spaces of 9 mm (see map 710); Card 2, lines 2.3 mm thick, square spaces of 7.7 mm (see map 720); Card 3, lines 3.7 mm thick, square spaces of 6.3 mm (see map 730); Card 4, lines 5.5 mm thick, square spaces of 4.5 mm (see map 740); and Card 5, black.

[079] The chart, as distributed by the Ministry of Mines and Energy, provides the shades of cards 1, 2, 3 and 4 on a single sheet, which are known as Ringelmann's No. 1, 2, 3, and 4, respectively.

[080] Several municipal, state, and federal guidelines prescribe smoke density limits based on the Ringelmann smoke chart, as published by the U.S. Department of Energy. While the chart still serves a useful purpose, it should be remembered that the data obtained through its use are empirical in nature. The apparent darkness or opacity of a column of Petition 870240067970, dated 09 / 08 / 2024, page 38 / 91 27 / 68 Smoke can depend on the concentration of particulate matter in the effluent, the size of the particulate matter, the depth of the smoke column that is seen, natural lighting conditions such as the direction of the sun relative to the observer, and the color of the particles. Since unburned carbon is a major coloring material in a smoke column from a furnace that uses coal or oil, the relative shade is a function of combustion efficiency.

[081] In various embodiments, a system, a method, etc., can provide operational situation characterization of a well test burner. As an example, one approach might utilize one or more color maps. For example, consider using an HSV color map for the purpose of determining combustion quality based at least in part on hue (e.g., a hue histogram, etc.) and determining a smoke index (e.g., by means of a Ringelmann smoke chart, etc.). In such an example, one or more RGB color map images could be acquired and processed using one or more machine learning (ML) models trained to output masks, which might include a flame mask to assess combustion quality and a smoke mask to assess smoke.In such an example, one or more RGB color map images can be transformed into one or more corresponding HSV color map images that can be evaluated using one or more masks emitted by a trained ML model. As an example, the operational situation of a well test burner can be used for well test burner control purposes. For example, consider that a combustion quality indicator is used to adjust one or more... Petition 870240067970, dated 09 / 08 / 2024, p. 39 / 91 28 / 68 plus operational parameters associated with fluid flow to a burner (e.g., gas, oil, water, air, etc.). In such an example, a feedback loop can be established that seeks to control how the material is burned, for example, to achieve one or more objectives (e.g., combustion, smoke, etc.).

[082] As an example, one method could be based on real-time video feed interpretation to generate status indicators and qualifiers of a well testing equipment burnout process. With respect to an ML model, for example, a convolutional neural network (CNN) could be trained and used to analyze, in real time, RGB frames generated by a network of cameras placed within sight of one or more burners. In such an approach, the trained CNN could detect and delineate objects, such as, for example, one or more flames, one or more smoke regions, and one or more protective water screens (e.g., sprinklers).

[083] As an example, a method may include defining a set of parameters for characterizing a flame or flames. For example, consider a method that defines and determines, through image interpretation, the status of one or more of the following parameters: flame presence and size; flame quality indicator; presence and size of white smoke; presence and size of gray smoke; presence and size of black smoke; and presence of a sprinkler. Such parameters may characterize the operational status of a burner, which may then optionally be controlled so that desired parameter values ​​are achieved.

[084] With regard to some examples of parameters Petition 870240067970, dated 09 / 08 / 2024, pp. 40 / 91 29 / 68 of flame parameters, which may be additional and / or alternative to one or more other flame parameters, consider one or more of the following: tip type, tip length, tip diameter, angle to the horizon, angle to the north, flow rate (or flow rates), calorific value, molecular weight (or molecular weights), heat release, radiated fraction, temperature, wind speed, wind direction, solar radiation, transmissivity, etc. As an example, a computational framework, such as the FLARESIM framework (Schlumberger Limited, Houston, Texas), can be used to determine values ​​of one or more flame parameters. As an example, a framework may include or be operationally coupled to an ML model, which may be an ML model framework that can output various types of data using one or more trained ML models.For example, one structure can access another structure through application programming interface (API) calls, where information (e.g., data, etc.) is returned in response to such calls. As explained, an ML model can be trained for image analysis of image data for one or more flames, etc., which can be fed into to generate output (see, for example, Figure 8, Figure 9, etc.).

[085] Monitoring oil and / or gas combustion in the field during the first days of well exploration can be very informative. An operator in the field may be called a “flame watcher,” who constantly observes a flame, on the one hand, to ensure that the flame is lit as oil and / or gas is expelled and, on the other hand, to adjust combustion driving parameters in case flame combustion is incomplete (e.g., Petition 870240067970, dated 09 / 08 / 2024, pp. 41 / 91 30 / 68 causing soot particles, toxic particles, etc.).

[086] As explained, a workflow can be implemented to detect a flame and characterize combustion based on real-time video monitoring using one or more machine learning techniques, which may include deep learning on data acquired by computer vision.

[087] As an example, a workflow can describe the state of a combustion process by detecting the presence of features such as flame, smoke, and a protective water screen. This workflow can further characterize combustion quality, for example, by analyzing the aspect and color distribution within a flame using computer vision. Regarding smoke, it can be classified into several categories; for example, consider the classification using a Ringelmann smoke graph approach, which can provide various classes ranging from white to black or various classes between white and black (e.g., various gray classes). A computerized approach can augment human observation or, for example, replace human observation.A computerized approach can help reduce subjectivity and increase the objectivity of operational analysis of a burner or burners, which can, in turn, improve the operational control of a burner or burners. Regarding subjectivity, in some cases, identification for the purpose of training a machine learning model (ML model) can be performed using experts. Petition 870240067970, dated 09 / 08 / 2024, page 42 / 91 31 / 68 humans. Alternatively or additionally, identification can be carried out using one or more computerized processes.

[088] As explained, cameras can be used to monitor well test burners. A camera can generate frames of a burning process where each frame corresponds to an exposure time and a field of view. For example, a camera may include a lens with a depth of field to focus an image onto an imaging plane, where the imaging plane may include a plurality of sensors that generate data that can be represented as pixels (e.g., figuration elements) using one or more color maps. As explained, a camera can generate a frame that is a pixelated image with a color intensity, which can be indicated by a bit intensity. As an example, a frame can be a pixelated image with a bit intensity corresponding to a particular color map, such as, for example, RGB or another suitable color map.Regarding exposure time and / or frame rate, consider an exposure time that is less than the reciprocity of the frame rate. For example, a frame rate of 24 frames per second would correspond to a maximum exposure time per frame of approximately 42 milliseconds or less.

