Method and device for detecting abnormality of pipe

Abstract

The present disclosure provides a device for detecting an abnormality of a pipe. The device comprises: a sensing unit including multiple sensors for sensing a signal dispersed through an optical fiber attached to one side of the pipe to obtain sensing information; and a control unit which determines one abnormality detection algorithm from among multiple abnormality detection algorithms on the basis of training completeness and determines whether the pipe is abnormal on the basis of an output value from the determined abnormality detection algorithm, the output value being derived by applying the sensing information as an input value.

Description

Method and device for detecting abnormalities in pipes

[0001] The present disclosure relates to a technique for detecting anomalies in piping.

[0002] The purpose of this study is to rapidly detect the location of blockages in pipes caused by lime slurry before and after the Mg process for concentrated brine treatment. The Mg process involves removing the magnesium component of brine by precipitating it as magnesium hydroxide using the auxiliary raw material slaked lime, which is then separated into solid and liquid. Clogging can occur frequently in the pipes where the slaked lime is injected. While clogging is particularly common at elbows, irregular clogging also occurs in straight sections of the pipe, necessitating precise location determination. Currently, without detection devices, pipes are cut and replaced at multiple locations based on estimated locations. Therefore, high-resolution detection of blockages is required by attaching optical fibers to the pipes.

[0003] The abnormalities of pipe blockages include changes in acoustics (vibration) caused by accumulated limescale and deformation of the pipe due to increased water pressure in the blocked area. Over time, limescale hardens inside the pipe, forming a slurry of salt water and limescale, blocking the pipe's internal structure. Pipe blockages result from the hardened limescale and slurry, which causes abnormal fluid flow.

[0004] Therefore, specific measures are required to minimize process downtime due to pipe abnormalities and to recognize the condition of the pipes and suggest measures to managers.

[0005] These embodiments are intended to provide a technique for detecting whether anomalies may occur in a pipe.

[0006] In addition, the present embodiments aim to provide a method and device for detecting pipe abnormalities that can recognize the condition of the pipe and suggest measures to a manager in order to minimize process downtime due to pipe abnormalities.

[0007] Additionally, the present embodiments aim to provide a step-by-step learning program that can advance the operation time of an anomaly detection algorithm by resolving the data bias problem.

[0008] In order to solve the above-described problem, one embodiment of the present disclosure provides a pipe abnormality detection device, which includes a sensing unit including a plurality of sensors that detect distributed signals through optical fibers attached to one side of a pipe to obtain sensing information, and a control unit that determines one abnormality detection algorithm among a plurality of abnormality detection algorithms based on a degree of learning completion, and determines whether there is an abnormality in the pipe based on an output value of the determined abnormality detection algorithm derived by applying the sensing information as an input value.

[0009] In addition, one embodiment can provide a method for detecting an abnormality in a pipe, including a step of detecting a signal dispersed through an optical fiber attached to one side of the pipe through a plurality of sensors to obtain sensing information, a step of determining one abnormality detection algorithm among a plurality of abnormality detection algorithms based on a degree of learning completion, and a step of determining whether there is an abnormality in the pipe based on an output value of the determined abnormality detection algorithm derived by applying the sensing information as an input value.

[0010] According to the present embodiment, a method and device for detecting a pipe abnormality can be provided that can recognize the condition of the pipe and suggest measures to a manager in order to minimize process downtime due to the pipe abnormality.

[0011] Additionally, by providing a step-by-step learning program that can address data bias issues and accelerate the operation of anomaly detection algorithms, the cost of losses due to process delays can be reduced.

[0012] FIG. 1 is a drawing for explaining the configuration of a pipe anomaly detection device according to one embodiment.

[0013] FIG. 2 is a drawing for explaining a pipe anomaly detection device applied to a pipe according to one embodiment.

[0014] FIG. 3 is a drawing for explaining the operation of a pipe anomaly detection device according to one embodiment.

[0015] FIG. 4 is a drawing for explaining a procedure of a pipe abnormality detection method according to one embodiment.

[0016] FIGS. 5 and 6 are drawings for explaining the procedure of a step-by-step anomaly detection method for a pipe according to one embodiment.

[0017] Hereinafter, some embodiments of the present disclosure will be described in detail with reference to exemplary drawings. When adding reference numerals to components in each drawing, identical components may have the same numerals as much as possible even if they are shown in different drawings. In addition, when describing the present embodiments, if it is determined that a detailed description of a related known configuration or function may obscure the gist of the technical idea of ​​the present invention, the detailed description may be omitted. When "includes," "has," "consists of," etc. are used in this specification, other parts may be added unless "only" is used. When a component is expressed in the singular, it may include a case in which the plural is included unless specifically stated otherwise.

[0018] Additionally, terms such as first, second, A, B, (a), (b), etc. may be used to describe components of the present disclosure. These terms are only intended to distinguish the components from other components, and the nature, order, sequence, or number of the components are not limited by the terms.

[0019] In a description of the positional relationship of components, when it is described that two or more components are "connected," "combined," or "connected," it should be understood that the two or more components may be directly "connected," "combined," or "connected," but that the two or more components may also be further "interposed" with another component to be "connected," "combined," or "connected." Here, the other component may be included in one or more of the two or more components that are "connected," "combined," or "connected" to each other.

[0020] In the description of the temporal flow relationship related to components, operation methods, or manufacturing methods, for example, when the temporal or flow relationship is described as “after”, “following”, “next to”, “before”, etc., it may also include cases where it is not continuous, unless “immediately” or “directly” is used.

[0021] Meanwhile, when numerical values ​​or corresponding information (e.g., levels, etc.) for components are mentioned, even without separate explicit description, the numerical values ​​or corresponding information may be interpreted as including an error range that may occur due to various factors (e.g., process factors, internal or external impact, noise, etc.).

