A Smart Inspection Fault Diagnosis Method and System Using Six-Dimensional Judgment

Abstract

This invention discloses an intelligent inspection fault diagnosis method and system employing six-dimensional judgment, belonging to the field of intelligent inspection technology for oilfield gathering and transportation. The data access module collects information from field sensors deployed at different stations during the oilfield gathering and transportation process. It performs protocol parsing, time alignment, and semantic mapping on the collected data, encapsulating the data collected by single-point sensors during the oilfield gathering and transportation process into a unified measurement point object (UPO). A standard data structure is constructed, including device ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanistic parameters, and upstream and downstream pointers. This invention's intelligent inspection fault diagnosis method and system employing six-dimensional judgment, by integrating multi-dimensional judgment, intelligent reasoning, and distributed collaborative fault diagnosis, improves the accuracy of anomaly identification, the efficiency of root cause localization, and the scientific nature of control decisions, providing strong support for the safe and stable operation of oilfield gathering and transportation systems.

Description

Technical Field

[0001] This invention belongs to the field of intelligent inspection technology for oilfield gathering and transportation, specifically an intelligent inspection fault diagnosis method and system that uses six-dimensional judgment. Background Technology

[0002] In oilfield gathering and transportation systems, the stable operation of key facilities such as pump sets, pipelines, and water mixing equipment is directly related to crude oil extraction efficiency and production safety. With the accelerated advancement of "less manned and unmanned inspection" in the intelligent construction of oil and gas fields, station operation faces multiple challenges such as complex equipment, harsh environment, fragmented data, and high response requirements.

[0003] Current diagnostic methods largely rely on single threshold judgments, identifying anomalies solely by monitoring whether parameters exceed preset ranges. This approach struggles to handle complex situations such as fluctuating operating conditions and interference signals, exhibits weak anti-disturbance capabilities, and is prone to misjudgments or missed diagnoses. Consequently, issues such as pump vibration deterioration and pipeline leaks cannot be predicted in advance, creating potential safety hazards. Furthermore, the root cause localization lacks a systematic approach, making it difficult to quickly trace the fault chain when an anomaly occurs. It often requires manual investigation based on experience, which is not only time-consuming and labor-intensive but may also lead to inaccurate root cause judgments due to differences in experience, affecting the effectiveness of subsequent control decisions. Therefore, these methods suffer from low practicality and functionality. Summary of the Invention

[0004] This invention aims to address the technical bottlenecks in high-risk, remote, and unmanned scenarios such as oilfield gathering and transportation, particularly in areas such as multi-source data fusion, accurate anomaly identification, root cause intelligent reasoning, predictive maintenance, and coordinated energy consumption and safety control. To this end, this invention proposes an intelligent inspection fault diagnosis method and system employing six-dimensional judgment, improving detection and processing methods to solve the aforementioned technical problems.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] An intelligent inspection fault diagnosis system employing six-dimensional judgment includes a data access module, an edge analysis module, an intelligent inference module, and an execution control model.

[0007] The data access module collects field sensor information deployed in equipment at different stations during the oilfield gathering and transportation process, and performs protocol parsing, time alignment, and semantic mapping on the collected data. It encapsulates the single-point sensor data collected during the oilfield gathering and transportation process into a unified measurement point object (UPO) and constructs a standard data structure that includes equipment ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters, and upstream and downstream pointers.

[0008] The edge analysis module dynamically constructs a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. It performs a six-dimensional judgment score calculation on the constructed single-point model, outputs a six-dimensional anomaly score vector, and performs Bayesian posterior update and logical voting fusion on the six-dimensional score vector to output anomaly level label.

[0009] The intelligent reasoning module maps abnormal measurement points to knowledge graph entities based on the anomaly level labels of the single-point model. It calculates the fault path score through causal chain weights, upstream and downstream influence paths and mechanism residual information, and outputs the Top-N root cause sequence using a reinforcement learning ranking model to form a priority list of problems to be dealt with. Combined with the current on-site operating conditions and operating goals of the equipment, it outputs the adjustment results.

[0010] The execution control module sends control commands to DCS, PLC or intelligent actuators through standard industrial protocols based on the output adjustment results, adjusts and controls abnormal equipment, and completes the closed-loop response to abnormal states.

[0011] Furthermore, the data access module collects information from field sensors deployed at different stations during the oilfield gathering and transportation process, and performs protocol parsing, time alignment, and semantic mapping on the collected data, specifically including the following steps:

[0012] For sensors deployed at different stations during the oilfield gathering and transportation process, data on flow rate, pressure, temperature, vibration, and electrical quantities are collected. Protocol parsing and matching are performed on key features of the initial sensor data streams to calculate the protocol matching score. ,in The weights represent the h-th feature. This represents the matching degree of the h-th feature; the value is 1 when there is a match and 0 when there is no match. This indicates that it belongs to the corresponding protocol;

[0013] A timestamp is added to the data collected by each sensor, the sensor data is time-aligned using interpolation, and semantic mapping is performed based on the sensor type and the meaning of the collected data.

[0014] Furthermore, the process of encapsulating single-point sensor data collected during oilfield gathering and transportation into a unified measurement point object (UPO) and constructing a standard data structure that includes design attributes, real-time values, historical sequences, alarm thresholds, mechanistic parameters, and upstream and downstream pointers specifically includes the following steps:

[0015] Define the structure of the Unified Measurement Point Object (UPO), which includes device ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters, and upstream and downstream pointers;

[0016] The data, after protocol parsing, time alignment, and semantic mapping, is encapsulated into UPO to construct a standard data structure. ,in This is a combination code of the device number and the sensor type at the measuring point. Represents the real-time value collected. Store a historical sequence of the N most recent real-time values ​​for a time window. This represents the rated value in the design attributes of the current measuring point object. For high reporting threshold and low reporting threshold, These are the upstream and downstream pointers, i.e., the IDs of the predecessor and successor nodes. These represent the parameters of the mechanism equation.

[0017] Furthermore, the edge analysis module dynamically constructs a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists, specifically including the following steps:

[0018] Based on the Unified Measurement Point Object (UPO) encapsulated by the data access module, a single-point model is dynamically constructed from the standard data structure, and the static parameters in the UPO are constructed into a static attribute sub-table, including design attributes and alarm thresholds.