[089] Actual camera operation may involve setting various parameters, such as, for example, a shutter speed mode, a shutter speed, a shutter angle, etc. Various cameras may include a set of compression circuits, which can compress frames using lossy or lossless compression. A stream Petition 870240067970, dated 09 / 08 / 2024, page 43 / 91 32 / 68 of the work may include using one or more types of image data, which can be represented as a frame, for example, for processing by a trained machine learning model (ML model).

[090] As an example, a deep learning approach can use one or more convolutional neural networks (CNNs) to segment a frame or frames into multiple categories (e.g., multiple masks). Segmentation via a CNN can provide pixel classification so that a pixel in a frame is identified as belonging to a class (e.g., flame, smoke, or water screen (or water screens)). In such an approach, one or more trained CNNs can provide, for each video frame, a probability map for each of a plurality of different object types (e.g., a flame-type object, a smoke-type object, a water screen-type object). As explained, the output of a trained CNN can be a plurality of masks that probabilistically provide indications that a pixel belongs to a corresponding object type from a plurality of object types.

[091] As an example, a method may include post-processing that uses masks resulting from one or more image processing techniques together with the application of one or more mathematical transformers of numerical values ​​of an RGB matrix to derive visual flame aspect indicators and smoke intensity (e.g., by means of a Ringelmann smoke chart scale, etc.).

[092] As an example, burner control may include video monitoring of a burner and evaluation. Petition 870240067970, dated 09 / 08 / 2024, page 44 / 91 33 / 68 of the burner's condition using various criteria, and this assessment can be performed using, for example, a trained machine learning model.

[093] One embodiment may involve using cameras to view a burner to generate RGB format frames as data from a burner combustion process, with the data being transmitted in real time to a computerized system that records and processes the data. The acquisition and analysis of such data may include visible light spectrum data and / or one or more other types of EM radiation. As an example, an embodiment may utilize one or more camera technologies, such as thermal and / or multispectral camera technologies.

[094] With regard to a burner operation, the combustion of hydrocarbons in the presence of air can generate a flame and smoke. A burner operation may use a water screen, which may be a type of sprinkler that uses water for cooling and / or one or more other purposes. As an example, a burner controller may implement burner control, where the burner controller may include features to characterize a burner operation by detecting the presence of flame, smoke and / or water screen and / or flame size, smoke and / or water screen.

[095] As explained, a CNN can be generated and trained to derive one or more probability masks, such as one or more of a flame mask, a smoke mask, and a water screen mask, which can be classes of objects in an image (e.g., or images). As an example, a CNN can provide segmentation of Petition 870240067970, dated 09 / 08 / 2024, p. 45 / 91 34 / 68 image, where a segmentation process allows classification of individual pixels with respect to the probability of belonging to one or more classes. As an example, a CNN can provide pixel classification into one of a plurality of classes, where the number of classes can be greater than or equal to two. Regarding a two-class approach, consider a flame class and a smoke class. Regarding a three-class approach, consider a flame class, a smoke class, and a water screen class, or, for example, a flame class, a smoke class, and a background class. Regarding a four-class approach, consider a flame class, a smoke class, a water screen class, and a background class. As an example, a background class can be for objects that do not belong to one of the flame, smoke, or water screen classes.

[096] As explained, a burner controller can be computerized when the set of circuits (e.g., hardware and / or software) operates to characterize a burner operation or burner operations (e.g., for control of multiple burners). With regard to some examples of characteristics, consider flame presence, smoke, and water screen indicators directly derived from the presence of a pixel blur, where pixels have corresponding probabilities that are greater than a given threshold. Although individual pixels are mentioned, an approach may consider neighboring pixels and / or regional pixels. For example, a flame pixel may, improbably, exist as a single pixel that does not have a or Petition 870240067970, dated 09 / 08 / 2024, p. 46 / 91 35 / 68 more neighboring flame pixels. As an example, a trained ML model can output one or more probability values ​​for individual pixels when a post-output process seeks to improve segmentation (e.g., mask smoothing, neighborhood analysis, contiguity analysis, etc.). Such post-output processing can be based, at least in part, on the laws of physics, such as, for example, the laws of flame physics, the laws of smoke physics, etc. When post-output processing is implemented, a burner controller can be considered a hybrid controller that uses non-physics-based machine learning through image analysis to generate probabilities for pixels relative to one or more classes and that uses physics-based analysis to output masks (e.g., filters, etc.).As an example, whether not based on the laws of physics or a hybrid approach, one could output data structures that can be applied to image data acquired via camera. For example, consider applying a mask to an RGB image frame so that a flame in the RGB image frame can be segmented for combustion analysis purposes (e.g., via hue analysis, etc.). As another example, consider applying a mask to an RGB image frame so that smoke in the RGB image frame can be segmented for smoke analysis purposes (e.g., via a Ringelmann smoke graph approach, etc.). Yet another example, consider applying a mask to an RGB image frame so that a water screen can be segmented for water screen analysis purposes. As a further example, consider applying a... Petition 870240067970, dated 09 / 08 / 2024, page 47 / 91 36 / 68 masking in an RGB image frame so that non-flame, non-smoke, and non-water screen pixels can be segmented. This approach can be used to reduce computational demands regarding additional processing, as the number of pixels in the RGB image frame to be processed can be reduced by ignoring non-object pixels (e.g., non-object image data).

[097] As an example, one method might involve processing an image by means of mask generation, where individual masks have corresponding classes, so that the image can be segmented into regions, each region being smaller than the entire image. In such an approach, individual masks can be applied to reduce computational demands for particular purposes. For example, a flame-based combustion analysis might proceed with fewer pixels than in an image where a flame mask is applied to the image to identify, at least in part, probabilistically, which pixels in the image correspond to a flame (e.g., a flame class). Similarly, a smoke-based analysis might proceed with fewer pixels than in an image where a smoke mask is applied to the image to identify, at least in part, probabilistically, which pixels in the image correspond to smoke (e.g., a smoke class).

[098] With regard to size detection, or size determination, a method may detect the size of a flame and smoke by the number of pixels that are likely to be larger than a defined threshold for the flame and / or smoke, since such Petition 870240067970, dated 09 / 08 / 2024, p. 48 / 91 37 out of 68 objects are detected.