[0022] Hereinafter, a device for detecting anomalies in a pipe according to embodiments of the present disclosure will be described in detail with reference to related drawings.

[0023] FIG. 1 is a drawing for explaining the configuration of a pipe anomaly detection device according to one embodiment.

[0024] Referring to FIG. 1, a pipe abnormality detection device (100) for detecting a pipe abnormality may include a sensing unit (110) including a plurality of sensors that detect scattered signals through optical fibers attached to one side of a pipe to obtain sensing information, and a control unit (120) that determines one of a plurality of abnormality detection algorithms based on a learning completion degree and determines whether a pipe abnormality exists based on an output value of the determined abnormality detection algorithm derived by applying the sensing information as an input value. The configuration of the pipe abnormality detection device (100) illustrated in FIG. 1 is an example and is not limited thereto, and other components may be further included as needed.

[0025] The magnesium process removes the magnesium component of brine by precipitating it as magnesium hydroxide using the auxiliary raw material slaked lime, which is then separated into solid and liquid. Consequently, clogging of the pipes through which the slaked lime is injected is a frequent problem.

[0026] The abnormalities of pipe blockages include changes in acoustics (vibration) caused by accumulated limescale and deformation of the pipe due to increased water pressure in the blocked area. Over time, limescale hardens inside the pipe, forming a slurry of salt water and limescale, blocking the pipe's internal structure. Pipe blockages result from the hardened limescale and slurry, which causes abnormal fluid flow.

[0027] Pipe blockages can be diagnosed by examining the time series or frequency changes in the acoustic (vibration) generated by the flowing fluid, due to the narrowed pipe diameter caused by hardened limestone. To measure this, optical fibers are attached to the pipe, and acoustic (vibration), temperature, and strain data can be acquired through an interrogator. The location of the optical fiber attachment is also important. It is advantageous to install the optical fiber inside the pipe so that it is protected from external factors while maximizing the fluid's influence. However, to detect the same anomaly in an existing pipe, it must be attached externally. In this case, exposure should be relatively reduced to minimize external influences, and since the vibration or sound associated with the location of the blockage should be measured, it is advantageous to attach it to the bottom of the pipe, where the fluid is constantly flowing and thus transmits the sound well.

[0028] To this end, the sensing unit (110) may include a Distributed Acoustic Sensing (DAS) that detects an acoustic signal distributed through an optical fiber to obtain sensing information, and a Distributed Strain Sensing (DSS) that detects a strain signal distributed through an optical fiber to obtain sensing information. The sensing unit (110) may obtain sensing information through each sensor.

[0029] In the case of DAS, the location resolution is at the level of 1m, so the blocked section can be diagnosed primarily using DAS. Afterwards, high-precision blockage location detection can be diagnosed secondarily using DSS with a location resolution of 0.1m. In the case of DSS, by measuring strain, anomalies can be detected in the section immediately before the blockage due to slaked lime due to the strain in the pipe caused by the increase in fluid pressure. In other words, DAS is used to monitor the location of a pipe that is gradually blocked by slaked lime, and DSS is used to detect the location when a specific section is completely blocked by slurry, etc.

[0030] The sensing unit (110) can transmit the acquired sensing information to the control unit (120). In one example, the sensing information may be configured to be continuously transmitted in real time during the aforementioned Mg process. Alternatively, the sensing information may be configured to be transmitted at the request of the control unit (120). Alternatively, the sensing information may be configured to be transmitted according to predetermined conditions, such as a preset time period or cycle before or after the Mg process.

[0031] The control unit (120) may be implemented in software, hardware, or a combination thereof in various devices capable of executing the method according to the technical idea of ​​the present disclosure, such as a processor, a computer, or other processing device.

[0032] The control unit (120) is connected to the sensing unit (110) in a communication manner, and can control the operation of the sensing unit (110) and monitor whether a malfunction occurs in the sensing unit (110). The control unit (120) can receive sensing information from the sensing unit (110). In one example, the control unit (120) can request the sensing unit (110) to transmit sensing information.

[0033] The control unit (120) can store the sensing information received from the sensing unit (110) in a database. The sensing information can be stored according to predetermined criteria, such as the sensed time information or the sensed location.

[0034] To detect and assess pipe blockages, an algorithm is needed to identify them based on data collected by interrogators connected to optical fibers. Representative algorithms that utilize collected time-series data to detect system anomalies include deep learning-based algorithms such as Long-Short Term Memory (LSTM) and Convolutional Neural Networks (CNNs). Utilizing these deep learning models for anomaly detection requires a sufficient dataset with unbiased labels. This is because the number of abnormal data is typically small compared to normal data, and the accuracy of predictions using deep learning algorithms is bound to decline until sufficient abnormal data is available. Therefore, it may be difficult to immediately apply deep learning-based anomaly detection algorithms in the field until sufficient pipe blockage abnormal data is generated.

[0035] Accordingly, the control unit (120) may determine one anomaly detection algorithm to derive a conclusion value based on the learning completion level of each anomaly detection algorithm that performs learning among the plurality of anomaly detection algorithms. For example, the plurality of anomaly detection algorithms may include an SVM (Support Vector Machine)-based anomaly detection algorithm and a deep learning-based anomaly detection algorithm. In this case, the SVM and deep learning algorithms may include SVM and deep learning algorithms that are known before or after the present disclosure, as long as they do not contradict the technical spirit of the present disclosure. In this case, anomaly classification is performed using an SVM (Support Vector Machine)-based algorithm, which is a simple algorithm learned from data created through experiments, and when the learning accuracy of deep learning for which sufficient data has been secured is higher than a predetermined target value (for example, 95%), the deep learning anomaly detection algorithm may be applied to the field.