[0019] By using real-time values ​​and historical sequence data in UPO objects, a dynamic status table is constructed. By calculating the statistical characteristics of the data, the trend slope of the measurement points is calculated in real time using a linear regression equation.

[0020] Based on the upstream and downstream association pointers in the UPO object, construct an upstream and downstream linked list, so that each node in the upstream and downstream linked list corresponds to a test point, and the upstream and downstream relationship is represented by pointer connection, that is, other test points associated with the current test point;

[0021] Mechanism parameters are extracted from UPO objects to construct a mechanism parameter sub-table. The static attribute sub-table, dynamic state sub-table, mechanism parameter table, and upstream and downstream linked lists are combined to form a decidable multi-dimensional model structure.

[0022] Furthermore, the step of calculating a six-dimensional judgment score on the constructed single-point model, outputting a six-dimensional anomaly score vector, and then performing Bayesian posterior update and logical voting fusion on the six-dimensional score vector to output anomaly level labels includes the following steps:

[0023] Based on single-point models constructed with different measurement points, a six-dimensional judgment score is calculated for different measurement point models, and a six-dimensional anomaly score vector is output. The six dimensions include design boundary, operational boundary, upstream and downstream coupling, mechanism residual, historical evolution, and high and low alarm thresholds, to form a normalized six-dimensional score vector. ,in Specifically, it includes the following steps:

[0024] Design boundary score Determine the normal range of design parameters through static attribute tables. Based on the current real-time values ​​of the measuring points Calculate the deviation of the design boundary. :

[0025] ;

[0026] Deviation of design boundary Mapped to Design boundary score obtained within the interval ;

[0027] The operational boundaries are determined by the statistical characteristics of historical data in the dynamic status table, based on the mean of the historical data. with standard deviation Obtain operating range ,in The constant is used to calculate the deviation from the running boundary. Obtain the running boundary score ;

[0028] Based on the current measuring point and related upstream and downstream measuring points, calculate the Pearson correlation coefficient between the real-time value of the current measuring point and the real-time values ​​of related upstream and downstream measuring points. ,in For the current measuring point, For other measurement points in the upstream and downstream related measurement point set, based on the normal correlation range Calculate the upstream and downstream coupling deviation. :

[0029] ;

[0030] in, The upstream-downstream coupling score is obtained by averaging the correlation coefficients between the current real-time value of the measuring point and the upstream and downstream measuring points. Based on the parameters in the parameter sub-table of the mechanism equation, the theoretical values ​​of the measuring points are calculated using the mechanism equation. And calculate the residuals. According to the set upper limit of the normal range of residuals Obtain the mechanism residual score ,when hour, ;

[0031] Trend slope obtained from historical data series According to the normal trend slope range The calculation can obtain the deviation from historical evolution. Receive historical evolution score :

[0032] ;

[0033] Based on the high and low reporting thresholds, when If the value falls within the high / low reporting threshold range, then a high / low reporting threshold score is applied. ,when If the report does not fall within the high / low reporting threshold range, then the high / low reporting threshold score is applied. ;

[0034] Based on the anomaly levels: Normal, Attention, Abnormal, and Severe, respectively... The prior probability of each anomaly level is obtained based on historical data statistics. ,and For each anomaly level Calculate the six-dimensional scoring vectors that appear under different anomaly levels. likelihood probability Calculate the posterior probability according to Bayes' theorem. ,in This represents the observed six-dimensional scoring vector. Subsequently, the current measuring point belongs to the abnormal level. The posterior probability is calculated, and when the scores of any four dimensions are greater than 0.7, a response is triggered simultaneously, the anomaly level is determined to be severe, and the anomaly level label is output and entered into the subsequent module.

[0035] Furthermore, the intelligent reasoning module includes the following steps:

[0036] By deploying an embedded Neo4j graph database, a five-element graph of "equipment-measuring point-process-fault-measure" is constructed. Based on the anomaly level labels of the single-point model, abnormal measuring points are mapped to knowledge graph entities. Through causal chain weights, upstream and downstream influence paths, and mechanism residual information, fault path scores are calculated, and a reinforcement learning ranking model is used to output a Top-N root cause sequence, forming a priority list of issues to be addressed. The specific steps are as follows:

[0037] Historical data related to equipment, measuring points, processes, faults, and corrective measures are collected, including basic equipment information, real-time data from measuring points, process flow and parameters, historical fault records, and corresponding control measures. Nodes and relationships are created in an embedded Neo4j graph database to construct a five-element graph of "equipment-measuring point-process-fault-corrective measures." Causal relationships between measuring points, processes, and faults are determined based on historical data, and the importance of these causal relationships is evaluated using the analytic hierarchy process (AHP) to obtain the weight of each causal chain. Based on the system's physical principles and process knowledge, mechanistic equations between measuring points and process parameters are established. ,in The theoretical values ​​representing process parameters, The measured value representing the measuring point, and the actual measured value of the process parameter. Compared with theoretical calculation values The residuals are obtained by performing difference calculations, and then the residuals are allocated to each measuring point to obtain the residual contribution value of each measuring point.

[0038] The identified abnormal measurement points are matched with measurement point nodes in the knowledge graph, mapping the abnormal information to the knowledge graph. Breadth-first search is used to search for all possible paths from the abnormal measurement points to the faulty nodes in the knowledge graph. Combining causal chain weights, upstream and downstream influence paths, and mechanism residual information, a score is calculated for each faulty path. :

[0039] ;

[0040] in, The fault path represents nodes and edges. Representing the Each node to The causal chain weight of each node, represent The residual contribution values ​​of each node are used to output the Top-N root cause sequence using an offline Q-learning architecture reinforcement learning ranking model, forming a priority list of problems to be addressed.