[099] As an example, a method for deriving a physical size of a flame might be based on a trigonometric calculation to derive a pixel's extent in meters (e.g., or feet) by knowing a distance between a camera and the flame and, for example, when appropriate, a camera's shooting angle (e.g., a geometric axis of the lens compared to a geometric axis of the flame). As an example, a method might be based on identifying an object of known dimensions in the vicinity of a flame in an image (e.g., consider a burner head, etc.) where such a method can deduce the area covered by the individual pixels at a depth of field.

[100] As an example, a burner may include one or more fiducials that can be used as references to determine size, angle, etc. For example, consider a burner component or components that are within a camera's field of view when at least some of the fiducials can be captured in an image. In such an approach, the fiducials may be emitters (e.g., LEDs, etc.), reflectors, etc.

[101] As an example, a trained ML model might include a fiducial class that can identify one or more fiducials in a pixelated image or images. For example, consider a two-camera approach where an image from a first camera captures two fiducials and an image from a second camera captures two fiducials. In such an approach, a trained ML model could output probabilities of an individual pixel being a fiducial (by Petition 870240067970, dated 09 / 08 / 2024, page 49 / 91 38 / 68 example, a member of a fiducial class). Such fiducial pixels may have corresponding coordinates, where the coordinates can be combined to reconstruct one or more multidimensional reference systems. In such an example, a first reference system may provide physical size determination of objects in the image from the first camera, and a second reference system may provide physical size determination of objects in the image from the second camera. Given the known data regarding the fiducials (e.g., spacing, size, orientation, etc.), the images can be compared and / or a three-dimensional model constructed.

[102] As mentioned, a trained ML model can be a multi-class model where multiple examples can include one or more of a flame class, a smoke class, a water screen class, a background class, and a fiducial class. As explained, a fiducial class can facilitate analysis, for example, with respect to size. For example, smoke size can be expressed as a ratio of flame size, which can be an actual flame size.

[103] As an example, one method may, once flame and smoke segments are detected, use temporal data from a sequence of images (e.g., video) to strengthen the presence accuracy of multiple objects. In such an approach, one or more computer vision techniques may be applied to derive flame quality and smoke classification (e.g., into multiple smoke categories).

[104] As mentioned in relation to Figure 5, RGB color model images can be transformed into Petition 870240067970, dated 09 / 08 / 2024, pp. 50 / 91 39 / 68 is an HSV (Hue, Saturation, Value) color model format in which data structures relating to Hue and Value can be used.

[105] Given the visible flame movement in time during real-time video, as an example, optical flow can be computed to find moving objects in a frame with a high concentration of orange and red in the Hue value color space (see, for example, Figure 6). Such data can be combined with a machine learning approach for flame detection that can enable the reduction of false detection and improve characterization accuracy.

[106] With regard to flame quality characterization, a flame mask can be applied to the Hue component of a flame in an image. In such an example, a method can use the results to derive a histogram (of Hue values) of the pixels that belong to the flame. The histogram can be further processed, for example, using a specialized filter with specifically designed coefficients, where applying the filter to histogram data can be used to derive a flame quality indicator, for example, according to a dominant flame color.

[107] Referring again to Figure 6, for example, consider a method that defines three colors, red, orange, and yellow, and two mixed color zones as red / orange and orange / yellow in the Hue map (see, for example, color map 610). With respect to the HSV color model, Saturation data can be used to divide data along the geometric axis. Petition 870240067970, dated 09 / 08 / 2024, pp. 51 / 91 40 / 68 Saturation to capture one or more bright flame zones. One method might include, for example, counting flame split pixels in each cell as a basis for color determination. In such an approach, when more than half of the pixels are within a box, the corresponding color (or colors) of that box can be considered the dominant flame color; otherwise, a final color can be determined as the average Hue value across the flame pixels.

[108] With regard to a smoke colorimetry indicator, consider flame characterization that uses a smoke mask to delineate boundaries of smoke detected in individual frames (e.g., individually acquired images).

[109] As mentioned, smoke classification can utilize various categories, such as categories in the Ringelmann smoke chart index. Again, the first level of the Ringelmann scale corresponds to white smoke, the 2nd and 3rd levels of the Ringelmann scale are for gray smoke, and the 4th level for the black smoke category. The index is derived from the average of the Value (V) values ​​of the image pixels that belong to the smoke mask inferred by a trained machine learning model (e.g., a trained CNN, etc.).

[110] Figure 8 shows an example of a method 800 that includes an acquisition block 810 for acquiring one or more images, an output block 814 for outputting image frame data (e.g., pixel image data in a color model space, such as RGB, etc.), and a processing block 818 for processing the image frame data using one or more machine learning models. Petition 870240067970, dated 09 / 08 / 2024, p. 52 / 91 41 / 68 trained machines where process block 818 emits a flame mask 822, a smoke mask 824, and a sprinkler mask 826. As mentioned, a mask can correspond to a class. Although the example in Figure 8 shows three masks corresponding to three separate classes (e.g., object classes), a method may include emitting a background mask, a fiducial mask, and / or one or more other mask types.

[111] As shown in the example in Figure 8, each of the masks 822, 824 and 826 can be applied to pixels, which can be pixels from the image frame data, to make determinations regarding whether a pixel belongs to a class or not, as indicated in blocks 823, 825 and 827. This approach can be called segmentation insofar as the image frame data is segmented into classes (e.g., pixel by pixel, etc.).

[112] As shown in the example in Figure 8, method 800 can include a generation block 830 to generate HSV picture frame data using the picture frame data from output block 814. In such an example, generation block 830 can perform a transformation that transforms picture frame data from a first color model space into an HSV color model space. As explained, hue and saturation data can be used for archiving and further classification regarding combustion while value data can be used to assess smoke.

[113] In the example in Figure 8, method 800 may include a generation block 840 to generate a hue histogram where an identification block 845 may provide Petition 870240067970, dated 09 / 08 / 2024, pp. 53 / 91 42 / 68 identification of one or more colors, as associated with combustion that generates a flame as present in at least a portion of the image frame data. As explained, such an approach can provide combustion quality determination. As shown in Figure 8, the method can include a generation block 860 to generate a combustion quality indicator through color identification by the identification block 845, as applied to a hue histogram generated by the generation block 840. For example, generation block 860 can generate a combustion quality indicator that represents flame color; note that overall combustion quality can be determined by a combination of factors, which may depend on smoke characterization and flame characterization (e.g., operations using at least one flame mask and operations using at least one smoke mask, etc.).

[114] As explained, the value (V) of the HSV color model can be used to characterize smoke. As shown in the example in Figure 8, method 800 can include a generation block 850 to generate a smoke index, which can include, for example, white smoke 852, gray smoke 854 (e.g., one or more gray smoke indicators) and black smoke 856, for example, as explained with regard to the Ringelmann smoke chart (see, for example, Figure 7).

