Artificial intelligence (AI) / machine learning (ML)-based fiber optic health management

Abstract

Aspects of the disclosure provide techniques and apparatus for performing predictive health management for a fiber optic monitoring system. An example technique includes obtaining an indication of one or more parameters of a fiber optic monitoring system for one or more well completions. The fiber optic monitoring system includes at least one monitoring device and one or more fiber optic cables coupled to the at least one monitoring device. At least one failure mode of the fiber optic monitoring system is predicted based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models. An indication of the at least one failure mode is provided.

Description

ARTIFICIAL INTELLIGENCE (AI)ZMACHINE LEARNING (ML)-BASED FIBER OPTIC HEALTH MANAGEMENTCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] The present application claims the benefit of United States Provisional Application No.63 / 674632 filed July 23, 2024, the entirety of which is incorporated by reference herein and should be considered part of this specification.BACKGROUNDField of the Disclosure

[0002] The present disclosure relates to fiber optic monitoring systems. More specifically, the present disclosure provides techniques and apparatus for performing predictive health management for a fiber optic monitoring system.Description of Related Art

[0003] Fiber optic-based sensing systems are used in variety of fields, such as oil and gas exploration, formation evaluation, production monitoring, and well integrity evaluations, among other fields. Some wells used in the production of hydrocarbons, for example, may include fiber optic systems to provide downhole measurement of one or more parameters, such as temperature, pressure, and fluid flowrate, as illustrative examples. A fiber optic-based sensing system may include optical connections between one or more fiber optic lines and other equipment, such as an optical monitoring device. The optical monitoring device can monitor various parameters (e g., temperature, strain, pressure, vibration, etc.) of the hardware environment. Various drilling, production, and remedial operations may be performed based on information derived from the monitored parameters.

[0004] In some cases, fiber optic-based sensing systems may be susceptible to conditions that adversely impact the performance of the fiber optic-based sensing system. In such cases, operators (or end users) may lose access to vital information regarding the monitored environment (e.g., oilwell, gas well, geothermal well, and injection well, among others), creating challenges for drilling, production, and remedial operations, as illustrative examples.SUMMARY

[0005] One embodiment of the present disclosure described herein is a method. The method generally includes obtaining an indication of one or more parameters of a fiber optic monitoring system for one or more well completions. The fiber optic monitoring system includes at least one monitoring device and one or more fiber optic cables coupled to the at least one monitoring device. The method also includes predicting at least one failure mode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models. The method further includes providing an indication of the at least one failure mode.

[0006] Another embodiment of the present disclosure described herein is a fiber optic monitoring system. The fiber optic monitoring system includes one or more fiber optic cables deployed within one or more well completions, and a computing system coupled to the one or more fiber optic cables. The computing system includes one or more memories collectively storing instructions, and one or more processors coupled to the one or more memories. The one or more processors are collectively configured to execute the instructions to cause the computing system to perform an operation. The operation includes obtaining an indication of one or more parameters of the fiber optic monitoring system for the one or more well completions. The operation also includes predicting at least one failure mode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models. The operation further includes providing an indication of the at least one failure mode.

[0007] Another embodiment of the present disclosure described herein is a non-transitory computer-readable medium. The non-transitory computer-readable medium includes computerexecutable instructions that, when executed by one or more processors of a computing system, cause the computing system to perform an operation. The operation includes obtaining an indication of one or more parameters of a fiber optic monitoring system for one or more well completions. The fiber optic monitoring system includes at least one monitoring device and one or more fiber optic cables coupled to the at least one monitoring device. The operation alsoincludes predicting at least one failure mode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models. The operation further includes providing an indication of the at least one failure mode.

[0008] Another embodiment of the present disclosure described herein is a method. The method generally includes obtaining one or more datasets, each comprising a first set of parameters associated with one or more conditions of one or more fiber optic monitoring systems. The method also includes training a plurality of machine learning (ML) models for predicting failure of a fiber optic monitoring system, each ML model of the plurality of ML models being trained on a respective dataset of the one or more datasets to predict a respective failure mode of a plurality of failure modes associated with the fiber optic monitoring system. The method further includes storing the plurality of ML models.

[0009] Another embodiment of the present disclosure described herein is a computing system. The computing system includes one or more memories collectively storing instructions, and one or more processors coupled to the one or more memories. The one or more processors are collectively configured to execute the instructions to cause the computing system to perform an operation. The operation includes obtaining one or more datasets, each comprising a first set of parameters associated with one or more conditions of one or more fiber optic monitoring systems. The operation also includes training a plurality of machine learning (ML) models for predicting failure of a fiber optic monitoring system, each ML model of the plurality of ML models being trained on a respective dataset of the one or more datasets to predict a respective failure mode of a plurality of failure modes associated with the fiber optic monitoring system. The operation further includes storing the plurality of ML models.

[0010] Another embodiment of the present disclosure described herein is a non-transitory computer-readable medium. The non-transitory computer-readable medium includes computerexecutable instructions that, when executed by one or more processors of a computing system, cause the computing system to perform an operation. The operation includes obtaining one or more datasets, each comprising a first set of parameters associated with one or more conditions of one or more fiber optic monitoring systems. The operation also includes training a plurality of machine learning (ML) models for predicting failure of a fiber optic monitoring system, each ML model ofthe plurality of ML models being trained on a respective dataset of the one or more datasets to predict a respective failure mode of a plurality of failure modes associated with the fiber optic monitoring system. The operation further includes storing the plurality of ML models.

[0011] The following description and the appended figures set forth certain features for purposes of illustration.BRIEF DESCRIPTION OF DRAWINGS

[0012] Various embodiments in accordance with the present disclosure will be described with reference to the drawings, where like designations denote like elements. Note that the appended drawings illustrate typical embodiments and are therefore not to be considered limiting; other equally effective embodiments are contemplated.

[0013] FIG. 1 is a schematic diagram of at least a portion of an example implementation of a system for performing predictive health management of a fiber optic monitoring system, according to various embodiments.

[0014] FIG. 2 depicts an example fiber optic monitoring system, according to various embodiments.

[0015] FIG. 3 depicts another example fiber optic monitoring system, according to various embodiments.

[0016] FIG. 4 further illustrates a monitoring device and predictive health management component of the system illustrated in FIG. 1, according to various embodiments.

[0017] FIG. 5 depicts an example workflow for training an AI / ML model to predict a failure mode of a fiber optic monitoring system, according to various embodiments.

[0018] FIG. 6 depicts different examples of health indicator transforms, according to various embodiments.

[0019] FIG. 7 depicts an example configuration of a laser module, according to various embodiments.

[0020] FIG. 8 illustrates an example optical signal obtained via a monitoring device, according to various embodiments.

[0021] FIG. 9 depicts a graph illustrating noise floor of a monitoring device as a function of fiber one-way attenuation, according to various embodiments.

[0022] FIG. 10 depicts a graph illustrating an example of reversible attenuation and irreversible attenuation in a fiber as a function of wavelength, according to various embodiments.

[0023] FIG. 11 depicts a graph illustrating attenuation growth in a fiber over time as a function of wavelength, according to various embodiments.

[0024] FIG. 12 depicts an example workflow for predicting a failure mode of a fiber optic monitoring system using an AI / ML model, according to various embodiments.

[0025] FIG. 13A depicts an example signal transform unit for an optical switch, according to various embodiments.

[0026] FIG. 13B depicts an example time series segmentation of the signal transform unit illustrated in FIG. 13A, according to various embodiments.

[0027] FIG. 14 is a flow diagram depicting an example operations for training one or more AI / ML models to perform predictive health management for a fiber optic monitoring system, according to various embodiments.

[0028] FIG. 15 is a flow diagram depicting an example operations for performing predictive health management for a fiber optic monitoring system, according to various embodiments.

[0029] FIG. 16 depicts an example computing device, according to various embodiments.DETAILED DESCRIPTION

[0030] The disclosure provides techniques, methods, systems, apparatus, and computer readable media for performing predictive health management for a fiber optic monitoring system.