[0041] Furthermore, the intelligent reasoning module, based on the priority list of problems to be addressed, combined with the current on-site equipment conditions and operational goals, outputs adjustment results, including the following steps:

[0042] Each root cause in the root cause list and the failure context information of historical cases are converted into semantic vectors. The text information is then encoded into vectors using the BERT model. Cosine similarity is used to calculate the similarity between the root cause list and the historical cases, where the root cause... The semantic vector is Historical cases The fault context semantic vector is cosine similarity for:

[0043] ;

[0044] Based on similarity threshold Root cause list root cause similarity greater than Historical cases are used to form a set of similar fault scenario cases. Corresponding historical handling paths are extracted from the set of similar fault scenario cases. The current on-site operating information of the equipment is recorded as a state vector. The operating goals of the equipment are clarified, including safety goals, energy consumption goals and response efficiency goals. The historical handling paths are used as actions for reinforcement learning. The merits of each action are evaluated through a reward function. The optimal strategy is obtained through Q-learning to output the adjustment results.

[0045] Furthermore, the execution control module, based on the output adjustment result, sends control commands to the DCS, PLC, or intelligent actuator via a standard industrial protocol to adjust and control the abnormal equipment, completing a closed-loop response to abnormal states, including the following steps:

[0046] For measurement points with a severe abnormality level label, control commands are sent to DCS, PLC or intelligent actuators in conjunction with the upstream and downstream pointers of the measurement point to perform safety interlocks on the measurement point and the upstream and downstream measurement points, and to perform equipment stop and start operations.

[0047] Based on the adjustment results output by the intelligent inference module, the abnormal equipment is adjusted and controlled according to the control instructions in the adjustment results. At the same time, the status feedback data of the measuring points is collected in real time and transmitted to the edge analysis module and the intelligent inference module.

[0048] A smart inspection fault diagnosis method using six-dimensional determination includes the following steps:

[0049] S1. Collect field sensor information deployed in equipment at different stations during the oilfield gathering and transportation process, and perform protocol parsing, time alignment and semantic mapping on the collected data. Encapsulate the single-point sensor data collected during the oilfield gathering and transportation process into a unified measurement point object (UPO), and construct a standard data structure that includes equipment ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters and upstream and downstream pointers.

[0050] S2. Dynamically construct a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. Perform a six-dimensional judgment score calculation on the constructed single-point model, output a six-dimensional anomaly score vector, and perform Bayesian posterior update and logical voting fusion on the six-dimensional score vector to output anomaly level label.

[0051] S3. Based on the anomaly level label of the single-point model, the abnormal measurement points are mapped to knowledge graph entities. Through causal chain weights, upstream and downstream influence paths and mechanism residual information, the fault path score is calculated and the Top-N root cause sequence is output using a reinforcement learning ranking model to form a priority list of problems to be dealt with. Combined with the current equipment site conditions and operating objectives, the adjustment results are output. Based on the output adjustment results, the control commands are sent to DCS, PLC or intelligent actuators through standard industrial protocols to adjust and control the abnormal equipment, completing the closed loop of response to abnormal states.

[0052] Compared with the prior art, the beneficial effects of the present invention are:

[0053] In this invention, a six-dimensional judgment link consisting of "design boundary + operation boundary + upstream and downstream coupling + mechanism residual + historical evolution + high and low reporting threshold" is constructed through the edge analysis module, forming a high-dimensional scoring matrix for each measuring point. Compared with the traditional judgment method based on a single threshold, it has stronger anti-disturbance capability and state discrimination accuracy, and can realize the early identification of oilfield gathering and transportation problems such as pump group vibration deterioration, abnormal water injection, and pipeline leakage, so as to form a more comprehensive risk prevention and control system.

[0054] In this invention, a knowledge graph reasoning engine for fault links is constructed through an intelligent reasoning module. Abnormal measurement points are mapped to fault modes and path ranking is performed by combining causal strength scores. A Top-N root cause sequence is output through reinforcement learning algorithm. It can independently output high-confidence root cause location results. Through reinforcement learning ranking and multi-objective optimization algorithm, risk level, energy consumption cost and response time are automatically evaluated and weighed. Finally, the optimal control command is recommended, which enhances practicality and functionality.

[0055] In this invention, by constructing a distributed collaborative system, the six-dimensional judgment and control logic is completed at the edge to execute the anomaly identification and linkage control chain, which meets the needs of high-risk scenarios such as oilfield well sites, remote unmanned stations, and field mobile operations, and enhances the system's risk resistance capability. Attached Figure Description

[0056] Figure 1 This is a block diagram of an intelligent inspection fault diagnosis system using six-dimensional judgment according to the present invention;

[0057] Figure 2 This is a diagram showing the structure of a single-point model of the UPO object dynamically constructed according to an intelligent inspection fault diagnosis system using six-dimensional judgment, as described in this invention. Detailed Implementation

[0058] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1: As Figure 1 As shown, an intelligent inspection fault diagnosis system employing six-dimensional judgment includes a data access module, an edge analysis module, an intelligent inference module, and an execution control model.

[0060] The data access module collects field sensor information deployed in equipment at different stations during the oilfield gathering and transportation process, and performs protocol parsing, time alignment, and semantic mapping on the collected data. It encapsulates the single-point sensor data collected during the oilfield gathering and transportation process into a unified measurement point object (UPO) and constructs a standard data structure that includes device ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters, and upstream and downstream pointers.

[0061] The data access module collects information from field sensors deployed at different stations during the oilfield gathering and transportation process, and performs protocol parsing, time alignment, and semantic mapping on the collected data. Specifically, this includes the following steps:

[0062] For sensors deployed at different stations during the oilfield gathering and transportation process, data on flow rate, pressure, temperature, vibration, and electrical quantities are collected. Protocol parsing and matching are performed on key features of the initial sensor data streams to calculate the protocol matching score. ,in The weights represent the h-th feature. This represents the matching degree of the h-th feature; the value is 1 when there is a match and 0 when there is no match. This indicates that it belongs to the corresponding protocol;

[0063] It should be noted that, among them The values ​​represent the even distribution of the number of h features. Unequal division can also be made based on the representativeness of features in different protocols, but it is necessary to ensure that the sum of all weights is 1. Key features include frame structure features and transport layer features, specifically including frame start symbol, frame end symbol, frame length range, port number, and message header identifier.

[0064] A timestamp is added to the data collected by each sensor, the sensor data is time-aligned using interpolation, and semantic mapping is performed based on the sensor type and the meaning of the collected data.