[115] In the example in Figure 8, method 800 includes several outputs, as represented by blocks 823, 825, 827, 852, 854, 856, and 860. Such output may characterize a burner operation, and such characterization may be used for burner control purposes. For example, Petition 870240067970, dated 09 / 08 / 2024, pp. 54 / 91 43 / 68 Consider using a combustion quality indicator from generation block 860 as a basis for adjusting one or more fluid flow rates to a burner in an effort to improve combustion quality. As another example, consider using one or more smoke indices from generation block 850 as a basis for adjusting one or more fluid flow rates to a burner in an effort to control what type of smoke, amount of smoke, etc., is generated during an ongoing burner operation.

[116] As explained in relation to Figure 3, a burner can be oriented by means of one or more mechanisms. For example, when an environmental condition impacts one or more aspects of a burner operation, a burner can be adjusted to improve a burner operation. For example, consider wind direction when adjusting the direction of a burner (e.g., by means of a burner rod, etc.) to improve combustion quality, etc.

[117] Figure 9 shows an example of a 900 machine learning model (ML 900 model). In the example in Figure 9, the ML 900 model includes a U-shaped architecture and can be termed a U-network architecture. In the example in Figure 9, several numbers are given to illustrate a modality; note that such numbers may depend on one or more factors, such as, for example, image frame data matrix size, etc. Specifically, in the example in Figure 9, an input image (e.g., image frame data) is shown as being 256 x 256 where operations are performed for 16 channels. As shown, the number of channels increases in Petition 870240067970, dated 09 / 08 / 2024, pp. 55 / 91 44 / 68 left leg of the U-shaped architecture and decreases in size on the right leg of the U-shaped architecture.

[118] The ML 900 model includes a U-shaped network architecture in which maximum grouping operations are performed in transitions to four layers, starting from a top layer, to reach a base layer of the U-shaped network architecture and in which multidimensional 2x2 transpose convolution operations are performed in transitions to the four layers, starting from the base layer, to return to the top layer.

[119] In the ML 900 model example, the first four layers involve concatenation through the legs of the U-shaped architecture. In each of the five layers, for both legs of the U-shaped architecture as well as for the base layer, multidimensional 3x3 convolutions are performed. Additionally, to emit three classes (e.g., three masks, etc.), the ML 900 model includes a 1x1 convolution. In the ML 900 model, hop connections act to improve accuracy compared to models such as a fully convolutional neural network (FCN) without hop connections.

[120] In the ML 900 model example, for a 256 x 256 RGB image (e.g., 256 x 256 x 3), the ML 900 model may be a thin U-network model with approximately 1.9 million weights. Training may have used pixel arrays identified as “real region” for the purpose of determining appropriate values ​​for the weights.

[121] In the exemplary ML 900 model, the output of the expansion path can be an array that is, for a 256 x 256 input image, a 256 x 256 array Petition 870240067970, dated 09 / 08 / 2024, pp. 56 / 91 45 / 68 x N (for example, a matrix), where N represents the number of classes, which may, for example, include one or more from a flame class, a smoke class, a water screen class (e.g., sprinkler), a background class, a fiducial class, etc. In such an approach, the array may include probability values ​​for each element (e.g., pixel) that may be in the range, for example, from zero (not a member of a class) to one (a member of a class). Such an array may be processed to generate binary masks, for example, consider using one or more thresholds to transform elements above a threshold into one (1) and elements below the threshold into zero (0).A plurality of binary masks can be applied to an input image to segment the input image into classes when a rule can be applied so that each pixel in the input image is assigned to an appropriate class through the application of the binary masks (for example, including a background class, which may include pixels that are not segmented into one or more other classes).

[122] As an example, a binary mask can be a probabilistic mask, for example, with zeros (0) to probabilistically indicate “not a flame”, “not smoke”, etc., and with ones (1) to probabilistically indicate “a flame”, “smoke”, etc.

[123] As an example, a pixelated image may include a color model bit intensity and may include an additional depth value, which may, for example, be used to encode a pixel as belonging, probabilistically, to a particular class. For example, Petition 870240067970, dated 09 / 08 / 2024, p. 57 / 91 46 / 68 When four classes exist, numbers such as 0, 1, 2, and 3 can be used in the additional depth value. As an example, a method, such as method 800 in Figure 8, can utilize the additional depth value for processing purposes related to flame, smoke, etc. As an example, when an array includes class values ​​directly, the use of such a class value can be considered by applying a corresponding mask. For example, the flame mask 822, the smoke mask 824, and the sprinkler mask 826 of method 800 can be a single two-dimensional array of values.

[124] As an example, consider a U-network architecture, as described in a paper by Ronneberger et al., “U-Net: Convolutional Networks for Biomedical Image Segmentation,” arXiv:1505.04597 (2015), which is incorporated by reference in this document. Such an architecture includes a shrinking path (see the left side of the ML 900 model) and an expanding path (see the right side of the ML 900 model). The Ronneberger et al. paper describes a shrinking path that follows a convolutional network architecture (e.g., a CNN) and includes repeated application of convolutions and use of a rectified linear unit (ReLU) along with a maximum clustering operation with a specified progression for decreasing sampling, where, in an expanding path, increasing sampling of a feature map can be performed followed by convolution.

[125] Specifically, Ronneberger et al. describe a network architecture that includes a contraction path (left side) and an expansion path (right side). Petition 870240067970, dated 09 / 08 / 2024, pp. 58 / 91 47 / 68 right), where the contraction path follows a type of convolutional network architecture and includes repeated applications of two 3x3 convolutions (unfilled convolutions), each followed by a rectified linear unit (ReLU) and a 2x2 maximum grouping operation with progress 2 for downsampling. And at each downsampling step, the number of feature channels is doubled. In Ronneberger et al., each step in the expansion path includes an upsampling of the feature map followed by a 2x2 convolution (upward convolution) that halves the number of feature channels, a concatenation with the correspondingly clipped feature map from the contraction path, and two 3x3 convolutions, each followed by a ReLU. The clipping handles edge pixel loss in each convolution. In the final layer, a 1x1 convolution is used to map each 64-component feature vector into the desired number of classes.In total, the network of Ronneberger et al. includes 23 convolutional layers.