[0036] The control unit (120) can determine the degree of learning completion based on the accuracy of the deep learning-based anomaly detection algorithm. The accuracy of the deep learning-based anomaly detection algorithm can be measured by comparing the model's predicted results with actual data.

[0037] In this case, the control unit (120) can receive an accurate value regarding whether the pipe is abnormal through the user interface unit. That is, the control unit (120) can input an accurate value regarding the actual state of the pipe (blockage or normal state) through the user interface unit. In one example, the control unit (120) can compare the output values ​​derived based on the sensing information with the accurate values ​​to obtain a precise ratio of cases that match in all cases.

[0038] Additionally, the deep learning anomaly detection algorithm can be preset with a target precision as a first target value. The first target value can be individually set as needed and is not limited to a specific value.

[0039] The control unit (120) can determine whether the current precision of the deep learning-based anomaly detection algorithm is higher than the first target value. If the precision of the deep learning-based anomaly detection algorithm is higher than the first target value, the control unit (120) can determine the deep learning-based anomaly detection algorithm as the anomaly detection algorithm from which to derive a conclusion value. That is, if the precision of the deep learning-based anomaly detection algorithm is higher than the first target value, the control unit (120) can determine whether there is an anomaly in the pipe based on the output value of the deep learning-based anomaly detection algorithm.

[0040] If the precision of the deep learning-based anomaly detection algorithm is lower than or equal to the first target value, the control unit (120) may determine the SVM-based anomaly detection algorithm as the anomaly detection algorithm for deriving a conclusion value. That is, if the precision of the deep learning-based anomaly detection algorithm is lower than or equal to the first target value, the control unit (120) may determine whether or not there is an anomaly in the pipe based on the output value of the SVM-based anomaly detection algorithm. In other words, although the deep learning-based anomaly detection algorithm and the SVM-based anomaly detection algorithm each derive output values ​​according to input values, the final selection of the output value of which algorithm is determined based on the precision of the deep learning-based anomaly detection algorithm.

[0041] The control unit (120) can output information to the user via the user interface regarding the presence or absence of a pipe abnormality. If the control unit (120) determines that a pipe abnormality has occurred, it can output an alarm regarding a pipe blockage. For example, the user interface unit is not limited to specific hardware, software, or a combination thereof, as long as it can input and output information according to user-authorized operations.

[0042] Since pipe anomaly data is acquired based on the assumption that actual pipe anomalies have occurred, it can take considerable time to balance the anomaly and normal data. Therefore, if a deep learning-based anomaly detection algorithm is directly applied to a less biased set of anomaly data, the accuracy may fall below the initial target value. Therefore, as described above, a step-by-step application of a deep learning-based anomaly detection algorithm and an SVM-based anomaly detection algorithm can be used to prioritize output prediction.

[0043] Time-series data and frequency-domain values ​​for each pipe section measured during the Mg process can be input to the SVM and deep learning-based anomaly detection algorithm. Furthermore, the state classifier of the SVM and deep learning-based anomaly detection algorithm can classify the closest pipe blockage state based on the output of the last layer. The output of the SVM and deep learning-based anomaly detection algorithm can output the status of the pipe classified by section.

[0044] As described above, the control unit (120) can receive an accurate value regarding the presence or absence of a pipe abnormality through the user interface unit. That is, information regarding the actual state of the pipe corresponding to the output value derived through the abnormality detection algorithm can be received. Based on the accurate value corresponding to the actual state of the pipe, the control unit (120) can determine whether the sensing information used to derive the result value corresponds to abnormal data of the pipe. The control unit (120) can store data identified as pipe abnormality data and use it for future learning.

[0045] The control unit (120) can reflect the correct answer value and perform learning of the deep learning-based anomaly detection algorithm and the SVM-based anomaly detection algorithm. That is, in a state of less bias in the abnormal data, the learning process of predicting the output value of the pipe abnormal data and reflecting the predicted result can be repeated every time an abnormality occurs in the pipe. Accordingly, as the bias in the abnormal data is resolved, the accuracy of the deep learning-based anomaly detection algorithm and the SVM-based anomaly detection algorithm can be increased.

[0046] The control unit (120) can perform learning through backpropagation for SVM-based anomaly detection algorithms and deep learning-based anomaly detection algorithms. In this case, backpropagation can be used to learn in the direction of minimizing a loss function by adjusting the weights and biases of the neural network model.

[0047] For example, the control unit (120) may derive output values ​​by applying the weights and activation functions of each layer through a forward propagation process in a deep learning-based anomaly detection algorithm. In this case, a loss function representing the difference between the predicted output value and the actual value may be derived. In the backpropagation process, the loss function may be differentiated with respect to the weights and biases to derive the gradients of each parameter. Using the gradients, the weights and biases may be adjusted through gradient descent or a similar optimization algorithm. These weight and bias updates may be performed in a direction that minimizes the loss function. If this process is repeated, the number of pipe anomaly data used for learning increases, the loss function is minimized, and the performance of the deep learning-based anomaly detection algorithm may be improved.

[0048] The control unit (120) can distribute the weights of the deep learning-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is higher than the first target value. In other words, when the biased state of the pipe anomaly data is resolved and the performance of the deep learning-based anomaly detection algorithm improves to a target precision or higher, the deep learning-based anomaly detection algorithm model is distributed and the presence or absence of a pipe anomaly can be determined through the algorithm. Accordingly, the deep learning-based anomaly detection algorithm can be applied immediately even when the pipe anomaly data is biased.

[0049] The control unit (120) performs the aforementioned process with priority on the sensing information obtained from the DAS, and when the pipe blockage state is classified as 'blocked', the same process can be performed on the sensing information obtained from the DSS for the relevant section to output the pipe blockage state.