[0065] It should be noted that when performing semantic mapping based on sensor type and the meaning of the collected data, the data collected by the corresponding sensor is mapped to the corresponding data according to the sensor type. For example, data from a temperature sensor is identified as "temperature," and data from a flow sensor is identified as "flow." At the same time, device ID, sensor measurement point type, and sensor value are mapped and associated for the sensor data sequence. , ,in Representing the timestamp and data value respectively, for Each time point in the sequence Find the two closest time points in the sequence , making The aligned values ​​are calculated using a linear interpolation formula. .

[0066] The data collected by single-point sensors during the oilfield gathering and transportation process is encapsulated into a unified measurement point object (UPO). A standard data structure is constructed that includes design attributes, real-time values, historical sequences, alarm thresholds, mechanistic parameters, and upstream and downstream pointers. The specific steps include:

[0067] Define the structure of the Unified Measurement Point Object (UPO), which includes device ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters, and upstream and downstream pointers;

[0068] The data, after protocol parsing, time alignment, and semantic mapping, is encapsulated into UPO to construct a standard data structure. ,in This is a combination code of the device number and the sensor type at the measuring point. Represents the real-time value collected. Store a historical sequence of the N most recent real-time values ​​for a time window. This represents the rated value in the design attributes of the current measuring point object. For high reporting threshold and low reporting threshold, These are the upstream and downstream pointers, i.e., the IDs of the predecessor and successor nodes. These represent the parameters of the mechanism equation.

[0069] It should be noted that, The system stores a historical sequence of the N most recent real-time values ​​within a time window, where N is initially 10 but can be adjusted based on actual conditions. The combination of device number and sensor type at the measurement point is unique to distinguish different measurement points. This can be determined through the equipment's design documents or user manual. The high and low reporting thresholds need to be set based on experience, taking into account the equipment characteristics at the measuring points and the actual application requirements. The parameters of the mechanism equation are determined based on physical models or empirical formulas for different measuring points. For example, for a temperature measuring point of a heat exchanger, the parameters of the mechanism equation include the heat exchange coefficient, the specific heat capacity of the fluid, etc.

[0070] Example 2: The edge analysis module dynamically constructs a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. It performs a six-dimensional judgment score calculation on the constructed single-point model, outputting a six-dimensional anomaly score vector. Furthermore, it performs Bayesian posterior update and logical voting fusion on the six-dimensional score vector, outputting anomaly level labels.

[0071] The edge analysis module dynamically constructs a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. Specifically, it includes the following steps:

[0072] Based on the Unified Measurement Point Object (UPO) encapsulated by the data access module, a single-point model is dynamically constructed from the standard data structure, and the static parameters in the UPO are constructed into a static attribute sub-table, including design attributes and alarm thresholds.

[0073] By using real-time values ​​and historical sequence data in UPO objects, a dynamic status table is constructed. By calculating the statistical characteristics of the data, the trend slope of the measurement points is calculated in real time using a linear regression equation.

[0074] Based on the upstream and downstream association pointers in the UPO object, construct an upstream and downstream linked list, so that each node in the upstream and downstream linked list corresponds to a test point, and the upstream and downstream relationship is represented by pointer connection, that is, other test points associated with the current test point;

[0075] Mechanism parameters are extracted from UPO objects to construct a mechanism parameter sub-table. The static attribute sub-table, dynamic state sub-table, mechanism parameter table, and upstream and downstream linked lists are combined to form a decidable multi-dimensional model structure.

[0076] It should be noted that the static attribute table is used to describe the design boundaries of different measuring point devices, such as the range of measurement and the parameter range of normal operation. In the dynamic status table, the statistical characteristics of the data, such as the mean and standard deviation, are calculated by combining historical sequence data. The real-time values ​​are dynamically analyzed based on the 3sigma principle to assist in the subsequent anomaly identification. The mechanism parameter sub-table is used to store the parameters in the physical model or empirical formula related to the measuring point.

[0077] The constructed single-point model is subjected to a six-dimensional judgment score calculation, outputting a six-dimensional anomaly score vector. The six-dimensional score vector is then fused with Bayesian posterior update and logical voting to output anomaly level labels. This process includes the following steps:

[0078] Based on single-point models constructed with different measurement points, a six-dimensional judgment score is calculated for different measurement point models, and a six-dimensional anomaly score vector is output. The six dimensions include design boundary, operational boundary, upstream and downstream coupling, mechanism residual, historical evolution, and high and low alarm thresholds, to form a normalized six-dimensional score vector. ,in Specifically, it includes the following steps:

[0079] Design boundary score Determine the normal range of design parameters through static attribute tables. Based on the current real-time values ​​of the measuring points Calculate the deviation of the design boundary. :

[0080] ;

[0081] Deviation of design boundary Mapped to Design boundary score obtained within the interval ;

[0082] It should be noted that the normal range of design parameters is obtained through the design attributes in the static attribute table. hour, ,when hour, .

[0083] The operational boundaries are determined by the statistical characteristics of historical data in the dynamic status table, based on the mean of the historical data. with standard deviation Obtain operating range ,in The constant is used to calculate the deviation from the running boundary. Obtain the running boundary score ;

[0084] It should be noted that, This is a constant for the operating range, with a value between 1 and 3. It is usually set to 2, but can also be set according to the actual usage to control the size of the operating range.

[0085] Based on the current measuring point and related upstream and downstream measuring points, calculate the Pearson correlation coefficient between the real-time value of the current measuring point and the real-time values ​​of related upstream and downstream measuring points. ,in For the current measuring point, For other measurement points in the upstream and downstream related measurement point set, based on the normal correlation range Calculate the upstream and downstream coupling deviation. :

[0086] ;

[0087] in, The upstream-downstream coupling score is obtained by averaging the correlation coefficients between the current real-time value of the measuring point and the upstream and downstream measuring points. Based on the parameters in the parameter sub-table of the mechanism equation, the theoretical values ​​of the measuring points are calculated using the mechanism equation. And calculate the residuals. According to the set upper limit of the normal range of residuals Obtain the mechanism residual score ,when hour, ;

[0088] Trend slope obtained from historical data series According to the normal trend slope range The calculation can obtain the deviation from historical evolution. Receive historical evolution score :

[0089] ;