[0031] For example, a predictive health management system according to one or moreembodiments described herein can monitor one or more operating parameters of a fiber optic monitoring system associated with one or more hardware installations, such as a land well and subsea well, as illustrative, non-limiting examples. Based on the monitoring, the predictive health management system may employ artificial intelligence (AI) / machine learning (ML) techniques to predict a future operating condition and / or parameter(s) of the fiber optic monitoring system. For example, the predictive health management system may predict whether and when the fiber optic monitoring system will be impacted by one or more failure modes, such as an optical switch failure within a monitoring device of the fiber optic monitoring system, laser failure, optical fiber loss, irreversible optical fiber darkening, and internal fan failure, as illustrative, non-limiting examples.

[0032] The predictive health management system may provide information associated with the prediction to an operator (or end user) responsible for managing the hardware installation. For example, such information may include an indication of the predicted operating conditions, a predicted lifetime until occurrence of the one or more failure modes, a suggested preventative action, or a combination thereof.

[0033] The techniques, methods, systems, apparatus, and computer readable media for performing predictive health management for a fiber optic monitoring system may provide various advantages. For example, the predictive health management system described herein may allow for early detection of conditions within the fiber optic monitoring system that may lead to potential failure of the fiber optic monitoring system. Providing early detection of such conditions may enable the operator to perform preventative maintenance and / or assess remaining lifetime of the system, as illustrative examples.

[0034] The following description includes embodiments of the best mode presently contemplated for practicing the described implementations. This description is not to be taken in a limiting sense, but rather is made merely for the purpose of describing the general principles of the implementations. The scope of the described implementations should be ascertained with reference to the issued claims.

[0035] Although the terms “first,” “second,” “third,” etc., may be used herein to describe various elements, components, regions, layers and / or sections, these elements, components, regions, layers and / or sections should not be limited by these terms. These terms may be only usedto distinguish one element, component, region, layer or section from another region, layer, or section. Terms such as “first,” “second,” and other numerical terms, when used herein, do not imply a sequence or order unless clearly indicated by the context. Thus, a first element, component, region, layer, or section discussed herein could be termed a second element, component, region, layer, or section without departing from the teachings of the example embodiments.

[0036] As used herein, a hyphenated form of a reference numeral refers to a specific instance of an element and the un-hyphenated form of the reference numeral refers to the collective element. Thus, for example, device “12-1” refers to an instance of a device class, which may be referred to collectively as devices “12” and any one of which may be referred to generically as a device “12”.Example System for Performing Predictive Health Management of a Hardw are Installation

[0037] FIG. l is a schematic diagram of at least a portion of an example implementation of a a system 100 that can be configured to perform predictive health management of a fiber optic monitoring system 150, according to various embodiments. As shown, the system 100 includes a predictive health management component 110 and the fiber optic monitoring system 150. The fiber optic monitoring system 150 includes a monitoring device 120, one or more fiber optic cables 116- 1 to 116-2, and one or more well completions 130-1 to 130-N.

[0038] Each well completion 130 is generally representative of a hardware installation that may be monitored and / or managed via the monitoring device 120. To reach a region of interest 124 (e.g., reservoir) for each well completion 130, a respective wellbore 126 is drilled through the surface 114 and the casing 118 is lowered into the wellbore 126. A production casing (or production tubing) 122 may be lowered into the wellbore 126 and installed within the casing 118.

[0039] Each fiber optic cable 116 is deployed within a respective wellbore 126. In some cases, the fiber optic cable 116 may be deployed within the production casing 122. In some cases, the fiber optic cable 116 may be deployed within the annulus between the production casing 122 and the casing 118. In some cases, at least a portion of the fiber optic cable 116 may be deployed within the production casing 122 and at least another portion of the fiber optic cable 16 may be deployed within the annulus between the production casing 122 and the casing 118.

[0040] The fiber optic cable 116 can be (or otherwise include) any of a variety of types ofoptical fibers. For example, the fiber optic cable 1 16 may be a hybrid cable including electrical conductor(s) and optical fibers, or may include optical fibers without electrical conductor(s). Additionally, the fiber optic cable 116 can be a multimode fiber or single mode fiber. In some cases, the fiber optic cable 116 can be configured so that it has enhanced sensitivity to vibration, such as by incorporating coatings or geometries that enhance the transfer of the incident vibration to the fiber optic core. Additionally, in some cases, the fiber optic cable 116 can be adapted by various means to increase backscatter. Additionally, the fiber optic cable 116 may have a dualended configuration (as depicted in FIG. 1) or a single-ended configuration.

[0041] The monitoring device 120 is generally representative of a variety of optical interrogation devices, including, for example, distributed temperature sensing (DTS) interrogators, distributed strain sensing (DSS) interrogators, distributed acoustic sensing (DAS) interrogators, optical time-domain reflectometers (OTDR), fiber optic pressure and temperature (P / T) gauge interrogators, and similar devices. The monitoring device 120 may be implemented using hardware, software, or combinations thereof. The monitoring device 120 may be located at the surface 114. In certain embodiments, at least some functionality (e.g., lasers, optical switches, receivers, etc.) of the monitoring device 120 is located at the surface 114 and other functionality (e.g., processing systems) of the monitoring device 120 is located elsewhere (e.g., another computing system coupled to the monitoring device 120, a cloud computing environment, etc.).

[0042] The monitoring device 120 is coupled to the fiber optic cables 116 and may use the fiber optic cables 116 to observe physical parameters associated with a region of interest 124. For example, the monitoring device 120 can transmit interrogating pulses to the fiber optic cables 116 and detect backscattered optical signals returned by the fiber optic cables 116. The detected backscatter can be processed by the monitoring device 120 to determine various parameters of the region of interest 124, such as vibration, noise, strain, temperature, and pressure, as illustrative, non-limiting examples. The measurements obtained via the monitoring device 120 and fiber optic cables 116 may provide information that may be used to manage the well completions 130. For example, the measurements may provide an indication of the characteristics of a production fluid, such as flow velocity, flow composition, and inflow location, or an indication of well integrity, such as leaks in the casing 118 and / or production casing 122.

[0043] In certain embodiments, the predictive health management component 110 is configured to monitor the health (e.g., operating condition) of the fiber optic monitoring system 150 (including one or more components thereof). In some cases, the predictive health management component 110 is communicatively coupled to the monitoring device 120 and may monitor one or more operating parameters of the monitoring device 120, optical signals transmitted from the monitoring device 120 via the fiber optic cables 116, optical signals reflected towards the monitoring device 120 via the fiberoptic cables 116, information derived from the reflected optical signals, or any combination thereof. As described in greater detail herein, the predictive health management component 110 may use the monitored information as input to one or more AI / ML models to predict one or more future conditions of the fiber optic monitoring system 150. Such future conditions, for example, may include one or more future operating parameters of the fiber optic monitoring system 150, one or more future failure modes of the fiber optic monitoring system 150, or a combination thereof. The predictive health management component 110 may provide information associated with the future conditions of the fiber optic monitoring system 150 to an operator (or end user) (e.g., via a computing system associated with the operator).

[0044] Note that while FIG. 1 depicts each well completion 130 as a cased vertical well structure, the techniques described herein are not limited to such configurations. For example, the techniques described herein can be used for uncased, open hole, gravel packed, deviated, horizontal, multi-lateral, deep sea or terrestrial surface injection and / or production wells, among others.

[0045] By way of example, FIG. 2 depicts a fiber optic monitoring system 250 that includes a dry -tree system 230, according to various embodiments. The dry -tree system 230 is an illustrative example of a hardware installation that can be monitored using the predictive health management component 110. The dry-tree system 230 may be installed on land or off-shore with a dry tree installed on platform. The dry-tree system 230 may include various hardware, such as well head outlet (WHO) 202, dry mate splice 204, downhole wet mate 206, fiber optic cable 208, and cable end 210, among other components.