[0050] For example, the control unit (120) may store additional information related to the pipe in the database. For example, if a pipe is determined to be defective and a portion of the pipe is replaced, the control unit (120) may store or update information regarding the portion of the pipe to be replaced and the timing of the replacement. The control unit (120) may learn patterns regarding the timing of pipe replacement and, if there is a risk of a pipe defect occurring, may output a replacement warning alarm in advance.

[0051] This approach can minimize process downtime due to piping abnormalities by recognizing the condition of the pipes and suggesting actions to managers. Furthermore, by providing a step-by-step learning program that addresses data bias and accelerates the implementation of anomaly detection algorithms, the cost of losses resulting from process delays can be reduced.

[0052] Below, each embodiment related to a device for detecting pipe abnormalities will be specifically described with reference to related drawings.

[0053] FIG. 2 is a drawing for explaining a pipe anomaly detection device applied to a pipe according to one embodiment. FIG. 3 is a drawing for explaining the operation of a pipe anomaly detection device according to one embodiment.

[0054] Referring to Fig. 2, a schematic structure of a pipe (10) and a pipe abnormality detection device (100) is illustrated. An optical fiber (111) may be installed at the bottom of the pipe (10). A portion (B) in which the pipe is blocked may occur due to foreign substances such as hardened lime (20) within the pipe (10).

[0055] A pipe abnormality detection device (100) for detecting a pipe blockage location may include a sensing unit (110) and a control unit (120).

[0056] Referring to FIG. 3, the sensing unit (110) includes a DAS (113) and a DSS (114), and can obtain sensing information by receiving a signal distributed from an optical fiber (111) through an interrogator (112).

[0057] The control unit (120) of the pipe anomaly detection device can be implemented as a computer such as a single board computer (SBC), and can include a signal processing unit (121), a data management unit (122), an algorithm unit (123), and a user interface unit (124).

[0058] The signal processing unit (121) may include an optical fiber data collection unit. The signal processing unit (121) may be connected to the DAS (113) and the DSS (114) in communication with each other, and may control the operation of the DAS (113) and the DSS (114) and monitor whether a failure occurs. The signal processing unit (121) may receive sensing information from the DAS (113) and the DSS (114). In one example, the signal processing unit (121) may request the transmission of sensing information to the DAS (113) and the DSS (114).

[0059] The data management unit (122) may include a database. The data management unit (122) may store sensing information received from the DAS (113) and the DSS (114). The sensing information may be stored according to predetermined criteria, such as sensing time information.

[0060] The user interface unit (124) may include monitoring, status input / output, and data query functions. For example, the user interface unit (124) may include various input interfaces for receiving user-authorized operations and various output interfaces for outputting information according to input information or control of the control unit (120).

[0061] The algorithm unit (123) may include a step-by-step anomaly detection learning algorithm for overcoming data insufficiency. The step-by-step anomaly detection learning algorithm for overcoming data insufficiency may include an SVM-based anomaly detection algorithm and a deep learning-based anomaly detection algorithm.

[0062] The algorithm section (123) can apply a step-by-step anomaly detection learning algorithm to overcome data insufficiency, evaluate the learning completion, detect the location of a pipe blockage, record the correct answer, perform learning, and perform distribution.

[0063] The algorithm unit (123) can evaluate the accuracy of the SVM-based anomaly detection algorithm and the accuracy of the deep learning-based anomaly detection algorithm in order to evaluate the learning completion rate.

[0064] The algorithm unit (123) may output the location of the pipe blockage through an SVM-based anomaly detector when the precision of the deep learning-based anomaly detection algorithm is lower than the target precision in order to detect the location of the pipe blockage. When the precision of the deep learning-based anomaly detector is lower than the target precision, the status of the pipe may be output through an SVM-based anomaly detector.

[0065] Time-series data and frequency-domain values ​​for each pipe section measured during the Mg process can be input to the SVM and deep learning-based anomaly detection algorithm. The state classifier of the SVM and deep learning-based anomaly detection algorithm can classify the closest pipe blockage state based on the output value of the last layer. The output of the SVM and deep learning-based anomaly detection algorithm can output the status of the pipe classified by section.

[0066] Training of SVM and deep learning-based anomaly detection algorithms can be performed through backpropagation.

[0067] The correct answer for the pipe condition can be input from the user through the status input / output function of the user interface until sufficient pipe blockage abnormality data is secured.

[0068] The algorithm unit (123) can distribute the weights of the SVM and / or deep learning-based anomaly detection algorithm when the precision of the SVM and / or deep learning-based anomaly detection algorithm is higher than the target precision.

[0069] First, if the anomaly detection algorithm of the DAS signal among the optical fiber data classifies the pipe blockage state as 'blockage', the anomaly detection algorithm of the DSS signal among the optical fiber data of the relevant section can output the pipe blockage state.

[0070] This approach can minimize process downtime due to piping abnormalities by recognizing the condition of the pipes and suggesting actions to managers. Furthermore, by providing a step-by-step learning program that addresses data bias and accelerates the implementation of anomaly detection algorithms, the cost of losses resulting from process delays can be reduced.

[0071]

[0072] Hereinafter, a method for detecting piping anomalies, which can perform some or all of the embodiments described with reference to FIGS. 1 through 3, will be described with reference to the drawings. The above description may be omitted to avoid redundant explanation, and in this case, the omitted content may be substantially equally applied to the following description, as long as it does not contradict the technical spirit of the invention.

[0073] FIG. 4 is a diagram illustrating a procedure of a pipe anomaly detection method according to one embodiment. FIG. 5 and FIG. 6 are diagrams illustrating a procedure of a step-by-step pipe anomaly detection method according to one embodiment.