[0090] Based on the high and low reporting thresholds, when If the value falls within the high / low reporting threshold range, then a high / low reporting threshold score is applied. ,when If the report does not fall within the high / low reporting threshold range, then the high / low reporting threshold score is applied. ;

[0091] It should be noted that the correlation coefficient range is determined by clustering the correlation coefficients in historical data using a clustering algorithm, dividing the correlation coefficients into different clusters, and selecting the cluster containing most normal data as the normal correlation range. The mechanistic equation needs to be determined based on the type of equipment at different measuring points and the types of relevant parameters involved. For the residuals, multiple measurements and calculations need to be performed on the system during the experiment, and the residual data needs to be recorded to analyze the distribution of the residuals and set an upper limit value. Alternatively, a large amount of simulated data can be generated by using computer simulation, and then the residuals can be calculated and their distribution analyzed. This needs to be set according to the specific actual situation of the system deployment. The slope range needs to be determined based on the physical characteristics, operating requirements, and operating conditions of the system's application environment. For example, in the temperature control system of oilfield gathering and transportation, the rate of temperature rise or fall is limited by the power of heating or cooling equipment. Therefore, the reasonable slope range of temperature change needs to be set according to the power of the relevant equipment. If the slope exceeds this range, it indicates that there is a fault at the current measuring point.

[0092] Based on the anomaly levels: Normal, Attention, Abnormal, and Severe, respectively... The prior probability of each anomaly level is obtained based on historical data statistics. ,and For each anomaly level Calculate the six-dimensional scoring vectors that appear under different anomaly levels. likelihood probability Calculate the posterior probability according to Bayes' theorem. ,in This represents the observed six-dimensional scoring vector. Subsequently, the current measuring point belongs to the abnormal level. The posterior probability is calculated, and when the scores of any four dimensions are greater than 0.7, a response is triggered simultaneously, the anomaly level is determined to be severe, and the anomaly level label is output and entered into the subsequent module.

[0093] It should be noted that, It sums the products of the likelihood probabilities and prior probabilities of all anomaly levels, and normalizes the summation term to ensure that the sum of all posterior probabilities is 1. By constructing dynamic modeling instances at each measuring point, and generating anomaly level results through parallel scoring and Bayesian fusion algorithms, it overcomes the problems of high false alarm rate and serious false negative rate of traditional single-threshold judgment. It achieves high-accuracy identification and early warning of typical oilfield gathering and transportation station faults such as pump vibration exceeding limits, heating anomalies, and water injection anomalies. The method and device structure of Bayesian fusion algorithm outputting anomaly levels are applicable to the status identification of equipment such as gathering and transportation, heating, and water injection.

[0094] Example 3: The intelligent reasoning module maps abnormal measurement points to knowledge graph entities based on the anomaly level labels of the single-point model. It calculates the fault path score through causal chain weights, upstream and downstream influence paths and mechanism residual information, and outputs the Top-N root cause sequence using a reinforcement learning ranking model to form a priority list of problems to be dealt with. Combined with the current equipment site conditions and operating goals, it outputs the adjustment results.

[0095] The intelligent reasoning module includes the following steps:

[0096] By deploying an embedded Neo4j graph database, a five-element graph of "equipment-measuring point-process-fault-measure" is constructed. Based on the anomaly level labels of the single-point model, abnormal measuring points are mapped to knowledge graph entities. Through causal chain weights, upstream and downstream influence paths, and mechanism residual information, fault path scores are calculated, and a reinforcement learning ranking model is used to output a Top-N root cause sequence, forming a priority list of issues to be addressed. The specific steps are as follows:

[0097] Historical data related to equipment, measuring points, processes, faults, and corrective measures are collected, including basic equipment information, real-time data from measuring points, process flow and parameters, historical fault records, and corresponding control measures. Nodes and relationships are created in an embedded Neo4j graph database to construct a five-element graph of "equipment-measuring point-process-fault-corrective measures." Causal relationships between measuring points, processes, and faults are determined based on historical data, and the importance of these causal relationships is evaluated using the analytic hierarchy process (AHP) to obtain the weight of each causal chain. Based on the system's physical principles and process knowledge, mechanistic equations between measuring points and process parameters are established. ,in The theoretical values ​​representing process parameters, The measured value representing the measuring point, and the actual measured value of the process parameter. Compared with theoretical calculation values The residuals are obtained by performing difference calculations, and then the residuals are allocated to each measuring point to obtain the residual contribution value of each measuring point.

[0098] It should be noted that the corresponding control measures include the parameter adjustment range and the adjustment parameters involved. Equipment, measuring points, processes, faults, and measures are treated as different types of nodes, and the relationships between them are treated as edges. For example, there is a "belongs to" relationship between measuring points and equipment, and an "influence" relationship between measuring points and processes. When using the analytic hierarchy process (AHP) to assess the importance of causal relationships, a target layer, a criterion layer, and a solution layer are constructed. For each criterion in the criterion layer, the elements of the solution layer are compared pairwise. The 1-9 scale method is used to determine the proportional scale of their relative importance, and the judgment matrix is ​​obtained through pairwise comparisons. For each judgment matrix, the sum-product method is used to solve for its largest eigenvalue and its corresponding eigenvector. The obtained eigenvector is the weight vector of the solution layer element relative to the criterion. When there are multiple criteria in the criterion layer, the weights of the solution layer elements relative to each criterion need to be summarized, and the combined weights need to be calculated.

[0099] The identified abnormal measurement points are matched with measurement point nodes in the knowledge graph, mapping the abnormal information to the knowledge graph. Breadth-first search is used to search for all possible paths from the abnormal measurement points to the faulty nodes in the knowledge graph. Combining causal chain weights, upstream and downstream influence paths, and mechanism residual information, a score is calculated for each faulty path. :

[0100] ;

[0101] in, The fault path represents nodes and edges. Representing the Each node to The causal chain weight of each node, represent The residual contribution values ​​of each node are used to output the Top-N root cause sequence using an offline Q-learning architecture reinforcement learning ranking model, forming a priority list of problems to be addressed.