[0046] By way of another example, FIG. 3 depicts a fiber optic monitoring system 350 that includes a sub-sea installation system 330, according to various embodiments. The sub-seainstallation system 330 is another illustrative example of a hardware installation that can be monitored using the predictive health management component 110. The sub-sea installation system 330 may include various hardware, such as surface cable 310, one or more junction boxes 314, fiber optic cable 308, and wet mate system 306, among other components.

[0047] FIG. 4 further illustrates the monitoring device 120 and predictive health management component 110 of the system 100 illustrated in FIG. 1, according to various embodiments. As shown, the monitoring device 120 includes, without limitation, an optical source 402 (e.g., a laser), a modulator 404, a directional coupler 406, a detector 408, a signal acquisition unit 410, and a processing system 414. The optical source 402 is configured to generate an optical signal (e.g., an optical pulse or series of pulses) to launch into the cable 116, which includes an optical fiber 452. The optical source 402 may be a narrowband laser that is followed by a modulator 404 that selects short pulses from the output of the laser. The optical signal or pulses generated by the optical source 402 are sent into the optical fiber cable 116 through the directional coupler 406. The directional coupler 406 is configured to separate outgoing and returning optical signals and direct the returning (backscattered) signals to the detector (e.g., optical receiver) 408. The directional coupler 406 may be implemented with a beam splitter, a fiber optic coupler, a circulator, or some other optical device.

[0048] The backscattered optical signals generated by the cable 116 in response to the interrogating optical signal may be detected and converted to an electrical signal at the detector 408. This electrical signal may be acquired by a signal acquisition unit 410 and then transferred as data representing the backscattered signals to the processing system 414. The processing system 414 may include various algorithms to process the data to determine characteristics of the data.

[0049] In certain embodiments, the processing system 414 sends equipment health condition data 440 (hereinafter “data 440”) to the predictive health management component 110. The data 440 may include multiple data subsets, each associated with a respective component of the fiber optic monitoring system 150. The predictive health management component 110 may use AI / ML techniques to predict one or more future conditions associated with the fiber optic monitoring system 150, based on the data 440.

[0050] As shown, the predictive health management component 110 includes an analysis tool 420 and an output tool 424, each of which may include hardware, software, or combinations thereof. The analysis tool 420 may use the data 440 (including one or more data subsets thereof) as input to one or more AI / ML models 422 (hereinafter, “models 422”), each of which is configured to predict whether and when the fiber optic monitoring system will encounter a respective failure mode, such as optical switch failure, laser failure, optical fiber darkening, an internal fan failure, as illustrative, non-limiting examples.

[0051] The analysis tool 420 may provide the output from the models 422 to the output tool 424, which provides an indication of prognostic data 450 to an operator (e.g., via a computing system associated with the operator). In this manner, the operator can determine future estimated conditions of one or more components of the fiber optic monitoring system, determine an estimated lifetime of one or more components of the fiber optic monitoring system, perform one or more preventative actions to improve operation of the fiber optic monitoring system, or any combination thereof. For example, the operator can monitor current and predicted health of the fiber optic monitoring system 150, based on the prognostic data 450. For instance, the operator can determine whether current operating conditions of the fiber optic monitoring system 150 are within predefined operating ranges associated with a target performance of the fiber optic monitoring system 150, determine whether predicted operating conditions of the fiber optic monitoring will be within predefined operating ranges associated with the target performance of the fiber optic monitoring system 150, or a combination thereof.

[0052] Note, the equipment health condition data 440 may include information associated with a variety of different components of a fiber optic monitoring system (e.g., fiber optic monitoring system 150, fiber optic monitoring system 250, fiber optic monitoring system 350, etc.). In certain embodiments, for example, the data 440 includes one or more operating parameters of the monitoring device 120, such as internal temperature, current draw, voltage draw, laser settings (e.g., output power), attenuator settings (e.g., output power), central processing unit (CPU) usage, and disk usage, as illustrative, non-limiting examples. Such operating parameters may be indicative of the health (e.g., operating conditions) of the monitoring device 120. Accordingly, in such embodiments, the predictive health management component 110 can use AI / ML techniques to predict, based on the operating parameters of the monitoring device 120, whether and when themonitoring device 120 will encounter a failure mode.

[0053] Additionally or alternatively, in certain embodiments, the data 440 includes information associated with optical signals reflected via the cable 116. Such information, for example, may include OTDR data, data associated with the Rayleigh band, Brillouin or Raman bands (including Stokes and Anti-Stokes bands), or any combination thereof. In such embodiments, the information may be indicative one or more fault conditions associated with one or more components of the fiber optic monitoring system 150. Such fault conditions may include, without limitation: (i) point losses, pinches, and / or bends in a surface cable (e.g., surface cable 310), (ii) splice losses, pinches, and / or bends in a junction box (e.g., junction box 314), (iii) splice losses, feedthrough losses, pinches, and / or bends in a WHO (e.g., WHO 202), (iv) fiber loss due to micro / macro-bending, fiber loss from chemically driven attenuation), fiber loss due to a stuck fiber and strain, pinches, bends, and / or excess strain associated with a cable (e.g., a permanent downhole cable, such as cable 116, cable 208, cable 306), (v) splice losses, pinches, and / or bends in a dry mate splice (e.g., dry mate splice 204), (vi) splice loss, connector loss, and / or connector reflection associated with a downhole wet mate (e.g., downhole wet mate 206, wet mate system 306, etc.), (vii) end reflection associated with a cable end (e.g., cable end 210), or (viii) any combination thereof.

[0054] In certain embodiments, the aforementioned information (or any combination thereof) is used to train one or more AI / ML models (e.g., models 422) to predict one or more failure modes of the fiber optic monitoring system 150. In certain embodiments, a respective AI / ML model is trained for each different failure mode of the fiber optic monitoring system 150 using a respective data subset corresponding to the failure mode. For example, a “first” model 422-1 may be trained to predict an optical switch failure, a “second” model 422-2 may be trained to predict laser failure (e.g., failure of optical source 402), a “third” model 422-3 may be trained to predict fiber loss, and a “fourth” model 422-4 may be trained to predict optical fiber darkening, among other models 422.

[0055] FIG. 5 depicts an example workflow 500 for training an AI / ML model (e.g., model 422) to predict a failure mode of a fiber optic monitoring system 150, according to various embodiments. The workflow 500 may be performed by a training component 510, which may include hardware, software, or combinations thereof. The training component 510 may beexecuted by one or more computing systems.

[0056] As shown, the training component 510 may obtain equipment health condition data 540 (hereinafter referred to as “data 540”) associated with one or more fiber optic monitoring systems (e g., fiber optic monitoring system 150, fiber optic monitoring system 250, fiber optic monitoring system 350, etc.). For example, the data 540 may include operating parameters of multiple monitoring devices (of the same type and / or different types), information associated with optical signals reflected from different cables of multiple wells, or any combination thereof. Additionally, the data 540 may be obtained from simulations, measurements in the field, or combinations thereof.

[0057] As shown, the training component 510 may include, without limitation, a labeled data generation tool 522, a feature extraction tool 524, a health indicator transform generation tool 526, and a supervised learning tool 528. In scenarios where the equipment health condition data 540 includes unlabeled data, the labeled data generation tool 522 may be used to assign labels to the “data 540” in order to generate a labeled dataset. The labeled dataset may then be provided to the feature extraction tool 524.

[0058] In certain embodiments, the feature extraction tool 524 extracts one or more features from the labeled dataset (e.g., a raw dataset). In an illustrative example, the feature extraction may involve identifying specific characteristics from the labeled dataset. For example, the feature extraction may involve determining segments of the labeled dataset that are indicative of “healthy” states, segments of the labeled dataset that are indicative of “faulty” states, or a combination thereof. In some cases, the feature extraction may further involve extracting correlated characteristics of the segments indicative of “healthy” states and correlated characteristics of the segments indicative of “faulty” states. The correlation may be based on metrics, such as standard deviation, peak-to-peak amplitude, and geometric mean, among others.