[0074] Referring to FIG. 4, the pipe anomaly detection device can obtain sensing information by detecting a signal distributed through an optical fiber attached to one side of the pipe through a plurality of sensors (S410).

[0075] Referring to Figure 5, an optical fiber can be attached to the pipe to acquire sensing information (S510). Pipe blockage abnormalities can be diagnosed through changes in the time series or frequency of sound (vibration) generated by the flowing fluid due to the pipe diameter being narrowed by hardened limestone. Therefore, the time series or frequency changes of sound (vibration) generated by the flowing fluid within the pipe can be measured through the optical fiber attached to the pipe.

[0076] First, the pipe anomaly detection device can detect dispersed acoustic signals via optical fiber and collect DAS signals, which acquire sensing information through an interrogator (S520). With a positional resolution of 1 m, DAS can be used to initially diagnose blocked sections. In other words, DAS can be used to monitor the location of pipes gradually becoming clogged by limescale.

[0077] A pipe anomaly detection device can control the operation of a sensing unit and monitor for malfunctions in the sensing unit. The pipe anomaly detection device can store the sensing information obtained through the sensing unit in a database. The sensing information can be stored according to predetermined criteria, such as the time of detection or the location of the detection.

[0078] Referring again to FIG. 4, the pipe anomaly detection device determines one anomaly detection algorithm among multiple anomaly detection algorithms based on the learning completion level (S420), and can determine whether there is an anomaly in the pipe based on the output value of the determined anomaly detection algorithm derived by applying sensing information as an input value (S430).

[0079] To detect and assess pipe blockages, an algorithm is needed to identify them based on data collected by interrogators connected to optical fibers. Representative algorithms that utilize collected time-series data to detect system anomalies include deep learning-based algorithms such as Long-Short Term Memory (LSTM) and Convolutional Neural Networks (CNNs). Utilizing these deep learning models for anomaly detection requires a sufficient dataset with unbiased labels. Therefore, it may be difficult to immediately apply deep learning-based anomaly detection algorithms in the field until sufficient pipe blockage data has been generated.

[0080] Accordingly, the pipe anomaly detection device can determine one anomaly detection algorithm to derive a conclusion value based on the learning completion level of each anomaly detection algorithm that performs learning among the plurality of anomaly detection algorithms. For example, the plurality of anomaly detection algorithms may include an SVM (Support Vector Machine)-based anomaly detection algorithm and a deep learning-based anomaly detection algorithm. In this case, anomaly classification is performed using an SVM (Support Vector Machine)-based algorithm, which is a simple algorithm learned from data created through experiments, and the deep learning anomaly detection algorithm can be applied to the field when the learning accuracy of deep learning with sufficient data is higher than a predetermined target value (e.g., 95%).

[0081] The pipeline anomaly detection device can determine the learning completion level based on the accuracy of the deep learning-based anomaly detection algorithm. The accuracy of the deep learning-based anomaly detection algorithm can be measured by comparing the model's predicted results with actual data.

[0082] In this case, the pipe anomaly detection device can receive an accurate value regarding the presence or absence of a pipe anomaly through the user interface. That is, the pipe anomaly detection device can receive an accurate value regarding the actual state of the pipe (blockage or normal state) through the user interface. For example, the pipe anomaly detection device can obtain a precise ratio of cases that match across all cases by comparing output values ​​derived based on sensing information with the accurate values.

[0083] Additionally, the deep learning anomaly detection algorithm can be preset with a target precision as a first target value. The first target value can be individually set as needed and is not limited to a specific value.

[0084] The pipe anomaly detection device can determine whether the current precision of the deep learning-based anomaly detection algorithm is higher than the first target value (S530). If the precision of the deep learning-based anomaly detection algorithm is higher than the first target value, the pipe anomaly detection device can determine the deep learning-based anomaly detection algorithm as the anomaly detection algorithm from which to derive a conclusion. That is, if the precision of the deep learning-based anomaly detection algorithm is higher than the first target value, the pipe anomaly detection device can determine whether the pipe has an anomaly based on the output value of the deep learning-based anomaly detection algorithm (S540).

[0085] If the precision of the deep learning-based anomaly detection algorithm is lower than or equal to the first target value, the pipe anomaly detection device may determine the SVM-based anomaly detection algorithm as the anomaly detection algorithm from which to derive a conclusion. That is, if the precision of the deep learning-based anomaly detection algorithm is lower than or equal to the first target value, the pipe anomaly detection device may determine whether or not the pipe is abnormal based on the output value of the SVM-based anomaly detection algorithm (S560).

[0086] Time-series data and frequency-domain values ​​for each pipe section measured during the Mg process can be input to the SVM and deep learning-based anomaly detection algorithm. Furthermore, the state classifier of the SVM and deep learning-based anomaly detection algorithm can classify the closest pipe blockage state based on the output of the last layer. The output of the SVM and deep learning-based anomaly detection algorithm can output the status of the pipe classified by section.

[0087] The pipe anomaly detection device can output information to the user via a user interface regarding the presence of an abnormality in the pipe. If the pipe anomaly detection device determines that an abnormality has occurred in the pipe based on a DAS signal, it can output an alarm indicating a pipe blockage. For example, the user interface is not limited to any specific hardware, software, or combination thereof, as long as it can input and output information based on user-authorized operations.

[0088] Since pipe anomaly data is acquired based on the assumption that actual pipe anomalies have occurred, it can take considerable time to balance the anomaly and normal data. Therefore, if a deep learning-based anomaly detection algorithm is directly applied to a less biased set of anomaly data, the accuracy may fall below the initial target value. Therefore, as described above, a step-by-step application of a deep learning-based anomaly detection algorithm and an SVM-based anomaly detection algorithm can be used to prioritize output prediction.