[0102] It should be noted that the specific steps of outputting the Top-N root cause sequence using the offline Q-learning architecture reinforcement learning ranking model are as follows: the score of the fault path, the weight of the causal chain, and the upstream and downstream influencing paths are used as the states of the reinforcement learning model; different fault paths are used as the actions of the reinforcement learning model; and the score of the fault path is used as the reward value, with higher scores resulting in larger rewards. The offline Q-learning reinforcement learning algorithm is used to ensure that the action selected in each state can obtain the maximum cumulative reward. Based on the policy learned by the reinforcement learning model, all fault paths are ranked, and the top N fault paths with the highest scores are selected as the root cause sequence, forming a priority list of problems to be addressed.

[0103] Example 4: The intelligent reasoning module, based on the priority list of problems to be addressed, combined with the current on-site equipment conditions and operational goals, outputs adjustment results, including the following steps:

[0104] Each root cause in the root cause list and the failure context information of historical cases are converted into a semantic vector. The text information is then encoded into a vector using the BERT model. Cosine similarity is used to calculate the similarity between the root cause list and the historical cases, where the semantic vector of the root cause is [vector], the failure context semantic vector of the historical cases is [vector], and the cosine similarity is [cosine similarity].

[0105] Based on the similarity threshold, historical cases with a root cause similarity greater than a certain threshold in the root cause list are used to form a set of similar fault scenario cases. Corresponding historical handling paths are extracted from the set of similar fault scenario cases. The current on-site operating information of the equipment is recorded as a state vector. The operating goals of the equipment are defined, including safety goals, energy consumption goals, and response efficiency goals. The historical handling paths are used as actions for reinforcement learning. The merits of each action are evaluated through a reward function. The optimal policy is obtained through Q-learning to output the adjustment result.

[0106] It should be noted that when obtaining the optimal policy through Q-learning to output the regulation result, a multi-objective function F(a)=(S(a),E(a),T(a)) can also be defined. The goal is to optimize safety, energy consumption, and response efficiency simultaneously under certain constraints. The multi-objective optimization algorithm NSGA-II is used to solve for the Pareto optimal solution set. That is, through iterative search, a set of non-dominated solutions is found in the solution space. Based on the reinforcement learning policy and the results of multi-objective optimization, multiple regulation schemes are generated. Each scheme corresponds to a historical disposal path. By controlling the weight ratio of S(a), E(a), and T(a), the comprehensive score of each regulation scheme is calculated according to the reward function. The regulation scheme with the highest comprehensive score is selected as the optimal control command recommendation output.

[0107] The execution control module, based on the output adjustment results, sends control commands to the DCS, PLC, or intelligent actuator via standard industrial protocols to adjust and control abnormal equipment, completing a closed-loop response to abnormal states. This includes the following steps:

[0108] For measurement points with a severe abnormality level label, control commands are sent to DCS, PLC or intelligent actuators in conjunction with the upstream and downstream pointers of the measurement point to perform safety interlocks on the measurement point and the upstream and downstream measurement points, and to perform equipment stop and start operations.

[0109] Based on the adjustment results output by the intelligent inference module, the abnormal equipment is adjusted and controlled according to the control instructions in the adjustment results. At the same time, the status feedback data of the measuring points is collected in real time and transmitted to the edge analysis module and the intelligent inference module.

[0110] It should be noted that after the control module executes the control commands, the system collects equipment status feedback in real time, judges whether the execution result has achieved the expected result, and automatically writes the result back to the edge analysis module and intelligent inference module to update the running boundary, residual threshold and reinforcement model parameters, forming a closed-loop self-learning path. Furthermore, the diagnostic data, root cause reasoning, strategy execution and response feedback of the complete process are structured and archived, which is convenient for subsequent model training sample extraction, event backtracking and rule base evolution. The system returns to the next monitoring cycle to continue running, realizing a complete closed loop of "execution → feedback → correction → evolution". The intelligent actuator interoperability interface supports point-to-point command issuance and status reading with various electric valves, frequency converters, pump start-stop units, burner control modules, etc.

[0111] Example 5: An intelligent inspection fault diagnosis method using six-dimensional determination, comprising the following steps:

[0112] S1. Collect field sensor information deployed in equipment at different stations during the oilfield gathering and transportation process, and perform protocol parsing, time alignment and semantic mapping on the collected data. Encapsulate the single-point sensor data collected during the oilfield gathering and transportation process into a unified measurement point object (UPO), and construct a standard data structure that includes equipment ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters and upstream and downstream pointers.

[0113] S2. Dynamically construct a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. Perform a six-dimensional judgment score calculation on the constructed single-point model, output a six-dimensional anomaly score vector, and perform Bayesian posterior update and logical voting fusion on the six-dimensional score vector to output anomaly level label.

[0114] S3. Based on the anomaly level label of the single-point model, the abnormal measurement points are mapped to knowledge graph entities. Through causal chain weights, upstream and downstream influence paths and mechanism residual information, the fault path score is calculated and the Top-N root cause sequence is output using a reinforcement learning ranking model to form a priority list of problems to be dealt with. Combined with the current equipment site conditions and operating objectives, the adjustment results are output. Based on the output adjustment results, the control commands are sent to DCS, PLC or intelligent actuators through standard industrial protocols to adjust and control the abnormal equipment, completing the closed loop of response to abnormal states.

[0115] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may exist in actual implementation. Modules described as separate components may or may not be physically separated, and components shown as modules may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of the method in this embodiment according to actual needs.