[0059] In certain embodiments, the health indicator transform generation tool 526 is configured to generate, from the extracted features, one or more health indicator transforms for at least one fiber optic monitoring system. Each health indicator transform is generally a representation of the health of the fiber optic monitoring system over time. In some cases, a health indicator transform may include a time series set of parameter values associated with the fiberoptic monitoring system, such as temperature values, strain values, pressure values, and noise values, as illustrative examples.

[0060] In certain embodiments, the supervised learning tool 528 implements a supervised learning algorithm with inputs that include at least one of the data 540, information output from the labeled data generation tool 522, information output from the feature extraction tool 524, information output from the health indicator transform generation tool 526, or any combination thereof. The supervised learning tool 528 can utilize any suitable supervised learning algorithm, such as random forest classifiers, naive bayes classifier algorithm, decision trees, nearest neighors, and logistic regression, as illustrative, non-limiting examples. The supervised learning tool 528 may output a trained AI / ML model 422 configured to predict a respective failure mode of the fiber optic monitoring system.

[0061] By way of example, in certain embodiments, the training component 510 trains a “first” model 422-1 to predict an optical switch failure. As noted, one potential fault condition of a fiber optic monitoring system 150 may involve failure of an optical switch (e.g., optical switch 412) within a monitoring device (e.g., monitoring device 120).

[0062] For example, in normal operation, the monitoring device may launch pulses of laser light into an optical fiber (e.g., optical fiber 452), and a proportion of that light may backscattered along the optical fiber. Parameters (e.g., temperature) along the optical fiber may be calculated via the ratio of backscattered power in the stokes and anti-stokes Raman bands. Similar to the mode of operation of an OTDR, the time of flight of the light may determine the distance that a particular section of fiber or component is from the monitoring device. The optical switch may control which optical fiber the backscattered light is being measured from.

[0063] When there is an optical switch failure, this failure may lead to corruption of a calculated trace obtained by the monitoring device due to TTS (anti-stokes band) data, NTS (stokes band) data, or both being read from the wrong connected fiber or switched part way through the acquisition of the trace data. In such a fault condition, TTS and NTS data may be measured from different fibers, resulting in corrupted data.

[0064] Optical switch failures can be partial and can grow progressively worse with time.From the end user’s perspective, the corruption may not be noticeable at early stages. As the corruption becomes progressively worse, decisions based on data obtained via the fiber optic monitoring system may become invalid, resulting in non-productive time and / or inefficient asset use.

[0065] In certain embodiments, the training component 510 may be configured to train a model 422 (e.g. model 422-1) to detect optical switch failures, based on temperature measurements from different fibers in the data 540. In some cases, the temperature measurements may be associated with fibers that have a double ended configuration. As noted, in a double ended configuration, both ends of the fiber under measurement are connected to the monitoring device and, at the furthest extent, the fiber is looped back upon itself. In a double ended configuration, pulses may be launched from each end of the fiber and backscatter may be measured from each end of the fiber. One potential advantage of using fibers with a double ended configuration is that data obtained via the fibers can be processed to correct for fiber and component attenuation, providing a more accurate temperature measurement.

[0066] In certain embodiments, a respective health indicator transform may be generated based on the temperature measurements (e.g., a respective data subset included within data 540) for each double ended configuration fiber. For example, the temperature measurements for a double ended configuration fiber may have symmetry under normal operation, and may lack symmetry in the event of an optical switch malfunction or failure. In such cases, generating a given health indicator transform may involve obtaining two vectors of temperature profile data (e.g., one vector for each end of the fiber), and calculating the ratio of the two vectors, yielding a single ratio data vector.

[0067] The standard deviation of the single ratio data vector may then be calculated and used as the health indicator transform for optical switches. For example, datasets for “healthy” optical switches may have symmetry (e.g., matching rolling standard ratios, resulting in smaller magnitude health indicators). On the other hand, datasets for “unhealthy” optical switches may have nonmatching rolling standard ratios, resulting in larger magnitude health indicators. As such, the failure of an optical switch may be predicted by monitoring health indicator transforms for optical switches over time.

[0068] In certain embodiments, model training for optical switch failures may performed bygenerating data (e.g., a respective data subset of data 540) that simulates a malfunctioning optical switch. The generated labeled data may then be processed to determine the health indicator transform as a time series and input to a supervised learning algorithm, which outputs a trained AI / ML model 422-1.

[0069] Note, in certain embodiments, the temperature measurements used to train the AI / ML model 422-1 may be associated with fibers that have a single ended configuration. In such embodiments, the generation of the health indicator transform may involve calculating the rolling ratio of consecutive temperature profiles, as opposed to calculating the rolling ratio between two ends of the same fiber. FIG. 6 depicts graphs 600-1 to 600-2 illustrating health indicator transforms for different single ended configuration fibers and graphs 600-3 to 6004 illustrating health indicator transforms for different double ended configuration fibers, according to various embodiments. In FIG. 6, graph 600-1 depicts two consecutive temperature measurement profiles 610, 620 of a cable 116-1 (e.g., fiber) in a fiber optic monitoring system with a “healthy” optical switch, graph 600-3 depicts temperature measurement profiles 630, 640 from two ends of the cable 116-1 in the fiber optic monitoring system with a “healthy” optical switch, graph 600-2 depicts two consecutive temperature measurement profiles 650, 660 of a cable 116-2 (e.g., fiber) in a fiber optic monitoring system with a “faulty” optical switch, and graph 600-4 depicts temperature measurement profiles 670, 680 from two ends of the cable 116-2 in the fiber optic monitoring system with a “faulty” optical switch.

[0070] Note, in some scenarios, the optical switch (e.g., optical switch 412) may not transition from a healthy state to a faulty state, but may begin in a faulty state. In order to detect such scenarios, the model 422-1 may be trained on data (e.g., field data and / or synthetically created data) where the anti-stokes / stokes information within the data is more unstable from one measurement to the next measurement in cases of single ended configurations due to, e.g., the natural noise of single ended detection in combination with a switch fault.

[0071] In certain embodiments, the training component 510 trains a “second” model 422-2 to predict laser failure. As noted, another potential fault condition of a fiber optic monitoring system 150 may involve failure of an optical source (e.g., optical source 402) within a monitoring device (e.g., monitoring device 120).

[0072] In certain embodiments, laser output power is an illustrative example of a metric that is indicative of the performance of the monitoring device, and in turn, the fiber optic monitoring system 150. Consider FIG. 7, which depicts an example configuration of a laser module 700, according to various embodiments. Note, the laser module 700 is an illustrative example of a configuration that may include or otherwise implement the optical source 402 illustrated in FIG. 4. As shown, the laser module 700 includes, without limitation, a laser drive chip 710, a pulsed laser 720, a gain medium (e.g., fiber or cable) 730, and a pump laser 740.

[0073] To generate a pulse, the pump laser current may be set to provide a sufficient gain in order to achieve a desired output power. The pulsed laser 720 and the pump laser 740 may be temperature stabilized. Because the gain of the gain medium 730 is temperature sensitive, the pump laser current may be adjusted relative to its default setting to compensate for the variation in the gain. The adjustment may be controlled by a processor (e g., processing system 414 and / or processor in the laser module 700 (not shown)). For example, the processor may use a thermistor (not shown) in the laser module 700 to determine the fiber temperature and use a preprogrammed coefficient to determine the target pump laser current.

[0074] In certain scenarios, such as over long periods of time, the laser output can degrade. Some potential causes of the degradation in the laser output include changes in efficiency of the lasers, the gain of the fiber, and the drive chip malfunctioning (e g., the laser driver chip being unable to drive the laser when triggered), as illustrative, non-limiting examples. In certain embodiments, the training component 510 may be configured to train a model 422 (e.g., model 422-2) to detect laser failure, based on output power measurements from different lasers in the data 540. By way of example, the output power from each respective laser can be monitored for long-term changes in behavior to determine when the laser power is no longer sufficient to achieve a target performance of the laser module 700 and / or fiber optic monitoring system 150.