[0089] As described above, the pipe anomaly detection device can receive and store an answer value for the presence or absence of a pipe anomaly through the user interface (S570). In other words, the device can receive information about the actual condition of the pipe corresponding to the output value derived through the anomaly detection algorithm. The pipe anomaly detection device can determine whether the sensing information used to derive the result value corresponds to pipe anomaly data based on the answer value corresponding to the actual condition of the pipe. The pipe anomaly detection device can store data identified as pipe anomaly data and use it for future learning.

[0090] The pipe anomaly detection device can train deep learning-based and SVM-based anomaly detection algorithms by reflecting the correct answer. In other words, with less bias in the anomaly data, the output value prediction for the pipe anomaly data and the learning process that reflects the predicted results can be repeated each time a pipe anomaly occurs. Accordingly, as the bias in the anomaly data is reduced, the accuracy of the deep learning-based and SVM-based anomaly detection algorithms can be improved.

[0091] The pipe anomaly detection device can train an SVM-based anomaly detection algorithm (S580) and a deep learning-based anomaly detection algorithm (S590) through backpropagation. In this case, backpropagation can be used to train the neural network model by adjusting its weights and biases to minimize the loss function.

[0092] For example, a pipe anomaly detection device can derive output values ​​by applying the weights and activation functions of each layer through the forward propagation process in a deep learning-based anomaly detection algorithm. This can yield a loss function representing the difference between the predicted output and the actual value. In the backpropagation process, the loss function can be differentiated with respect to the weights and biases to derive the gradients of each parameter. Using these gradients, the weights and biases can be adjusted using gradient descent or a similar optimization algorithm. These weight and bias updates can be performed in a direction that minimizes the loss function. If this process is repeated, the amount of pipe anomaly data used for training increases, the loss function is minimized, and the performance of the deep learning-based anomaly detection algorithm can be improved.

[0093] If the pipe abnormality detection device classifies the pipe condition of a certain section of the pipe as 'blocked' based on the DAS signal (S550), it can perform the same process on the sensing information acquired from the DSS for the section to output the pipe blockage condition.

[0094] That is, referring to Fig. 6, secondarily, the pipe anomaly detection device can detect a distributed deformation signal through an optical fiber and collect a DSS signal that acquires sensing information through an interrogator (S610). In the case of DSS, the position resolution is at the level of 0.1 m, and the pipe anomaly detection device can secondarily diagnose a high-precision blockage location using DSS. In other words, DSS can be used to detect the location when a specific section is completely blocked by slurry, etc.

[0095] The pipe anomaly detection device can determine whether the current accuracy of the deep learning-based anomaly detection algorithm is higher than the first target value (S620). For example, the first target value may be set to the same value (95%) as illustrated in FIGS. 5 and 6 . In this case, in FIG. 6 , the anomaly detection operation may be terminated after steps S640 and S650 following step S620.

[0096] Alternatively, unlike what is illustrated in FIGS. 5 and 6, even if the current precision of the deep learning-based anomaly detection algorithm is lower than the first target value, if the pipe is classified as being clogged based on the output of the deep learning-based anomaly detection algorithm, the operation of FIG. 6 for the DSS signal may be performed. That is, since the current precision of the deep learning-based anomaly detection algorithm is lower than the first target value, the final judgment for the DAS signal is based on the output of the SVM-based anomaly detection algorithm, but whether or not to operate for the DSS signal may be determined based on the output of the deep learning-based anomaly detection algorithm.

[0097] Alternatively, according to another example, the first target value may be set higher for the DSS signal than for the DAS signal. For example, the first target value may be set to 90% for the DAS signal and 95% for the DSS signal. In this case, the operation of FIG. 5 described above may be performed primarily until the accuracy of the deep learning-based anomaly detection algorithm for the DAS signal exceeds 90%. Accordingly, when the accuracy of the deep learning-based anomaly detection algorithm exceeds 90%, the output value of the deep learning-based anomaly detection algorithm is used to determine whether a pipe abnormality has occurred. If the pipe is determined to be clogged, anomaly detection may be performed secondarily for the DSS signal. In this case, the operation of FIG. 6 below may be performed until the accuracy of the deep learning-based anomaly detection algorithm exceeds 95%.

[0098] If the precision of the deep learning-based anomaly detection algorithm is higher than the first target value, the pipe anomaly detection device can determine the deep learning-based anomaly detection algorithm as the anomaly detection algorithm for deriving a conclusion value. That is, if the precision of the deep learning-based anomaly detection algorithm is higher than the first target value, the pipe anomaly detection device can determine whether there is an anomaly in the pipe based on the output value of the deep learning-based anomaly detection algorithm for the DSS signal (S640).

[0099] If the precision of the deep learning-based anomaly detection algorithm is lower than the first target value, the pipe anomaly detection device may determine the SVM-based anomaly detection algorithm as the anomaly detection algorithm from which to derive a conclusion. That is, if the precision of the deep learning-based anomaly detection algorithm is lower than the first target value, the pipe anomaly detection device may determine whether or not there is an anomaly in the pipe based on the output value of the SVM-based anomaly detection algorithm (S630).

[0100] The pipe anomaly detection device can output information to the user via a user interface regarding the presence of an abnormality in the pipe. If the pipe anomaly detection device determines that an abnormality has occurred in the pipe based on a DSS signal, it can output an alarm for a pipe blockage.

[0101] As described above, the pipe anomaly detection device can receive and store an answer value for the presence or absence of a pipe anomaly through the user interface (S660). That is, the device can receive information on the actual condition of the pipe corresponding to the output value derived through the anomaly detection algorithm. The pipe anomaly detection device can determine whether the sensing information used to derive the result value corresponds to pipe anomaly data based on the answer value corresponding to the actual condition of the pipe. The pipe anomaly detection device can store data identified as pipe anomaly data and use it for future learning.