[126] Regarding training, consider an approach by Ronneberger et al., where input images and their corresponding segmentation maps are used to train the network with the stochastic gradient source implementation of the CAFFE framework. Due to unfilled convolutions, the output image is smaller than the input image by a constant edge width. To minimize overhead and maximize GPU memory usage, Ronneberger et al. favored large input blocks over a large batch size and then reduced the batch to a single image where a high boost (0.99) was used so that a high number of previously viewed training samples were used. Petition 870240067970, dated 09 / 08 / 2024, pp. 59 / 91 48 / 68 determined the update in a current optimization step.

[127] The exemplary ML model 900 in Figure 9 can be used for mask generation, for example, to generate one or more masks from image data that can segment the image data, which can be in a raw format, a transformed format, a processed format, etc.

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

[129] As an example, a training method may include several actions that can operate on a dataset to train an ML model. For example, a dataset may be divided into training data and test data, where the test data may provide evaluation. A method may include parameter cross-validation and best parameters, which may be provided for Petition 870240067970, dated 09 / 08 / 2024, pages 60 / 91 49 / 68 model training.

[130] The TENSORFLOW framework can 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.) based operating system platforms.

[131] TENSORFLOW computations can be expressed as stateful data flow graphs; note that the name TENSORFLOW derives from the operations that such neural networks perform on multidimensional data arrays. Such arrays can be called tensors.

[132] Figure 10 shows a 1000 architecture of a framework, such as the TENSORFLOW framework. As shown, the 1000 architecture includes several features. For example, in the terminology of the 1000 architecture, a client can define a computation as a data flow graph and, for example, can initiate graph execution using a session. For example, a distributed master can prune a specific subgraph of the graph, as defined by the arguments to Session.run(); divide the subgraph into multiple chunks that are executed in different processes and devices; distribute the graph chunks across worker services; and initiate graph chunk execution by worker services. Regarding worker services (e.g., one per task), for example, Petition 870240067970, dated 09 / 08 / 2024, pages 61 / 91 50 / 68 can schedule the execution of graphics operations using kernel implementations appropriate to the available hardware (CPUs, GPUs, etc.) and, for example, send and receive operation results to and from other worker services. Regarding kernel implementations, these can, for example, perform computations for individual graphics operations.

[133] Figure 10 also shows some examples of machine learning model types 1010, 1020, 1030 and 1040, one or more of which can be used. As explained, a model-based ML approach can include receiving image data which can be spatial image data, which can be 1D, 2D or 3D spatial image data. As an example, time can be a dimension so that image data can be both spatial and temporal. As an example, a convolutional neural network and / or one or more other types of neural networks can be used for spatial and / or spatiotemporal image analysis.

[134] As an example, an ML model can approach video (spatiotemporal image data) as sequences of short clips (e.g., frame N) (e.g., consider that N is greater than one and less than approximately 100). In such an approach, various aspects of changes with respect to time can be related to one or more types of physical phenomena associated with hydrocarbon combustion, which, once analyzed, can be used for burner control. In such an example, frame rate (e.g., frames per second) and number of frames in a sequence (e.g., a series) can cover a desired amount of time, which can be for behavior tracking. Petition 870240067970, dated 09 / 08 / 2024, pages 62 / 91 51 / 68 dynamic of one or more burner operations. For example, consider 8 frames at a frame rate of 24 frames per second, which can be a time period of approximately one-third of a second. As an example, analyses can be performed using different sequences, using a different number of frames, using different frame rates, using different processing techniques in relation to the raw frame data, using different frame resolutions, etc. Such variables can be adapted to provide processing efficiency, which can help provide real-time control (e.g., near real-time control with latency of approximately half an hour or less for 2D spatial and 1D temporal processing).

[135] With respect to 3D, defined by 2D in space and 1D in time, a separable convolution approach can be used or a more standard 3D convolution. A separable 3D convolution approach can factor a standard 3D convolution into a 3D convolution in channels and a 3D convolution in points with a 1x1x1 kernel. A convolution in channels can apply a specific filter to each input channel, then the convolution in points can combine the outputs of the convolution in channels through a 1x1x1 convolution. Such factorization can reduce computation and model size.

[136] In Figure 10, the ML models include a 2D convolution model 1010, a 3D convolution model 1020, a 3D spatial and temporal convolution model 1030, and a 3D channel convolution model 1040. In Figure 10, the 1030 model may be an R2plus1D model that factors the 3D convolution into spatial and temporal convolutions, and the Petition 870240067970, dated 09 / 08 / 2024, pp. 63 / 91 The 52 / 68 model 1040 may be a separable 3D convolution that includes two convolution processes: a channel convolution and a point convolution.

[137] As an example, a spatiotemporal architecture may include several features, such as one or more features of an architecture that includes an encoder (feature extractor), a pyramid clustering module, and a decoder to gradually cover spatiotemporal details. In such an example, the encoder may receive pixel values ​​as input, extract features layer by layer, and generate rich contextual features as output. In such an example, intermediate results may be combined with a decoder module. As an example, a pyramid clustering module may connect an encoder module and a decoder module. Pyramid clustering may vary a receptive field size, for example, by modifying the spatial progression.With respect to a separable 3D convolution, a block of frame-level features may involve reduced average grouping and duplication to generate frame-level features (e.g., with the same dimension as the other branches, etc.). A paper by Hou et al., titled “An efficient 3D CNN for action / object segmentation in video”, arXiv:1907.08895v1 (July 21, 2019) is incorporated by reference in this document.

[138] Figure 11 shows an example of a method 1100 that includes an operation block 1101 for operating a burner system 1110 that includes one or more cameras 1112-1 and 1112-2 that can acquire image data from various objects, which may include one or more of a flame 1114, Petition 870240067970, dated 09 / 08 / 2024, pp. 64 / 91 53 / 68 smoke 1115, a water screen 1116 and fiducials 1117. As shown, method 1100 includes an image block 1102 for imaging during burner system operation 1110 by operation block 1101, wherein image block 1102 can generate image data (e.g., image frame data) 1120-1 and 1120-2 for one or more time periods. As shown, method 1100 includes a mask block 1103 to generate one or more masks 1122-1, 1122-2 to 1122N and / or 1132-1, 1132-2 to 1132-M using the image data generated from image block 1102. In such an example, N and M may be different or the same. For example, a first camera's field of view might capture fiducials while a second camera's field of view might not; therefore, a fiducial mask might be for the first camera, which could result in N being greater than M.