[0075] By way of example, FIG. 8 depicts a graph 800 illustrating an example raw NTS (stokes band) signal, according to various embodiments. The raw NTS signal may be indicative of the laser output power for a given monitoring device. For example, the signal level in the receiver coil portion of the raw NTS signal may be indicative of the laser output power. This respective portion of the raw NTS signal for respective monitoring devices may be monitored over a periodof time and used to generate a dataset for training the model 422-1 . Based on this information, the model 422-2 can be trained to output a forecast for when the laser power of a given monitoring device will no longer be sufficient to achieve a target performance of the fiber optic monitoring system 150.

[0076] In certain embodiments, the training component 510 trains a “third” model 422-3 to predict fiber and component loss. As noted, another potential fault condition of a fiber optic monitoring system 150 may involve fiber loss, which may be caused by splice losses, connector losses, pinches, bends, and excess strain in the cable, among other causes. In some cases, fiber loss may be reflected in OTDR traces. OTDR traces are often employed to identify and categorize local events and / or distributed events that impact a fiber optic monitoring system. A local event may include a splice or a connector, and a distributed event may include fiber loss due to strain, poor cabling, or chemically induced attenuation. In certain embodiments, the monitoring device 120 includes OTDR functionality that enables the monitoring device 120 to perform an OTDR measurement, for example, to acquire an OTDR trace of an installed fiber optic cable, such as cable 116.

[0077] In such embodiments, the training component 510 may be configured to train a model 422 (e.g., model 422-3) to detect fiber and component losses, based on OTDR traces of multiple fiber optic cables. For example, a training dataset may be generated based on periodic monitoring of OTDR traces and based on processing resulting local and distributed loss events. In certain embodiments, the model 422-3 is further trained to predict the impact on the fiber loss on the fiber optic monitoring system 150. For example, a fiber optic monitoring system may have a defined loss budget under which data quality is assured. If the fiber optic monitoring system exceeds that loss budget, then data quality may reduce, e.g., in terms of temperature resolution (DTS, production monitoring, flow profiling) or minimum detectable events (DAS, micro-seismic, integrity monitoring). By way of example, FIG. 9 depicts a graph 900 illustrating noise floor of a monitoring device (e.g., monitoring device 120) as a function of fiber one-way attenuation, according to various embodiments.

[0078] In certain embodiments, the training component 510 trains a “fourth” model 422-4 to predict optical fiber darkening. One aspect of fiber loss that has a significant impact on assetoperations (e.g., oil and gas completions) is that of chemically induced attenuation. For example, in cable and fiber structures, reversible and irreversible attenuation may be produced by diffusion and interaction of molecular hydrogen with silica optical fiber. By way of example, FIG. 10 depicts a graph 1000 illustrating an example of reversible attenuation and irreversible attenuation in a fiber as a function of wavelength, according to various embodiments. As shown in FIG 10, peak reversible attenuation (e.g., due to presence of molecular hydrogen in an optical fiber) may occur at wavelengths (X) of approximately 1244 nanometers (nm) and irreversible attenuation (e g., hydroxide production due to hydrogen reacting with the silica in the fiber) may occur over a broad band of wavelengths but centered approximately 1380-1390 nm. Thus hydrogen induced darkening can affect the operation of DTS, DSS, DAS and other interrogation systems with laser sources in the 1000 to 1600nm region.

[0079] Over time, there may be an increase in the attenuation. By way of example, FIG. 11 depicts a graph 1100 illustrating attenuation growth in a fiber over time as a function of wavelength, according to various embodiments. The rate of growth may be caused by factors, such as temperature, hydrogen partial pressure, optical fiber type, optical cable construction, well hydrogen sulfide content, completion materials, well operations (e.g., injection, production, flow rates), produced / injected fluid composition, annulus fluid composition and pressure, and cable encapsulation, as illustrative, non-limiting examples.

[0080] In certain embodiments, a training dataset with one or more of the aforementioned factors for different fibers may be generated and used to train a model (e.g., model 422-4) to predict optical fiber darkening. The training dataset, for example, may indicate, for each fiber, the nature of loss growth along the fiber, and the monotonic loss growth with time. In certain embodiments, the trained model 422-4 may be used to recognize and classify fiber loss growth, perform regression analysis for end of life estimation, or a combination thereof.

[0081] In certain embodiments, information including one or more of the aforementioned parameters may be collected from multiple fibers across multiple well completions to aid in cable and fiber selection. In this manner, reliability information can be obtained for different installed optical fiber and cable types, which can be used for planning fiber and cable installations.

[0082] In certain embodiments, failure of the fiber optic monitoring system 150 may bedetected through unusual system optimizations. In certain monitoring devices 120, the optimal optical power in the fiber is often determined using an algorithm which varies optical power and measures power wasted through non-linear effects, such as self-phase modulation (SPM). Because SPM is a cumulative process through the fiber, if power levels are too high, a manifestation of the effect is an increase in noise floor toward the end of the fiber being measured. Monitoring devices 120 (e.g., DAS interrogators) may alter power by changing pump power in an amplifier loop or by altering the value of a variable optical attenuator (VOA) on the output of the instrument. Thus, over time, the monitoring device can periodically optimize output power.

[0083] Thus, in certain embodiments, the pump power value and / or VOA value may be included within the equipment health condition data (e.g., data 440, data 540, or a combination thereof). In certain embodiments, equipment health condition data, including pump power values and / or VOA values, may form time-series data that can be analyzed and tracked through life of an installation, such as a well. AI / ML models, such as a model 422, can be trained on the dataset to classify events, such as step changes or monotonic increases in optimized power. The model(s) 422 can then be used as part of regression analysis to predict the system end-of-life.

[0084] As noted, one or more of the trained AI / ML models 422 may be used to predict one or more future conditions associated with the fiber optic monitoring system 150. By way of example, FIG. 12 depicts an example workflow 1200 for predicting a failure mode of a fiber optic monitoring system 150 using a respective AI / ML model (e.g., model 422) associated with the failure mode, according to various embodiments. The workflow 1200 may be performed by the predictive health management component 110 (including one or more components thereof). The predictive health management component 110 (including one or more components thereof) may be executed by one or more computing systems.

[0085] As shown, the analysis tool 420 includes a signal transform unit tool 1210, a time series segmentation tool 1220, a feature extraction tool 1230, and one or more models 422. The signal transform unit tool 1210 is generally configured to perform preprocessing of the data 540 (e.g., raw data) and to output a signal transform unit, which is generally a processed version of the data 540. The preprocessing may include various techniques, such as feature extraction, normalization, and / or data reduction, as illustrative, non-limiting examples. By way of example, FIG. 13A depictsan example signal transform unit 1310 for an optical switch, according to various embodiments. Here, the signal transform unit 1310 may be single ratio data vector over time that is generated by obtaining two vectors of temperature profile data (e.g., one vector for each end of the fiber), and calculating the ratio of the two vectors.

[0086] Referring back to FIG. 12, the signal transform unit (e.g., signal transform unit 1310) output from the signal transform unit tool 1210 may be provided to the time series segmentation tool 1220. In certain embodiments, the time series segmentation tool 1220 separates the values of the signal transform unit into time segments with homogenous behavior. By way of example, FIG. 13B depicts an example time series segmentation 1320 of the signal transform unit 1310 into segments Si to Sn. The time series segmentation tool 1220 may perform the segmentation using an autoregressive model or a change point detection algorithm, such as pruned exact linear timer (PELT), as an illustrative example.

[0087] Referring back to FIG. 12, the time series segmentation (e.g., time series segmentation 1320) output from the time series segmentation tool 1220 may be provided to the feature extraction tool 1230. In certain embodiments, the feature extraction tool 1230 extracts one or more features from each segment of the time series segmentation. For example, for each segment, the feature extraction tool 1230 may extract the characteristics (or features) most correlated with the label corresponding to the level of corruption. Examples of such characteristics (or features) can include, but are not limited to standard deviation, peak-to-peak amplitude, geometric mean, among others. The extracted features may be provided as input to the trained model(s) 422.