[0102] The pipe anomaly detection device can train deep learning-based and SVM-based anomaly detection algorithms by reflecting the correct answer. In other words, with less bias in the anomaly data, the output value prediction for the pipe anomaly data and the learning process that reflects the predicted results can be repeated each time a pipe anomaly occurs. Accordingly, as the bias in the anomaly data is reduced, the accuracy of the deep learning-based and SVM-based anomaly detection algorithms can be improved.

[0103] The pipe anomaly detection device can train an SVM-based anomaly detection algorithm (S670) and a deep learning-based anomaly detection algorithm (S680) through backpropagation. In this case, backpropagation can be used to train the neural network model by adjusting its weights and biases to minimize the loss function.

[0104] The pipe anomaly detection device can distribute the weights of the deep learning-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is higher than the first target value (S650). When the biased state of the pipe anomaly data is resolved and the performance of the deep learning-based anomaly detection algorithm improves to a target precision or higher, the deep learning-based anomaly detection algorithm model is distributed and the presence or absence of anomalies in the pipe can be determined through the algorithm. Accordingly, the deep learning-based anomaly detection algorithm can be applied immediately even when the pipe anomaly data is biased.

[0105] In the above, it has been described that secondary diagnosis based on the DSS signal is performed sequentially after the primary diagnosis based on the DAS signal. However, this is only an example and is not limited thereto. If necessary, the diagnosis of the DAS signal and the DSS signal may be performed independently.

[0106] This approach can minimize process downtime due to piping abnormalities by recognizing the condition of the pipes and suggesting actions to managers. Furthermore, by providing a step-by-step learning program that addresses data bias and accelerates the implementation of anomaly detection algorithms, the cost of losses resulting from process delays can be reduced.

[0107] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will readily appreciate that the present invention can be readily modified into other specific forms without altering the technical spirit or essential characteristics of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single entity may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined manner.

[0108] The scope of the present invention is indicated by the claims described below rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and their equivalent concepts should be interpreted as being included in the scope of the present invention.

[0109] The above-described methods and / or various embodiments may be realized by digital electronic circuits, computer hardware, firmware, software, and / or a combination thereof. Various embodiments of the present disclosure may be implemented as a computer program that is executed by a data processing device, for example, one or more programmable processors and / or one or more computing devices, or stored on a computer-readable recording medium and / or a computer-readable recording medium. The above-described computer program may be written in any form of programming language, including a compiled language or an interpreted language, and may be distributed in any form, such as a standalone program, a module, a subroutine, etc. The computer program may be distributed through a single computing device, multiple computing devices connected through the same network, and / or multiple computing devices distributed to be connected through multiple different networks.

[0110] The methods and / or various embodiments described above may be performed by one or more processors configured to execute one or more computer programs that process, store, and / or manage any function, function, etc. by operating on the basis of input data or generating output data. For example, the methods and / or various embodiments of the present disclosure may be performed by special purpose logic circuits such as Field Programmable Gate Arrays (FPGAs) or Application Specific Integrated Circuits (ASICs), and an apparatus and / or system for performing the methods and / or embodiments of the present disclosure may be implemented as special purpose logic circuits such as FPGAs or ASICs.

[0111] The one or more processors executing the computer program may include a general-purpose or special-purpose microprocessor and / or one or more processors of any type of digital computing device. The processor may receive instructions and / or data from each of read-only memory and random-access memory, or may receive instructions and / or data from the read-only memory and the random-access memory. In the present disclosure, components of a computing device performing the methods and / or embodiments may include one or more processors for executing instructions, and one or more memory devices for storing instructions and / or data.

[0112] According to one embodiment, the computing device can transmit and receive data to and from one or more mass storage devices for storing data. For example, the computing device can receive and / or transfer data from a magnetic disc or an optical disc. A computer-readable storage medium suitable for storing instructions and / or data associated with a computer program may include, but is not limited to, any form of non-volatile memory, including semiconductor memory devices such as Erasable Programmable Read-Only Memory (EPROM), Electrically Erasable PROM (EEPROM), flash memory devices, and the like. For example, the computer-readable storage medium may include a magnetic disk such as an internal hard disk or a removable disk, a magneto-optical disk, a CD-ROM, and a DVD-ROM disk.

[0113] To provide interaction with a user, a computing device may include, but is not limited to, a display device (e.g., a cathode ray tube (CRT), a liquid crystal display (LCD), etc.) for providing or displaying information to the user, and a pointing device (e.g., a keyboard, a mouse, a trackball, etc.) for allowing the user to provide input and / or commands to the computing device. That is, the computing device may further include any other types of devices for providing interaction with the user. For example, the computing device may provide any form of sensory feedback to the user, including visual feedback, auditory feedback, and / or tactile feedback, for interaction with the user. In this regard, the user may provide input to the computing device through various gestures, such as visual, vocal, or motion.

[0114] In the present disclosure, various embodiments may be implemented in a computing system that includes a backend component (e.g., a data server), a middleware component (e.g., an application server), and / or a front-end component. In this case, the components may be interconnected via any form or medium of digital data communication, such as a communications network. For example, the communications network may include a Local Area Network (LAN), a Wide Area Network (WAN), and the like.