[139] As explained, mask generation can be performed using a trained machine learning model (a trained ML model), which can segment image data into defined classes, which, for example, can correspond to one or more object types (e.g., flame, smoke, water screen, fiducial, etc.) and / or antecedent (e.g., a class that does not include multiple defined objects). As explained, a mask can be specific to a particular physical phenomenon, such as combustion or smoke. As shown, an analysis block 1104 can provide one or more analysis types 11421, 1142-2 to 1142-P. As an example, P can be equal to N and / or M or it can be different. As explained, a separate type of analysis can be applied to mask results (e.g., segmentation results). For example, a method Petition 870240067970, dated 09 / 08 / 2024, pages 65 / 91 54 / 68 applying a flame mask to generate flame results may include analyzing flame results for combustion quality (e.g., by means of hue and saturation of flame image pixels, etc.); while applying a smoke mask to generate smoke results may include analyzing smoke results for one or more smoke characteristics, such as a Ringelmann smoke graph index value (e.g., by means of the (V) value of an HSV color model for smoke image pixels, etc.). As explained, a fiducial mask can be applied to generate fiducial results, which can be for spatial analysis, e.g., to determine the size of a flame, smoke size, etc. As mentioned, fiducial analysis can provide image correlation, e.g., to compare features (e.g., objects) in two or more images.

[140] As shown in the example in Figure 11, method 1100 may include a control block 1105 to control the operation of burner system 1110, wherein one or more controls 1152 may correspond to one or more operating parameters, such as one or more fluid flow rates, one or more valve settings, one or more nozzle settings, one or more rod directions, etc. Although the example in Figure 11 shows a single burner system (burner system 1110), method 1100 may be used to characterize and / or control more than one burner system.

[141] As an example, method 1100 can be used to monitor burner operation through start-up, shutdown and continuous firing operations, and can be Petition 870240067970, dated 09 / 08 / 2024, pages 66 / 91 55 / 68 replicated through behavior cloning. For example, when a model is trained and tested, and generates low errors in predicting air control, the model can be installed in a control circuit and used to control a burner system. As an example, a trained ML model can apply tolerances to various inputs; note that certain signatures in image data or sensor data may indicate weak or deteriorated combustion. When one or more problems arise, a controller can perform corrective action, such as increasing or decreasing airflow, fuel flow, or air-to-fuel ratio.

[142] Figure 12 shows an example of a method 1200 and an example of a system 1290. As shown, method 1200 includes an acquisition block 1210 to acquire image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; a process block 1220 for processing the image data using at least one machine learning model trained to generate a flame mask and a smoke mask; a characterization block 1230 to characterize the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and a control block 1240 to, based on the characterization, control the operation of the burner system.

[143] Method 1200 is shown including several blocks of computer-readable storage media (CRM) 1211, 1221, 1231 and 1241 which may include processor-executable instructions that may instruct a computing system, which may be a control system, to perform Petition 870240067970, dated 09 / 08 / 2024, pages 67 / 91 56 / 68 one or more of the actions described in relation to method 1200.

[144] In the example in Figure 12, the system 1290 includes one or more information storage devices 1291, one or more computers 1292, one or more networks 1295, and instructions 1296. With respect to the one or more computers 1292, each computer may include one or more processors (e.g., or processing cores) 1293 and memory 1294 for storing instructions 1296, for example, executable by at least one of the one or more processors 1293 (see, for example, blocks 1211, 1221, 1231, and 1241). 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.

[145] As an example, method 1200 can be a workflow that can be implemented using one or more frameworks that may be within a framework environment. As an example, system 1290 can include local and / or remote resources. For example, consider a browser application running on a client device as a local resource with respect to a user of the browser application and a cloud-based computing device as a remote resource with respect to the user. In such an example, the user can interact with the client device through the browser application where information is transmitted to the cloud-based computing device(s) and where information can be received in response and rendered to a display operationally coupled to the client device (e.g., via Petition 870240067970, dated 09 / 08 / 2024, pages 68 / 91 57 / 68 services, APIs, etc.).

[146] As an example, a method might include acquiring image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; processing the image data using at least one machine learning model trained to generate a flame mask and a smoke mask; characterizing the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and, based on the characterization, controlling the operation of the burner system. In such an example, the at least one trained machine learning model might include a single trained machine learning model that generates at least the flame mask and the smoke mask.

[147] As an example, a trained machine learning model may be or include a U-network machine learning model architecture.

[148] As an example, a trained machine learning model might include a contraction trajectory that receives image data and an expansion trajectory that emits a flame mask and a smoke mask.

[149] As an example, image data may include pixel data, where a flame mask is applied to the image data to probabilistically identify flame pixels in the image data, and where a smoke mask is applied to the image data to probabilistically identify smoke pixels in the image data.

[150] As an example, one method might include characterizing combustion by applying a flame mask to image data by applying the flame mask. Petition 870240067970, dated 09 / 08 / 2024, pp. 69 / 91 58 / 68 in the image data in a hue, saturation, and value (HSV) color model to generate at least flame-masked image data for hue. In such an example, the method might include generating a combustion quality indicator to characterize combustion using at least flame-masked image data for hue. As an example, a method might include generating a combustion quality indicator to characterize combustion using at least one flame-masked image data for hue and for saturation.

[151] As an example, a method might include characterizing combustion by applying a smoke mask to image data by applying the smoke mask to the image data in a hue, saturation, and value (HSV) color model to generate at least smoke-masked-for-value image data. In such an example, the method might include generating a smoke indicator to characterize combustion using at least the smoke-masked-for-value image data. In such an example, the smoke indicator might be a Ringelmann smoke plot indicator.

[152] As an example, one method might include characterizing combustion by applying a smoke mask to image data by using a Ringelmann smoke plot indicator.

[153] As an example, one method might include processing image data using at least one machine learning model trained to generate a flame mask, a smoke mask, and a sprinkler mask.

[154] As an example, one method might include processing image data using at least one machine learning model trained to generate a mask of Petition 870240067970, dated 09 / 08 / 2024, pp. 70 / 91 59 / 68 flame, a smoke mask, and a fiducial mask. In such an example, the method might include applying the fiducial mask to the image data to identify fiducials and spatially characterizing the flame using the identified fiducials. In such an example, spatially characterizing the flame might include determining a flame size (e.g., in one or more dimensions).

[155] As an example, one method might include characterizing combustion by generating a flame-to-smoke ratio.

[156] As an example, one method might include acquiring real-time video image data during the operation of a burner system and, for example, controlling the operation of the burner system in real time.