[0116] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An intelligent inspection fault diagnosis system employing six-dimensional judgment, characterized in that: It includes a data access module, an edge analysis module, an intelligent inference module, and an execution control module. The data access module collects field sensor information deployed in equipment at different stations during the oilfield gathering and transportation process, and performs protocol parsing, time alignment, and semantic mapping on the collected data. It encapsulates the single-point sensor data collected during the oilfield gathering and transportation process into a unified measurement point object (UPO) and constructs a standard data structure that includes equipment ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters, and upstream and downstream pointers. The edge analysis module dynamically constructs a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. It performs a six-dimensional judgment score calculation on the constructed single-point model, outputs a six-dimensional score vector, and performs Bayesian posterior update and logical voting fusion on the six-dimensional score vector to output an anomaly level label. The six dimensions include design boundary, operational boundary, upstream and downstream coupling, mechanism residual, historical evolution, and high / low reporting thresholds, forming a normalized six-dimensional scoring vector. ,in Design boundary scoring , To design boundary deviation; Running boundary scoring , Operating boundary deviation; Upstream and downstream coupling score , The deviation of upstream and downstream coupling; Mechanism residual score ,when hour , For residuals, This is the upper limit of the normal range for the residuals. Historical Evolution Rating , Deviation from historical evolution; when If the value falls within the high / low reporting threshold range, then a high / low reporting threshold score is applied. ,when If the report does not fall within the high / low reporting threshold range, then the high / low reporting threshold score is applied. , This is the real-time value at the current measuring point; The intelligent reasoning module maps abnormal measurement points to knowledge graph entities based on the anomaly level labels of the single-point model. It calculates the fault path score through causal chain weights, upstream and downstream influence paths and mechanism residual information, and outputs the Top-N root cause sequence using a reinforcement learning ranking model to form a priority list of problems to be dealt with. Combined with the current on-site operating conditions and operating goals of the equipment, it outputs the adjustment results. The execution control module sends control commands to the DCS or PLC through a standard industrial protocol based on the output adjustment results, adjusts and controls abnormal equipment, and completes the closed-loop response to abnormal states.

2. The intelligent inspection fault diagnosis system using six-dimensional determination according to claim 1, characterized in that: The data access module collects information from field sensors deployed at different stations during the oilfield gathering and transportation process, and performs protocol parsing, time alignment, and semantic mapping on the collected data. Specifically, this includes the following steps: For sensors deployed at different stations during the oilfield gathering and transportation process, data on flow rate, pressure, temperature, vibration, and electrical quantities are collected. Protocol parsing and matching are performed on key features of the initial sensor data streams to calculate the protocol matching score. ,in The weights represent the h-th feature. This represents the matching degree of the h-th feature; the value is 1 when there is a match and 0 when there is no match. This indicates that it belongs to the corresponding protocol; A timestamp is added to the data collected by each sensor, the sensor data is time-aligned using interpolation, and semantic mapping is performed based on the sensor type and the meaning of the collected data.

3. The intelligent inspection fault diagnosis system using six-dimensional determination according to claim 2, characterized in that: The process of encapsulating single-point sensor data collected during oilfield gathering and transportation into a unified measurement point object (UPO) and constructing a standard data structure that includes design attributes, real-time values, historical sequences, alarm thresholds, mechanistic parameters, and upstream and downstream pointers specifically includes the following steps: Define the structure of the Unified Measurement Point Object (UPO), which includes device ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters, and upstream and downstream pointers; The data, after protocol parsing, time alignment, and semantic mapping, is encapsulated into UPO to construct a standard data structure. ,in This is a combination code of the device number and the sensor type at the measuring point. Represents the real-time value collected. Store a historical sequence of the N most recent real-time values ​​for a time window. This represents the rated value in the design attributes of the current measuring point object. For high reporting threshold and low reporting threshold, These are the upstream and downstream pointers, i.e., the IDs of the predecessor and successor nodes. These represent the parameters of the mechanism equation.

4. The intelligent inspection fault diagnosis system using six-dimensional determination according to claim 3, characterized in that: The edge analysis module dynamically constructs a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. Specifically, this includes the following steps: Based on the Unified Measurement Point Object (UPO) encapsulated by the data access module, a single-point model is dynamically constructed from the standard data structure, and the static parameters in the UPO are constructed into a static attribute sub-table, including design attributes and alarm thresholds. By using real-time values ​​and historical sequence data from UPO objects, a dynamic state sub-table is constructed. The trend slope of the measurement points is calculated in real time by calculating the statistical characteristics of the data and using a linear regression equation. Based on the upstream and downstream association pointers in the UPO object, construct an upstream and downstream linked list, so that each node in the upstream and downstream linked list corresponds to a test point, and the upstream and downstream relationship is represented by pointer connection, that is, other test points associated with the current test point; Mechanism parameters are extracted from UPO objects to construct a mechanism parameter table. The static attribute sub-table, dynamic state sub-table, mechanism parameter table and upstream and downstream linked lists are combined to form a decidable multidimensional model structure.

5. The intelligent inspection fault diagnosis system using six-dimensional determination according to claim 4, characterized in that: The process of calculating a six-dimensional score for the constructed single-point model, outputting a six-dimensional score vector, and then performing Bayesian posterior update and logical voting fusion on the six-dimensional score vector to output an anomaly level label includes the following steps: Based on single-point models constructed with different measurement points, a six-dimensional judgment score is calculated for different measurement point models, and a six-dimensional score vector is output. The six dimensions include design boundary, operational boundary, upstream and downstream coupling, mechanism residual, historical evolution, and high and low reporting thresholds, to form a normalized six-dimensional score vector. ,in Specifically, it includes the following steps: Design boundary score Determine the normal range of design parameters through static attribute tables. Based on the current real-time values ​​of the measuring points Calculate the deviation of the design boundary. : ； Deviation of design boundary Mapped to Design boundary score obtained within the interval ; The operational boundaries are determined by the statistical characteristics of historical data in the dynamic status sub-table, based on the mean of the historical data. with standard deviation Obtain operating range ,in The constant is used to calculate the deviation from the running boundary. Obtain the running boundary score ; Based on the current measuring point and related upstream and downstream measuring points, calculate the Pearson correlation coefficient between the real-time value of the current measuring point and the real-time values ​​of related upstream and downstream measuring points. ,in For the current measuring point, For other measurement points in the upstream and downstream related measurement point set, based on the normal correlation range Calculate the upstream and downstream coupling deviation. : ； in, The upstream-downstream coupling score is obtained by averaging the correlation coefficients between the current real-time value of the measuring point and the upstream and downstream measuring points. Based on the parameters in the mechanism parameter table, the theoretical values ​​at the measuring points are calculated using the mechanism equation. And calculate the residuals. According to the set upper limit of the normal range of residuals Obtain the mechanism residual score ,when hour, ; Trend slope obtained from historical data series According to the normal trend slope range The calculation can obtain the deviation from historical evolution. Receive historical evolution score : ； Based on the high and low reporting thresholds, when If the report falls within the high / low reporting threshold range, then a high / low reporting threshold score is applied. ,when If the report does not fall within the high / low reporting threshold range, then the high / low reporting threshold score is applied. ; Based on the anomaly levels: Normal, Attention, Abnormal, and Severe, respectively... The prior probability of each anomaly level is obtained based on historical data statistics. ,and For each anomaly level Calculate the six-dimensional scoring vectors that appear under different anomaly levels. likelihood probability Calculate the posterior probability using Bayes' theorem. ,in This represents the observed six-dimensional scoring vector. Subsequently, the current measuring point belongs to the abnormal level. The posterior probability is calculated, and when the scores of any four dimensions are greater than 0.7, a response is triggered simultaneously, the anomaly level is determined to be severe, and the anomaly level label is output and entered into the subsequent module.