[0088] In certain embodiments, the trained model(s) 422 may output a set of prognostic data 450, based on the extracted features. As noted, in some embodiments, the model 422 may perform classification (e.g., the model may be a classifier, such as a random forest classifier). In the case of optical switch failure detection, for example, the model 422 may be a random forest classifier (or other suitable classifier model) that outputs an indication of whether the optical switch is in a faulty state or healthy state based on the data 540. In some embodiments, the model 422 may perform regression (e.g., the model 422 may be a regression model). In such embodiments, the model 422 may output an estimate of when the fiber optic monitoring system (including components thereof, such as an optical switch) will be in a faulty state.

[0089] Advantageously, the techniques presented herein may be used to predict operating conditions of a fiber optic monitoring system, allowing for early detection of conditions within the fiber optic monitoring system that may lead to potential failure of the fiber optic monitoring system. Such information including, for example, the identification or classification of the condition as well as the location of the condition may be communicated to the end user, allowing the end user to perform preventative maintenance. In some cases, the information may also include an indication of the estimated time remaining for the end user to use the fiber optic monitoring system before the fiber optic monitoring system will encounter a failure mode.Example Operations

[0090] FIG. 14 is a flow diagram depicting an example operations 1400 for training one or more AI / ML models (e.g., model(s) 422) to perform predictive health management for a fiber optic monitoring system (e.g., fiber optic monitoring system 150), according to various embodiments. The operations 1400 may be performed, for example, by a training component (e.g., training component 510). The operations 1400 may be implemented as software components that are executed and run on one or more processors (e.g., central processing unit (CPU) 1605 of computing device 1600).

[0091] The operations 1400 may involve, at block 1402, obtaining one or more datasets (e.g., data 540), each comprising a first set of parameters associated with one or more conditions of one or more fiber optic monitoring systems (e.g., fiber optic monitoring system 150).

[0092] The operations 1400 may involve, at block 1404, training a plurality of ML models (e.g., model(s) 422) for predicting failure of a fiber optic monitoring system, each ML model of the plurality of ML models being trained on a respective dataset of the one or more datasets to predict a respective failure mode of a plurality of failure modes associated with the fiber optic monitoring system.

[0093] The operations 1400 may further involve, at block 1406, storing the plurality of ML models (e.g., in a storage system, such as storage 1660 of computing device 1600).

[0094] In certain embodiments, at least one of the plurality of ML models is used during a runtime of the fiber optic monitoring system to predict an occurrence of the respective failure modeassociated with the fiber optic monitoring system.

[0095] In certain embodiments, the plurality of failure modes include (i) a failure of an optical switch within the fiber optic monitoring system, (ii) an amount of signal loss associated with one or more fiber optic cables of the fiber optic monitoring system being greater than a first threshold, (iii) an output power of a laser source of the fiber optic monitoring system being less than a second threshold, (iv) an attenuation metric associated with the one or more fiber optic cables satisfying a predetermined condition, or (v) any combination thereof.

[0096] In certain embodiments, training the plurality of ML models includes, for at least one ML model of the plurality of ML models: (i) extracting, from the respective dataset associated with the at least one ML model, one or more features associated with the respective failure mode for the at least one ML model; (ii) generating, based on the one or more features, a health indicator for the fiber optic monitoring system; and (iii) using the health indicator to train the at least one ML model. In some cases, the respective dataset includes data obtained via a single-ended configuration of a fiber optic cable of the fiber optic monitoring system. In some cases, the respective dataset includes data obtained via a double-ended configuration of a fiber optic cable of the fiber optic monitoring system.

[0097] FIG. 15 is a flow diagram depicting an example operations 1500 for performing predictive health management for a fiber optic monitoring system (e.g., fiber optic monitoring system 150), according to various embodiments. The operations 1500 may be performed, for example, by a predictive health management component (e.g., predictive health management component 110). The operations 1500 may be implemented as software components that are executed and run on one or more processors (e.g., CPU 1605 of computing device 1600).

[0098] The operations 1500 may involve, at block 1502, obtaining an indication of one or more parameters (e.g., data 440) of a fiber optic monitoring system (e.g., fiber optic monitoring system 150) for one or more well completions (e.g., well completions 130). The fiber optic monitoring system includes at least one monitoring device (e.g., monitoring device 120) and one or more fiber optic cables (e g., cables 116) coupled to the at least one monitoring device.

[0099] The operations 1500 may also involve, at block 1504, predicting at least one failuremode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained ML models (e.g., models 422).

[0100] The operations 1500 may further involve, at block 1506, providing an indication (e g., prognostic data 450) of the at least one failure mode.

[0101] In certain embodiments, the at least one monitoring device includes an optical switch (e g., optical switch 412) configured to select a set of the one or more fiber optic cables for monitoring by the at least one monitoring device. In such embodiments, the at least one failure mode may include a failure of the optical switch.

[0102] In certain embodiments, the at least one monitoring device includes a laser source (e g., optical source 402), and the at least one failure mode includes an output power of the laser source being less than a threshold.

[0103] In certain embodiments, the at least one failure mode includes an attenuation metric associated with at least one of the one or more fiber optic cables satisfying a predetermined condition.

[0104] In certain embodiments, the at least one failure mode includes an amount of signal loss associated with at least one of the one or more fiber optic cables being greater than a threshold.

[0105] In certain embodiments, the at least one failure mode includes a number of power adjustments of the at least one monitoring device being greater than a threshold.

[0106] In certain embodiments, providing the indication (in block 1506) may involve providing an indication of when the at least one failure mode will occur.

[0107] In certain embodiments, predicting the at least one failure mode includes predicting multiple failure modes of the fiber optic monitoring system using a respective trained ML model, of the one or more trained ML models, associated with the failure mode.

[0108] In certain embodiments, the one or more parameters include one or more operating parameters of the at least one monitoring device, one or more parameters indicative of a condition of the one or more fiber optic cables, or a combination thereof.

[0109] In certain embodiments, the at least one monitoring device includes a DTS interrogator, a DSS interrogator, or a DAS interrogator.Example Computing Device

[0110] FIG. 16 illustrates an example computing device 1600 configured to perform predictive health management for a fiber optic monitoring system, according to various embodiments. In certain embodiments, the computing device 1600 may be configured to perform operations 1400 illustrated in FIG. 14, operations 1500 illustrated in FIG. 15, or any other technique or combination of techniques described herein.[0U1] As shown, the computing device 1600 includes, without limitation, a central processing unit (CPU) 1605, a network interface 1615, a memory 1620, and storage 1660, each connected to a bus 1617. The computing device 1600 may also include an I / O device interface 1610 connecting I / O devices 1612 (e.g., keyboard, display and mouse devices) to the computing device 1600. The computing device 1600 is generally under the control of an operating system (not shown).

[0112] The CPU 1605 retrieves and executes programming instructions stored in the memory 1620 as well as stored in the storage 1660. The bus 1617 is used to transmit programming instructions and application data between the CPU 1605, I / O device interface 1610, storage 1660, network interface 1615, and memory 1620. Note, CPU 1605 is included to be representative of a single CPU, multiple CPUs, a single CPU having multiple processing cores, and the like, and the memory 1620 is generally included to be representative of a random access memory. The storage 1660 may be a disk drive or flash storage device. Although shown as a single unit, the storage 1660 may be a combination of fixed and / or removable storage devices, such as fixed disc drives, removable memory cards, optical storage, network attached storage (NAS), or a storage areanetwork (SAN).