[0115] A computing device based on the present embodiments may be implemented using hardware and / or software configured to interact with a user, including a user device, a user interface (UI) device, a user terminal, or a client device. For example, the computing device may include a portable computing device such as a laptop computer. Additionally or alternatively, the computing device may include, but is not limited to, Personal Digital Assistants (PDAs), tablet PCs, game consoles, wearable devices, Internet of Things (IoT) devices, virtual reality (VR) devices, augmented reality (AR) devices, and the like. The computing device may further include other types of devices configured to interact with a user. In addition, the computing device may include a portable communication device (e.g., a mobile phone, a smart phone, a wireless cellular phone, etc.) suitable for wireless communication over a network such as a mobile communication network. The computing device may be configured to communicate wirelessly with a network server using wireless communication technologies and / or protocols, such as Radio Frequency (RF), Microwave Frequency (MWF), and / or Infrared Ray Frequency (IRF).

[0116] The above description is merely an illustrative example of the technical idea of ​​the present disclosure, and those skilled in the art to which the present disclosure pertains will appreciate that various modifications and variations can be made without departing from the essential characteristics of the technical idea of ​​the present disclosure. In addition, the present embodiments are not intended to limit the technical idea of ​​the present disclosure but rather to explain it, and therefore the scope of the technical idea of ​​the present disclosure is not limited by these embodiments. The scope of protection of the present disclosure should be interpreted by the claims below, and all technical ideas within a scope equivalent thereto should be interpreted as being included within the scope of the rights of the present disclosure.

[0117]

[0118] CROSS-REFERENCE TO RELATED APPLICATION

[0119] This patent application claims priority under 35 USC § 119(a) to Korean Patent Application No. 10-2023-0183265, filed December 15, 2023, the entire contents of which are incorporated herein by reference. Furthermore, this patent application claims priority in countries other than the United States for the same reasons, the entire contents of which are incorporated herein by reference.

Claims

1. In a device for detecting abnormalities in pipes, A sensing unit including a plurality of sensors that detect scattered signals through optical fibers attached to one side of a pipe to obtain sensing information; and A control unit that determines one anomaly detection algorithm among multiple anomaly detection algorithms based on the learning completion level, and determines whether the pipe is abnormal based on an output value of the determined anomaly detection algorithm derived by applying the sensing information as an input value; A pipe anomaly detection device comprising:

2. In paragraph 1, The above sensing part, A pipe anomaly detection device including a Distributed Acoustic Sensing (DAS) that detects an acoustic signal distributed through the optical fiber to obtain sensing information and a Distributed Strain Sensing (DSS) that detects a strain signal distributed through the optical fiber to obtain sensing information.

3. In paragraph 1, The above multiple anomaly detection algorithms are: A pipe anomaly detection device including an SVM (Support Vector Machine)-based anomaly detection algorithm and a deep learning-based anomaly detection algorithm.

4. In paragraph 3, The above control unit, A pipe anomaly detection device that determines the completion of learning based on the accuracy of a deep learning-based anomaly detection algorithm.

5. In paragraph 4, The above control unit, A pipe anomaly detection device that determines whether there is an anomaly in the pipe based on the output value of the deep learning-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is higher than the first target value.

6. In paragraph 5, The above control unit, A pipe anomaly detection device that determines whether there is an anomaly in the pipe based on the output value of the SVM-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is lower than or equal to the first target value.

7. In paragraph 3, The above control unit, A pipe abnormality detection device that receives an answer value regarding whether or not the pipe is abnormal through a user interface section.

8. In paragraph 7, The above control unit, A pipe anomaly detection device that performs learning of the SVM-based anomaly detection algorithm and the deep learning-based anomaly detection algorithm by reflecting the above correct answer value.

9. In paragraph 5, The above control unit, A pipe anomaly detection device that distributes weights of the deep learning-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is higher than the first target value.

10. In a method for detecting anomalies in pipes, A step of acquiring sensing information by detecting a signal distributed through an optical fiber attached to one side of a pipe through a plurality of sensors; A step of determining one anomaly detection algorithm among multiple anomaly detection algorithms based on the learning completion level; and A step of determining whether there is an abnormality in the pipe based on the output value of the determined abnormality detection algorithm derived by applying the sensing information as an input value; A method for detecting pipe abnormalities, comprising:

11. In Article 10, The above multiple sensors are, A method for detecting anomalies in a pipe, comprising: Distributed Acoustic Sensing (DAS) for detecting acoustic signals distributed through the optical fiber to obtain sensing information; and Distributed Strain Sensing (DSS) for detecting strain signals distributed through the optical fiber to obtain sensing information.

12. In paragraph 10, The above multiple anomaly detection algorithms are: A method for detecting anomalies in a pipeline, comprising an anomaly detection algorithm based on a support vector machine (SVM) and an anomaly detection algorithm based on deep learning.

13. In paragraph 12, The step of determining the above one anomaly detection algorithm is: A method for detecting pipe anomalies by judging the completion of learning based on the accuracy of a deep learning anomaly detection algorithm.

14. In paragraph 13, The steps for determining whether the above pipe is abnormal are: A pipe anomaly detection method for determining whether the pipe is abnormal based on the output value of the deep learning-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is higher than the first target value.

15. In paragraph 14, The steps for determining whether the above pipe is abnormal are: A pipe anomaly detection method for determining whether the pipe is abnormal based on the output value of the SVM-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is lower than or equal to the first target value.

16. In paragraph 12, A method for detecting a pipe abnormality, further comprising: a step of receiving an answer value regarding whether or not the pipe is abnormal through a user interface unit; 17. In paragraph 16, A method for detecting anomalies in a pipe, further comprising: a step of performing learning of the SVM-based anomaly detection algorithm and the deep learning-based anomaly detection algorithm by reflecting the above correct answer value; 18. In paragraph 14, A method for detecting an anomaly in a pipe, further comprising: a step of distributing weights of the deep learning-based anomaly detection algorithm when the precision of the deep learning-based anomaly detection algorithm is higher than the first target value;

Images