6. The intelligent inspection fault diagnosis system using six-dimensional determination according to claim 1, characterized in that: The intelligent reasoning module includes the following steps: By deploying an embedded Neo4j graph database, a five-element graph of "equipment-measuring point-process-fault-measure" is constructed. Based on the anomaly level labels of the single-point model, abnormal measuring points are mapped to knowledge graph entities. Through causal chain weights, upstream and downstream influence paths, and mechanism residual information, fault path scores are calculated, and a reinforcement learning ranking model is used to output a Top-N root cause sequence, forming a priority list of issues to be addressed. The specific steps are as follows: Historical data related to equipment, measuring points, processes, faults, and corrective measures are collected, including basic equipment information, real-time data from measuring points, process flow and parameters, historical fault records, and corresponding control measures. Nodes and relationships are created in an embedded Neo4j graph database to construct a five-element graph of "equipment-measuring point-process-fault-corrective measures." Causal relationships between measuring points, processes, and faults are determined based on historical data, and the importance of these causal relationships is evaluated using the analytic hierarchy process (AHP) to obtain the weight of each causal chain. Based on the system's physical principles and process knowledge, mechanistic equations between measuring points and process parameters are established. ,in The theoretical values ​​representing process parameters, The measured value representing the measuring point, and the actual measured value of the process parameter. Compared with theoretical calculation values The residuals are obtained by performing difference calculations, and then the residuals are allocated to each measuring point to obtain the residual contribution value of each measuring point. The identified abnormal measurement points are matched with measurement point nodes in the knowledge graph, mapping the abnormal information to the knowledge graph. Breadth-first search is used to search for all possible paths from the abnormal measurement points to the faulty nodes in the knowledge graph. Combining causal chain weights, upstream and downstream influence paths, and mechanism residual information, a score is calculated for each faulty path. : ； in, The fault path represents nodes and edges. Representing the Each node to The causal chain weight of each node, represent The residual contribution values ​​of each node are used to output the Top-N root cause sequence using an offline Q-learning architecture reinforcement learning ranking model, forming a priority list of problems to be addressed.

7. The intelligent inspection fault diagnosis system using six-dimensional determination according to claim 6, characterized in that: The intelligent reasoning module, based on the priority list of issues to be addressed and combined with the current equipment operating conditions and operational goals, outputs adjustment results, including the following steps: Each root cause in the root cause list and the failure context information of historical cases are converted into semantic vectors. The text information is then encoded into vectors using the BERT model. Cosine similarity is used to calculate the similarity between the root cause list and the historical cases, where the root cause... The semantic vector is Historical cases The fault context semantic vector is cosine similarity for: ； Based on similarity threshold Root cause list root cause similarity greater than Historical cases are used to form a set of similar fault scenario cases. Corresponding historical handling paths are extracted from the set of similar fault scenario cases. The current on-site operating information of the equipment is recorded as a state vector. The operating goals of the equipment are clarified, including safety goals, energy consumption goals and response efficiency goals. The historical handling paths are used as actions for reinforcement learning. The merits of each action are evaluated through a reward function. The optimal strategy is obtained through Q-learning to output the adjustment results.

8. The intelligent inspection fault diagnosis system using six-dimensional determination according to claim 1, characterized in that: The execution control module, based on the output adjustment result, sends control commands to the DCS or PLC via a standard industrial protocol to adjust and control the abnormal equipment, completing a closed-loop response to the abnormal state, including the following steps: For measurement points with a severe abnormality level label, control commands are sent to the DCS or PLC in conjunction with the upstream and downstream pointers of the measurement point to perform safety interlocks on the measurement point and its upstream and downstream measurement points, and to perform equipment stop and start operations. Based on the adjustment results output by the intelligent inference module, the abnormal equipment is adjusted and controlled according to the control instructions in the adjustment results. At the same time, the status feedback data of the measuring points is collected in real time and transmitted to the edge analysis module and the intelligent inference module.

9. A method for intelligent inspection fault diagnosis employing six-dimensional judgment, characterized in that, This method employs an intelligent inspection fault diagnosis system with six-dimensional determination as described in any one of claims 1-8, and includes the following steps: S1. Collect field sensor information deployed in equipment at different stations during the oilfield gathering and transportation process, and perform protocol parsing, time alignment and semantic mapping on the collected data. Encapsulate the single-point sensor data collected during the oilfield gathering and transportation process into a unified measurement point object (UPO), and construct a standard data structure that includes equipment ID, design attributes, real-time values, historical sequences, alarm thresholds, mechanism parameters and upstream and downstream pointers. S2. Dynamically construct a single-point model for each UPO object, forming a static attribute sub-table, a dynamic state sub-table, a mechanism parameter table, and upstream and downstream linked lists. Perform a six-dimensional judgment score calculation on the constructed single-point model, output a six-dimensional score vector, and perform Bayesian posterior update and logical voting fusion on the six-dimensional score vector to output an anomaly level label. S3. Based on the anomaly level label of the single-point model, the abnormal measurement points are mapped to knowledge graph entities. Through causal chain weights, upstream and downstream influence paths and mechanism residual information, the fault path score is calculated and the Top-N root cause sequence is output using a reinforcement learning ranking model to form a priority list of problems to be dealt with. Combined with the current on-site operating conditions and operating objectives of the equipment, the adjustment results are output. Based on the output adjustment results, the control commands are sent to the DCS or PLC through the standard industrial protocol to adjust and control the abnormal equipment, thus completing the closed loop of response to the abnormal state.

Images