[157] As an example, a method might involve processing image data using at least one trained machine learning model, where the image data comprises spatiotemporal image data (see, for example, 1020, 1030, and 1040 models in Figure 10). For example, a series of image frames acquired with respect to a period of time might be received as input and processed in a multidimensional manner with time as a dimension. In such an example, the output might be, for example, time masks which might be “video masks” that are time-dependent. For example, a flame mask might be time-dependent and include a series of frames for individual points in time, a smoke mask might be time-dependent and include a series of frames for individual points in time, etc. As an example, temporal processing might provide the trend of an or Petition 870240067970, dated 09 / 08 / 2024, pp. 71 / 91 60 / 68 plus parameters, which may be indicative of flame character, smoke character, combustion character, etc. In such an example, a trend can be identified and used to control a burner system. For example, a trend indicator can provide control with respect to a favorable trend or control with respect to an unfavorable trend. For example, when a series of frames generates a temporal smoke mask that indicates an unfavorable smoke trend over time associated with the series of frames, the trend may be related to flow rates, composition, environmental conditions, etc., for that time period, which can provide dynamic characterization over individual time periods. Such an approach can characterize control dynamics, which include, for example, lags that may exist between the emission of a control signal to adjust an operation and changes in flame imaging, smoke, etc.

[158] As an example, a system might include a processor; processor-accessible memory; processor-executable instructions stored in memory and executable by the processor to instruct the system to: acquire image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; process the image data using at least one machine learning model trained to generate a flame mask and a smoke mask; characterize the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and, based on the combustion characterization, control Petition 870240067970, dated 09 / 08 / 2024, pp. 72 / 91 61 / 68 burner system operation.

[159] As an example, one or more computer-readable storage media may include executable computer instructions to instruct a computing system to: acquire image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; process the image data using at least one machine learning model trained to generate a flame mask and a smoke mask; characterize the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and, based on the combustion characterization, control the operation of the burner system.

[160] As an example, a method may be implemented, in part, using computer-readable media (CRM), for example, as a module, a block, etc., which includes information such as 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. As an example, a single medium may be configured with instructions to enable, at least in part, the performance of several actions of a method. As an example, a computer-readable medium (CRM) may be a computer-readable storage medium (e.g., a non-transient medium) that is not a carrier wave.

[161] According to one embodiment, one or more computer-readable media may include computer-executable instructions for instructing a system of Petition 870240067970, dated 09 / 08 / 2024, pp. 73 / 91 62 / 68 computing to output information to control a process. For example, such instructions can provide output for a detection process, an injection process, a drilling process, an extraction process, an extrusion process, a pumping process, a heating process, a burning process, an analysis process, etc.

[162] In some embodiments, a method or methods may be executed by a computing system. Figure 13 shows an example of a 1300 system that may include one or more 1301-1, 1301-2, 1301-3 and 13014 computing systems, which may be operationally coupled by means of one or more 1309 networks, which may include wired and / or wireless networks.

[163] As an example, a system may include an individual computer system or an array of distributed computer systems. In the example in Figure 13, computer system 1301-1 may include one or more modules 1302, which may be or include processor-executable instructions, for example, executables to perform various tasks (e.g., receive information, request information, process information, simulation, output information, etc.).

[164] As an example, a module may run independently of, or in conjunction with, one or more processors 1304, which are operationally coupled to one or more storage media 1306 (for example, wired, wirelessly, etc.). As an example, one or more of the processors 1304 may be operationally coupled to at least one of one or more network interfaces 1307. In such an example, the system of Petition 870240067970, dated 09 / 08 / 2024, pp. 74 / 91 63 / 68 computer 1301-1 can transmit and / or receive information, for example, through one or more networks 1309 (for example, consider one or more of the Internet, a private network, a cellular network, a satellite network, etc.).

[165] As an example, computer system 1301-1 may receive from and / or transmit information to one or more other devices, which may be or include, for example, one or more of computer systems 1301-2, etc. A device may be located at a physical location that differs from that of computer system 1301-1. As an example, a location may be, for example, a processing facility location, a data center location (e.g., server tower, etc.), a platform location, a well site location, a well interior location, etc.

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

[167] As an example, storage media 1306 can be implemented as one or more computer-readable or machine-readable storage media. As an example, storage can be distributed within and / or across multiple internal and / or external enclosures of a computing system and / or additional computing systems.

[168] As an example, a storage medium or storage media may include one or more different forms of memory that include semiconductor memory devices, such as random access memories. Petition 870240067970, dated 09 / 08 / 2024, pp. 75 / 91 64 / 68 dynamic or static (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 disks, floppy disks and removable disks, other magnetic media including tape, optical media, such as compact discs (CDs) or digital video discs (DVDs), Blu-ray discs or other types of optical storage, or other types of storage devices.

[169] As an example, a storage medium or media may be located on a machine that executes machine-readable instructions, or located at a remote site from which machine-readable instructions can be downloaded over a network for execution.

[170] 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 hardware and software (for example, including firmware), which includes one or more specific integrated circuits for signal processing and / or application.

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

[172] Figure 14 shows components of a computing system 1400 and a network system 1410. The 1400 system includes one or more processors 1402, memory and / or storage components 1404, one or more input and / or output devices 1406, and a bus 1408. According to a Petition 870240067970, dated 09 / 08 / 2024, pp. 76 / 91 In embodiment 65 / 68, instructions can be stored on one or more computer-readable media (e.g., memory / storage components 1404). Such instructions can be read by one or more processors (e.g., the processor(s) 1402) via a communication bus (e.g., the bus 1408), which can be wired or wireless. The one or more processors can execute such instructions to implement (completely or partially) one or more attributes (e.g., as part of a method). A user can view the output of and interact with a process via an I / O device (e.g., the device 1406). According to one embodiment, a computer-readable medium can be a storage component, such as a physical memory storage device, e.g., a chip, a chip in a package, a memory card, etc.

[173] According to one embodiment, components may be distributed, as in the 1410 network system. The 1410 network system includes 1422-1, 1422-2, 14223, . . . 1422-N components. For example, the 1422-1 component may include the 1402 processor (or processors) while the 1422-3 component (or components) may include memory accessible by the 1402 processor (or processors). Furthermore, the 1422-2 component (or components) may include an I / O device for display and optionally interaction with a method. The network may be or include the Internet, an intranet, a cellular network, a satellite network, etc.