[0113] Illustratively, the memory 1620 includes the predictive health management component 110 and the training component 510, which are discussed in greater detail above. Further, storage 1660 includes data 440, data 540, prognostic data 450, AI / ML model(s) 422, or any combination thereof, which are also discussed in greater detail above. Note, while FIG. 16 depicts the predictive health management component 110 and the training component 520 within memory 1620, whichis generally representative of volatile memory (e.g., random access memory), in certain embodiments, the predictive health management component 110 and the training component 520 are included in persistent (e.g., non-volatile) memory or persistent (e.g., non-volatile) storage, such as storage 660 of FIG. 16.Example Clauses

[0114] Implementation examples are described in the following numbered clauses:

[0115] Clause 1 : A computer-implemented method comprising: obtaining an indication of one or more parameters of a fiber optic monitoring system for one or more well completions, the fiber optic monitoring system comprising at least one monitoring device and one or more fiber optic cables coupled to the at least one monitoring device; predicting at least one failure mode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models; and providing an indication of the at least one failure mode.

[0116] Clause 2: The computer-implemented method of claim 1, wherein: the at least one monitoring device comprises an optical switch configured to select a set of the one or more fiber optic cables for monitoring by the at least one monitoring device; and the at least one failure mode comprises a failure of the optical switch.

[0117] Clause 3: The computer-implemented method of claim 1, wherein: the at least one monitoring device comprises a laser source; and the at least one failure mode comprises an output power of the laser source being less than a threshold.

[0118] Clause 4: The computer-implemented method of claim 1, wherein the at least one failure mode comprises an attenuation metric associated with at least one of the one or more fiber optic cables satisfying a predetermined condition.

[0119] Clause 5: The computer-implemented method of claim 1, wherein the at least one failure mode comprises an amount of signal loss associated with at least one of the one or more fiber optic cables being greater than a threshold.

[0120] Clause 6: The computer-implemented method of claim 1, wherein the at least onefailure mode comprises a number of power adjustments of the at least one monitoring device being greater than a threshold.

[0121] Clause 7: The computer-implemented method of claim 1, wherein providing the indication of the at least one failure mode comprises providing an indication of when the at least one failure mode will occur.

[0122] Clause 8: The computer-implemented method of claim 1, wherein predicting the at least one failure mode comprises predicting a plurality of failure modes of the fiber optic monitoring system using a respective trained ML model, of the one or more trained ML models, associated with the failure mode.

[0123] Clause 9: The computer-implemented method of claim 1, wherein the one or more parameters comprise one or more operating parameters of the at least one monitoring device, one or more parameters indicative of a condition of the one or more fiber optic cables, or a combination thereof.

[0124] Clause 10: The computer-implemented method of claim 1, wherein the at least one monitoring device comprises a distributed temperature sensing interrogator, a distributed strain sensing interrogator, a distributed acoustic sensing interrogator, or a combination thereof.

[0125] Clause 11 : A computer-implemented method comprising: obtaining one or more datasets, each comprising a first set of parameters associated with one or more conditions of one or more fiber optic monitoring systems; training a plurality of machine learning (ML) models for predicting failure of a fiber optic monitoring system, each ML model of the plurality of ML models being trained on a respective dataset of the one or more datasets to predict a respective failure mode of a plurality of failure modes associated with the fiber optic monitoring system; and storing the plurality of ML models.

[0126] Clause 12: The computer-implemented method of claim 11, wherein at least one of the plurality of ML models is used during a run-time of the fiber optic monitoring system to predict an occurrence of the respective failure mode associated with the fiber optic monitoring system.

[0127] Clause 13: The computer-implemented method of claim 11, wherein the plurality offailure modes comprise (i) a failure of an optical switch within the fiber optic monitoring system, (ii) an amount of signal loss associated with one or more fiber optic cables of the fiber optic monitoring system being greater than a first threshold, (iii) an output power of a laser source of the fiber optic monitoring system being less than a second threshold, (iv) an attenuation metric associated with the one or more fiber optic cables satisfying a predetermined condition, or (v) any combination thereof.

[0128] Clause 14: The computer-implemented method of claim 11, wherein training the plurality of ML models comprises, for at least one ML model of the plurality of ML models: extracting, from the respective dataset associated with the at least one ML model, one or more features associated with the respective failure mode for the at least one ML model; generating, based on the one or more features, a health indicator for the fiber optic monitoring system; and using the health indicator to train the at least one ML model.

[0129] Clause 15: The computer-implemented method of claim 14, wherein the respective dataset comprises data obtained via a single-ended configuration of a fiber optic cable of the fiber optic monitoring system.

[0130] Clause 16: The computer-implemented method of claim 14, wherein the respective dataset comprises data obtained via a double-ended configuration of a fiber optic cable of the fiber optic monitoring system.

[0131] Clause 17: A fiber optic monitoring system comprising: one or more fiber optic cables deployed within one or more well completions; and a computing system coupled to the one or more fiber optic cables, the computing system comprising: one or more memories collectively storing instructions; and one or more processors coupled to the one or more memories, the one or more processors being collectively configured to execute the instructions to cause the computing system to perform an operation comprising: obtaining an indication of one or more parameters of the fiber optic monitoring system for the one or more well completions; predicting at least one failure mode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models; and providing an indication of the at least one failure mode.

[0132] Clause 18: The fiber optic monitoring system of claim 17, wherein: the computing system further comprises an optical switch configured to select a set of the one or more fiber optic cables for monitoring by the computing system; and the at least one failure mode comprises a failure of the optical switch.

[0133] Clause 19: The fiber optic monitoring system of claim 17, wherein the at least one failure mode comprises an attenuation metric associated with at least one of the one or more fiber optic cables satisfying a predetermined condition.

[0134] Clause 20: The fiber optic monitoring system of claim 17, wherein the at least one failure mode comprises an amount of signal loss associated with at least one of the one or more fiber optic cables being greater than a threshold.

[0135] Clause 21 : A computing system comprising: one or more memories collectively storing executable instructions; and one or more processors coupled to the one or more memories, the one or more processors being collectively configured to execute the executable instructions and cause the computing system to perform a method in accordance with any of Clauses 1-10.

[0136] Clause 22: A computing system comprising: one or more memories collectively storing executable instructions; and one or more processors coupled to the one or more memories, the one or more processors being collectively configured to execute the executable instructions and cause the computing system to perform a method in accordance with any of Clauses 11-16.

[0137] Clause 23: An apparatus comprising means for performing a method in accordance with any of Clauses 1-10.

[0138] Clause 24: An apparatus comprising means for performing a method in accordance with any of Clauses 11-16.

[0139] Clause 25: A non-transitory computer-readable medium comprising computerexecutable instructions that, when executed by one or more processors of a computing system, cause the computing system to perform a method in accordance with any of Clauses 1-10.

[0140] Clause 26: A non-transitory computer-readable medium comprising computerexecutable instructions that, when executed by one or more processors of a computing system,cause the computing system to perform a method in accordance with any of Clauses 1 1-16.Additional Considerations

[0141] The preceding description is provided to enable any person skilled in the art to practice the various aspects described herein. The examples discussed herein are not limiting of the scope, applicability, or aspects set forth in the claims. Various modifications to these aspects will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other aspects. For example, changes may be made in the function and arrangement of elements discussed without departing from the scope of the disclosure. Various examples may omit, substitute, or add various procedures or components as appropriate. For instance, the methods described may be performed in an order different from that described, and various actions may be added, omitted, or combined. Also, features described with respect to some examples may be combined in some other examples. For example, an apparatus may be implemented or a method may be practiced using any number of the aspects set forth herein. In addition, the scope of the disclosure is intended to cover such an apparatus or method that is practiced using other structure, functionality, or structure and functionality in addition to, or other than, the various aspects of the disclosure set forth herein. It should be understood that any aspect of the disclosure disclosed herein may be embodied by one or more elements of a claim.