[174] As an example, a device can be a mobile device that includes one or more network interfaces for information communication. For example, a device Petition 870240067970, dated 09 / 08 / 2024, pp. 77 / 91 A 66 / 68 mobile device may include a wireless network interface (e.g., operable via IEEE 802.11, ETSI GSM, BLUETOOTH, satellite, etc.). For example, a mobile device may include components such as a main processor, memory, a display, a set of graphic display circuits (e.g., optionally including a set of touch-sensitive and gesture-sensitive circuits), a SIM card, an audio / video circuit, a set of motion processing circuits (e.g., accelerometer, gyroscope), a wireless LAN circuit, a smart card circuit, a transmitter circuit, a GPS circuit, and a battery. For example, a mobile device may be configured as a cell phone, a tablet computer, etc. For example, a method may be implemented (e.g., completely or in part) using a mobile device.For example, a system may include one or more mobile devices.

[175] As an example, a system can 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 communicating information via one or more of the Internet (for example, 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 (for example, wholly or partly as a cloud-based service). Petition 870240067970, dated 09 / 08 / 2024, pp. 78 / 91 67 / 68

[176] As an example, information can be output from a display (e.g., consider a touch screen), output to a display, or both. As an example, information can be output to a projector, a laser device, a printer, etc., so that the information can be visualized. As an example, information can be output stereographically or holographically. With regard to a printer, consider a 2D or 3D printer. As an example, a 3D printer can incorporate one or more substances that can be output to construct a 3D object. For example, data can be fed to a 3D printer to construct a 3D representation of an underground formation. As an example, layers can be constructed in 3D (e.g., horizons, etc.), geobodies constructed in 3D, etc. As an example, holes, fractures, etc., can be constructed in 3D (e.g., as positive structures, as negative structures, etc.).

[177] Although only a few examples have been described in detail above, those skilled in the art will readily verify that various modifications are possible in the examples. Consequently, all such modifications are intended to be included within the scope of this disclosure, as defined in the following claims. In the claims, the device-plus-function clauses are intended to cover the structures described herein insofar as they perform the aforementioned function and not only structural equivalents, but also equivalent structures. Thus, although a nail and a screw may not be structural equivalents because a nail Petition 870240067970, dated 09 / 08 / 2024, pp. 79 / 91 68 / 68 employs a cylindrical surface to fasten wooden parts, while a screw employs a helical surface; in the context of fastening wooden parts, a nail and a screw can be considered equivalent structures. It is the applicant's clear intention not to invoke Title 35 of USC § 112, paragraph 6, for any limitation of any claims herein, except for those where the claim expressly uses the words "means to" along with an associated function. Petition 870240067970, dated 09 / 08 / 2024, pp. 80 / 91

Claims

1 / 5 CLAIMS 1. A method for burner control, characterized in that it comprises: acquiring image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; processing the image data using at least one machine learning model trained to generate a flame mask and a smoke mask; characterizing the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and based on the characterization, controlling the operation of the burner system, wherein the at least one trained machine learning model comprises a single trained machine learning model that generates at least the flame mask and the smoke mask.

2. A method according to claim 1, characterized in that at least one trained machine learning model comprises a U-Net machine learning model architecture.

3. A method according to claim 1, characterized in that at least one trained machine learning model comprises a contraction trajectory that receives the image data and an expansion trajectory that emits the flame mask and the smoke mask.

4. Method according to claim 1, Petition 870260073813, dated 24 / 07 / 2026, p. 10 / 19 2 / 5 characterized in that the image data comprises pixel data, wherein the flame mask is applied to the image data to probabilistically identify flame pixels in the image data, and wherein the smoke mask is applied to the image data to probabilistically identify smoke pixels in the image data.

5. Method, according to claim 1, characterized in that characterizing combustion by applying flame masking to image data comprises applying flame masking to image data in a hue, saturation, and value (HSV) color model to generate at least flame masked image data for hue.

6. Method, according to claim 5, characterized in that it comprises generating a combustion quality indicator to characterize combustion using at least flame mask-for-hue image data.

7. Method, according to claim 5, characterized in that it comprises generating a combustion quality indicator to characterize combustion using at least flame mask image data for hue and saturation.

8. Method, according to claim 1, characterized in that characterizing combustion by applying smoke masking to image data comprises applying smoke masking to image data in a hue, saturation, and value (HSV) color model to generate at least hue-masked smoke masked image data.

9. Method according to claim 8, Petition 870260073813, dated 07 / 24 / 2026, page 11 / 19 3 / 5 characterized in that it comprises generating a smoke indicator to characterize combustion using at least smoke mask-for-value image data.

10. Method according to claim 9, characterized in that the smoke indicator comprises a Ringelmann smoke chart indicator.

11. Method according to claim 1, characterized in that characterizing combustion by applying the smoke mask to image data comprises the use of a Ringelmann smoke chart indicator.

12. Method, according to claim 1, characterized in that processing image data using at least one trained machine learning model generates a flame mask, a smoke mask, and a sprinkler mask.

13. Method, according to claim 1, characterized in that processing image data using at least one trained machine learning model generates a flame mask, a smoke mask, and a fiducial mask.

14. Method, according to claim 13, characterized in that it comprises applying the fiducial mask to image data to identify fiducials and spatially characterizing the flame using the identified fiducials.

15. Method according to claim 14, characterized in that spatially characterizing the flame comprises determining a flame size.

16. Method according to claim 1, Petition 870260073813, dated 07 / 24 / 2026, page 12 / 19 4 / 5 characterized in that characterizing the combustion comprises generating a flame-to-smoke ratio.

17. Method, according to claim 1, characterized in that the acquisition acquires real-time video image data during the operation of the burner system and in that the control controls the operation of the burner system in real time.

18. Burner control system, characterized in that it comprises: a processor; processor-accessible memory; processor-executable instructions stored in memory and executable by the processor to instruct the system to: acquire image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; process the image data using at least one trained machine learning model to generate a flame mask and a smoke mask; characterize the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and based on the combustion characterization, control the operation of the burner system, wherein the at least one trained machine learning model comprises a single trained machine learning model that generates at least the flame mask and the smoke mask.

19. One or more computer-readable storage media for burner control, characterized in that they comprise executable computer instructions to instruct a computing system to: acquire image data during the operation of a burner system that, via hydrocarbon combustion, generates a flame and smoke; process the image data using at least one trained machine learning model to generate a flame mask and a smoke mask; characterize the combustion by applying the flame mask to the image data and applying the smoke mask to the image data; and based on the combustion characterization, control the operation of the burner system, wherein the at least one trained machine learning model comprises a unique trained machine learning model that generates at least the flame mask and the smoke mask.Petition 870260073813, dated 07 / 24 / 2026, pp. 14 / 19.

Images