[0142] The various illustrative logical blocks, modules and circuits described in connection with the present disclosure may be implemented or performed with a general purpose processor, a digital signal processor (DSP), an ASIC, a field programmable gate array (FPGA) or other programmable logic device (PLD), discrete gate or transistor logic, discrete hardware components, or any combination thereof designed to perform the functions described herein. A general-purpose processor may be a microprocessor, but in the alternative, the processor may be any commercially available processor, controller, microcontroller, or state machine. A processor may also be implemented as a combination of computing devices, e.g., a combination of a DSP and a microprocessor, a plurality of microprocessors, one or more microprocessors in conjunction with a DSP core, a system on a chip (SoC), or any other such configuration.

[0143] As used herein, a phrase referring to “at least one of’ a list of items refers to any combination of those items, including single members. As an example, “at least one of: a, b, or c” is intended to cover a, b, c, a-b, a-c, b-c, and a-b-c, as well as any combination with multiples of the same element (e.g., a-a, a-a-a, a-a-b, a-a-c, a-b-b, a-c-c, b-b, b-b-b, b-b-c, c-c, and c-c-c or any other ordering of a, b, and c).

[0144] As used herein, “a processor,” “at least one processor,” or “one or more processors” generally refer to a single processor configured to perform one or multiple operations or multiple processors configured to collectively perform one or more operations. In the case of multiple processors, performance of the one or more operations could be divided amongst different processors, though one processor may perform multiple operations, and multiple processors could collectively perform a single operation. Similarly, “a memory,” “at least one memory,” or “one or more memories” generally refer to a single memory configured to store data and / or instructions or multiple memories configured to collectively store data and / or instructions.

[0145] As used herein, the term “determining” encompasses a wide variety of actions. For example, “determining” may include calculating, computing, processing, deriving, investigating, looking up (e.g., looking up in a table, a database or another data structure), ascertaining and the like. Also, “determining” may include receiving (e.g., receiving information), accessing (e.g., accessing data in a memory) and the like. Also, “determining” may include resolving, selecting, choosing, establishing and the like.

[0146] The methods disclosed herein comprise one or more actions for achieving the methods. The method actions may be interchanged with one another without departing from the scope of the claims. In other words, unless a specific order of actions is specified, the order and / or use of specific actions may be modified without departing from the scope of the claims. Further, the various operations of methods described above may be performed by any suitable means capable of performing the corresponding functions. The means may include various hardware and / or software component(s) and / or module(s), including, but not limited to a circuit, an application specific integrated circuit (ASIC), or processor.

[0147] The following claims are not intended to be limited to the aspects shown herein, but are to be accorded the full scope consistent with the language of the claims. Within a claim,reference to an element in the singular is not intended to mean “one and only one” unless specifically so stated, but rather “one or more.” Unless specifically stated otherwise, the term “some” refers to one or more. No claim element is to be construed under the provisions of 35 U.S.C. §112(f) unless the element is expressly recited using the phrase “means for”. All structural and functional equivalents to the elements of the various aspects described throughout this disclosure that are known or later come to be known to those of ordinary skill in the art are expressly incorporated herein by reference and are intended to be encompassed by the claims. Moreover, nothing disclosed herein is intended to be dedicated to the public regardless of whether such disclosure is explicitly recited in the claims.

Claims

CLAIMS1. A computer-implemented method comprising: obtaining an indication of one or more parameters of a fiber optic monitoring system for one or more well completions, the fiber optic monitoring system comprising at least one monitoring device and one or more fiber optic cables coupled to the at least one monitoring device; predicting at least one failure mode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models; and providing an indication of the at least one failure mode.

2. The computer-implemented method of claim 1, wherein: the at least one monitoring device comprises an optical switch configured to select a set of the one or more fiber optic cables for monitoring by the at least one monitoring device; and the at least one failure mode comprises a failure of the optical switch.

3. The computer-implemented method of claim 1, wherein: the at least one monitoring device comprises a laser source; and the at least one failure mode comprises an output power of the laser source being less than a threshold.

4. The computer-implemented method of claim 1, wherein the at least one failure mode comprises an attenuation metric associated with at least one of the one or more fiber optic cables satisfying a predetermined condition.

5. The computer-implemented method of claim 1, wherein the at least one failure mode comprises an amount of signal loss associated with at least one of the one or more fiber optic cables being greater than a threshold.

6. The computer-implemented method of claim 1, wherein the at least one failure mode comprises a number of power adjustments of the at least one monitoring device being greater than a threshold.

7. The computer-implemented method of claim 1, wherein providing the indication of the at least one failure mode comprises providing an indication of when the at least one failure mode will occur.

8. The computer-implemented method of claim 1, wherein predicting the at least one failure mode comprises predicting a plurality of failure modes of the fiber optic monitoring system using a respective trained ML model, of the one or more trained ML models, associated with the failure mode.

9. The computer-implemented method of claim 1, wherein the one or more parameters comprise one or more operating parameters of the at least one monitoring device, one or more parameters indicative of a condition of the one or more fiber optic cables, or a combination thereof.

10. The computer-implemented method of claim 1, wherein the at least one monitoring device comprises a distributed temperature sensing interrogator, a distributed strain sensing interrogator, a distributed acoustic sensing interrogator, or a combination thereof.

11. A computer-implemented method comprising: obtaining one or more datasets, each comprising a first set of parameters associated with one or more conditions of one or more fiber optic monitoring systems; training a plurality of machine learning (ML) models for predicting failure of a fiber optic monitoring system, each ML model of the plurality of ML models being trained on a respective dataset of the one or more datasets to predict a respective failure mode of a plurality of failure modes associated with the fiber optic monitoring system; and storing the plurality of ML models.

12. The computer-implemented method of claim 11, wherein at least one of the plurality of ML models is used during a run-time of the fiber optic monitoring system to predict an occurrence of the respective failure mode associated with the fiber optic monitoring system.

13. The computer-implemented method of claim 11 , wherein the plurality of failure modes comprise (i) a failure of an optical switch within the fiber optic monitoring system, (ii) an amount of signal loss associated with one or more fiber optic cables of the fiber optic monitoring system being greater than a first threshold, (iii) an output power of a laser source of the fiber optic monitoring system being less than a second threshold, (iv) an attenuation metric associated with the one or more fiber optic cables satisfying a predetermined condition, or (v) any combination thereof.

14. The computer-implemented method of claim 11, wherein training the plurality of ML models comprises, for at least one ML model of the plurality of ML models: extracting, from the respective dataset associated with the at least one ML model, one or more features associated with the respective failure mode for the at least one ML model; generating, based on the one or more features, a health indicator for the fiber optic monitoring system; and using the health indicator to train the at least one ML model.

15. The computer-implemented method of claim 14, wherein the respective dataset comprises data obtained via a single-ended configuration of a fiber optic cable of the fiber optic monitoring system.

16. The computer-implemented method of claim 14, wherein the respective dataset comprises data obtained via a double-ended configuration of a fiber optic cable of the fiber optic monitoring system.

17. A fiber optic monitoring system comprising: one or more fiber optic cables deployed within one or more well completions; and a computing system coupled to the one or more fiber optic cables, the computing system comprising: one or more memories collectively storing instructions; and one or more processors coupled to the one or more memories, the one or more processors being collectively configured to execute the instructions to cause the computing system to perform an operation comprising: obtaining an indication of one or more parameters of the fiber optic monitoring system for the one or more well completions; predicting at least one failure mode of the fiber optic monitoring system, based at least in part on evaluating the one or more parameters with one or more trained machine learning (ML) models; and providing an indication of the at least one failure mode.

18. The fiber optic monitoring system of claim 17, wherein: the computing system further comprises an optical switch configured to select a set of the one or more fiber optic cables for monitoring by the computing system; and the at least one failure mode comprises a failure of the optical switch.

19. The fiber optic monitoring system of claim 17, wherein the at least one failure mode comprises an attenuation metric associated with at least one of the one or more fiber optic cables satisfying a predetermined condition.

20. The fiber optic monitoring system of claim 17, wherein the at least one failure mode comprises an amount of signal loss associated with at least one of the one or more fiber optic cables being greater than a threshold.

Images